Qatar Foundation Annual Research Forum Volume 2011 Issue 1

We have permanently deployed Hala; the world first's English and Arabic Robot Receptionist for 500+ days in an uncontrolled multi-cultural/multi-lingual environment within Carnegie Mellon Qatar.

Hala serves as a research testbed for studying the influence of socio-cultural norms and the nature of interactions during human-robot interaction within a multicultural, yet primarily ethnic Arab, setting.

Hala, as a platform, owes its uptime to several independently engineered components for modeling user interactions, syntactic and semantic language parsing, inviting users with a laser, handling facial animations, text-to-speech and lip synchronization, error handling and reporting, post dialogue analysis, networking/interprocess communication, and a rich client interface.

We conjecture that disparities in discourse, appearances, and non-verbal gestures amongst interlocutors of different cultures and native tongues. By varying Hala's behavior and responses, we gain insight into these disparities (if any) and therefore we've calculated: rate of thanks after the robot's answer amongst cultures, user willingness to answer personal questions, correlation between language and acceptance of robot invites, the duration of conversations, effectiveness of running an open-ended experiment versus surveys.

We want to understand if people communicate with a robot (rather, an inanimate object with human-like characteristics) differently than amongst themselves. Additionally, we want to extrapolate these differences/similarities while accounting for culture and language. Our results indicate that users in Qatar thanked Hala less frequently than their counterparts in the US. The robot often answered personal questions and inquiries (for ex: her marital status, job satisfaction, etc); however, only 10% of the personal questions posed by the robot were answered by users. We observed a 34% increase in interactions when the robot initiated the conversation by inviting nearby users, and the subsequent duration of the conversation also increased by 30%. Upon bringing language into the mix, we observed that native Arabic speakers were twice more likely to accept an invite from the robot and they also tended to converse for 25% longer than other cultures.

These results indicate a disparity in interaction across English and Arabic users thereby encouraging the creation of culture specific dialogues, appearances and non-verbal gestures for an engaging social robot with regionally relevant applications.

Recent advances in computer hardware and signal processing have made possible the use of EEG signals or “brain waves” for communication between humans and computers. Locked-in patients have now a way to communicate with the outside world, but even with the last modern techniques, such systems still suffer communication rates on the order of 2-3 tasks/minute. In addition, existing systems are not likely to be designed with flexibility in mind, leading to slow systems that are difficult to improve.

This Thesis is classifying different mental tasks through the use of the electroencephalogram (EEG). EEG signals from several subjects through channels (electrodes) have been studied during the performance of five mental tasks: Baseline task for which the subjects were asked to relax as much as possible, Multiplication task for which the subjects were given nontrivial multiplication problem without vocalizing or making any other movements, Letter composing task for which the subject were instructed to mentally compose a letter without vocalizing (imagine writing a letter to a friend in their head),Rotation task for which the subjects were asked to visualize a particular three-dimensional block figure being rotated about its axis, and Counting task for which the subjects were asked to imagine a blackboard and to visualize numbers being written on the board sequentially.

vvvThe work presented here maybe a part of a larger project, with a goal to classify EEG signals belonging to a varied set of mental activities in a real time Brain Computer Interface, in order to investigate the feasibility of using different mental tasks as a wide communication channel between people and computers.

Although various access control models have been proposed, access control configurations are error prone. There is no assurance of the correctness of access control configurations. When we find errors in an access control configuration, we take immediate actions to repair the configuration. The repairing is difficult, largely because arbitrary changes to the configuration may result in no less threat than errors do. In other words, constraints are placed on the repaired configuration. The presence of constraints makes a manual error-and-trial approach less attractive. There are two main shortcomings with the manual approach. Firstly, it is not clear whether the objectives are reachable at all; if not, we waste time trying to repair an error prone configuration. Secondly, we have no knowledge of the quality of the solution such as correctness of the repair.

In order to address these shortcomings, we aim to develop an automated approach to the repairing task of access control configurations. We have utilized answer set programming (ASP), a declarative knowledge representation paradigm, to support such an automated approach. The rich modeling language of ASP enables us to capture and express the repairing objectives and the constraints. In our approach, the repairing instance is translated into an ASP, and the ASP solvers are invoked to evaluate it.

Although the applications of ASP follow the general “encode-compute-extract” approach, they differ in the representations of the problems in ASP. In our case, there are two principal factors which render the proposed problem and approach non-trivial. Firstly, we need to identify constraints which are not only amendable to ASP interpretation, but also expressive enough to capture common idioms of security and business requirements; there is a trade-off to make. Secondly, our ASP programs should model the quality measure of repairs—when more than one repair is possible, the reported one is optimized in terms of the quality measure. We have also undertaken extensive experiments on both real-world and synthetic data-sets. The experiment results validate the effectiveness and efficiency of the automated approach.

Keyphrase extraction is a process by which the set of words or phrases that best describe a document is specified. The phrases could be extracted from the document words itself, or they could be external and specified from an ontology for a given domain. Extracting keyphrases from documents is critical for many applications such as information retrieval, document summarization or clustering. Many keyphrase extractors view the problem as a classification problem and therefore they need training documents (i.e. documents which their keyphrases are known in advance). Other systems view keyphrase extraction as a ranking problem. In the latter approach, the words or phrases of a document are ranked based on their importance and phrases with high importance (usually located at the beginning of the list) are recommended as possible keyphrases for a document.

This abstract explains Shihab; a system for extracting keyphrases from Arabic documents. Shihab views keyphrase extraction as a ranking problem. The list of keyphrases is generated by clustering the phrases of a document. Phrases are built from words which appear in the document. These phrases consist of 1-, 2- or 3-words. The idea is to group phrases which are similar into one cluster. The similarity between phrases is determined by calculating the Dice value of their corresponding contexts. A phrase context is the sentence in which that phrase appears. Agglomerative hierarchical clustering is used in the clustering phase. Once the clusters are ready, then each cluster will nominate a phrase to the set of candidate keyphrases. This phrase is called cluster representative and is determined according to a set of heuristics. Shihab results were compared with other existing keyphrase extractors such as KP-Miner and Arabic-KEA and the results were encouraging.

In the recent years, several mathematical concepts were successfully explored in computer science domain, as basis for finding original solutions for complex problems related to knowledge engineering, data mining, information retrieval, etc.

Thus, Relational Algebra (RA) and Formal Concept Analysis (FCA) may be considered as useful mathematical foundations that unified data and knowledge in information retrieval systems. As for example, some elements in a fringe relation (related to the RA domain) called isolated points were successfully of use in FCA as formal concept labels or composed labels. Once associated to words, in a textual document, these labels constitute relevant features of a text. Here, we propose the GenCoverage algorithm for covering a Formal Context (as a formal representation of a text) based on isolated labels and we use these labels (or text features) for categorization, corpus structuring and micro-macro browsing as an advanced functionality in the information retrieval task.

The main thrust of the introduced approach heavily relies on the snugness connection between isolated points and minimal generators (MGs). MGs stand at the antipodes of the closures within their respective equivalence classes. Relying on the fact the minimal generators are the smallest elements within an equivalence class, so their detection/traversal is largely eased and permits a swift building of the coverage. Thorough carried out experiments provide empirical evidences about the performances of our approach.

The exponential growth of financial news coming from different sources makes getting effective benefit from them very hard. Business decision makers who reply to these news, are unable to follow them accurately in real time. They need always to be alerted immediately for any potential financial events that may affect their businesses whenever they occur. Many important news can simply be embedded into thousands of lines of financial data and cannot be detected easily. Such kind of news may have sometimes a major impact on businesses and the key decision makers should be alerted about them. In this work, we propose an alert system that screens structured financial news and trigger alerts based on the users profiles. These alerts have different priority levels: low, medium and high. Whenever the alert priority level is high, a quick intervention should be taken to avoid potential risks on businesses. Such events are considered as tasks and should be treated immediately. Matching users profiles with news events can sometimes be straightforward. It can also be challenging especially whenever the keywords in the users profiles are just synonyms of the events keywords. In addition, alerts can sometimes be dependable on the combination of correlated news events coming from different sources of information. This makes their detection a computationally challenging problem. Our system allows the user to define their profiles in three different ways: (1) selecting from a list of keywords that are related to events properties; (2) providing free keywords; and (3) entering simple short sentences. The system triggers alerts immediately whenever some news events related to the users profiles occur. It takes into consideration the correlated events and the concordance of the events keywords with the synonymous of the users profiles keywords. The system uses the vector space model to match keywords with the news events words. As consequence, the rate of false-positive alerts is low whereas the rate of false-negative alerts is relatively high. However, enriching the dictionary of synonym words would reduce the false-negative alerts rate.

The Cognitive Dimensions of Notations are a heuristic framework created by Thomas Green for analysing the usability of notational systems. Microsoft used this framework as a vocabulary for evaluating the usability of their C# and .NET development tools. In this research we used this framework to compare the evaluation of the Constraint Diagrams and the evaluation of the Natural Language by running a usability study. The result of this study will help in determining if users would be able to use constraint diagrams to accomplish a set of tasks. From this study we can predicate difficulties that may be faced when working on these tasks. Two steps were required. The first step is to decide what generic activities a system is desired to support. An Activity is described at a rather abstract level in terms of the structure of information and constraints on the notational environment. Cognitive dimensions constitute method to theoretically evaluate the usability of a system. Its dimensional checklist approach is used to improve different aspects of the system. Each improvement will be associated with a trade-off cost on other aspects. Each generic activity has its own requirements in terms of cognitive dimensions, so the second step is to scrutinize the system and determine how it lies on each dimension. If the two profiles match, all is well. Every dimension should be described with illustrative examples, case studies, and associated advice for designers. In general, an activity such as exploratory design where software designers make changes at different levels is the most demanding activity. This means that dimensions such as viscosity and premature-commitment must be low while visibility and role-expressiveness must be high.

Over the past several years ORC processes have become very promising for power production from low grade heat sources: solar, biomass, geothermal and waste heat. The key challenge in the design process is the selection of an appropriate working fluid. A large number of authors used pure components as working fluid, and assess ORC performance.

ORC systems that use single working fluid component have two major shortcomings. First, the majority of applications involve temperatures of the heat sink and source fluid varying during the heat transfer process, whereas the temperatures of the working fluid during evaporation and condensing remains constant. As a consequence a pinch point is encountered in the evaporator and condenser, giving rise to large temperature differences at one end of the heat exchanger. This leads to irreversibility that in turn reduces process efficiency. A similar situation is also encountered in the condenser. A second shortcoming of the Rankine cycle is lack of flexibility.

These shortcomings result from a mismatch between thermodynamic properties of pure working fluids, the requirements imposed by the Rankine cycle and the particular application. In contrast, when working fluid mixtures are used instead of single component working fluids, improvements can be obtained in two ways, through inherent properties of the mixture itself, and through cycle variations which, become available with mixtures. The most obvious positive effect is decrease in energy destruction, since the occurrence of a temperature glide during a phase change provides a good match of temperature profiles in the condenser and evaporator.

This paper presents detailed simulations and economic analyses of Organic Rankine Cycle processes for energy conversion of low heat sources. The paper explores the effect of mixture utilization on common ORC performance assessment criteria in order to demonstrate advantages of employing mixtures as working fluid as compared to pure fluids. We illustrate these effects based on of zeotropic mixtures of paraffins, as ORC working fluids.

Major challenges towards the development of next generation 4G wireless networks include fulfilling the foreseeable increase in power demand of future mobile terminals (MTs) in addition to meeting the high throughput and low latency requirements of emerging multimedia services. Studies show that the high-energy consumption of battery operated MTs will be one of the main limiting factors for future wireless communication systems. Emerging multimedia applications require the MTs’ wireless interfaces to be active for long periods while downloading large data sizes. This leads to draining the power of the batteries.

The evolution of MTs with multiple wireless interfaces helps to deal with this problem. This results in a heterogeneous network architecture with MTs that actively use two wireless interfaces: one to communicate with the base station (BS) or access point over a long-range (LR) wireless technology (e.g., UMTS/HSPA, WiMAX, or LTE) and one to communicate with other MTs over a short-range (SR) wireless technology (e.g., Bluetooth or WLAN). Cooperative wireless networks proved to have a lot of advantages in terms of increasing the network throughput, decreasing the file download time, and decreasing energy consumption at MTs due to the use of SR mobile-to-mobile collaboration (M2M). However, the studies in the literature apply only to specific wireless technologies in specific scenarios and do not investigate optimal strategies.

In this work, we consider energy minimization in content distribution with M2M collaboration and derive the optimal solution in a general setup with different wireless technologies on the LR and SR. Scenarios with multicasting and unicasting are investigated. Content distribution delay is also analyzed. Practical implementation aspects of the cooperative techniques are studied and different methods are proposed to overcome the practical limitations of the optimal solution. Simulation results with different technologies on the LR and SR are generated, showing significant superiority of the proposed techniques. Ongoing work is focusing on incorporating quality of service constraints in the energy minimization problem and in designing a testbed validating the proposed methods.

Hierarchical structures are found in many natural and man-made materials [1]. This structural hierarchy play an important role in determining the overall mechanical behavior of the structure. It has been suggested that increasing the hierarchical level of a structure will result in a better performing structure [2]. Besides, honeycombs are well known structures for lightweight and high strength applications [3]. In this work, we have studied the mechanical properties of honeycombs with hierarchical organization using theoretical, numerical, and experimental methods. The hierarchical organization is made by replacing the edges of a regular honeycomb structure with smaller regular honeycomb. Our results showed that honeycombs with structural hierarchy have superior properties compared to regular honeycombs. The results show that a relatively broad range of elastic properties, and thus behavior, can be achieved by tailoring the structural organization of hierarchical honeycombs, and more specifically the two dimension ratios. Increasing the level of hierarchy provides a wider range of achievable properties. Further optimization should be possible by also varying the thickness of the hierarchically introduced cell walls, and thus the relative distribution of the mass, between different hierarchy levels. These hierarchical honeycombs can be used in development of novel lightweight multifunctional structures, for example as the cores of sandwich panels, or development of lightweight deployable energy systems.