Qatar Foundation Annual Research Conference Proceedings Volume 2016 Issue 1

We present a mixed hybrid finite element formulation for modelling subsurface flow and transport. The formulation is fully implicit in time and employs tetrahedron elements for the spatial discretization for the subsurface domain. It comprises all the main physics that dictate the flow behaviour for subsurface flow and transport, since it is developed on, and inherits them from the Automatic Differentiation General Purpose Research Simulator (AD-GPRS) of Stanford University Petroleum Research Institute (SUPRI-B).

Traditionally, the finite volume formulation is the method employed for the computation of fluid dynamics and reservoir simulation, thanks to its local conservation of mass and energy, and straight-forward implementation. However, it requires the use of structural grids and fails in handling high anisotropy inside the material properties of the domain. Also, the method is of a low computational order; the computed local solution in the gird is piecewise constant.

Here, we use the mixed hybrid finite element formulation which is of high order and can handle the high anisotropy for the material properties. It solves the momentum and mass balance equations simultaneously, hence the name mixed. This strongly coupled scheme facilitates the use of unstructured grids which are important for modelling the complex geometry of the subsurface reservoirs. The Automatic Differentiation library of AD-GPRS automatically differentiates the computational variables needed for the construction the Jacobian matrix which consists of the momentum and mass balance unknowns and any presence of wells.

We use two types of tetrahedron elements, Raviart Thomas (RT0) and Brezzi-Douglas-Marini (BDM1), low and high order respectively. The RT0 has one momentum equation per interface, and the BDM1 has three momentum equations per interface assuring second-order flux approximation. Therefore, when compared to the finite volume approach where the Jacobian consists of the mass balance and well unknowns only, the mixed hybrid formulation will eventually have a larger Jacobian (a one order of magnitude for the high order element) which is computationally expensive. However, none the less, the formulation converges numerically and physically better than the finite volume approach, as we show.

The full system is solved implicitly in time to account for the non-linear behaviour of the flow and transport at the subsurface level which is highly pressure, volume, and temperature (PVT) dependent. Therefore, we make use of the already robust PVT formulations in AD-GPRS. We present a carbon dioxide (CO2) sequestration case for the Johnson formation and discuss the numerical and computational results.

This is of crucial important for Qatar and the Middle East where effective reservoir modelling and management requires a robust representation of the flow and transport at the subsurface level using state of the art formulations.

In the literature, Wheeler et al. (2010) employ a multipoint flux mixed finite element approach to eliminate the momentum balance equation form the Jacobian and substitute it by the well-established multipoint flux approximation (MPFA) in the mass balance equation. Since it is based on MPFA, it will still suffer from convergence issues where high anisotropy is present in the material properties. They have recently expanded their work to compositional modelling of fluid, Singh & Wheeler (2014), however they solve the system sequentially in time, where in our method we solve the system fully implicit in time. Sun & Firoozabadi (2009) solve the pressure implicitly and the fluid properties explicitly in time by further decoupling the mass balance equations which decreases the physical representation of the non-linear behaviour of the flow and transport at the subsurface level.

Introduction

Pedestrians are vulnerable road users. In Japan, more than one-third of the fatalities in traffic crashes are pedestrians and most accidents occur as the pedestrians cross a road. To evaluate alternative countermeasures effectively, recently traffic simulation is considered as one of the powerful decision support tools (Shahdah et al. 2015). A very important requirement for a reliable utilization of traffic simulations for the safety assessments is the proper representation of road user behaviors at potential conflict areas. Severe conflicts usually occur when road users fail to predict other users’ decisions and properly react to it. The widely varying behaviors and maneuvers of vehicles and pedestrians may lead to misunderstanding their decisions, which can result in severe conflicts. So far, most existing studies assume constant walk speeds for pedestrians and complete obedience to traffic rules when crossing roads as if they are at walkways. However, it is known that pedestrian behave differently at crosswalks compared to other walking facilities such as sidewalks and walkways. Pedestrians tend to walk faster at crosswalks (Montufar et al. 2007). Furthermore, their compliance to traffic signals vary by traffic conditions and other factors (Wang et al. 2011). Although many studies have analyzed pedestrian behavior including speed at crosswalks, most of them are based on the average crossing speed without considering the speed profile of the crossing process and the variations within. Iryo-Asano et al. (2015) observed from empirical data that pedestrians may suddenly and significantly change their speed on crosswalks as a reaction to surrounding conditions. Such speed changes cannot be predicted by drivers, which can lead to safety hazards. A study of the speed change maneuvers is critical for representing the potential collisions in the simulation systems and evaluating the probability and severity of collisions reasonably. The objective of this study is to quantitatively model the pedestrian speed change maneuvers and integrate the model into traffic simulation for assessing traffic safety.

Pedestrian speed change events as critical maneuver Figure 1 shows an observed pedestrian trajectory with sudden speed change. If there is a turning vehicle approaching the conflict area, the driver may behave based on his expectation of pedestrian arrival time to the conflict area. If the pedestrian suddenly changes his/her speed close to the conflict area, drivers will not be able to predict the new arrival time, which might lead to severe conflicts. Figure 1 demonstrates a real observed example of such speed change. The pedestrian suddenly increased his speed at the beginning of the conflict area, which yielded to the early arrival to the conflict area by 2.0 seconds (Tdif) than expected time assuming that the pedestrian will continue with his/her speed. A turning vehicle cannot predict this early arrival if it exists at the same time. Furthermore, these 2 seconds are large in terms of collision avoidance Iryo-Asano et al. (2015) showed that timings and locations of pedestrians speed changes mainly occur at the entrance to the pedestrian-vehicle conflict area and 2) when there is a large gap between pedestrian's current speed and his/her necessary speed to complete crossing before the end of pedestrian flashing green interval. In this study, further in-depth analysis is conducted by combining the pedestrian data and the information of approaching vehicle trajectories to identify the influencing factors on pedestrians’ sudden speed change events. The probability of speed change is quantitatively modeled as functions of the remaining green time, the remaining length to cross, the current walking speed and other related variables.

Simulation integration for safety assessment

The proposed pedestrian maneuver model is implemented into an integrated simulation model by combining it with a comprehensive turning vehicle maneuver model (Dang et al. 2012). The vehicle maneuver model is dedicated to represent probabilistic nature of drivers’ reaction to road geometry and surrounding road users in order to evaluate user behavior upon traffic safety. It produces speed profiles of turning vehicles considering the impacts of geometry (i.e. intersection angles, setback distance of the crosswalks) and the gap between the expected arrival time of the vehicle and that of the pedestrians at the conflict area. The proposed model allows us to study the dependencies and the interactions between pedestrians and turning vehicle s at crosswalks. Using the integrated traffic simulation, pedestrian-vehicle conflicts are generated and surrogate safety measures, such as Post Encroachment Time and the vehicle speeds at conflict points, are estimated. These measures are used to evaluate the probability and severity of pedestrian-vehicle conflicts. To verify the characteristics of the simulated conflicts, estimated and observed surrogate safety measures at a selected signalized crosswalk are compared through statistical tests.

Conclusions

The consideration of sudden speed change behavior of pedestrians in the simulation environment generates more reliable and realistic pedestrian maneuvers and turning vehicle trajectories, which enables more accurate assessment of pedestrian-vehicle conflicts. This enables the assessment of improvements in the signal control settings and the geometric layout of crosswalks towards safer and more efficient operations. Furthermore, the model is useful for the real-time hazardous conflict event detection, which can be applied to the vehicle safety assistance systems.

This research is supported by JSPS KAKENHI Grant No. 15H05534. The authors are grateful to Prof. Hideki Nakamura and Ms. Xin Zhang for providing video survey data.

References

Dang M.T., et al. (2012). Development of a Microscopic Traffic Simulation Model for Safety

Assessment at Signalized Intersections, Transportation Research Record, 2316, pp. 122?131.

Iryo-Asano, M., Alhajyaseen, W., Zhang, X. and Nakamura, H. (2015) Analysis of Pedestrian Speed Change Behavior at Signalized Crosswalks, 2015 Road Safety & Simulation International Conference, October 6th–8th, Orlando, USA.

Montufar, J., Arango, J., Porter, M., and Nakagawa, S. (2007), The Normal Walking Speed of Pedestrians and How Fast They Walk When Crossing The Street, Proceedings of the 86th Annual Meeting of the Transportation Research Board, Washington D. C., USA.

Shahdah U,. et al. (2015), Application of traffic microsimulation for evaluating safety performance of urban signalized intersections, Transportation Research Part C, 60, pp. 96?104.

Wang, W., Guo, H., Gao, Z., and Bubb, H. (2011) Individual Differences of Pedestrian Behaviour in Midblock Crosswalk and Intersection, International Journal of Accident worthiness, Vol. 16, No. 1, pp. 1–9.

People who are Blind or Visually Impaired (BVI) have one goal in common: navigate through unfamiliar indoor environments without the intervention of a human guide. The number of blind people in the world is not accurately known at the present, however, based on the 2010 Global Data on Visual Impairments, World Health Organization, approximately 285 million people are estimated to be visually impaired worldwide: 39 million are blind and 246 have low vision, 90% in developing countries, with 82% of blind people aged 50 and above. Available extrapolated statistics about blindness in some countries in the Middle East show ∼102,618 in Iraq, ∼5,358 in Gaza strip, ∼22,692 in Jordan, ∼9,129 in Kuwait, ∼104,321 in Saudi Arabia, and ∼10,207 in the United Arab Emirates. These statistics reveal the importance of developing a useful, accurate, and easy to use navigation system to help this large population of disabled people in their everyday lives. Various commercial products are available to navigate BVI people in outdoor environments based on the Global Positioning System (GPS) where the receiver must has a clear view of the sky. Indoor geo-location, on the other hand, is much more challenging because objects surrounding the user can block or interfere with the GPS signal.

In this paper, we present a centralized wireless indoor navigation system to aid the BVI. The system is designed not only to accurately locate, track, and navigate the user, but to also find the safest travel path and easily communicate with the BVI. A centralized approach is adopted because of the lack of research in this area. Some proposed navigation systems require users to inconveniently carry heavy navigation devices; some require administrators to install a complex network of sensors throughout a building; and others are simply impractical in practice. The system consists of four major components: 1) Wireless Positioning Subsystem, 2) Visual Indoor Modeling Interface, 3) Guidance and Navigation Subsystem, and 4) Path-Finding Subsystem. The system is designed not only to accurately locate, track, and navigate the user, but to also find the safest travel path and easily communicate with the BVI.

A significant part of the navigation system is the virtual modeling of the building and the design of the path-finding algorithms, which will be the main focus of this research. Ultimately, the proposed system provides the design and building blocks for a fully functional package that can be used to build a complete centralized indoor navigation system, from creating the virtual models for buildings to tracking and interacting with BVI users over the network.

We propose to present a video demonstration of LiveAR at the ARC'16 conference. For this purpose, we have prepared three demo videos, which can be found in the submission files. These video demos show the effectiveness and efficiency of LiveAR running on video streams containing a diverse set of human actions. Additionally, the demo also exhibits important system performance parameters such as latency and resource usage.

LiveAR is a novel system for recognizing human actions, such as running and fighting, in a video stream in real time, backed by a massively-parallel processing (MPP) platform. Although action recognition is a well-studied topic in computer vision, so far most attention has been devoted to improving accuracy, rather than efficiency. To our knowledge, LiveAR is the first that achieves real-time efficiency in action recognition, which can be a key enabler in many important applications, e.g., video surveillance and monitoring over critical infrastructure such as water reservoirs. LiveAR is based on a state-of-the-art method for offline action recognition which obtains high accuracy; its main innovation is to adapt this base solution to run on an elastic MPP platform to achieve real-time speed at an affordable cost.

The main objectives in the design of LiveAR are to (i) minimize redundant computations, (ii) reduce communication costs between nodes in the cloud, (iii) allow a high degree of parallelism and (iv) enable dynamic node additions and removals to match the current workload. LiveAR is based on an enhanced version of Apache Storm. Each video manipulation operation is implemented as a bolt (i.e., logical operator) executed by multiple nodes, while the input frame arrive at the system via a spout (i.e., streaming source). The output of the system is presented on screen using FFmpeg.

Next we briefly explain the main operations in LiveAR. The dense point extraction bolt is a first step for video processing, which has two input streams: the input video frame and the current trajectories. The output of this operator consists of dense points sampled in the video frame that are not already on any of the current trajectories. In particular, LiveAR partitions the frame into different regions, and assigns one region to a dense point evaluator, each running in a separate thread. Then, the sampled coordinates are grouped according to the partitioning, and routed to the corresponding dense point evaluator. Meanwhile, coordinates on current trajectories are similarly grouped by a point dispatcher, and routed accordingly. Such partitioning and routing minimizes network transmissions as each node is only fed the pixels and trajectory points it needs.

The optic flow generation operator is executed by multiple nodes in parallel similarly to the dense point extractor. An additional challenge here is that the generation of optic flows involves (i) comparing two frames at consecutive time instances and (ii) multiple pixels in determining the value of the flow in each coordinate. (i) means that the operator is stateful, i.e., each node must store the previous frame and compare with the current one. Hence, node additions and removals (necessary for elasticity) become non-trivial as a new node does not immediately possess the necessary states to work (i.e., pixels on the previous frame) on its inputs. Regarding (ii), each node cannot simply handle a region in the frame, as is the case in the dense point extractor, as the computation at one coordinates relies on the surrounding pixels. Our solution in LiveAR is to split the frame into overlapping patches; each patch contains a partition of the frame, as well as the pixels surrounding the partition. This design effectively reduces the amount of network transmissions, thus improving system scalability.

Lastly, the trajectory tracking operator involves three inputs: the current trajectories, the dense points detected from the input frame, and the optic flows of the input frame. The main idea of this operator is to “grow” a trajectory, either an existing one or a new one starting at a dense point, by adding one more coordinate computed from the optic flow. Note that it is possible that the optic flow indicates that there is no more coordinate on this trajectory in the input frame, ending the trajectory. The parallelization of this operator is similar to that of the dense point extractor, except that each node is assigned trajectories rather than pixels and coordinates. Grouping of the trajectories is performed according to their last coordinates (or the newly identified dense points for new trajectories).

Alice is a visualization software for introducing computational thinking and programming concepts in the context of creating 3D animations. Our research aims to introduce computational thinking and problem solving skills in the middle schools in Qatar. To make this aim accessible, we have adapted the Alice software for a conservative Middle Eastern culture, developed curricular materials, and provided professional development workshops for teachers and students in the Middle East. There is a trend for countries, to evaluate curriculum from other cultures, and then try to bring the successful curriculum to their own school systems. This culture is a result of societies beginning to realize the importance of education and knowledge. Qatar's efforts towards building knowledge-based society and upgrading their higher education infrastructure are proofs of this realization. The challenge is to recognize that although a strong curriculum is necessary, simply porting a successful curriculum to a different environment is not sufficient to guarantee success. Here we share our attempt to take a tool with associated curriculum that has been very successful in several countries in the West, and apply it in an environment with very different cultures and social values.

The Alice ME project is targeted at middle school (grades 6–8) teachers and students in the Middle East. The overall goal of the project is to adapt the Alice 2 software to the local cultures, develop new instructional materials appropriate for local systems, and test the effectiveness of the Alice approach at the middle school level. The “Alice approach” – using program visualization to teach/learn analytic, logical, and computational thinking, problem solving skills and fundamental programming concepts in the context of animation – remains the same.

In the formative phase of this project, our goal was to understand the environment and local culture and evaluate the opportunities and challenges. The lessons learned in this phase are being used to formulate the future direction of our research. Although the Middle Eastern countries are rapidly modernizing, the desire to maintain traditional norms is strong. For this reason, we compiled two lists of models. One list was of existing models in the Alice gallery that are not appropriate for the Middle Eastern culture. Qatar (and Middle East in general) is a religious society that follows conservative norms in dress for both men and women. The second was a list of models that would be interesting and relevant to the Qatari society. These two lists helped us determine those models that might be modified, removed, or added to the gallery. We found that Qatar is a cultural society with a lot of emphasis on local customs. Local hospitality, religion, and traditional professions like pearl diving, fishing, and police officers have special place in society. We also discovered that people in general and boys in particular, have a special respect for camels and the desert.

We created the curriculum in collaboration with one private school and the Supreme Education Council. Creating artifacts in isolation and expecting the educational systems to adopt them is not a prudent approach. Due to this collaboration, we learned that a majority of existing ICT and computing curriculum is based on step-by-step instructions that students are expected to follow and reproduce. There is a lack of stress on student learning, creativity, and application.

Most ICT teachers in Qatar, both in public and private schools, are trained ICT professionals. At the same time, most of these teachers are not familiar with program visualization tools such as Alice and have not taught fundamental programming concepts in their classes. As a result, the need for professional development workshops is urgent. We have conducted several workshops for teachers to help them use Alice during the pilot study of the curriculum. During these workshops, we focus on two main concepts – learning to use Alice as a tool, and learning to teach computational thinking using Alice.

We have piloted the curriculum, instructional materials, and the new 3-D models in the Alice gallery for middle school students in one private English school and two public Arabic schools. The pilot test involved more than 400 students in the three schools combined. During the pilot testing, we conducted a survey to obtain initial feedback regarding the 3D models from the Middle East gallery (students have access to all models that are part of Alice's core gallery). Through these surveys, we learned that those objects that students use in everyday life were more popular when it came to using models in Alice.

As part of the curriculum to teach computational thinking using Alice as a tool, we have created several artifacts which are made available to the schools. These items include:

Academic plan for the year

Learning out comes for each lecture, PowerPoint presentation, class exercises, and assessment questions

Student textbook – One English book for grade 8, one Arabic textbook for grade 8 and one Arabic textbook for grade 11.

One of the most important skills that is essential for building the future generation is critical thinking. Although, we are currently only looking at the acceptability of the newly created 3-D models and usability of our curriculum and instructional material, we are still curious about the effectiveness of this curriculum in teaching computational thinking. We analyzed the results of an exam conducted at a local school and it was observed that students in grade 7 with Alice based curriculum performed better than those in grade 9 on the same exam. This exam was designed to measure the critical thinking skill in problem solving without any reference to Alice. We hope that this result is directly related to the students' experience with Alice, as it works on making the students think about the problem from different perspectives. We acknowledge that still more formal work needs to be done in order to support our hypothesis.

This academic year (2015–2016), Alice based curriculum is being used in Arabic in six independent schools and in English in four private English schools. There are more than 1400 students currently studying this content.

Due to the universal use of technology and its pervasive connection to the world, organisations have become more exposed to frequent and various threats (Rotvold, 2008).Therefore, organisations today are giving more attention to information security as it has become a vital and challenging issue. Mackay (2013) noted that the significance of information security, particularly information security policies and awareness, is growing due to the increasing use of IT and computerization. Accordingly, information security presents a key role in the internet era of technology. Gordon & Loep (2006) stated that information security involves a group of actions intended to protect information and information systems. It involves software, hardware, physical security and human factors, where each element has its own features. Information security not only secures the organisation's security but the complete infrastructure that enables the information's use. Organisations are facing an increase in daily security breaches, especially information that is more accessible to the public as the threat becomes greater. Therefore security requirements need to be tightened.

Information security policies control employees' behavior as well as securing the use of hardware and software. Organisations benefit from implementing information security policies as it helps them to classify their information assets and define the importance of the information assets to the organisation (Canavan, 2003). Information security policy as a number of principles, regulations, methodologies, procedures and tools created to secure the organisation from threats. Boss and Kirsch (2007) stated that employees' compliance with information security polices has become an important socio-organizational resource. Information security policies are applied in organisations to provide the employees with guidelines to guarantee information security.

Herold (2010) expressed the importance for organisations to have constant training programmes and educational awareness to attain the required result from the implementation of an information security policy. Security experts' emphasise the importance of security awareness programmes and how they improve information security as a whole. Nevertheless, implementing security awareness in organisations is a challenging process as it requires actively interacting with an audience that usually does not know the importance of information security (Manke, 2013). Organisations tend to use advanced security technologies and constantly train their security professionals, while paying little attention on enhancing the security awareness of employees and users. This makes employees and users the weakest link in any organization (Warkentin & Willison, 2009).

In the last ten years, the state of Qatar has witnessed remarkable growth and development of its civilization, having embraced information technology as a base for innovation and success. The country has perceived tremendous improvement in the sectors of health care, education and transport (Al-Malki, 2015). Information technology plays a strategic role in building the country's knowledge based economy. Due to the country's increasing use of internet and being connected to the global environment, Qatar needs to adequately address the global threats arising from the internet. The global role of Qatar in world politics has led Qatar to not just face the traditional threats from hackers, but more malicious performers such as terrorists, organized criminal networks and foreign government spying. Qatar has faced a lot of discomfort with some countries which try to breach the county's security. Qatar Computer Emergency Response Team (Q-CERT) who is responsible of addressing the state's Information Security needs stated “As Qatar's dependence on cyberspace grows, its resiliency and security become even more critical, and hence the needs for a comprehensive approach that addresses this need” (Q-CERT, 2015). Therefore Q-CERT established National Information Assurance policy (NIA), which is an information security policy designed to help both government and private sectors in Qatar, to protect their information and enhance their security. Nevertheless the NIA policy has not been implemented still in any organization in Qatar. This is due to the barriers and challenges of information security in Qatar such as culture and awareness, which make the implementation of information security policies a challenging approach.

As a result, the scope of this research is to investigate information security in Qatar. There are many solutions for information security, some are technical and others are non-technical, such as security policies and information security awareness. This research focusses on enhancing information security through non-technical solutions, in particular information security policy. The aim of this research is to enhance information security in organizations in Qatar by developing a comprehensive Information Security Management System (ISMS) that considers the country-specific and cultural factors of Qatar. ISMS is a combination of policies and frameworks which ensure information security management (Rouse, 2011). This information security management approach is unique to Qatar as it considers Qatar culture and country specific factors. Although there are a lot of international information security policies available, such as ISO27001 but this research shows that sometimes these do not address the security needs particular to the culture of the country. Therefore there was a need to define a unique ISMS approach for Qatar.

To accomplish the aim of this research the following objectives must be achieved.

1. To review literature on information security in general and in Qatar in particular.

2. To review international and local information security standards and policies.

3. To explore the NIA policy in Qatar and compare it with others in the region and internationally.

4. To define problems with implementing information security policies and NIA policy in particular in organisations in Qatar.

5. To provide recommendations for the new version of the NIA policy.

6. To assess the awareness of employees on information security.

7. To assess the information security process in organisations in Qatar.

8. To identify the factors which affect information security in Qatar including culture and country specific factors.

9. To propose an ISMS for Qatari organisations taking into consideration the above factors.

10. To define a process for organisations to maintain the ISMS.

11. To evaluate the effectiveness of the proposed ISMS.

To achieve the aim of this research, different research methodologies, strategies and data collection methods will be used, such as literature review, surveys, interviews and case study. The research undergoes three phases, currently the researcher has completed phase one of the research which analyses the field of information security and highlights the gaps in the literature that can be investigated further in this research. It also examines the country factors that affect information security and the implementation of such information security policies. While undertaking interviews with experts in the field of information technology, information security, culture and law, to identify the situation of Information Security in Qatar, and the factors which might affect the development of Information Security in Qatar including the cultural effect, legal and political issues. In the following two years, the researcher will complete phase two and three of the research. During phase two the researcher will measure the awareness of employees and their knowledge of information security and information security policies in particular. The finding will help the researcher in completing phase three which involves investigating further the NIA policy and a real implementation of ISMS in an organisation in Qatar, and analyses the main findings to finally providing recommendations for improving NIA policy.

In conclusion, the main contribution of this research is to investigate the NIA policy and the challenges facing its implementation, and then define an ISMS process for the policy to assist organisations in Qatar in implementing and maintaining the NIA policy. The research is valuable since it will perform the first real implementation of the NIA policy in an organisation in Qatar taking advantage of the internship the researcher had with ICT. The research will move the policy from paper-based form into a real ISMS system and oversees it in reality in one of the organizations.

Information security, National Information Assurance policy, Information Security Management System, Security Awareness, Information Systems

References

Rotvold, G. (2008). How to Create a Security Culture in Your Organization. Available: http://content.arma.org/IMM/NovDec2008/How_to_Create_a_Security_Culture.aspx. Last accessed 1st Aug 2015.

Manke, S. (2013). The Habits of Highly Successful Security Awareness Programs: A Cross-Company Comparison. Available: http://www.securementem.com/wp-content/uploads/2013/07/Habits_white_paper.pdf. Last accessed 1st Aug 2015.

Al-Malki. (2015). Welcome to Doha, the pearl of the Gulf. Available: http://www.itma-congress-2015.com/Welcome_note_2.html. Last accessed 4th May 2015.

Rouse, M. (2011). Information security management system (ISMS). Available: http://searchsecurity.techtarget.in/definition/information-security-management-system-ISMS. Last accessed 22th Aug 2015.

Q-CERT. (2015). About Q-CERT. Available: http://www.qcert.org/about-q-cert. Last accessed 1st Aug 2015.

Warkentin, M., and Willison, R. (2009). “Behavioral and Policy Issues in Information Systems Security: The Insider Threat,” European Journal of Information Systems (18:2), pp. 101–105.

Mackay, M. (2013). AN EFFECTIVE METHOD FOR INFORMATION SECURITY AWARENESS RAISING INITIATIVES. International Journal of Computer Science & Information Technology. 5 (2), p 63–71.

Gordon, L. A. & Loep, M. P. (2006). Budgeting Process for Information Security Expenditures. Communications of the ACM. 49 (1), p 121–125.

Herold. R (2010). Managing an Information Security and Privacy Awareness and Training Program. New York: CRC Press.

Boss, S., & Kirsch, L. (2007). The Last Line of Defense: Motivating Employees to Follow Corporate Security Guidelines. International Conference on Information Systems. unknown (unknown), p 9–12.

Canavan, S. (2003). An Information Security Policy Development Guide for Large Companies. SANS Institute.

In broadband communications, the long channel delay spread, defined as the duration in time, or samples, over which the channel impulse response (CIR) has significant energy, is too long and results in a highly-frequency-selective channel frequency response. Hence, a long CIR can spread over tens, or even hundreds, of symbol periods and causes impairments in the signals that have passed through such channels. For instance, a large delay spread causes inter-symbol interference (ISI) and inter-carrier interference (ICI) in multi-carrier modulation (MCM). Therefore, long finite impulse response (FIR) equalizers have to be implemented at high sampling rates to avoid performance degradation. However, the implementation of such equalizers is prohibitively expensive as the design complexity of FIR equalizers grows proportional to the square of the number of nonzero taps in the filter. Sparse equalization, where only few nonzero coefficients are employed, is a widely-used technique to reduce complexity at the cost of a tolerable performance loss. Nevertheless, reliably determining the locations of these nonzero coefficients is often very challenging.

In this work, we first propose a general framework that transforms the problem of design of sparse single-input single-output (SISO) and multiple-input multiple-output (MIMO) linear equalizers (LEs) into the problem of sparsest-approximation of a vector in different dictionaries. In addition, we compare several choices of sparsifying dictionaries under this framework. Furthermore, the worst-case coherence of these dictionaries, which determines their sparsifying effectiveness, are analytically and/or numerically evaluated. Second, we extend our framework to accommodate SISO and MIMO non-linear decision-feedback equalizers (DFEs). Similar to the sparse FIR LEs design problem, the design of sparse FIR DFEs can be cast into one of sparse approximation of a vector by a fixed dictionary whose solution can be obtained by using either greedy algorithms, such as Orthogonal Matching Pursuit (OMP), or convex-optimization-based approaches, with the former being more desirable due to its low complexity. Third, we further generalize our sparse design framework to the channel shortening setup. Channel shortening equalizers (CSEs) are used to ensure that the cascade of a long CIR and the CSE is approximately equivalent to a target impulse response (TIR) with much shorter delay spread. Channel shortening is essential for communication systems operating over highly-dispersive broadband channels with large channel delay spread. Fourth, as an application of recent practical interest for power-line communication (PLC) community, we consider channel shortening for the impulse responses of medium-voltage power-lines (MV-PLs) with length of 10 km and 20 km to reduce the cyclic prefix (CP) overhead in orthogonal frequency-division multiplexing (OFDM) and, hence, improves the data rate accordingly. For all design problems, we propose reduced-complexity sparse FIR SISO and MIMO linear and non-linear equalizers by exploiting the asymptotic equivalence of Toeplitz and circulant matrices, where the matrix factorizations involved in our design analysis can be carried out efficiently using the fast Fourier transform (FFT) and inverse FFT with negligible performance loss as the number of filter taps increases.

Finally, the simulation results show that allowing for a little performance loss yields a significant reduction in the number of active filter taps, for all proposed LEs and DFEs design filters, which in turn results in substantial complexity reductions. The simulation results also show that the CIRs of MV-PLs with length of 10 km and 20 km can be shortened to fit within the broadband PLC standards. Additionally, our simulations validate that the sparsifying dictionary with the smallest worst-case coherence results in the sparsest FIR filter design. Furthermore, the numerical results demonstrate the superiority of our proposed approach compared to conventional sparse FIR filters in terms of both performance and computational complexity. Acknowledgment: This work was made possible by grant number NPRP 06-070-2-024 from the Qatar National Research Fund (a member of Qatar Foundation). The statements made herein are solely the responsibility of the authors.

Due to rapid growth in traffic demands and the number of subscribers, the transmit energy consumption becomes critical, both environmentally and economically. Increasing energy efficiency for wireless networks is the main goal for the 5G network research. The research community has proposed promising solutions supporting green communication techniques. However, energy efficiency can also be enhanced in a different way. We can get energy from the renewable sources, which can compensate (totally or partially) the traditional power consumption from the power grid. Energy harvesting has emerged as a promising technique which helps to increase the sustainability of wireless networks. In this paper, we investigate energy efficient antenna selection schemes for a MIMO relay powered by a hybrid energy source (from grid or through RF energy harvesting). We try to utilize the large number of antennas efficiently for both data decoding and energy transfer. Then, we formulate an optimization problem and provide the optimum antenna selection scheme such that the joint power consumption (source and relay power) is minimized while meeting the rate requirements. The problem is categorized as a mixed non-linear integer and non-convex, i.e., prohibitively complex. We propose two special cases of the general problem, Fixed Source Power Antenna Selection (FSP-AS), in which we assume a fixed source power and control the antenna selection only, and All Receive Antenna Selection (AR-AS), in which we choose to turn all receiving antennas ON. Finally, we introduce two less complex heuristics, Decoding Priority Antenna Selection (DP-AS) and Harvesting Priority Antenna Selection (HP-AS). Finally we compare our work with the Generalized Selection Combiner (GSC) scheme used in some previous works.

The main contributions of our work can be summarized as follows:

(1) We introduce the energy harvesting technique as an effective way to improve the energy efficiency by using it as a substitute for the grid energy.

(2) In addition to the transmitted energy, we model the circuit power as an important part of the total energy consumption which can affect the energy efficiency.

(3) We make a possibility to turn each antenna ON or OFF individually, so we can turn off only the antennas we don't need to save the energy as much as possible.

(4) We introduce two special case schemes, each of them care about a special type of energy consumption, FSP-AS scheme cares more about the circuit energy, while the

AR-AS concentrates mainly on the transmitted energy.

(5) We also propose two heuristics to accommodate the complexity of the target problem. We evaluate the performance for the proposed schemes numerically. Our key performance indicator (KPI) is the joint power consumed in both the source and the relay. The simulation results show the gain of our optimal scheme in terms of energy efficiency, which can be up to 80% as compared to solutions proposed in the literature. Our developed heuristics show reasonable performance at small rate with almost no gap with the optimal scheme at higher target rates. In our future work, we will consider modeling more than one source and destination nodes and extend this model to include interference scenarios.

Harvesting of vibration energy from the ambient environment, such as vibrations experienced by bridges due to vehicle movements, wind, earthquakes, has become an essential area of study by many scientists aiming to design new systems which can improve self-powered network sensors in wireless sensor networks (WSN), thus providing a more efficient system that does not require the human involvement.

One of the essential components of WSN systems is the sensor node. It is used to continuously send/receive information to monitor a certain behavior targeted by the application; for example, to monitor bridge infrastructure's health. Sometimes, sensors are programmed and adjusted to send useful data for monitoring 24 hours a day, seven days a week. This configuration harms the sensors' batteries and shortens their lives, since sending/receiving data consumes power and leads to the reduction of the batteries' voltage levels. Due to this fact, energy harvesting is critical to maintaining long-term batteries that can recharge themselves from the available ambient harvested energy and eliminate the need for human involvement in replacing or recharging them in their specified locations in the network.

Recent structural health monitoring systems (SHM), in civil infrastructure environments, have focused heavily on the use of wireless sensor networks (WSNs) due to their efficient use of wireless sensor nodes. Such nodes can be fixed onto any part of the infrastructure, such as bridges, to collect data remotely for monitoring and further processing. However, the drawback of using such sensor networks relies mainly on the finite life-time of their batteries. Due to this problem, the concept of harvesting energy from the ambient environment became more important. Ensuring efficient battery usage would have a great benefit in maximizing overall systems functionality time and ensures efficient use of natural energy resources like solar, wind and vibration energies.

This work aims to study the feasibility of using a piezoelectric vibration energy harvester to extend overall battery life using a single, external, super-capacitor component which is serving as a storage unit for the harvested energy. The methodology followed in this work states the general direction of the flow of energy in a sensor node which can be summarized into the following:

1 Piezoelectric Vibration Energy Harvester: This module was used to convert mechanical energy of the vibrations from the ambient environment to electrical energy.

2 Energy Harvesting Circuit: This circuit is responsible for the power conditioning, enabling the circuit to output energy to the sensors under certain threshold criteria.

3 Energy Storage: This the super-capacitor served to store harvested energy.

4 Energy Management Scheme: The scheme proposed by this work under the energy requirements and constraints of the sensor nodes in order to conserve batteries voltage level to extend sensors' batteries lives.

5 Wireless Sensors Nodes: Each sensor node type has specific energy requirements that must be recognized so that it can be adequately powered and turned on using the harvested energy.

The main contribution of this work is a proposal of an energy management scheme which ensures that the harvested energy being provided to the harvester circuit must be greater than the energy output that is going to be consumed by the sensor. This proposed scheme has proved the feasibility of using impact vibrations for efficient energy harvesting and subsequently increase the battery life time needed to turn on the wireless sensor nodes.

Furthermore, as a future direction of work, to increase the amount of harvested energy, hybrid power sources can be explored by combining more than one energy source from the ambient environment, such as solar and vibration energy.

Critical and public safety operations require real-time communication from the incident area(s) to the distant operations command center going through the evacuation and medical support areas. Data transmitted throughout such type of network is extremely useful for decisions makers and operations' conducting. Therefore, any delay in communication may cause lives' loss. Above all, the existing infrastructure communication systems (PSTN, WiFi, 4/5 G, etc.) can be damaged and is often not available solution. An alternate option is to deploy autonomous tactical network at unpredictable location and time. However, in this context there are many challenge especially how to effectively rout or disseminate the information. In this paper, we present behavior of varied multi-hops routing protocols evaluated in a disaster-simulated scenario with different communication technologies (i.e. WiFi IEEE 802.11; WSN IEEE 802.15.4; WBAN IEEE 802.15.6). Studied routing strategies are classified: Ad hoc proactive and reactive protocols, geographic-based and gradient-based protocols. To be realistic, we have conducted our simulations by considering a Mall in Doha city in the State of Qatar. Moreover, we have generated a mobility trace to model the rescue teams and crowd movements during the disaster. In conclusion, extensive simulations showed that WiFi IEEE 802.11 is the best wireless technology to consider in an emergency urban with the studied protocols. On the other hand, gradient based routing protocol performed much better especially with WBAN IEEE 802.15.6.

Keyword: Tactical Ad-hoc Networks; Public Safety and Emergency; Routing Protocols; IEEE 802.11; IEEE 802.15.4; IEEE 802.15.6; Performance Evaluation

I. Introduction

Public safety is a worldwide governments' concern. It is a special continuous reactive set of studies, operations and actions in order to predict plan and perform a successful reaction in a disaster case. Coupled with the raise of number and variety of disasters, not only the economies and infrastructures are affected, but significant number of human lives. With regards to the emergency response to these disasters, the role of existing Public Safety Network (PSN) infrastructures (e.g. TETRA, LTE) is extremely vital. However, it is anticipated that, during and after the disasters, existing PSN infrastructures can be flawed, oversaturated or completely damaged. Therefore, there is a growing demand for ubiquitous emergency response system that could be easily and rapidly deployed at unpredictable location and time. Wherefore Wearable Body Area Networks (W-BAN) is a relevant candidate that can play a key role to monitor the physiological status of involved workforces and the affected individuals. Composed by small and low-power devices connected to a coordinator, WBAN communication architecture relies on three levels: On-Body (or intra-Body), Body-to-Body (or inter-Body) and off-Body communication networks.

In disaster scenarios, the network connectivity and data is a challenging problem due to the dynamic mobility and harsh environment [1]. It is envisioned that, in case of unavailable or out-of-range network infrastructures, the WBANs coordinators along with WBANs sensors can exploit cooperative and multi-hop body-to-body communications to extend the end-to-end network connectivity. The Opportunistic and Mobility Aware Routing (OMAR) scheme is an on-body routing protocol proposed in one of our earlier works [2].

A realistic mobility model is also a major challenge related to the simulations. To the best knowledge of the authors, no comparable study in disaster context is conducted by considering a realistic disaster mobility pattern.

In this paper, we investigate varied classes of multi-hop Body-to-Body routing protocols using a realistic mobility modeling software provided by Bonn University in Germany [3]. The mobility pattern is exploited by the WSNET simulator as a mobility trace of the nodes moving during and after the disaster. Note here that individuals are considered mobile nodes in the scenario. In the conducted simulations, each iteration, one communication technology configuration is selected (i.e. WiFi IEEE 802.11; WSN IEEE 802.15.4; WBAN IEEE 802.15.6), then simulations are ran with the routing protocols (i.e. proactive, reactive, gradient-based and geographical-based). This strategy provides a vision not only on the behavior of the routing protocols in the disaster context, but evaluates the communication technologies in such case also. For proactive, reactive, gradient-based and geographical-based routing protocols, Optimized Link State Routing protocol version 2 (OLSRv2) [4], Ad-hoc On-Demand Distance Vector (AODVv2) [5], Directed Diffusion (DD) [6] and Greedy Perimeter Stateless Routing (GPSR) [7] protocols are selected respectively

The remainder of this abstract is organized as follows. In section II, we present briefly the disaster scenario considered. In section III, we explain the results of the simulations. Finally, in Section IV, we conclude and discuss perspectives.

II. Landmark disaster scenario

we are investigating a disaster scenario (fire triggering) in the “Landmark” shopping mall. The mobility model used is generated by the BonnMotion tool. We assume that the incident is caused by fire into two different locations. Rescuers are immediately called to intervene. We consider that firefighters are divided into 3 groups of vehicles with 26 firefighters in each group. Medical emergency teams that probably could reach the mall just after the incident, are consisting of 6 ambulances with 5 medical staff in each ambulance (30 rescuers in total).

Civilians could also help rescuers and they are also considered in the mobility trace generation. Injured individuals are transported from the incident areas to the patients waiting for treatment areas to get first aids. Then, they will be transported to the patients clearing areas where they will be either put under observation or evacuated to hospitals by ambulance or helicopter. A tactical command center conducting the operations is represented by WSNET is an event-driven simulator for wireless networks. WSNET is able to design, configure and simulate a whole communication stack from the Physical to the Application Layer. We benefit from these features to vary the payload with the selected MAC and routing layer in each iteration. These combinations provided a deep review on the possible communication architecture to consider in disaster operations. The following section describes the outcome of these extensive simulations.

III. Performance evaluation

The main difference between the disaster and any other scenario is the emergency aspect. All flowing data in the network is considered highly important. The probability of packets that did not reach the destination must be zero. For this reason, our evaluation is regarded to the Packet Reception rate (PRR). Likewise, a delayed packet is such like an unreceived packet. That's why we consider the delay as decisive factor. Similarly, the energy consumption is also observed.

The following table summarizes the obtained results.

In terms of average PRR, WiFi IEEE 802.11 is convincingly better than the two counterparts combined with all the routing protocols. GPSR has a considerable PRR with WBAN IEEE 802.15.6, but the location information of the nodes is considered as known. Regarding the delay, DD is particularly efficient with WiFi and WBAN.

To conclude, DD is an efficient routing protocol to consider in case of Indoor operations, while GPSR is most relevant in Outdoor operations where locations can be obtained from GPS.

IV. Conclusion

Disasters are growing remarkably worldwide. The existing communication infrastructures are not considered a part of the communication recuing system. Consequently, to monitor deployed individuals (rescue teams, injured individuals, etc.) data should be forwarded throughout these individuals (WBANs) to reach a distant command center. In order to evaluate the performance of diverse multi-hop routing protocols, we have conducted extensive simulations on WNSET using a realistic mobility model. The simulations showed that all evaluated routing protocols (i.e.AODVv2, OLSRv2, DD and GPSR) has the best PRR with the WiFi technology. While, DD was found to be the most efficient with the WBAN technology. GPSR also could be considered when the location information is available.

Acknowledgment

The work was supported by NPRP grant #[6-1508-2-616] from the Qatar National Research Fund which is a member of Qatar Foundation. The statements made herein are solely the responsibility of the authors.

References

[1] M. M. Alam, D. B. Arbia, and E. B. Hamida, “Device-to-Device Communication in Wearable Wireless Networks,” 10th CROWNCOM Conf., Apr-2015.

[2] E. B. Hamida, M. M. Alam, M. Maman, and B. Denis, “Short-Term Link Quality Estimation for Opportunistic and Mobility Aware Routing in Wearable Body Sensors Networks,” WIMOB 2014 2014 IEEE 10th Int. Conf. Wirel. Mob. Comput. Netw. Commun. WiMob, pp. 519–526, Oct-2014.

[3] N. Aschenbruck, “BonnMotion: A Mobility Scenario Generation and Analysis Tool.” University of Osnabruuck, Jul-2013.

[4] T. Clausen, C. Dearlove, P. Jacquet, and U. Herberg, “RFC7181: The Optimized Link State Routing Protocol Version 2” Apr-2014.

[5] C. Perkins, S. Ratliff, and J. Dowdell, “Dynamic MANET On-demand (AODVv2) Routing draft-ietf-manet-dymo-26.” Feb-2013.

[6] C. Intanagonwiwat, R. Govindan, and D. Estrin, “Directed diffusion: a scalable and robust communication paradigm for sensor networks,” pp. 56–67, 2000.

[7] B. Karp and H. T. Kung, “GPSR: Greedy Perimeter Stateless Routing for Wireless Networks,” Annu. ACMIEEE Int. Conf. Mob. Comput. Netw. MobiCom 2000, no. 6, 2000.

Background

As the number of handheld devices increases, wireless data traffic is expanding exponentially. With the ever increasing demand of higher data rate, it will be very challenging for the system designers to meet the requirement using limited Radio Frequency (RF) communication spectrum. One of the possible remedies to overcome this problem is the use of freely available visible light spectrum [1].

Introduction

This paper proposes an indoor communication system by utilizing Visible Light Communication (VLC). VLC technology utilizes visible light spectrum (750–380 nm) not only for illumination but with an additive advantage of data communication [2]. Visible Light Communication exploits high frequency switching capabilities of Light Emitting Diodes (LEDs) to transmit data. A receiver generally containing a Photo Diode receives signals from optical source and can easily decode the information being transmitted. In practical systems, usage of CMOS imager as a receiver containing an array of Photo Diodes is preferred over a single Photo Diode. Such receiver will enable multi-target detection and multi-channel communication resulting in more robust transceiver architecture [3].

Method

This work demonstrates a real-time transceiver implementation for Visible Light Communication on FPGA. A Pseudo Noise (PN) sequence is generated that will act as input data for the transmitter. Direct Digital Synthesizer (DSS) is implemented for the generation of carrier signal for modulation purpose [4]. Transmitter utilizes On-Off-Keying (OOK) for modulation of incoming data due to its simplicity [5]. The modulated signal is then converted into analog form using Digital-to-Analog (DAC) converter. An analog driver circuit is connected with digital transmitter which is capable of driving an array of Light Emitting Diodes (LEDs) for data transmission. Block level architecture of VLC Transmitter is shown in Fig. 1.

The receiver architecture uses analog circuitry including Photo Diodes for optical detection and Operational Amplifiers for amplification of received signal. Analog-to-Digital (ADC) conversion is performed before transmitting data back to FPGA for demodulation and data reconstruction. Figure 2 demonstrates the architecture of VLC Receiver.

Results and Conclusion

The system is implemented and tested using Xilinx Spartan 3A series FPGA [6]. Basic transceiver implementation utilizes data rate of 1Mbps with a carrier frequency of 5 MHz. However, in VLC, data rate and carrier frequency directly affects the optical characteristics including color and intensity of LEDs. Therefore, different data rates and modulation frequencies are evaluated for optimum data transmission with minimal effects on optical characteristics of LEDs. System complexity in terms of hardware resources and performance analysis including Bit Error Rate (BER) under varying conditions is also compared.

Results demonstrate that it is feasible to establish a low data rate communication link for indoor applications ranging up to 10 m using commercially available LEDs. Integrating a CMOS imager at receiver end will enable a VLC based Multiple-Input-Multiple-Output (MIMO) communication link that can serve multiple channels, maximizing to 1 channel per pixel [3]. Higher data rates are also achievable by utilizing high data rate modulation techniques (OFDM) at the expense of computational complexity and hardware resource utilization [7].

One of the possible implications of this work could be the implementation of VLC based Indoor positioning and navigation system. It can be a potential benefit for large constructions involved in public interactions including but not limited to hospitals, customer support centers, public facilitation offices, shopping malls and libraries. The system will largely utilize existing infrastructure of indoor illumination with added advantage of data communication.

The study also proposes extension of this work for utilization of VLC in outdoor applications. However, more robust algorithms are required for outdoor communication due to the presence of optical noise and interference caused by weather and atmospheric conditions. Robustness of existing algorithm can be increased by integrating Direct Sequence Spread Spectrum (DSSS) together with OOK for modulation. However, further research work is required to evaluate the performance, complexity and robustness of this system under realistic conditions.

References

[1]Cisco Visual Networking Index, “Global Mobile Data Traffic Forecast Update, 2012–2017,” CISCO, White Paper, Feb. 2013.

[2]Terra, D. Inst. de Telecomun., Univ. de Aveiro, Aveiro, Portugal, Kumar, N. Lourenco, N. Alves, L.N.,” Design, development and performance analysis of DSSS-based transceiver for VLC”, EUROCON - International Conference on Computer as a Tool (EUROCON), 2011 IEEE.

[3]“Image Sensor Communication”. VLCC Consortium.

[4]Xilinx DDS Compiler IP Core “http://www.xilinx.com/products/intellectual property/

dds_compiler.html#documentation”

[5]Nuno Lourenço, Domingos Terra, Navin Kumar, Luis Nero Alves, Rui L Aguiar, “Visible Light Communication System for Outdoor Applications”, 8th IEEE, IET International Symposium on Communication Systems, Networks and Digital Signal Processing

[6]Xilinx Spartan-3A Starter Kit “http://www.xilinx.com/products/boards-and-kits/

hw-spar3a-sk-uni-g.html”

[7]Liane Grobe, Anagnostis Paraskevopoulos, Jonas Hilt, Dominic Schulz, Friedrich Lassak, Florian Hartlieb, Christoph Kottke, Volker Jungnickel, and Klaus-Dieter Langer, “High Speed Visible Light Communication Systems”, IEEE Communications Magazine, December 2013.

One of the characteristics of writing in Modern Standard Arabic (MSA) is that the commonly used orthography is mostly consonantal and does not provide full vocalization of the text. It sometimes includes optional diacritical marks (henceforth, diacritics or vowels).

Arabic script consists of two classes of symbols: letters and diacritics. Letters comprise long vowels such as A, y, w as well as consonants. Diacritics on the other hand comprise short vowels, gemination markers, nunation markers, as well as other markers (such as hamza, the glottal stop which appears in conjunction with a small number of letters, dots on letters, elongation and emphatic markers) which in all, if present, render a more or less exact precise reading of a word. In this study, we are mostly addressing three types of diacritical marks: short vowels, nunation, and shadda (gemination).

Diacritics are extremely useful for text readability and understanding. Their absence in Arabic text adds another layer of lexical and morphological ambiguity. Naturally occurring Arabic text has some percentage of these diacritics present depending on genre and domain. For instance, religious text such as the Quran is fully diacritized to minimize chances of reciting it incorrectly. So are children's educational texts. Classical poetry tends to be diacritized as well. However, news text and other genre are sparsely diacritized (e.g., around 1.5% of tokens in the United Nations Arabic corpus bear at least one diacritic (Diab et al., 2007)).

In general, building models to assign diacritics to each letter in a word requires a large amount of annotated training corpora covering different topics and domains to overcome the sparseness problem. The currently available diacritized MSA corpora are generally limited to the newswire genres (those distributed by the LDC) or religion related texts such as Quran or the Tashkeela corpus. In this paper we present a pilot study where we annotate a sample of non-diacritized text extracted from five different text genres. We explore different annotation strategies where we present the data to the annotator in three modes: basic (only forms with no diacritics), intermediate (basic forms–POS tags), and advanced (a list of forms that is automatically diacritized). We show the impact of the annotation strategy on the annotation quality.

It has been noted in the literature that complete diacritization is not necessary for readability Hermena et al. (2015) as well as for NLP applications, in fact, (Diab et al., 2007) show that full diacritization has a detrimental effect on SMT. Hence, we are interested in discovering the optimal level of diacritization. Accordingly, we explore different levels of diacritization. In this work, we limit our study to two diacritization schemes: FULL and MIN. For FULL, all diacritics are explicitly specified for every word. For MIN, we explore what is a minimum and optimal number of diacritics that needs to be added in order to disambiguate a given word in context and make a sentence easily readable and unambiguous for any NLP application.

We conducted several experiments on a set of sentences that we extracted from five corpora covering different genres. We selected three corpora from the currently available Arabic Treebanks from the Linguistic Data Consortium (LDC). These corpora were chosen because they are fully diacritized and had undergone significant quality control, which will allow us to evaluate the anno tation accuracy as well as our annotators understanding of the task. We select a total of 16,770 words from these corpora for annotation. Three native Arabic annotators with good linguistic background annotated the corpora samples. Diab et al. (2007), define six different diacritization schemes that are inspired by the observation of the relevant naturally occurring diacritics in different texts. We adopt the FULL diacritization scheme, in which all the diacritics should be specified in a word. Annotators were asked to fully diacritize each word.

The text genres were annotated following the different strategies:

- Basic: In this mode, we ask for annotation of words where all diacritics are absent, including the naturally occurring ones. The words are presented in a raw tokenized format to the annotators in context.

- Intermediate: In this mode, we provide the annotator with words along with their POS information. The intuition behind adding POS is to help the annotator disambiguate a word by narrowing down on the diacritization possibilities.

- Advanced: In this mode, the annotation task is formulated as a selection task instead of an editng task. Annotators are provided with a list of automatically diacritized candidates and are asked to choose the correct one, if it appears in the list. Otherwise, if they are not satisfied with the given candidates, they can manually edit the word and add the correct diacritics. This technique is designed in order to reduce annotation time and especially reduce annotator workload. For each word, we generate a list of vowelized candidates using MADAMIRA (Pasha et al., 2014). MADAMIRA is able to achieve a lemmatization accuracy 99.2% and a diacritization accuracy of 86.3%. We present the annotator with the top three candidates suggested by MADAMIRA, when possible. Otherwise, only the available candidates are provided.

We also provided annotators with detailed guidelines, describing our diacritization scheme and specifying how to add diacritics for each annotation strategy. We described the annotation procedure and specified how to deal with borderline cases. We also provided in the guidelines many annotated examples to illustrate the various rules and exceptions.

In order to determine the most optimized annotation setup for the annotators, in terms of speed and efficiency, we test the results obtained following the three annotation strategies. These annotations are all conducted for the FULL scheme. We first calculated the number of words annotated per hour, for each annotator and in each mode. As expected, following the Advanced mode, our three annotators could annotate an average of 618.93 words per hour which is double those annotated in the Basic mode (only 302.14 words). Adding POS tags to the Basic forms, as in the Intermediate mode, does not accelerate the process much. Only − 90 more words are diacritized per hour compared to the basic mode.

Then, we evaluated the Inter-Annotator Agreement (IAA) to quantify the extent to which independent annotators agree on the diacritics chosen for each word. For every text genre, two annotators were asked to annotate independently a sample of 100 words.

We measured the IAA between two annotators by averaging WER (Word Error Rate) over all pairs of words. The higher the WER between two annotations, the lower their agreement. The results obtained show clearly that the Advanced mode is the best strategy to adopt for this diacritization task. It is the less confusing method on all text genres (with WER between 1.56 and 5.58).

We also conducted a preliminary study for a minimum diacritization scheme. This is a diacritization scheme that encodes the most relevant differentiating diacritics to reduce confusability among words that look the same (homographs) when undiacritized but have different readings. Our hypothesis in MIN is that there is an optimal level of diacritization to render a text unambiguous for processing and enhance its readability. We showed the difficulty in defining such a scheme and how subjective this task can be.

Acknowledgement

This publication was made possible by grant NPRP-6-1020-1-199 from the Qatar National Research Fund (a member of the Qatar Foundation).

Introduction

This research aims to enhance the culture of data in education which is in the middle of a major transformation by technology and Big Data Analytics. The core purpose of schools is providing an excellent education to every learner; data can be the leverage of that mission. Big data analytics is the process of examining large sets containing a variety of data types to uncover hidden patterns, unknown correlations, market trends, customer preferences and other useful business information. Valuable lessons can be learnt from other industries when considered in terms of their practicality for public education. Hence, Big Data Analytics, also known as Education Data Mining and Learning Analytics, develop capacity for quantitative research in response to the growing need for evidence-based analysis related to education policy and practice. However, education has been slow to follow the Data Analytics evolution due to difficulties surrounding what data to collect, how to collect those data and what they might mean. Our research works identify, quantify, and measure the qualitative teaching practices, the learning performances, and track the learners' academic progress. Teaching and learning databases are accumulated from quantitative “measures” done through indoors classroom visits within academic institutions, online web access learners' questionnaires answers, paper written statements' analysis of academic exams in mathematics, science, and literacy disciplines, and online elementary grades seizure from written traces of learners' performances within mathematics, science, and literacy exams. The project's data mining strategy will support and develop teachers' expertise, enhance and scaffold students' learning, improve and raise education system's performance. The supervisor expertise will mentor the researcher for information and educational knowledge extraction from collected data. As consequence, the researcher will acquire the wisdom of knowledge use to translate it into more effective training sessions on educational concrete policies.

State-of-the-art

Anne Jorro says: “evaluate is necessarily considering how we will support, advise, exchange, to give recognition to encourage the involvement of the actor giving him the means to act”. PISA report states that many of the world's best-performing education systems have moved from bureaucratic “command and control” environments towards school systems in which the people at the frontline have much more control. Making teaching and learning data available leads to information then knowledge extraction. As advised by PISA report, the effective use of extracted knowledge drives decision making to wisdom. Linda Darling-Hammond and Charles E. Ducommun underscore the important assumption that, undoubtedly, teachers are the fulcrum that has the biggest impact and makes any school initiative leads toward success or failure. Rivkin and al ensure that a teacher's classroom instructional practice is perhaps one of the most important yet least understood factors contributing to teacher effectiveness. As consequence, many classroom observation tools are designed to demystify effective teaching practices. The Classroom Assessment Scoring System is a well-respected classroom climate observational system. The CLASS examines three domains of behaviour including, firstly, emotional support (positive classroom climate, teacher sensitivity, and regard for student perspectives). Secondly, it includes classroom organization (effective behaviour management, productivity, and instructional learning formats). Thirdly, it contains instructional supports (concept development, quality of feedback, and language modelling). The Framework for Teaching Method for Evaluation Classroom Observation is lastly releasing as 2013 edition. It divides the complex activity of teaching into 22 components clustered into four domains of teaching responsibility. This last tool's edition was conceived to respond to the instructional implications of the American Common Core State Standards. Those standards envision, for literacy and mathematics initially, deep engagement by students with important concepts, skills, and perspectives. They emphasize active, rather than passive, learning by students. In all areas, they place a premium on deep conceptual understanding, thinking and reasoning, and the skill of argumentation. Heather Hill from Harvard University, and Deborah Loewenberg Ball from university of Michigan, had developed “Mathematical Quality of Instruction (MQI)”. Irving Hamer is an education consultant and a former deputy superintendent for academics, technology, and innovation for school system.

Objectives

Our project's wider objective is to improve teaching and learning effectiveness within K12 classes by exploiting data mining methods for educational knowledge extraction. The researcher realizes three daily visits to mathematics, science, and literacy courses. Using his interactive educational grid, an average of 250 numerical data were stored as quantified teaching and learning practices for one classroom visit for every teacher. At the same time through, and in parallel with on-field activities, distance interactivity via website is processed. At the beginning and for once, each learner from planned classes to be visited fills an individual questionnaire form for learning style identification. He seizes, on another website form, every elementary grade on each question from his maths, science and literacy exams' answer sheets. Those exams statements were previously analysed and saved by the researcher on website. Averages of 150 numerical data were stored as quantified learning performances for every learner. Meetings at partner University for data analytics and educational knowledge extraction were done followed by meetings at inspectorate headquarters for in-depth data. Then, in partner schools, training sessions were the theatres of constructive reflections and feedbacks on major findings about teaching and learning effectiveness. Those actions were reiterated for months. Each year, averagely the performance of 1000 students and the educational practices of 120 teachers will be specifically and tracked. Within summer's months, workshops, seminars, and an international conference will be organised for stakeholders from educational fields. Thus, among project's actions three specific objectives shall be achieved. First, sufficient data on students' profiles and performances related to educational weaknesses and strengths will be provided. Second, teachers' practices inside classrooms at each partner school will be statistically recorded. Third, a complete data mining centre for educational research will be conceived and cognitively interpreted by researchers' teams then findings are exposed for teachers' reflexive thoughts, and discussions within meetings and training sessions

Research methodology and approach

1.Introduction

Social media gives us instant access to a continuous stream of information generated by users around the world. This enables real-time monitoring of users’ behavior (Abbar et al., 2015), events’ life-cycles (Weng and Lee. 2010), and large-scale analysis of human interactions in general. Social media platforms are also used to propagate influence, spread content, and share information about events happening in real-time. Detecting the location of events directly from user-generated text can be useful in different contexts, such as humanitarian response, detecting the spread of diseases, or monitoring traffic. In this abstract, we define a system that can be used for any of the purposes described above, and illustrate its usefulness with an application for locating traffic-related events (e.g., traffic jams) in Doha.

The goal of this project is to design a system that, given a social-media post describing an event, predicts whether or not the event belongs to a specific category (e.g., traffic accidents) within a specific location (e.g. Doha). If the post is found to belong to the target category, the system proceeds with the detection of all possible mentions of locations (e.g. “Corniche”, “Sports R/A”, “Al Luqta Street.”, etc.), landmarks (“City Center”, “New Al-Rayyan gas station”, etc.), and location expressions (e.g. “On the Corniche between the MIA park and the Souq)”. Finally, the system geo-localizes (i.e. assigns latitude and longitude coordinates) to every location expression used in the description of the event. This makes it useful for placing the different events onto a map; a downstream application will use these coordinates to monitor real-time traffic, and geo-localize traffic-related incidents.

2.System Architecture

In this section we present an overview of our system. We first describe its general “modular” architecture, and then proceed with the description of each module.

2.1. General view

The general view of the system is depicted in Figure 1. The journey starts by listening to some social media platforms (e.g., Twitter, Instagram) to catch relevant social posts (e.g., tweets, check-ins) using a list of handcrafted keywords related to the context of the system (e.g., road traffic). Relevant posts are then pushed through a three-steps pipeline in which we double-check the relevance of the post using an advanced binary classifier (Content Filter). We then extract location names mentioned in the posts if any. Next, we geo-locate the identified locations to their accurate placement on the map. This process allow to filter undesirable posts, and augment the relevant once with precise geo-location coordinates which are finally exposed for consumption via a restful API. We provide below details on each of the aforementioned modules.

Figure 1: Data processing pipeline.

2.2. Content filter

The Content Filter consists of a binary classifier that, given a tweet deemed to be about Doha, decides whether the tweet is a real-time report about traffic in Doha or not. The classifier receives as input tweets that have been tweeted from a location enclosed in a geographic rectangle (or bounding box) that roughly corresponds to Doha, and that contain one or more keywords expected to refer to traffic-related events (e.g., “accident”, “traffic”, “jam”, etc.). The classifier is expected to filter out those tweets that are not real-time reports about traffic (e.g., tweets that mention “jam”’ as a type of food, tweets that complain about the traffic in general, etc.). We build the classifier using supervised learning technology; in other words, a generic learning process learns, from a set of tweets that have been manually marked as being either real-time reports about traffic or not, the characteristics that a new tweet should have in order to be considered a real-time report about traffic. For our project, 1000 tweets have been manually marked for training purposes. When deciding about a new tweet, the classifier looks for “cues” that, in the training phase, have been found to be “discriminative”, i.e., helpful in taking the classification decision. In our project, we used the Stanford Maximum Entropy Classifier (Manning and Klein, 2003) to perform the discriminative training. In order to generate candidate cues, the tweet is preprocessed via a pipeline of natural language analysis tools, including a social-media-specific tokenizer (O'Connor et al., 2010) which splits words, and a rule-based Named-Entity Simplifier which substitutes mentions of local entities by their corresponding meta-categories (for example, it substitutes “@moi_qatar” or “@ashghal” for “government_entity”).

2.3.NLP components

The Location Expression Extractor is a module that identifies (or extracts) location expressions, i.e., natural language expressions that denote locations (e.g., “@ the Slope roundabout”, “right in front of the Lulu Hypermarket”, “on Khalifa”, “at the crossroads of Khalifa and Majlis Al Taawon”, etc.). A location expression can be a complex linguistic object, e.g., “on the Corniche between the MIA and the underpass to the airport”. A key component of the Location Expression Extractor is the Location Named Entity Extractor, i.e., a module that identifies named entities of Location type (e.g. “the Slope roundabout”) or Landmark type (e.g., “the MIA”). For our purposes, a location is any proper name in the Doha street system (e.g., “Corniche”, “TV roundabout”, “Khalifa”, “Khalifa Street”); landmarks are different from locations, since the locations are only functional to the Doha street system, while landmarks have a different purpose (e.g., the MIA is primarily a museum, although its whereabouts may be used as a proxy of a specific location in the Doha street system – i.e., the portion of the Corniche that is right in front of it).

The Location Named Entity Extractor receives as input the set of tweets that have been deemed to be about some traffic-related event in Doha, and returns the same tweet where named entities of type Location or of type Landmark have been marked as such. We generate a Location Named Entity Extractor via (again) supervised learning technology. In our system, we used the Stanford CRF-based Named Entity Recognizer (Finkel et al., 2005) to recognize named entities of type Location or of type Landmark using a set of tweets where such named entities have been manually marked. From these “training” tweets the learning system automatically recognizes the characteristics that a natural language expression should have in order to be considered a named entity of type Location or of type Landmark. Again, the learning system looks for “discriminative” cues, i.e., features in the text that may indicate the presence of one of the sought named entities. To improve the accuracy over tweets, we used a tweet-specific tokenizer (O'Connor et al., 2010), a tweet-specific Part-of-Speech tagger (Owoputi et al., 2013) and an in-house gazetteer of locations related to Qatar.

2.4.Resolving location expression onto the map

Once location entities are extracted using the NLP components, we use the APIs of Google, Bing and Nominatim to request the geographic coordinates of the map location entities into geographic coordinates. Each location entity is geo-coded by the Google Geolocation API, Bing Maps REST API and Nomination gazetteer individually. We use multiple geo-coding sources to increase the robustness of our application, as a single API might fail to retrieve geo-coding data. Given a location entity, the result of the geo-coding retrieval is formatted as a JSON object containing the name of the location entity, its address, and the corresponding geo-coding results from Bing, Google or Nominatim. The geo-coding process is validated by comparing the results of the different services used. We first make sure that the location returned falls within Qatar's bounding box. We then compute the pairwise distance between the different geographic coordinates to ensure their consistency.

2.5.Description of the Restful API

In order to ease the consumption of the relevant geo-located posts and make it possible to integrate these posts in a comprehensive way with other platforms, we have built a Restful API. In the context of our system, this refers to using HTTP verbs (GET, POST, PUT) to retrieve relevant social posts stored by our back-end processing.

Our API exposes two endpoints: Recent and Search. The former endpoint provides an interface to request the latest posts identified by our system. It supports two parameters: Count (maximum number of posts to return) and Language (the language of posts to return i.e., English or Arabic.) The later endpoint enables querying the posts for specific keywords and return only posts matching them. This endpoint supports three parameters: Query (list of keywords), Since (date-time of the oldest post to retrieve), From-To (two date-time parameters to express the time interval of interest.) In the case of a road traffic application, one could request tweets about “accidents” that occurred in West-Bay since the 10th of October.

3.Target: single architecture for multiple applications

Our proposed platform is highly modular (see Figure 1). This guarantees that relatively simple changes in some modules can make the platform relevant to any applicative context where locating user messages on a map is required. For instance, the content classifier – the first filtering element in the pipeline – can be oriented to mobility problems in a city: accident or congestion reporting, road blocking or construction sites, etc. With the suitable classifier, our platform will collect traffic and mobility tweets, and geo-locate them when possible. However, there are many other contexts in which precise location is needed. For instance, in natural disaster management, it is well admitted that people involved in catastrophic events (floods, typhoons, etc.) use social media as a means to create awareness, demand help or medical attention (Imran et al., 2013). Quite often, these messages may contain critical information for relief forces, who may not have enough knowledge of the affected place and/or accurate information of the level of damage in buildings or roads. Often, the task to read, locate on a map and mark is crowd-sourced to volunteers; we foresee that, in such time-constrained situations, our proposed technology would represent an advancement. Likewise, the system may be oriented towards other applications: weather conditions, leisure, etc.

4.System Instantiation

We have instantiated the proposed platform to the problem of road traffic in Doha. Our objective is to sense in real-time the traffic status in the city using social media posts only. Figure 2 shows three widgets of the implemented system. First, the Geo-mapped Tweets Widget shows a Doha map with different markers: the yellow markers symbolize the tweets geo-located by the users, the red markers represent the tweets geo-located by our system; the large markers come from tweets that have an attached photo, while the small markers represent the text-only tweets. Second, the Popular Hashtags Widget illustrates hashtags mentioned by the users, where the large font size shows the most frequent one. Third, the Tweets Widget lists the traffic-related tweets which are collected by our system.

Figure 2: Snapshot of some System's frontend widgets.

5.References

This paper addresses an important problem in the aircraft traffic management caused by the rapid growth of air traffic. The air route traffic control center has to deal with different plans of airlines in which they specify a requested entry time of their aircraft to the airspace. Each flight has to be assigned to a track and a level in order to ensure the Federal Aviation Administration (FAA) safety standards. When two flights are assigned to the same track and level, a minimum separation time has to be ensured. If this condition could not be satisfied, one of the flight will be delayed. This solution is undesirable for many reasons such as missing the connecting flight, decrease in the passengers' satisfaction, etc.

The problem of track-level scheduling can be defined as follows. Given a set of flights, each flight has to be assigned to one track and one level. To ensure the separation time between two flights assigned to the same track and level, it is possible to delay the requested departure time of a flight. The objective is to minimize the overall flight delay.

To deal with this problem, we propose a mixed integer programming formulation to find a flight plan that minimizes the objective function, while ensuring the FAA safety standards. In particular, this model considers an aircraft-dependent separation time: the separation time depends on the type of the aircraft assigned to the same track and level. However, some problems are too large to be solved in a reasonable time with the proposed model using a commercial solver. In this study, we developed a scatter search (SS) to deal with larger instances. SS is an evolutionary heuristic and the feature to be a problem-independent structure. This metaheuristic has been efficiently applied to a variety of optimization problems. Initially, SS starts with a set of solutions (reference set) that is constantly updated through two procedures (solution generation and combination) in the aim to produce high-quality solutions.

In order to assess the quality of the exact method and the scatter search, we carried out an experimental study on a set of instances that are generated from a real case data. This includes small (80 to 120 flights), medium (200 to 220 flights), and large (400 to 420 flights) instances. The mathematical model has been solved using CPLEX 12.6 and the scatter search has been coded using C language under Microsoft Visual Studio v12 environment. The tests were conducted under a Windows 7 machine with an Intel Core i7 and 8 GB of RAM. The model was tested on each instance with 1 hour time limit. The results show that no instances have been solved to optimality. For small instances, the model and the scatter search provide comparable results; however, for medium and large instances, scatter search gives the best results.

This conference was made possible by the UREP award [UREP 13 - 025 - 2 - 010] from the Qatar National Research Fund (a member of The Qatar Foundation).

Simultaneous wireless information and power transfer (SWIPT) is a promising solution to increase the lifetime of wireless nodes and hence alleviate the energy bottleneck of energy constrained wireless networks. To recent days, there are three different designs of SWIPT system which includes integrated SWIPT, closed-loop SWIPT, decoupled SWIPT. Integrated SWIPT is the simplest design where power and information are extracted by the mobile from the same modulated microwave transmitted by a base station (BS). For this scheme, the information transfer (IT) and power transfer (PT) distances are equal. For the closed loop scenario, it splits IT and PT between uplink and downlink wherein PT is in downlink and IT is for dedicated for uplink. The last one is to add additional special base station (BS) called a power beacon PB in which PT and IT are orthogonalized by using different frequency bands or time slots to avoid interference. Therefore, powering a cognitive radio networks through RF energy harvesting can be efficient in terms of spectrum usage and energy limits for wireless networking. The RF energy harvesting technique also is applicable in cooperative networks wherein an energy constrained relay with limited battery depends on external charging mechanism to assist the transmission of source information to the destination. In an effort to further improve spectrum sharing network performance, a number of works has suggested the idea of incorporating the multiple antenna technique into cognitive relaying. In particular, transmit antenna selection with receive maximal ratio combining (TAS/MRC) is adopted as a low complexity and power efficient approach which achieves full transmit/receive diversity.

Since the SUs and PUs share the same frequency band, there will be inevitably interference between the SUs and PUs. Therefore, reducing the effect of PU interference on the performance of secondary receiver is of significance important. Consequently, smart antennas can be employed to mitigate the PU interference. With knowledge of the direction of arrival (DoA), the receive radiation pattern can be shaped to place deep nulls in the directions of some of interfering signals. By doing so, two null-steering algorithms were proposed in the literature i.e., dominant interference reduction algorithm, and adaptive arbitrary interference reduction algorithm. The first algorithm requires perfect predication and statistical ordering of the interference signals instantaneous power, and the later algorithm does not need prior knowledge of the statistical properties of interfering signals. In this work, we limit our analysis to the dominant interference reduction algorithm.

In this work, we consider a dual-hop relaying with amplify-and-forward (AF) scheme where the source, relay, and the destination are equipped with multiple antennas. The relay node is experiencing co-channel interference. The purpose of array processing at the relay is to provide interference cancellation. Therefore, the energy constrained relay collects energy from ambient RF signals, cancel CCI, and then forward the information to the destination. In particular, we provide a comprehensive analysis for the system assuming selection at the source, and the destination. We derive the end-to-end exact and asymptotic outage probability for the proposed system model. A key parameters are also obtained featuring the diversity and coding gains.

The field of plasmonics is continuously attracting research in the area of integrated photonics development, to highly integrate photonic components, devices and detectors in a single photonic chip, just as electronic chips containing many electronic components. The interesting properties that plasmonics offer include the electromagnetic fields enhancement and the confinement of the propagating surface plasmon polaritons to sub-100 nm size features at metal dielectric interfaces. Thereby, the field of plasmonics is very promising for minimizing the photonic components to smaller sizes that cannot be experienced using conventional optics and in particular the Silicon photonics industry. Many applications based on plasmonics are being increasingly developed and studied such as electromagnetic field enhancement for surface enhanced spectroscopy, wave guiding, sensing, modulation, switching, and photovoltaic applications.

We hereby propose a novel compact plasmonic resonator that can be utilized for different applications that depend on optical resonance phenomena in the Near Infrared spectral range, a very interesting range for a variety of applications, including sensing and modulation. The resonator structure consists of a gold layer which is etched to form a metal-insulator-metal wave guide and a rectangular cavity. The rectangular cavity and the wave guide are initially treated as a dielectric material. The strong reflectivity of gold at frequencies higher than the plasma frequency is the origin of the Fabry Perot resonator behavior of the rectangular cavity. The fano resonance was produced successfully and controlled by varying the rectangular cavity dimensions. The fano profile is generated in as the result of the redistribution of the electromagnetic field in the rectangular cavity as depicted by the plasmonic mode distribution in the resonator. The fano resonance is characterized by its sharp spectral line which attracts applications requiring sharp spectral line shapes such as sensing and modulation applications.

Optical modulators are key components in the modern communication technology. The research trend on optical modulators aims at achieving compact designs, low power consumption and large bandwidth operation. Plasmonic modulators emerge as promising devices since they can have high modulation speeds and very compact designs.

The operation mechanism of our introduced plasmonic modulator is as follows: instead of a constant refractive index dielectric, an electro-optic polymer that has its refractive index is dependent on some controlled phenomena, is filled in the metal insulator metal waveguide and the rectangular cavity. Efficient modulation was achieved by changing the applied voltage (DC Signal) on the metal contacts which then changes the refractive index of the polymer, thereby shifting the resonant wavelength position in the resonator, leading to signal modulation. Our modulator is operational at the telecom wavelength 1.55 μm, thereby suitable for the modern communication technology.

Finite Difference Time Domain (FDTD) simulations were conducted to design the modulator structure and run the simulations experiments, and to study the resonance effects of the structure and optimize its response to the desired results, the most important results however are the efficient modulation of the optical energy at the wavelengths required in the modern communication technology, around 1.5 μm, all results were carried on using the commercially available Lumerical FDTD software.

Handled devices and software applications are susceptible to ameliorate learning strength, awareness, and career development. Many mobile-based learning applications are obtainable from the market but Arabic learning shortage is not taken in consideration. We conduct an Arabic Text-to-Picture (TTP) mobile educational application which performs knowledge extraction and concept analysis to generate pictures that represent the content of the Arabic text. The knowledge extraction is based on Arabic semantic models cover important scopes for young children and new Arabic learners (i.e., grammar, nature, animals). The concept analysis uses semantic reasoning, semantic rules, and Arabic natural text processing (NLP) tool to identify word-to-word relationships. The retrieval of images is done spontaneously from local repository and online search engine (i.e., Google or Bing). The instructor can select the Arabic educational content, get semi-automatic generated pictures, and use them for explanation. Preliminary results show improvement in Arabic learning strength and memorization.

Keywords

Mobile Learning, Natural Language Processing, Ontology, Multimedia Learning, Engineering Education.

I. Introduction

Nowadays, mobile learning environment has been excessively used in diverse fields and has become a common matter in educational movement. In such an environment, learners are able to reach online educational materials from any location. Learners of Arabic language suffer from the lack of adequate resources. In fact, most of the educational software, tools, and web sites use classical techniques of introducing the concepts and explaining the vocabulary. We present in this paper a text-to-picture (TTP) educational mobile system that promotes Arabic children stories through semi-automatic generated pictures to illustrate their contents in an attractive manner. Preliminary results show that the system enhances the Arabic learners' comprehension, deduction and realization.

II. Background

Natural language processing (NLP) stresses the extraction of useful information and mining natural text. These information can be used to identify the scope of the text in order to generate summaries, classify contents and teach vocabulary. Diverse NLP-based systems that illustrate the text to images have been developed recently [1, 2, 3]. In general, these systems divide the text into segments and single words, access local multimedia resources, or explore the web to get pictures and images to illustrate the content.

All the proposed systems and techniques do not include the Arabic language. In this paper, we propose an Arabic TTP educational system using multimedia technology to teach children in an attractive way. Our proposal generates the multimedia tutorials dynamically by using Arabic text processing, entities relationship extraction, multimedia ontology, and online extraction of multimedia contents fetched from Google search engine.

III. Methodology:

In order to develop our system, we have created first the general system artwork, set the end user graphical user interface, design the semantic model that will store all semantic information about terms, and collect educational stories and analyze them. We have gathered 30 educational stories, annotated terms, and associated illustrations manually. Illustrations were gathered from the Internet and educational repository. The semantic model is developed using “Protégé editor”, a free open source ontology editor developed by Stanford [4]. The semantic model is composed from many classes that are referred to as concepts.

IV. The proposed system

The proposed system is a client-server application. When the server launched, it loads its packages and components, it loads the defined ontology, text parser components, and finally it opens a connection to listen for users' requests. Upon an effective connection trial, the user will be eligible to enter or open existing Arabic stories and process them. On the client side, the processing request and response of the story is done in a different thread, to keep the user able to continue his work without any interruption. Finally, server reply will be displayed for the user on his mobile device which consists from the processed Arabic story, related images, and different questions about an animal.

V. Conclusion

This study presents a complete system that automatically generates illustrations for Arabic stories through text processing, Arabic ontology, relationship extraction, and illustration generation. The proposed system belongs to learning technology which can be on mobile devices to teach children in an attractive and non-traditional style. Preliminary results demonstrate that the system improved learners' comprehension and realization.

References

[1] Bui, Duy, Carlos Nakamura, Bruce E Bray, and Qing Zeng-Treitler, “Automated illustration of patients instructions,” in AMIA Annual Symposium Proceedings, vol. 2012, pp. 1158, 2012.

[2] Li, Cheng-Te, Chieh-Jen Huang, and Man-Kwan Shan, “Automatic generation of visual story for fairy tales with digital narrative,” in Web Intelligence, vol. 13, pp. 115–122, 2015.

[3] Ustalov, Dmitry and R Kudryavtsev, “An Ontology Based Approach to Text to Picture Synthesis Systems,” in Proceedings of the 2nd International Workshop on Concept Discovery in Unstructured Data (CDUD 2012), 2012.

[4] Protégé. Ontology Editor Software. Available from: http://protege.stanford.edu, Accessed: September 2015.

Our research is focused on expanding the reach and impact of Arabic language news articles by attracting more readers. In pursuit of this research goal, we analyze attributes that result in certain news articles becoming viral, relative to other news articles that do not become viral or so viral. Specifically, we focus on Arabic language news articles, as Arabic language articles have unique linguistic, cultural, and social constrains relative to most Western languages news stories. In order to understand virality, we take two approaches, a time series and linguistical, in an Arabic language data of more than 1,000 news articles with associated temporal traffic data. For data collection, we select (Kasra, “a breaking”) (http://kasra.co/) is an Arabic language online news site that targets Arabic language speakers worldwide, but particularly in the Middle East North Africa (MENA) region. We gathered more than 3,000 articles, originally, then gathered traffic data for this set of articles, reducing the set to more than 1,000 with complete traffic data. We focus first on the temporal attributes in order to categorize clusters of virality with this set of articles. Then, with topical analysis, we seek to identify linguistical aspects common to articles within each virality cluster identified by time series. Based on results from the time series analysis, we cluster articles based on common temporal characteristics of traffic access. Once clustered by time series, we analyze each cluster for content attributes, topical and linguistical, in order to identify specific attributes that may be causing the virality of articles within each times-series cluster. To compute dissimilarity for time-series, we utilize and evaluate the performance of several state-of-the-art time series dissimilarity-based clustering approaches, such as dynamic time warping, discrete wavelet transformation, and others. To identify the dissimilarity algorithm with the most discriminating power, we conduct a principal component analysis (PCA), which is a statistical technique used to highlight variations and patterns in a dataset. Based on findings from our PCA, we select discrete wavelet transformation-based dissimilarity as the best times-series algorithm for our research because the resulting principal axes explain more proportion of variability (75.43 percent) relative to the other time-series algorithms that we had employed. We identify five virality clusters using times series. For topic modeling, we employ Latent Dirichlet allocation (LDA) for this portion of the research. LDA is a generative probabilistic model for collections of discrete data, such as text, LDA explains similarities among groups of observations within a data set. For text modeling, the topic probabilities of LDA provide an explicit representation of a document. For the topical classification analysis, we use Linguistic Inquiry and Word Count (LIWC), which is a sentiment analysis tool. LIWC is a text processing program based on occurrences of words in several categories covering writing style and psychological meaning. Prior empirical work shows the value of a LIWC linguistic analysis for detecting meanings in various experimental settings, including attention focus, thinking style, and social relationships. In terms of results, surprising, the article topic is not predictive of virality of Arabic language news articles. Instead we find that linguistical aspects and style of the news article is the most predictive attribute for predicting virality for Arabic news articles. In analyzing the attributes of virality in Arabic language news articles, our research finds that, perhaps counter intuitively, the topic of the article does not impact the virality. Instead, we find that style of the article is the most impactful attribute for predicting virality for Arabic news articles. Building on these findings, we will leverage aspects of the news articles with other factors to develop tools to assist content creators to more effectively reach their user segment. Our research results will assist in understanding the virality of Arabic news and ultimately improve readership and dissemination of Arabic language news articles.

Education of deaf children in the developing world is very dire and there is a dearth of sign language interpreters to assist them with translation for sign language-dependent students in the classrooms. Illiteracy within the deaf population is rampant. Over the past several years, a unique team of Moroccan and American deaf and hearing researchers have united to enhance the literacy of deaf students by creating tools that incorporate Moroccan Sign Language (MSL) and American Sign Language (ASL) under funding grants from USAID and the National Science Foundation (NSF). MSL is a gestural language distinct from both the spoken languages and written language of Morocco and has no text representation. Accordingly, translation is quite challenging and requires representation of MSL in graphics and video.

Many deaf and hard of hearing people do not have good facility with their native spoken language because they have no physiological access to it. Because oral languages depend, to a great extent, upon phonology, reading achievement of deaf children usually falls far short of that of hearing children of comparable abilities. And, by extension, reading instructional techniques that rely on phonological awareness, letter/sound relationships, and decoding, all skills proven essential for reading achievement, have no sensory relevance. Even in the USA, where statistics are available and education of the deaf is well advanced, on average, deaf high school graduates have a fourth grade reading level; only 7–10% of deaf students read beyond a seventh to eighth grade reading level; and approximately 20% of deaf students leave school with a second grade or below reading level (Gallaudet University's national achievement testing programs (1974, 1983, 1990, and 1996); Durnford, 2001; Braden, 1992; King & Quigley, 1985; Luckner, Sebald, Cooney, Young III, & Muir, 2006; Strong, & Prinz, 1997).

Because of spoken language inaccessibility, many deaf people rely on a sign language. Sign language is a visual/gestural language that is distinct from spoken Moroccan Arabic and Modern Standard/written Arabic and has no text representation. It can only be depicted via graphics, video, and animation.

In this presentation, we present an innovative technology Clip and Create, a tool for automatic creation of sign language supported instructional material. The technology has two tools– Custom Publishing and Instructional Activities Templates, and the following capabilities:

(1)Automatically constructs customizable publishing formats;

(2)Allows users to import Sign Language clip art and other graphics;

(3)Allows users to draw free-hand orusere-sizable shapes;

(4)Allows users to incorporate text, numbers, and scientific symbols in various sizes, fonts, and colors;

(5)Saves and prints published products;

(6)Focuses on core vocabulary, idioms, and STEM content;

(7)Incorporates interpretation of STEM symbols into ASL/MSL;

(8)Generates customizable and printable Instructional Activities that reinforce vocabulary and concepts found in instructional content using Templates:

a. Sign language BINGO cards,

b. Crossword puzzles,

c. Finger spelling/spelling scrambles,

d. Word searches (in finger spelling and text),

e. Flashcards (with sign, text, and concept graphic options), and

f. Matching games (i.e., Standard Arabic-to-MSL and English-to-ASL).

(cf. Figure 1: Screenshots from Clip and Create)

The ability of this tool to efficiently create bilingual (i.e., MSL and written Arabic and ASL and English) educational materials will have a profound positive impact on the quantity and quality of sign-supported curricular materials teachers and parents are able to create for young deaf students. And, as a consequence, deaf children will show improved vocabulary recognition, reading fluency, and comprehension.

A unique aspect of this software is that written Arabic is used by many Arab countries even though the spoken language varies. Though there are variations in signs as well, there is enough consistency to make this product useful in other Arab-speaking nations as is. Any signing differences can easily be adjusted by swapping sign graphic images.

Mobile learning can help in evolving students learning strength and comprehension skills. A connection is required to enable such devices communicate between each other. Brain-Computer Interface (BCI) can read brain signals and transform them into readable information. For instance, an instructor can use such device to track interest, stress level, and engagement of his students. We propose in this paper a mobile learning system that can transpose text-to-picture (TTP) to illustrate the content of Arabic stories and synchronize information with connected devices in a private wireless mesh network. The device of the instructor can connect the internet to download further illustrative information. It shares Wireless and Bluetooth connection with at least one student. Therefore, students can share the connection between each other to synchronize information on their devices. BCI devices are used to navigate, answer questions, and get to track students' performance. The aim of our educational system is to establish a private wireless mesh network that can operate in a dynamic environment.

Keywords

Mobile Learning, Arabic Text Processing, Brain Computer Interface, Engineering Education, Wireless Mesh Network.

I. Introduction

Nowadays mobile devices and collaborative work opened a new horizon for collaborative learning. As most people own handheld private portable smart phones, this has become the main means of connectivity and communication between people. Using smart phones for learning is beneficial and more attractive as learners can access educational resources at any time. Different eLearning systems available provide different options for collaborative classroom environment. However, they do not consider effective needs to adapt learning performance. They do not provide dynamic communication, automatic feedback, and other required classroom events.

II. Backgrounds

There are several collaborative learning applications have been proposed. BSCW [1] enables for online sharing of workspaces between distant people. Lotus Sametime Connect [2] also provides services for collaborative multiway chat, web conferencing, location awareness, and so. Saad et al. [3] proposed an intelligent collaborative system that can enable small range of mobile devices to communicate using WLAN, and uses the Bluetooth in case of power outage. The architecture of the proposed system is central where client can connect the server. Saleem [4] proposed a Bluetooth Assessment System (BAS) to use Bluetooth as an alternative to transfer questions and answers between the instructor and students. As many systems have been proposed in the domain of collaborative learning, several of them support mobile technology while others do not. BCI is not considered as part of the mobile educational system that can surround the environment with reading mental signals.

III. The proposed system

Our proposed system provides educational content and and synchronize it in a Wireless and Bluetooth wireless mesh network. It can be used in classrooms independent from the public network. The primary device broadcasts messages to enable users follow the explanation of the instructor on their mobile devices. The proposed system covers: 1) the establishment of wireless mesh network between mobile devices, 2) reading BCI data, 3) message communications, and 4) performance analysis between Wireless and Bluetooth technologies through device-to-device communication.

References

[1] K. Klö, “BSCW: cooperation support for distributed workgroups”, Parallel, Distributed and Network-based Processing. In 10th Euromicro Workshop, pp. 277–282, 2002.

[2] Lotus Sametime Connect. (2011, Feb. 17). Available: http://www.lotus.com/sametime.

[3] T. Saad, A. Waqas, K. Mudassir, A. Naeem, M. Aslam, S. Ayesha, A. Martinez-Enriquez, and P.R. Hedz, “Collaborative work in class rooms with handheld devices using bluetooth and WLAN,” IEEE 27th Canadian Conference on Electrical and Computer Engineering (CCECE), 2014, pp. 1–6.

[4] N.H., Saleem, “Applying Bluetooth as Novel Approach for Assessing Student Learning,” Asian Journal of Natural & Applied Sciences, vol 4, no. 2, 2015.

Motivation Chronic kidney disease (CKD) refers to the loss of kidney functions over time which is primarily to filter blood. Based on its severity it can be classified into various stages with the later ones requiring regular dialysis or kidney transplant. Chronic kidney disease mostly affects patients suffering from the complications of diabetes or high blood pressure and hinders their ability to carry out day-to-day activities. In Qatar, due to the rapidly changing lifestyle there has been an increase in the number of patients suffering from CKD. According to Hamad Medical Corporation [2], about 13% of Qatar's population suffers from CKD, whereas the global prevalence is estimated to be around 8–16% [3]. CKD can be detected at an early stage and can help at-risk patients from a complete kidney failure by simple tests that involve measuring blood pressure, serum creatinine and urine albumin [1]. Our goal is to use machine learning techniques and build a classification model that can predict if an individual has CKD based on various parameters that measure health related metrics such as age, blood pressure, specific gravity etc. By doing so, we shall be able to understand the different signals that identify if a patient at risk of CKD and help them by referring to preventive measures.

Dataset Our dataset was obtained from the UCI Machine Learning repository, which contains about 400 individuals of which 250 had CKD and 150 did not. The dataset was obtained from a hospital in southern India over a period of two months. In total there are 24 fields, of which 11 are numeric and 13 are nominal i.e. can take on only one of many categorical values. Some of the numerical fields include: blood pressure, random blood glucose level, serum creatinine level, sodium and potassium in mEq/L. Similarly, examples of nominal fields are answers to yes/no type questions such as whether the patient suffers from hypertension, diabetes mellitus, coronary artery disease. There was missing data values in a few rows which was addressed by imputing them with the mean value of the respective column feature. This ensures that the information in the entire dataset is leveraged to generate a model that best explains the data.

Approach We use two different machine learning tasks to approach this problem, namely: classification and clustering. In classification we built a model that can accurately classify if a patient has CKD based on their health parameters. And in order to understand if people can be grouped together based on the presence of CKD we have performed clustering on this dataset. Both these approaches provide good insights into the patterns present in the underlying data. Classification This problem can be modeled as a classification task in machine learning where the two classes are: CKD and not CKD which represents if a person is suffering from chronic kidney disease or not respectively. Each person is represented as a set of features provided in the dataset described earlier. We also have ground truth as to if a patient has CKD or not, which can be used to train a model that learns how to distinguish between the two classes. Our training set consists of 75% of the data and the remaining 25% is used for testing. The ratio of CKD to non-CKD persons in the test dataset was maintained to be approximately the similar to the entire dataset to avoid the problems of skewness. Various classification algorithms were employed such as logistic regression, Support Vector Machine (SVM) with various kernels, decision trees and Ada boost so as to compare their performance. While training the model, a stratified K-fold cross validation was adopted which ensures that each fold has the same proportion of labeled classes. Each classifier has a different methodology for learning. Some classifiers assign weights to each input feature along with a threshold that determines the output and updates them accordingly based on the training data. In the case of SVM, kernels map input features into a different dimension which might be linearly separable. Decision tree classifiers have the advantage that it can be easily visualized since it is analogous to a set of rules that need to be applied to an input feature vector. Each classifier has a different generalization capability and the efficiency depends on the underlying training and test data. Our aim is to discover the performance of each classifier on this type of medical information.

Clustering Clustering involves organizing a set of items into groups based on a pre-defined similarity measure. This is an unsupervised learning method that doesn't use the labeled information. There are various popular clustering algorithms and we use k-means and hierarchical clustering to analyze our data. K-means involves specifying the number of classes and the initial class means which are set to random points in the data. We vary the number of groups from 2 to 5 to figure out which maximizes the quality of clustering. Clustering with more than 2 groups also might allow to quantify the severity of Chronic Kidney Disease (CKD) for each patient instead of the binary notion of just having CKD or not. In each iteration of k-means, each person is assigned to a nearest group mean based on the distance metric and then the mean of each group is calculated based on the updated assignment. After a few iterations, once the means converge the k-means is stopped. Hierarchical clustering follows another approach whereby initially each datapoint is an individual cluster by itself and then at every step the closest two clusters are combined together to form a bigger cluster. The distance metric used in both the methods of clustering is Euclidean distance. Hierarchical clustering doesn't require any assumption about the number of clusters since the resulting output is a tree-like structure that contains the clusters that were merged at every time-step. The clusters for a certain number of groups can be obtained by slicing the tree at the desired level. We evaluate the quality of the clustering based on a well known criteria known as purity. Purity measures the number of data points that were classified correctly based on the ground truth which is available to us [5].

Principal Component Analysis Principal Component Analysis (PCA) is a popular tool for dimensionality reduction. It reduces the number of dimensions of a vector by maximizing the eigenvectors of the covariance matrix. We carry out PCA before using K-Means and hierarchical clustering so as to reduce it's complexity as well as make it easier to visualize the cluster differences using a 2D plot.

Results Classification In total, 6 different classification algorithms were used to compare their results. They are: logistic regression, decision tree, SVM with a linear kernel, SVM with a RBF kernel, Random Forest Classifier and Adaboost. The last two classifiers fall under the category of ensemble methods. The benefit of using ensemble methods is that it aggregates multiple learning algorithms to produce one that performs in a more robust manner. The two types of ensemble learning methods used are: Averaging methods and Boosting methods [6].

The averaging method typically outputs the average of several learning algorithms and one such type we used is random forest classifier. On the other hand, a boosting method “combines several weak models to produce a powerful ensemble” [6]. Ada boost is an example of boosting method that we have used.

We found that the SVM with linear kernel performed the best with 98% accuracy in the prediction of labels in the test data. The next best performance was by the two ensemble methods: Random Forest Classifier with 96% and Adaboost 95% accuracy. The next two classifiers were: Logistic regression with 91% and Decision tree with 90%. The classifier with the least accuracy was SVM with a RBF kernel which has about 60% accuracy. We believe that RBF gave lower performance because the input features are already high dimensional and don't need to be mapped into a higher dimensional space by RBF or other non-linear kernels. A Receiver Operating Characteristic (ROC) curve can also be plotted to compare the true positive rate and false positive rate. We also plan to compute other evaluation metrics such as precision, recall and F-score. The results are promising as majority of the classifiers have a classification accuracy of above 90%.

After classifying the test dataset, feature analysis was performed to compare the importance of each feature. The most important features across the classifiers were: albumin level and serum creatinine. Logistic regression classifier also included the ‘pedal edema’ feature along with the previous two features mentioned. Red blood cell feature was included as an important feature by Decision tree and Adaboost classifier.

Clustering After performing clustering on the entire dataset using K-Means we were able to plot it on a 2D graph since we used PCA to reduce it to two dimensions. The purity score of our clustering is 0.62. A higher purity score (max value is 1.0) represents a better quality of clustering. The hierarchical clustering plot provides the flexibility to view more than 2 clusters since there might be gradients in the severity of CKD among patients rather than the simple binary representation of having CKD or not. Multiple clusters can be obtained by intersecting the hierarchical tree at the desired level.

Conclusions We currently live in the big data era. There is an enormous amount of data being generated from various sources across all domains. Some of them include DNA sequence data, ubiquitous sensors, MRI/CAT scans, astronomical images etc. The challenge now is being able to extract useful information and create knowledge using innovative techniques to efficiently process the data. Due to this data deluge phenomenon, machine learning and data mining have gained strong interest among the research community. Statistical analysis on healthcare data has been gaining momentum since it has the potential to provide insights that are not obvious and can foster breakthroughs in this area.

This work aims to combine work in the field of computer science and health by applying techniques from statistical machine learning to health care data. Chronic kidney disease (CKD) affects a sizable percentage of the world's population. If detected early, its adverse effects can be avoided, hence saving precious lives and reducing cost. We have been able to build a model based on labeled data that accurately predicts if a patient suffers from chronic kidney disease based on their personal characteristics.

Our future work would be to include a larger dataset consisting of of thousands of patients and a richer set of features that shall improve the richness of the model by capturing a higher variation. We also aim to use topic models such as Latent Dirichlet Allocation to group various medical features into topics so as to understand the interaction between them. There needs to be a greater encouragement for such inter-disciplinary work in order to tackle grand challenges and in this case realize the vision of evidence based healthcare and personalized medicine.

References

[1] https://www.kidney.org/kidneydisease/aboutckd

[2] http://www.justhere.qa/2015/06/13-qatars-population-suffer-chronic-kidney-disease-patients-advised-take-precautions-fasting-ramadan/

[3] http://www.ncbi.nlm.nih.gov/pubmed/23727169

[4] https://archive.ics.uci.edu/ml/datasets/Chronic_Kidney_Disease

[5] http://nlp.stanford.edu/IR-book/html/htmledition/evaluation-of-clustering-1.html

[6] http://scikit-learn.org/stable/modules/ensemble.html

1. Introduction

Twitter is the micro blogging social media platform which has at most variety of content. The open access to Twitter data with the usage of Twitter APIs has made it a important area of research. Twitter has a useful feature called “Trends” which displays Hot topics or trending information differing for every location. This trending information is evolved by tweets being shared on Twitter in a particular location. But Twitter limits the trending information to current tweets, as the algorithm for finding trends is concentrated on generating trends in real time, rather than trends summarization of hot topics on daily basis. Thus a clear summarization of the contemporary trending information is missing and is much needed. Latest Twitter Trends - our application discussed in this paper, is built to get the aggregate of hot topics on Twitter for Arab Countries and the World. This is a real time application with summarization of Hot topics over time. It enables users to study the summarization of twitter trends by location with the help of a Word Cloud. The tool also enables the user to click on the particular trend, which will allow the user to navigate and search through Twitter Search - also in real time. This tool also overcomes a drawback of Twitter trending information, in addition to the Twitter trend algorithm. The trends differ for different languages in different locations and are often mixed. For eg, if #Eid-ul-Adha is trending in Arab countries, عيد الأضحى# is also trending. This application focuses on consolidating the trends in Arabic and English, which have the same meaning and display only one trending topic, instead of two same topics in different language. This application also gives an estimation of the different kind of Twitter users, analyzing the percentage of tweets made by Male and Female in that location.

2. Trends data gathering

Twitter APIs give developers access to the real time data comprising of tweets & trends. The Twitter REST API is used by the tool - Latest Twitter Trends, to connect and get trending data from Twitter. The API is used to authenticate and establish connection with Twitter and also returns Twitter trending data in JSON format. Python programming language is used to write scripts to gather data from Twitter. A Data Crawling Script is developed for connecting with Twitter API by authenticating the credentials generated by twitter on creating an application from the app.twitter.com. The Customer Key, Customer Secret, Access Token, Access Token Secret are the credentials used to perform authentication by Twitter. The data returned by Twitter is in JSON (JavaScript Object Notation) format and the Python Data Crawling Script is commanded to handle the JSON files and create a CSV database. This High Level Gathering of Data comprises of the following: Python data crawling script connects and authenticates with Twitter API and gets trending places data in JSON format from Twitter. The data in JSON format is stored in to our tool database as a CSV file. The Twitter data gathered is all the trending location/places with the WOIED (Where On Earth ID). The WOEID is used as a key to get Twitter trending topics location by location - in real time using the Twitter REST API. The trends for every location are also returned to the tool in JSON format, which is again changed converted to CSV for saving in the tool database. This CSV file for Trends is appended every time a new trending data is collected from twitter. Another CSV file is maintained in the Database which holds only the current information for all trending places - for later use. Natural language processing is done on trends by location CSV data, dictionary, to consolidate and consider Arabic and English Trending topics as one. The results are stored in CSV file and will be used for the hot topic identification.

3. Hot topic identification

After the High level Data Gathering, CSV files containing data are used as a Database for generating Word Cloud using D3.js. This trending data is processed by calculating the number of occurrences to give an estimate of which trending topic was trending for a long time. The frequency is taken as the count value for trending topic and a word cloud is generated. This algorithm for calculating frequency is a python script, written mainly for Word Cloud Data Crawling. This word cloud data crawling script takes the Trends by Location data as input and generates a huge database of trends by cities in JSON files. This word cloud crawling script gives output in the JSON files to be stored with key as the trend topic and value as the frequency of the trend occurrence.

4. Architecture

Figure 1: Latest Trends in Twitter Application Architecture The python scripts for data crawling and word cloud crawling are sued to connect with Twitter, gather data, process and store in a database. The D3.js and Google fusion table API are used for displaying the application results. Google Fusion Table API is used to create a Map containing current trends by location – geo-tagged on the map. Java program is used as a dedicated project to connect & authenticate with Google API and clear old fusion table data to import new updated rows in to the Google Fusion Table. Python script Tagcloud.py is used to generate cities. JSON with trending topics from the Trends.csv file. These files from the database for generating word cloud using D3.js, individually for every city/location. Fusion table is used to visualize the trending information from Twitter. A java program along with Google API is used to authenticate and connect. Also to delete previous information in fusion table and update/import new records of data.

5. Results

The data crawling script establishes connection with Twitter and returns a JSON format as in Fig. 2. This data is processed and saved as a CSV in to our application database for later use. Figure 2: Trends Data Output from Twitter in JSON format The word cloud crawling script generates key value pairs of processed trending data from the database. The key containing trending topic and the value containing the frequency of the trending topic's occurrence. The Fig. 3 displays the JSON dataset used for generating word cloud. Figure 3: JSON data of the processed trending data The word cloud is generated using the D3.js library and is used to display summarized trending data to the user. Figure 4 shows the word cloud result for London country. Figure 4: Word cloud for trending data.

Mobile-health (m-health) systems leverage wireless and mobile communication technologies to promote new ways to acquire, process, transport, and secure the raw and processed medical data. M-health systems provide the scalability needed to cope with the increasing number of elderly and chronic disease patients requiring constant monitoring. However, the design and operation of such pervasive health monitoring systems with Body Area Sensor Networks (BASNs) is challenging in twofold. First, limited energy, computational and storage resources of the sensor nodes. Second, need to guarantee application level Quality of Service (QoS). In this paper, we integrate wireless network components, and application-layer characteristics to provide sustainable, energy-efficient and high-quality services for m-health systems. In particularly, we propose an Energy-Cost-Distortion (E-C-D) solution, which exploits the benefits of medical data adaptation to optimize transmission energy consumption and cost of using network services. However, at large scale networks and due to heterogeneity of wireless m-health systems, centralized approach becomes less efficient. Therefore, we present a distributed cross-layer solution, which is suitable for heterogeneous wireless m-health systems and scalable with the network size. Our scheme leverages Lagrangian duality theory and enables us to find efficient trade-off among energy consumption, network cost, and vital signs distortion, for delay sensitive transmission of medical data over heterogeneous wireless environment. In this context, we propose a solution that enables energy-efficient high-quality patient health monitoring to facilitate remote chronic disease management. We propose a multi-objective optimization problem that targets different QoS metrics, namely, signal distortion, delay, and Bit Error Rate (BER), as well as monetary cost and transmission energy. In particularly, we aim to achieve the optimal trade-off among the above factors, which exhibit conflicting trends.

The main contributions of our work can be summarized as follows:

•   (1) We design a system for EEG health monitoring systems that achieves high performance by properly combining network functionalities and EEG application characteristics.
• (2) We formulate a cross-layer multi-objective optimization model that aims at adapting and minimizing, at each PDA, the encoding distortion and monetary cost at the application layer, as well as the transmission energy at the physical layer, while meeting the delay and BER constraints.
• (3) We use geometric program transformation to convert the aforementioned problem into a convex problem, for which an optimal, centralized solution can be obtained.
• (4) By leveraging Lagrangian duality theory, we then propose a scalable distributed solution. The dual decomposition approach enables us to decouple the problem into a set of sub-problems that can be solved locally, leading to a scalable distributed algorithm that converges to the optimal solution.
• (5) The proposed distributed algorithm for EEG based m-health systems is analyzed and compared to the centralized approach.

Our results show the efficiency of our distributed solution, its ability to converge to the optimal solution and to adapt to varying network conditions. In particular, simulation results show that the proposed scheme achieves the optimal trade-off between energy efficiency and QoS requirements of health monitoring systems. Moreover, it offers significantly savings in the objective function (i.e., E-C-D utility function), compared to solutions based on equal bandwidth allocation.

Every day, humans interact with text spanning from different sources such as news, literature, education, and even social media. While reading, humans process text word by word, accessing the meaning of a particular word from the lexicon, and when needed, changing its meaning to match the context of the text (Harley, 2014). The process of reading can induce a range of emotions, such as engagement, confusion, frustration, surprise or happiness. For example, when readers come across unfamiliar jargon, this may confuse them, as they try to understand the text.

In the past, scientists have addressed the emotion in text from a writer's perspective. For example the field of Sentiment Analysis, aims to detect the emotional charge of words, to infer the intentions of the writer. However, here we propose the reverse approach: detect emotions produced on readers while processing text.

Detecting which emotions are induced by reading a piece of text can give us insights about the nature of the text itself. A word-emotion detector can be used to assign specific emotions experienced by readers to specific words or passages of text. This area of research has never been explored before.

There are many potential applications to a word-emotion detector. For example, a word-emotion detector can be used to analyze how passages in books, news or social media are perceived by readers. This can guide stylistic choices to cater for a particular audience. In a learning environment, it can be used to detect the affective states and emotions of students, so as to infer their level of understanding which can be used to provide assistance to students over difficult passages. In a commercial environment, it can be used to detect reactions to wording in advertizements. In the remainder of this report, we detail the first steps we followed to build a word-emotion detector. Moreover, we present the details of our system developed during QCRI's 2015 Hot Summer Cool Research internship program, as well as the initial experiments. In particular, we describe our experimental setup in which viewers watch a foreign language video with modified subtitles containing deliberate emotion inducing changes. We analyze the results and provide discussion about the future work.

The Pokerface System

A Pokerface is an inscrutable face that reveals no hint of a person's thoughts or feelings. The goal of the ‘Pokerface’ project is to build a word-emotion detector that works even if no facial movements are present. To do so, the Pokerface system uses a unique symbiose of the latest consumer-level technologies such as: eye-tracking to detect words that are being read; electroencephalography (EEG) to detect brain activity of the reader; and facial-expression recognition (FER) to detect movement in a reader's face. We then classify the brain activity and facial movements detected into emotions using Neural Networks.

In this report, we present the details of our Pokerface system, as well as the initial experiments done during QCRI's 2015 Hot Summer Cool Research internship program. In particular, we describe the setup in which viewers watch a foreign language video with subtitles containing deliberate emotion inducing changes.

Methodology

To detect emotions experienced by readers as they read text, we used different technologies. FER and EEG are used to detect emotional reactions through changes in facial expressions and brainwaves, while eye-tracking is used to identify the stimulus (text) to the reaction detected. A video interface was created to run the experiments. Below we describe each of them independently, and how we used them in the project.

EEG

EEG is the recording of electrical activity along the scalp (Niedermeyer and Lopes da Silva, 2005). EEG measures voltage fluctuations resulting from ionic current flows within the neurons of the brain. EEG is one of the few non-intrusive techniques available that provides a window on physiological brain activity. EEG averages the response from many neurons as they communicate, measuring the electrical activity by surface electrodes. We can then use the brain activity of a user to detect their emotional status.

Data Gathering

In our experiments, we used the Emotiv | EEG EPOC neuroheadset (2013), which has 14 EEG channels plus two references, inertial sensors, and two gyroscopes. The raw data from the neuroheadset was parsed with the timestamps for each sample.

Data Cleaning and Artifact Removal

After retrieving the data from the EEG, we need to remove “artifacts” which are changes in the signals that do not originate from neurons (Vidal, 1977), such as ocular movements, muscular movements, as well as technical noise. To do so, we used the open source toolbox EEGlab (Delorme & Makeig, 2004) to remove artifacts and filtering (removing the 4–45 Hz line noise).

ERP Collection

We decided to consider remaining artifacts as random noise and move forward with extracting Event Related Potentials (ERPs), since all of other options we had found required some level of manual intervention. ERPs are the relevant sections of our EEG data with regards to stimuli and the subjects' reaction time. To account for random effects form the artifacts, we averaged the ERPs over different users and events. To do so, we used EEGlab's plugin ERPlab, to add events codes to our continuous EEG data based on stimulus time.

Events

Our events were defined as textual modifications in subtitles, designed to induce emotions of confusion, frustration or surprise. The time at which the subject looks at a word was marked to be the stimulus time (st) for that word, and the reaction time was marked to be st+800 ms, because we rarely see a reaction to a stimulus 800 ms after its appearance (Fischler and Bradley, 2006).

The ERPs were obtained as the average of different events corresponding to the same condition (control or experimental).

Eye-Tracking

An eye-tracker is an instrument to detect the movements of the eye. Based on the nature of the eye and human vision, the eye-tracker identifies where a user is looking by shining a light that will be reflected into the eye, such that the reflection will be captured by image sensors. The eye-tracker will then measure the angle between the cornea and pupil reflections to calculate a vector and identify the direction of the gaze.

In this project, we used the EyeTribe eye-tracker to identify the words a reader looked at while reading. It was set up in a Windows machine. Before an experiment, the user needs to calibrate the eye-tracker. Recalibration is necessary every time the user changes their sitting position. While the eye-tracker is running, Javascript and NodeJS were used to create a function that extracts and parses the data from the machine and prints it into a text file. This data includes the screen's x and y coordinates of the gaze; the timestamp; and an indicator of whether the gaze point is a fixation or not. The data is received at a rate of 60fps. The gaze points are used to determine which words are looked at at any specific time.

Video Interface

In our experiments, each user was presented with a video with subtitles. To create the experimental interface, we made different design choices based on previous empirical research. Therefore, we used Helvetica font, given its consistency across all platforms (Falconer, 2011), and used the font size 26 given that it improves the readability of subtitles on large desktops (Franz, 2014). We used Javascript to detect the location of each word that was displayed on the screen.

After gathering the data the experiment, we used an off-line process to detect the “collisions” between the eye-tracker gaze points and the words displayed to the user. To do so, we used both time information and coordinate information. The result was a series of words annotated with the specific time-spans in which they were looked at.

FER

Facial Expression Recognition (FER) is the process of detecting an individual's emotion by accessing their facial expressions in an image or video. In the past, FER has been used for various purposes, including psychological studies, tiredness detection, facial animation and robotics, etc.

Data Gathering

We used the Microsoft Kinect with the Kinect SDK 2.0 for capturing the individual's face. The data extracted from the Kinect provided us with color and infrared images, as well as depth data. However, for this project we only worked with the color data. The data from the Kinect was saved as a sequence of color images, recorded at a rate of 30 frames per second (fps). The code made use of the process of multithreading to ensure high fps, and low memory usage. Each image frame was assigned a timestamp in milliseconds, which was saved in a text file.

Feature Extraction

After having extracted the data from the Kinect, the images were processed so as to locate the facial landmarks. We tested the images with Face++ which is a free API for face detection, recognition and analysis. Using Face++, we were able to locate 83 facial landmarks on the images. The data obtained from the API was the name of the landmark along with it's x and y-coordinates.

The next step involved obtaining Action Units (AUs) by using the facial landmarks located through Face++. Action Units are the actions of individual muscles or groups of muscles, such as, raising the outer eyebrow, or stretching of lips etc (Cohn et al. 2001). To determine which AUs to use for FER, as well as how to calculate them, Tekalp and Ostermann's (2000) was taken as a reference.

Classification

The final step of the process was classifying the image frames into one of the eight different emotions - happiness, sadness, fear, anger, disgust, surprise, neutral and confused. We used the MATLAB Neural Network toolkit (MathWorks, Inc., 2015) and designed a simple feed-forward neural network with backpropagation.

Pilot results

EEG

In our pilot classification study, we used we experimented with the Alcoholism data used in Kuncheva and Rodriguez (2012) paper, from the UC Irvine (UCI) machine learning repository (2010) which contains ERP raw data for 10 alcoholic subjects and 10 sober subjects. We extracted the features using the interval feature extraction. Total of 96K Features were extracted from each subject's data. We Achieved around 98% accuracy with the training data.

FER

We experimented on three different individuals as well as several images from the Cohn-Kanade facial expressions database. It was found that the application had roughly 75–80% accuracy, and this accuracy could further be improved by adding more data into the training set. It was also observed that the classifier was more accurate when classifying certain emotions than some others. For example, the images depicting happiness were classified accurately more than images that depicted any other emotion. The classifier had difficulty distinguishing between fear, anger and sadness.

Conclusion

In this paper, we presented Pokerface, a word-emotion detector that can detect the emotion of users as they read text. We built a video interface that displays subtitled videos and used the EyeTribe eye-tracker to identify the word a user is looking at a certain time in the subtitles. We used the Emotiv Epoc headset to obtain EEG brainwaves from the user, and the Microsoft Kinect to obtain their facial expressions, and extracted features from both. We used Neural Networks to classify both the facial expressions and the EEG brainwaves into emotions. In the future.

Future directions of work include to improve the accuracy of the FER and EEG emotion classification components. Furthermore, the EEG results can be improved by exploring additional artifact detection and removal techniques. Furthermore, we want to integrate the whole pipeline in a seamless application that allows effortless experimentation.

Once the setup is streamlined, Pokerface can be used to explore many different applications to optimize users' experiences in education, news, advertizing, etc. For example, the word-emotion detector can be utilized in Computer-assisted learning to provide students with virtual affective support, such as detecting confusion and providing clarifications.