Text analytics of social media: Sentiment analysis, event detection and summarization

In the last decade, large numbers of social media services have emerged and been widely used in people's daily life as important information sharing and acquisition tools. With a substantial amount of user-contributed text data on social media, it becomes a necessity to develop methods and tools for text analysis for this emerging data, in order to better utilize it to deliver meaningful information to users. ^ Previous work on text analytics in last several decades is mainly focused on traditional types of text like emails, news and academic literatures, and several critical issues to text data on social media have not been well explored: 1) how to detect sentiment from text on social media; 2) how to make use of social media's real-time nature; 3) how to address information overload for flexible information needs. ^ In this dissertation, we focus on these three problems. First, to detect sentiment of text on social media, we propose a non-negative matrix tri-factorization (tri-NMF) based dual active supervision method to minimize human labeling efforts for the new type of data. Second, to make use of social media's real-time nature, we propose approaches to detect events from text streams on social media. Third, to address information overload for flexible information needs, we propose two summarization framework, dominating set based summarization framework and learning-to-rank based summarization framework. The dominating set based summarization framework can be applied for different types of summarization problems, while the learning-to-rank based summarization framework helps utilize the existing training data to guild the new summarization tasks. In addition, we integrate these techneques in an application study of event summarization for sports games as an example of how to better utilize social media data. ^

Text Analytics of Social Media: Sentiment Analysis, Event Detection and Summarization

In the last decade, large numbers of social media services have emerged and been widely used in people's daily life as important information sharing and acquisition tools. With a substantial amount of user-contributed text data on social media, it becomes a necessity to develop methods and tools for text analysis for this emerging data, in order to better utilize it to deliver meaningful information to users. Previous work on text analytics in last several decades is mainly focused on traditional types of text like emails, news and academic literatures, and several critical issues to text data on social media have not been well explored: 1) how to detect sentiment from text on social media; 2) how to make use of social media's real-time nature; 3) how to address information overload for flexible information needs. In this dissertation, we focus on these three problems. First, to detect sentiment of text on social media, we propose a non-negative matrix tri-factorization (tri-NMF) based dual active supervision method to minimize human labeling efforts for the new type of data. Second, to make use of social media's real-time nature, we propose approaches to detect events from text streams on social media. Third, to address information overload for flexible information needs, we propose two summarization framework, dominating set based summarization framework and learning-to-rank based summarization framework. The dominating set based summarization framework can be applied for different types of summarization problems, while the learning-to-rank based summarization framework helps utilize the existing training data to guild the new summarization tasks. In addition, we integrate these techneques in an application study of event summarization for sports games as an example of how to better utilize social media data.

Actionable pattern analysis in computing and business intelligence systems

With advances in science and technology, computing and business intelligence (BI) systems are steadily becoming more complex with an increasing variety of heterogeneous software and hardware components. They are thus becoming progressively more difficult to monitor, manage and maintain. Traditional approaches to system management have largely relied on domain experts through a knowledge acquisition process that translates domain knowledge into operating rules and policies. It is widely acknowledged as a cumbersome, labor intensive, and error prone process, besides being difficult to keep up with the rapidly changing environments. In addition, many traditional business systems deliver primarily pre-defined historic metrics for a long-term strategic or mid-term tactical analysis, and lack the necessary flexibility to support evolving metrics or data collection for real-time operational analysis. There is thus a pressing need for automatic and efficient approaches to monitor and manage complex computing and BI systems. To realize the goal of autonomic management and enable self-management capabilities, we propose to mine system historical log data generated by computing and BI systems, and automatically extract actionable patterns from this data. This dissertation focuses on the development of different data mining techniques to extract actionable patterns from various types of log data in computing and BI systems. Four key problems—Log data categorization and event summarization, Leading indicator identification , Pattern prioritization by exploring the link structures , and Tensor model for three-way log data are studied. Case studies and comprehensive experiments on real application scenarios and datasets are conducted to show the effectiveness of our proposed approaches.

Business process quantitative analysis and optimization for service industries

The total time a customer spends in the business process system, called the customer cycle-time, is a major contributor to overall customer satisfaction. Business process analysts and designers are frequently asked to design process solutions with optimal performance. Simulation models have been very popular to quantitatively evaluate the business processes; however, simulation is time-consuming and it also requires extensive modeling experiences to develop simulation models. Moreover, simulation models neither provide recommendations nor yield optimal solutions for business process design. A queueing network model is a good analytical approach toward business process analysis and design, and can provide a useful abstraction of a business process. However, the existing queueing network models were developed based on telephone systems or applied to manufacturing processes in which machine servers dominate the system. In a business process, the servers are usually people. The characteristics of human servers should be taken into account by the queueing model, i.e. specialization and coordination. ^ The research described in this dissertation develops an open queueing network model to do a quick analysis of business processes. Additionally, optimization models are developed to provide optimal business process designs. The queueing network model extends and improves upon existing multi-class open-queueing network models (MOQN) so that the customer flow in the human-server oriented processes can be modeled. The optimization models help business process designers to find the optimal design of a business process with consideration of specialization and coordination. ^ The main findings of the research are, first, parallelization can reduce the cycle-time for those customer classes that require more than one parallel activity; however, the coordination time due to the parallelization overwhelms the savings from parallelization under the high utilization servers since the waiting time significantly increases, thus the cycle-time increases. Third, the level of industrial technology employed by a company and coordination time to mange the tasks have strongest impact on the business process design; as the level of industrial technology employed by the company is high; more division is required to improve the cycle-time; as the coordination time required is high; consolidation is required to improve the cycle-time. ^

Statistical dialog management for health interventions

Research endeavors on spoken dialogue systems in the 1990s and 2000s have led to the deployment of commercial spoken dialogue systems (SDS) in microdomains such as customer service automation, reservation/booking and question answering systems. Recent research in SDS has been focused on the development of applications in different domains (e.g. virtual counseling, personal coaches, social companions) which requires more sophistication than the previous generation of commercial SDS. The focus of this research project is the delivery of behavior change interventions based on the brief intervention counseling style via spoken dialogue systems. ^ Brief interventions (BI) are evidence-based, short, well structured, one-on-one counseling sessions. Many challenges are involved in delivering BIs to people in need, such as finding the time to administer them in busy doctors' offices, obtaining the extra training that helps staff become comfortable providing these interventions, and managing the cost of delivering the interventions. Fortunately, recent developments in spoken dialogue systems make the development of systems that can deliver brief interventions possible. ^ The overall objective of this research is to develop a data-driven, adaptable dialogue system for brief interventions for problematic drinking behavior, based on reinforcement learning methods. The implications of this research project includes, but are not limited to, assessing the feasibility of delivering structured brief health interventions with a data-driven spoken dialogue system. Furthermore, while the experimental system focuses on harmful alcohol drinking as a target behavior in this project, the produced knowledge and experience may also lead to implementation of similarly structured health interventions and assessments other than the alcohol domain (e.g. obesity, drug use, lack of exercise), using statistical machine learning approaches. ^ In addition to designing a dialog system, the semantic and emotional meanings of user utterances have high impact on interaction. To perform domain specific reasoning and recognize concepts in user utterances, a named-entity recognizer and an ontology are designed and evaluated. To understand affective information conveyed through text, lexicons and sentiment analysis module are developed and tested.^

Statistical Dialog Management for Health Interventions

Research endeavors on spoken dialogue systems in the 1990s and 2000s have led to the deployment of commercial spoken dialogue systems (SDS) in microdomains such as customer service automation, reservation/booking and question answering systems. Recent research in SDS has been focused on the development of applications in different domains (e.g. virtual counseling, personal coaches, social companions) which requires more sophistication than the previous generation of commercial SDS. The focus of this research project is the delivery of behavior change interventions based on the brief intervention counseling style via spoken dialogue systems. Brief interventions (BI) are evidence-based, short, well structured, one-on-one counseling sessions. Many challenges are involved in delivering BIs to people in need, such as finding the time to administer them in busy doctors' offices, obtaining the extra training that helps staff become comfortable providing these interventions, and managing the cost of delivering the interventions. Fortunately, recent developments in spoken dialogue systems make the development of systems that can deliver brief interventions possible. The overall objective of this research is to develop a data-driven, adaptable dialogue system for brief interventions for problematic drinking behavior, based on reinforcement learning methods. The implications of this research project includes, but are not limited to, assessing the feasibility of delivering structured brief health interventions with a data-driven spoken dialogue system. Furthermore, while the experimental system focuses on harmful alcohol drinking as a target behavior in this project, the produced knowledge and experience may also lead to implementation of similarly structured health interventions and assessments other than the alcohol domain (e.g. obesity, drug use, lack of exercise), using statistical machine learning approaches. In addition to designing a dialog system, the semantic and emotional meanings of user utterances have high impact on interaction. To perform domain specific reasoning and recognize concepts in user utterances, a named-entity recognizer and an ontology are designed and evaluated. To understand affective information conveyed through text, lexicons and sentiment analysis module are developed and tested.

The role of individualism and collectivism as predictors of attributions for unethical work behavior: An empirical examination across two culturally diverse groups

This dissertation addresses how the cultural dimensions of individualism and collectivism affect the attributions people make for unethical behavior at work. The moderating effect of ethnicity is also examined by considering two culturally diverse groups: Hispanics and Anglos. The sample for this study is a group of business graduate students from two universities in the Southeast. A 20-minute survey was distributed to master's degree students at their classroom and later on returned to the researcher. Individualism and collectivism were operationalized as by a set of attitude items, while unethical work behavior was introduced in the form of hypothetical descriptions or scenarios. Data analysis employed multiple group confirmatory factor analysis for both independent and dependent variables, and subsequently multiple group LISREL models, in order to test predictions. Results confirmed the expected link between cultural variables and attribution responses, although the role of independent variables shifted, due to the moderating effect of ethnicity, and to the nuances of each particular situation. ^

Team diversity and leadership: Tests of model of process and direct effects

A model was tested to examine relationships among leadership behaviors, team diversity, and team process measures with team performance and satisfaction at both the team and leader-member levels of analysis. Relationships between leadership behavior and team demographic and cognitive diversity were hypothesized to have both direct effects on organizational outcomes as well as indirect effects through team processes. Leader member differences were investigated to determine the effects of leader-member diversity leader-member exchange quality, individual effectiveness and satisfaction.^ Leadership had little direct effect on team performance, but several strong positive indirect effects through team processes. Demographic Diversity had no impact on team processes, directly impacted only one performance measure, and moderated the leadership to team process relationship.^ Cognitive Diversity had a number of direct and indirect effects on team performance, the net effects uniformly positive, and did not moderate the leadership to team process relationship.^ In sum, the team model suggests a complex combination of leadership behaviors positively impacting team processes, demographic diversity having little impact on team process or performance, cognitive diversity having a positive net impact impact, and team processes having mixed effects on team outcomes.^ At the leader-member level, leadership behaviors were a strong predictor of Leader-Member Exchange (LMX) quality. Leader-member demographic and cognitive dissimilarity were each predictors of LMX quality, but failed to moderate the leader behavior to LMX quality relationship. LMX quality was strongly and positively related to self reported effectiveness and satisfaction.^ The study makes several contributions to the literature. First, it explicitly links leadership and team diversity. Second, demographic and cognitive diversity are conceptualized as distinct and multi-faceted constructs. Third, a methodology for creating an index of categorical demographic and interval cognitive measures is provided so that diversity can be measured in a holistic conjoint fashion. Fourth, the study simultaneously investigates the impact of diversity at the team and leader-member levels of analyses. Fifth, insights into the moderating impact of different forms of team diversity on the leadership to team process relationship are provided. Sixth, this study incorporates a wide range of objective and independent measures to provide a 360$\sp\circ$ assessment of team performance. ^

The Use of the Military for Law Enforcement Activities in Central America

Faced with the violence, criminality and insecurity now threatening peace and democratic governance in Central America, the region’s governments have decided to use the Armed Forces to carry out actions in response to criminal actions, looking to improve their performance. Although public demand for including the Armed Forces in these functions takes place within a legally legitimate framework, it is motivated by tangible circumstances such as increased levels of violence, delinquency and crime. Despite being coupled with the perception of institutional weakness within the security and judicial system (particularly police) and the recognition of prestige, efficiency, discipline and severity in fulfilling the Armed Forces’ missions, these arguments are insufficient to legitimize the use of the military as a police force. Within this context, this paper reflects on the implications or consequences of the use of the Armed Forces in duties traditionally assigned to the police in the Central American region with the goal of contributing to the debate on this topic taking place in the Americas. To achieve this end, first we will focus on understanding the actual context in which a decision is made to involve the Armed Forces in security duties in the region. Second, we will examine the effects and implications of this decision on the Armed Forces’ relations within their respective societies. Third and finally, considering this is already a reality in the region, this paper will provide recommendations. The main findings of this research, resulting from the application of an analyticaldescriptive and historically based study, are organized in three dimensions: the political dimension, by implication referring to the relationship between the ultimate political authority and the Armed Forces; the social dimension, by implication the opinion of citizens; and other implications not only affecting the structural and cultural organization of armies and police but also the complementary operational framework within a context of comprehensive response by the State. As a main conclusion, it poses there is an environment conducive to the use of the Armed Forces in citizen’s security, in view of the impact of threats provoked by criminal structures of a military nature currently operating in Central America. However, this participation creates an inevitable social and political impact if implemented in isolation or given a political leading role and/or operational autonomy. This participation poses risks to the institutions of the Armed Forces and the police as well. Finally, this paper identifies an urgent need for the Armed Forces’ role to be more clearly defined with regard to security matters, limiting it to threats that impact States’ governability and existence. Nonetheless, Central American States should seek a COMPREHENSIVE response to current crime and violence, using all necessary institutions to confront these challenges, but with defined roles and responsibilities for each and dynamic coordination to complement their actions.

Analysis and comparison of the moral development of students required to graduate with an ethics course

Accounting students become practitioners facing ethical decision-making challenges that can be subject to various interpretations; hence, the profession is concerned with the appropriateness of their decisions. Moral development of these students has implications for a profession under legal challenges, negative publicity, and government scrutiny. Accounting students' moral development has been studied by examining their responses to moral questions in Rest's Defining Issues Test (DIT), their professional attitudes on Hall's Professionalism Scale Dimensions, and their ethical orientation-based professional commitment and ethical sensitivity. This study extended research in accounting ethics and moral development by examining students in a college where an ethics course is a requirement for graduation. ^ Knowledge of differences in the moral development of accounting students may alert practitioners and educators to potential problems resulting from a lack of ethical understanding as measured by moral development levels. If student moral development levels differ by major, and accounting majors have lower levels than other students, the conclusion may be that this difference is a causative factor for the alleged acts of malfeasance in the profession that may result in malpractice suits. ^ The current study compared 205 accounting, business, and nonbusiness students from a private university. In addition to academic major and completion of an ethics course, the other independent variable was academic level. Gender and age were tested as control variables and Rest's DIT score was the dependent variable. The primary analysis was a 2 x 3 x 3 ANOVA with post hoc tests for results with significant p-value of less than 0.05. ^ The results of this study reveal that students who take an ethics course appear to have a higher level of moral development (p = 0.013), as measured by the (DIT), than students at the same academic level who have not taken an ethics course. In addition, a statistically significant difference (p = 0.034) exists between freshmen who took an ethics class and juniors who did not take an ethics class. For every analysis except one, the lower class year with an ethics class had a higher level of moral development than the higher class year without an ethics class. These results appear to show that ethics education in particular has a greater effect on the level of moral development than education in general. Findings based on the gender specific analyses appear to show that males and females respond differently to the effects of taking an ethics class. The male students do not appear to increase their moral development level after taking an ethics course (p = 0.693) but male levels of moral development differ significantly (p = 0.003) by major. Female levels of moral development appear to increase after taking an ethics course (p = 0.002). However, they do not differ according to major (p = 0.097). ^ These findings indicate that accounting students should be required to have a class in ethics as part of their college curriculum. Students with an ethics class have a significantly higher level of moral development. The challenges facing the profession at the current time indicate that public confidence in the reports of client corporations has eroded and one way to restore this confidence could be to require ethics training of future accountants. ^

Self-other rating discrepancy and personality antecedents

This study explored individual difference factors to help explain the discrepancy that has been found to exist between self and other ratings in prior research. Particularly, personality characteristics of the self-rater were researched in the current study as a potential antecedent for self-other rating agreement. Self, peer, and supervisor ratings were provided for global performance as well as five competencies specific to the organization being examined. Four rating tendency categories, over-raters, under-raters, in-agreement (good), and in-agreement (poor), established in research by Atwater and Yammarino were used as the basis of the current research. The sample for rating comparisons within the current study consisted of 283 self and supervisor dyads and 275 for self and peer dyads from a large financial organization. Measures included a custom multi-rater performance instrument and the personality survey instrument, ASSESS, which measures 20 specific personality characteristics. MANCOVAs were then performed on this data to examine if specific personality characteristics significantly distinguished the four rating tendency groups. Examination of all personality dimensions and overall performance uncovered significant findings among rating groups for self-supervisor rating comparisons but not for self-peer rating comparisons. Examination of specific personality dimensions for self-supervisory ratings group comparisons and overall performance showed Detail Interest to be an important characteristic among the hypothesized variables. For self-supervisor rating comparisons and specific competencies, support was found for the hypothesized personality dimension of Fact-based Thinking which distinguished the four rating groups for the competency, Builds Relationships. For both self-supervisor and self-peer rating comparisons, the competencies, Builds Relationships and Leads in a Learning Environment, were found to have significant relationship with several personality characteristics, however, these relationships were not consistent with the hypotheses in the current study. Several unhypothesized personality dimensions were also found to distinguish rating groups for both self-supervisor and self-peer comparisons on overall performance and various competencies. Results of the current study hold implications for the training and development session that occur after a 360-degree evaluation process. Particularly, it is suggested that feedback sessions may be designed according to particular rating tendencies to maximize the interpretation, acceptance and use of evaluation information. ^

A retrospective-longitudinal examination of the relationship between apportionment of seat time in community-college algebra courses and student academic performance

During the past decade, there has been a dramatic increase by postsecondary institutions in providing academic programs and course offerings in a multitude of formats and venues (Biemiller, 2009; Kucsera & Zimmaro, 2010; Lang, 2009; Mangan, 2008). Strategies pertaining to reapportionment of course-delivery seat time have been a major facet of these institutional initiatives; most notably, within many open-door 2-year colleges. Often, these enrollment-management decisions are driven by the desire to increase market-share, optimize the usage of finite facility capacity, and contain costs, especially during these economically turbulent times. So, while enrollments have surged to the point where nearly one in three 18-to-24 year-old U.S. undergraduates are community college students (Pew Research Center, 2009), graduation rates, on average, still remain distressingly low (Complete College America, 2011). Among the learning-theory constructs related to seat-time reapportionment efforts is the cognitive phenomenon commonly referred to as the spacing effect, the degree to which learning is enhanced by a series of shorter, separated sessions as opposed to fewer, more massed episodes. This ex post facto study explored whether seat time in a postsecondary developmental-level algebra course is significantly related to: course success; course-enrollment persistence; and, longitudinally, the time to successfully complete a general-education-level mathematics course. Hierarchical logistic regression and discrete-time survival analysis were used to perform a multi-level, multivariable analysis of a student cohort (N = 3,284) enrolled at a large, multi-campus, urban community college. The subjects were retrospectively tracked over a 2-year longitudinal period. The study found that students in long seat-time classes tended to withdraw earlier and more often than did their peers in short seat-time classes (p < .05). Additionally, a model comprised of nine statistically significant covariates (all with p-values less than .01) was constructed. However, no longitudinal seat-time group differences were detected nor was there sufficient statistical evidence to conclude that seat time was predictive of developmental-level course success. A principal aim of this study was to demonstrate—to educational leaders, researchers, and institutional-research/business-intelligence professionals—the advantages and computational practicability of survival analysis, an underused but more powerful way to investigate changes in students over time.

A Retrospective-Longitudinal Examination of the Relationship between Apportionment of Seat Time in Community-College Algebra Courses and Student Academic Performance

During the past decade, there has been a dramatic increase by postsecondary institutions in providing academic programs and course offerings in a multitude of formats and venues (Biemiller, 2009; Kucsera & Zimmaro, 2010; Lang, 2009; Mangan, 2008). Strategies pertaining to reapportionment of course-delivery seat time have been a major facet of these institutional initiatives; most notably, within many open-door 2-year colleges. Often, these enrollment-management decisions are driven by the desire to increase market-share, optimize the usage of finite facility capacity, and contain costs, especially during these economically turbulent times. So, while enrollments have surged to the point where nearly one in three 18-to-24 year-old U.S. undergraduates are community college students (Pew Research Center, 2009), graduation rates, on average, still remain distressingly low (Complete College America, 2011). Among the learning-theory constructs related to seat-time reapportionment efforts is the cognitive phenomenon commonly referred to as the spacing effect, the degree to which learning is enhanced by a series of shorter, separated sessions as opposed to fewer, more massed episodes. This ex post facto study explored whether seat time in a postsecondary developmental-level algebra course is significantly related to: course success; course-enrollment persistence; and, longitudinally, the time to successfully complete a general-education-level mathematics course. Hierarchical logistic regression and discrete-time survival analysis were used to perform a multi-level, multivariable analysis of a student cohort (N = 3,284) enrolled at a large, multi-campus, urban community college. The subjects were retrospectively tracked over a 2-year longitudinal period. The study found that students in long seat-time classes tended to withdraw earlier and more often than did their peers in short seat-time classes (p < .05). Additionally, a model comprised of nine statistically significant covariates (all with p-values less than .01) was constructed. However, no longitudinal seat-time group differences were detected nor was there sufficient statistical evidence to conclude that seat time was predictive of developmental-level course success. A principal aim of this study was to demonstrate—to educational leaders, researchers, and institutional-research/business-intelligence professionals—the advantages and computational practicability of survival analysis, an underused but more powerful way to investigate changes in students over time.