Times of crisis and prosperity may be the most defining moments of leadership and therefore one of the most important contexts in which to study leadership processes. In the present research we explored linguistic patterns of cognitive complexity, social representations and social coordination of Mao Zedong speeches during different socio-historical contexts, namely times of prosperity and crisis. The texts of Mao Zedong were analyzed using a computerized text analysis tool, Linguistic Inquiry Word Count ( LIWC ), to explore how his linguistic style was influenced by the social climate. The Pearson’s correlations and structural equation modeling results show, during times of prosperity, Chairman Mao’s linguistic style increased in cognitive complexity, social representations and social coordination.

The automated analysis of facial expressions has been widely used in different research areas, such as biometrics or emotional analysis. Special importance is attached to facial expressions in the area of sign language, since they help to form the grammatical structure of the language and allow for the creation of language disambiguation, and thus are called Grammatical Facial Expressions (GFEs). In this paper we outline the recognition of GFEs used in the Brazilian Sign Language. In order to reach this objective, we have captured nine types of GFEs using a KinectTM sensor, designed a spatial-temporal data representation, modeled the research question as a set of binary classification problems, and employed a Machine Learning technique.

We study in this paper the cohesion of a leader's speeches over time. This is part of a larger project that aims at investigating the language of leaders and how their language changes over their stay in power. Here, we analyze the speeches of a leader who stayed in power for a long period of time, i.e. more than 30 years. We measure cohesion of speeches in the original language, which is Arabic in our case, as well as in English, based on human translations of the original speeches. The cohesion is measured in two different ways: using word overlap and Latent Semantic Analysis. Because of the morphological complexity of Arabic, the word overlap measure of cohesion becomes challenging in Arabic. Latent Semantic Analysis, which is totally unsupervised, is applied similarly for Arabic and English. The results show that cohesion has a general down trend over time and that during and after major crises the leader's speeches exhibit an increas in cohesion which can be explained as an attempt on leader's behalf to make his policies more clear, most likely as a form of post-crisis management.

We have designed an academic advising online system to advise college students using natural language conversations. The system embeds knowledge of current and future teaching schedules, degree requirements, course prerequisites and various administrative procedures. While this information can be found by searching several university websites and catalogs, students continually ask human advisors these questions during their limited face-to-face time, which limits deeper developmental and educative advising that is only available from human advisors. Our system enhances the advising experience by offering a source for instant academic advice that does not require student training or additional human resources. The system contains a pattern-matching dialog management system with access via a web browser. We describe the motivation for our system, the design requisites, our approach for deployment, and analyze results from real-world field tests.

Electronic discovery is an interesting subproblem of information retrieval in which one identifies documents that are potentially relevant to issues and facts of a legal case from an electronically stored document collection (a corpus). In this paper, we consider representing documents in a topic space using the well-known topic models such as latent Dirichlet allocation and latent semantic indexing, and solving the information retrieval problem via finding document similarities in the topic space rather doing it in the corpus vocabulary space. We also develop an iterative SMART ranking and categorization framework including human-in-the-loop to label a set of seed (training) documents and using them to build a semi-supervised binary document classification model based on Support Vector Machines. To improve this model, we propose a method for choosing seed documents from the whole population via an active learning strategy. We report the results of our experiments on a real dataset in the electronic discovery domain.

In general, the goal of the track is to bring together an international group of scientists to present current research, design, and empirical evaluations of their tutoring systems. This track is meant to inform researchers on the recent developments in both the design and evaluation of tutoring sys- tems.

K-means is undoubtedly the most widely used partitional clustering algorithm. Unfortunately, this algorithm is highly sensitive to the initial selection of the cluster centers. Numerous initialization methods have been proposed to address this drawback. Many of these methods, however, have superlinear complexity in the number of data points, which makes them impractical for large data sets. On the other hand, linear methods are often random and/or sensitive to the order in which the data points are processed. These methods are generally unreliable in that the quality of their results is unpredictable. In this paper, we propose a linear, deterministic, and order-invariant initialization method based on multidimensional histograms. Experiments on a diverse collection of data sets from the UCI Machine Learning Repository demonstrate the superiority of our method over the well-known maximin method.

The existing applications in Natural Language Process (NLP) weren’t reliable. Indeed, many complex phenomena are not treated completely such as relatives, juxtaposition, ellipsis and the coordination, essentially for Arabic grammar. In fact, the problems encountered relate to the choice of the formalism and the parser validating the constructed grammar. Therefore, our work aims to appreciate the different forms of Arabic coordination. Moreover, we plan to represent them with an adequate formalism, the Head-driven Phrase Structure (HPSG). The constructed grammar was validated with Linguistic Knowledge Builder (LKB), a parser generator system.

We report the results of an empirical evaluation of structural simplification of Bayesian networks by removing weak arcs. We conduct a series of experiments on six networks built from real data sets selected from the UC Irvine Machine Learning Repository. We systematically remove arcs from the weakest to the strongest, relying on four measures of arc strength, and measure the classification accuracy of the resulting simplified models. Our results show that removing up to roughly 20 percent of the weakest arcs in a network has minimal effect on its classification accuracy. At the same time, structural simplification of networks leads to significant reduction of both the amount of memory taken by the clique tree and the amount of computation needed to perform inference.

Construction of agent based model systems is often difficult considering the dynamic and complex nature of real world problems and various implicit factors affecting their behavior. This presents a problem in building accurate and valid systems for use as decision support tools. In this paper, we present an approach for handling some of these factors, specifically acceptance rate, retention rate and social influence, and enable simulated social agent models to evolve in a dynamic environment.The objective is to achieve a more reliable behavior and outcome for the agents in the simulation, despite unpredictable environmental changes. The agents’ knowledge, in terms of their observed responses to the environment and corresponding outcomes, is captured in a semantic tree. A metric is used to detect changes in the environment and the threshold of response of the agent, thereby triggering the agent to adapt its decision tree to maintain a reasonable response beyond its historical knowledge. The results reveal the ability of agents to detect changes in the environment more quickly and with better accuracy, using a case study, and as a result learn to adapt by modifying their decision tree under the influence of considered factors.