Il-Chul Moon

Mining social networks for personalized email prioritization

Email is one of the most prevalent communication tools today, and solving the email overload problem is pressingly urgent. A good way to alleviate email overload is to automatically prioritize received messages according to the priorities of each user. However, research on statistical learning methods for fully personalized email prioritization (PEP) has been sparse due to privacy issues, since people are reluctant to share personal messages and importance judgments with the research community. It is therefore important to develop and evaluate PEP methods under the assumption that only limited training examples can be available, and that the system can only have the personal email data of each user during the training and testing of the model for that user. This paper presents the first study (to the best of our knowledge) under such an assumption. Specifically, we focus on analysis of personal social networks to capture user groups and to obtain rich features that represent the social roles from the viewpoint of a particular user. We also developed a novel semi-supervised (transductive) learning algorithm that propagates importance labels from training examples to test examples through message and user nodes in a personal email network. These methods together enable us to obtain an enriched vector representation of each new email message, which consists of both standard features of an email message (such as words in the title or body, sender and receiver IDs, etc.) and the induced social features from the sender and receivers of the message. Using the enriched vector representation as the input in SVM classifiers to predict the importance level for each test message, we obtained significant performance improvement over the baseline system (without induced social features) in our experiments on a multi-user data collection. We obtained significant performance improvement over the baseline system (without induced social features) in our experiments on a multi-user data collection: the relative error reduction in MAE was 31% in micro-averaging, and 14% in macro-averaging.

Efficient extraction of domain specific sentiment lexicon with active learning

Proposed a graphical model to extract a sentiment lexicon with document annotations.Applied an active learning to extract a sentiment lexicon to reduce the annotation.Suggested and experimented four distinct initialization methods for active learners.Proposed lexicon coverage analysis algorithm to initialize the active learner. Recent research indicates that a sentiment lexicon focusing on a specific domain leads to better sentiment analyses compared to a general-purpose sentiment lexicon, such as SentiWordNet. In spite of this potential improvement, the cost of building a domain-specific sentiment lexicon hinders its wider and more practical applications. To compensate for this difficulty, we propose extracting a sentiment lexicon from a domain-specific corpus by annotating an intelligently selected subset of documents in the corpus. Specifically, the subset is selected by an active learner with initializations from diverse text analytics, i.e. latent Dirichlet allocation and our proposed lexicon coverage algorithm. This active learning produces a better domain-specific sentiment lexicon which results in a higher accuracy of the sentiment classification. Subsequently, we evaluate extracted sentiment lexicons by observing (1) the increased F1 measure in sentiment classifications and (2) the increased similarity to the sentiment lexicon with the full annotation. We expect that this contribution will enable more accurate sentiment classification by domain-specific sentiment lexicons with less sentiment tagging efforts.

Personalized Email Prioritization Based on Content and Social Network Analysis

The proposed system combines unsupervised clustering, social network analysis, semisupervised feature induction, and supervised classification to model user priorities among incoming email messages.

New insight into doctrine via simulation interoperation of heterogeneous levels of models in battle experimentation

As the complexity of military operations increases, defense modeling and simulation (DM&S) has contributed in analytically improving doctrines at various levels. To date, defense modelers concentrate on the best representation model of their targeted system at their targeted level with their targeted doctrine. However, insights into the doctrine from the battle experiment using such models are limited by the represented world of the model, and the modelers are missing potential insights into the doctrine that they might have gained if they had included more features in the models. Hence, this paper illustrates a battle experiment framework via the simulation interoperation of the heterogeneous levels of models. Our application is developing a mission-level doctrine for naval air defense scenarios, but a mission-level model alone does not represent the whole picture of the scenarios, and the model only represents the command and control (C2) procedures in detail, not the mechanical- and the engagement level features. On the other hand, an engagement-level model depicts some of the missing parts of the scenarios in the mission-level model. Our finding is that we can gain new insights, that is, an optimal decision-making timing of C2, into the mission-level doctrine from performing battle experiments by interoperating two such models at the mission and the engagement levels. We expect that this work will provide a new methodology for battle experiments by extending the limitation of single-model representation of the real world.

Detailed Analysis of Factors Affecting Team Success and Failure in the America's Army Game

We analyzed an extensive data trace of the on-line multi-player first-person-shooter game America’s Army to understand the traits of the social and dynamic networks present in the game. Analyses were performed at the player level, team level, and clan level. Statistical analysis methods are used to examine the data at those three levels. In addition, the dynamic social networks of the teams are examined using a variety of social network analysis methods. Particular focus is given to discovering and explaining winning strategies employed by game players. From the analyses, some ways to win the game are revealed: top America’s Army players’ distinct behaviors, the optimum size of an America’s Army team, the importance of fire volume toward opponent, the recommendable communication structure and content, and the contribution of the unity among the team members. Also, the analyses are compared to squad-level military research, and some similarities and differences are found.

Simulation-based analyses of an evacuation from a metropolis during a bombardment

The bombardment of a metropolis is considered a nightmare scenario. To reduce losses from such an assault, big cities have developed evacuation policies in case of bombardment. However, to build efficient evacuation policies, much footing data is required that considers both military and civilian views. Agent-based modeling and simulation could be utilized as a method to obtain the footing data. In this paper, we develop an evacuation agent-based model that describes a massive evacuation through the road network of a metropolis during a bombardment. In particular, our model took account of bombing strategies (i.e. the military view) as well as the characteristics of roads and evacuation agents (i.e. the civilian view) in order to analyze evacuations from both military and civilian perspectives. Moreover, we applied real data from a target region to calibrate parameters and initial conditions of the evacuation agent-based models, which increased the reliability of simulation results. Using the evacuation agent-based model, we designed and performed virtual experiments with varying military and civilian factors. Through the various analyses on the experiment results, we showed that our model could be a framework that provides footing data to develop efficient evacuation policies and preparations.

Analyzing social media in escalating crisis situations

The rapid diffusion of information and opinions through social media, such as web forums and micro-blogs, is affecting the development of crisis situations, such as the Iranian presidential election, the Egyptian protest, and the ROKS Cheonan sinking. Understanding this rapid widespread diffusion, and assessing what information is spreading, what ideas are becoming common, and who is talking about what, is critical for crisis management. This paper presents a computational system for social media assessing the flow of ideas on the web and changes in who is talking about what. This system, given raw social media data, identifies the key topics, the key paths by which topics evolve, the key individuals who contribute to the topic, and the key influence relations between the contributors. We present this system implemented with the Author-Topic model, the meta-network model, and various computational techniques to find and filter the heavy contributors and influences. We demonstrate the performance of the system, by applying it to social media data surrounding the ROKS Cheonan sinking. We describe the results of assessing the initial and changing perceptions of the event using this system.

Collaborative Work in Domain-Specific Discrete Event Simulation Software Development: Fleet Anti-air Defense Simulation Software

Modeing and simulation (M&S) is an important method to evaluate numerous designs and operational concepts for a real-world system. If a system to be modeled is domain-specific, developing the simulation software of the system will require domain knowledge about the system as well as understanding the M&S methodology. This paper describes M&S stakeholders and proposes a collaborative work process in the development of domain-specific simulation software. M&S stakeholders are persons with their own professional knowledge: subject matter experts (SME), domain engineers, M&S engineers, platform engineers, and simulation data analysts. The M&S process consists of eight activities from defining modeling objectives to analyzing simulation data, and diverse M&S stakeholders work together in each activity. The M&S process is applied to develop a real-world M&S software development experience in a military domain. Through the proposed collaborative work process, the capabilities of the M&S stakeholders can be utilized maximally by seamlessly separating yet correlating their works.

Simulation Experiment of Disaster Response Organizational Structures With Alternative Optimization Techniques

Disaster response operations are critical for decreasing the devastating impacts that result in casualties and property damages. Since these operations require cooperation in dynamic and complex situations, the responding organizations require a solid organizational structure collectively. This article introduces computational designs and evaluations of alternative organizational structures for disaster responses to resolve the disconnections between resource demands and supplies. In particular, this research consists of (1) organizational structure designs with two optimization techniques, (2) agent-based simulations that virtually replicate disaster response contexts, and (3) social network analysis to interpret the relations between the structures and the performances from the network perspectives. We applied this approach to log records of Hurricane Katrina, and our evaluations suggest that alternative organizations would improve operation outcomes, that is, increase the successful resource delivery counts and reduce a number of organizational conflicts. This computational approach could be further utilized in designing and evaluating organizations under complex and dynamic situations.

Temporal Issue Trend Identifications in Blogs

Many blog posts deal with current issues, so much attention has been paid to identifying topic trends in blogs. This paper suggests a new metric of selecting topic words. We empirically tested the accuracy and the performance of the metric with a massive blog corpus. First, we created blog site groups to their indegree influence. Second, we ran the metric with blog posts of each group. The test was encouraging because the metric identified key issues matching to the headlines of New York Times when it is applied to the top indegree blog group. We expect that this metric and the source grouping methods will be developed to a new topic analysis framework of a large blog corpus.