Search

Online health communities are a major source for patients and their informal caregivers in the process of gathering information and seeking social support. The Cancer Survivors Network of the American Cancer Society has many users and presents a large number of user interactions with regards to coping with cancer. Sentiment analysis is an important process in understanding members' needs and concerns and the impact of users' responses on other members. It aims to determine the participants' subjective attitude and reflect their emotions. Analyzing the sentiment of posts in online health communities enables the investigation of various factors such as what affects the sentiment change and discovery of sentiment change patterns. Since each writer has his or her own personality, and temporal emotional state, behavioral traits can be reflected in the writer's writing style. Pronouns are function-words which often convey some unique styling patterns into the texts. Drawing on a lexical approach to emotions, we conduct factor analysis on the use of pronouns in self-descriptions texts. Our analysis shows that the usage of pronouns has an effect on sentiment classification. Moreover, we evaluated the use of pronouns in our domain, and found it different than standard English usage.

As cyber-attacks become more sophisticated, cyber-attack analysts are required to process large amounts of network data and to reason under uncertainty with the aim of detecting cyber-attacks. Capturing and studying the fine-grained analysts' cognitive processes helps researchers gain deep understanding of how they conduct analytical reasoning and elicit their procedure knowledge and experience to further improve their performance. However, it's very challenging to conduct cognitive task analysis studies in cyber-attack analysis. To address the problem, we propose an integrated computer-aided data collection method for cognitive task analysis (CTA) which has three building blocks: a trace representation of the fine-grained cyber-attack analysis process, a computer tool supporting process tracing and a laboratory experiment for collecting traces of analysts' cognitive processes in conducting a cyber-attack analysis task. This CTA method integrates automatic capture and situated self-reports in a novel way to avoiding distracting analysts from their work and adding much extra work load. With IRB approval, we recruited thirteen full-time professional analysts and seventeen doctoral students specialized in cyber security in our experiment. We mainly employ the qualitative data analysis method to analyze the collected traces and analysts' comments. The results of the preliminary trace analysis turn out highly promising.

Sentiment analysis has been widely researched in the domain of online review sites with the aim of getting summarized opinions of product users about different aspects of the products. However, there has been little work focusing on identifying the polarity of sentiments expressed by users in online health communities such as cancer support forums, etc. Online health communities act as a medium through which people share their health concerns with fellow members of the community and get social support. Identifying sentiments expressed by members in a health community can be helpful in understanding dynamics of the community such as dominant health issues, emotional impacts of interactions on members, etc. In this work, we perform sentiment classification of user posts in an online cancer support community (Cancer Survivors Network). We use Domain-dependent and Domain-independent sentiment features as the two complementary views of a post and use them for post classification in a semi-supervised setting using the co-training algorithm. Experimental results demonstrate effectiveness of our methods.

Citation counts have been widely used in a digital library for purposes such as ranking scientific publications and evaluating patents. This paper demonstrates that distinguishing different types of citations could rank better for these purposes. We differentiate patent citations along two dimensions (assignees and technologies) into four types, and propose a weighted citation approach for assessing and ranking patents. We investigate five weight learning methods and compare their performance. Our weighted citation method performs consistently better than simple citation counts, in terms of rank correlations with patent renewal status. The estimated weights on different citations are consistent with economic insights on patent citations. Our study points to an interesting and promising research line on patent citation and network analysis that has not been explored.

Identifying a good set of features is critical to the performance of learning algorithms such as classifiers. Previous methods have focused on either selecting a subset of features or transforming features using principle components analysis. In this paper, we propose a genetic algorithm approach that searches for a good feature transformation function over a subset of features using a novel representation scheme with novel reproduction operators. Preliminary experimental results using the UCI data set show promising results.

Motivation - To investigate ways to support human-automation teams with real-world, imperfect automation where many system failures are the result of systematic failure.Research approach - An experimental approach was used to investigate how variance in agent reliability may influence human's trust and subsequent reliance on agent's decision aids. Sixty command and control (C2) teams, each consisting of a human operator and two cognitive agents, were asked to detect and respond to battlefield threats in six ten-minute scenarios. At the end of each scenario, participants completed the SAGAT queries, followed by the NASA TLX queries.Findings/Design - Results revealed that teams with experienced human operators accepted significantly less inappropriate recommendations from agents than teams with inexperienced operators. More importantly, the knowledge of agent's reliability and the ratio of unreliable tasks have significant effects on human's trust, as manifested in both team performance and human operators' rectification of inappropriate recommendations from agents.Originality/Value - It represents an important step toward uncovering the nature of human trust in human-agent collaboration.Take away message - This research has shown that given even minimal basis for understanding when the operator should and should not trust the agent recommendations allows operators to make better AUDs, to have better situation awareness on the critical issues associated with automation error, and to establish better trust in intelligent agents.

Topic detection and tracking and topic segmentation play an important role in capturing the local and sequential information of documents. Previous work in this area usually focuses on single documents, although similar multiple documents are available in many domains. In this paper, we introduce a novel unsupervised method for shared topic detection and topic segmentation of multiple similar documents based on mutual information (MI) and weighted mutual information (WMI) that is a combination of MI and term weights. The basic idea is that the optimal segmentation maximizes MI (or WMI). Our approach can detect shared topics among documents. It can find the optimal boundaries in a document, and align segments among documents at the same time. It also can handle single-document segmentation as a special case of the multi-document segmentation and alignment. Our methods can identify and strengthen cue terms that can be used for segmentation and partially remove stop words by using term weights based on entropy learned from multiple documents. Our experimental results show that our algorithm works well for the tasks of single-document segmentation, shared topic detection, and multi-document segmentation. Utilizing information from multiple documents can tremendously improve the performance of topic segmentation, and using WMI is even better than using MI for the multi-document segmentation.

Human team members often develop shared expectations to predict each other's needs and coordinate their behaviors. In this paper the concept "Shared Belief Map" is proposed as a basis for developing realistic shared expectations among a team of Human-Agent-Pairs (HAPs). The establishment of shared belief maps relies on inter-agent information sharing, the effectiveness of which highly depends on agents' processing loads and the instantaneous cognitive loads of their human partners. We investigate HMM-based cognitive load models to facilitate team members to "share the right information with the right party at the right time". The shared belief map concept and the cognitive/processing load models have been implemented in a cognitive agent architecture---SMMall. A series of experiments were conducted to evaluate the concept, the models, and their impacts on the evolving of shared mental models of HAP teams.

There has been much research investigating team cognition, naturalistic decision making, and collaborative technology as it relates to real world, complex domains of practice. However, there has been limited work in incorporating naturalistic decision making models for supporting distributed team decision making. The aim of this research is to support human decision making teams using cognitive agents empowered by a collaborative Recognition-Primed Decision model. In this paper, we first describe an RPD-enabled agent architecture (R-CAST), in which we have implemented an internal mechanism of decision-making adaptation based on collaborative expectancy monitoring, and an information exchange mechanism driven by relevant cue analysis. We have evaluated R-CAST agents in a real-time simulation environment, feeding teams with frequent decision-making tasks under different tempo situations. While the result conforms to psychological findings that human team members are extremely sensitive to their workload in high-tempo situations, it clearly indicates that human teams, when supported by R-CAST agents, can perform better in the sense that they can maintain team performance at acceptable levels in high time pressure situations.

Techniques that support effective communication during teamwork processes are of particular importance. Psychological study shows that an effective team often can anticipate information exchange among the team and communicate relevant information proactively. Proactive communication is crucial for understanding and sharing common goals and for cooperative actions. Communication can be valuable if it assists agents with new and timely information; it also has cost because it consumes network resources such as bandwidth. To address these issues, we present a new model that uses information production and need to capture the complex multi-agent communication process and a dynamic decision-theoretic determination of communication strategies. We also introduce a generic utility function and an algorithm, DTPC (Decision-Theoretic Proactive Communication), that focuses on representing information production and need of team members and resolving decision interactions among them for making decisions.

Keeping track of changes in user interests from a document stream with a few relevance judgments is not an easy task. To tackle this problem, we propose a novel method that integrates (1) pseudo-relevance feedback mechanism, (2) assumption about the persistence of user interests and (3) incremental method for data clustering. This approach has been empirically evaluated using Reuters-21578 corpus in a setting for information filtering. The experiment results reveal that it significantly improves the performances of existing user-interest-tracking systems without requiring additional, actual relevance judgments.

Effective human teams often benefit from proactivity through members' capability of anticipating different needs of teammates. In this paper, we focus on three issues related to the behavior of proactive information delivery. Two types of information needs are identified, the intentional semantics of ProInform is given, and priliminary experiments are carried out to study how different strategies for choosing ProInform impact team performance. The work presented in this paper provide a sound and practical framework that enables further studies regarding proactive information delivery.

In a multi-agent system, how agents accomplish a goal task is usually specified by multi-agent plans built from basic actions (e.g. operators) of which the agents are capable. A critical problem with such an approach is how can the designer make sure the plans are reliable. To tackle this problem, this paper presents a formal approach for modeling and analyzing multi-agent behaviors using Predicate/Transition (PrT) nets, a high-level formalism of Petri nets. We construct a multi-agent model by representing agent capabilities as transitions. To verify a multi-agent PrT model, we adapt the planning graphs as a compact structure for the reachability analysis. We also demonstrate that, based on the PrT model, whether parallel actions specified in multi-agent plans can be executed in parallel and whether the plans guarantee the achievement of the goal can be verified by analyzing the dependency relations among the transitions.

In this paper, we describe a new scheme to learn dynamic user's interests in an automated information filtering and gathering system running on the Internet. Our scheme is aimed to handle multiple domains of long-term and short-term user's interests simultaneously, which is learned through positive and negative user's relevance feedback. We developed a 3-descriptor approach to represent the user's interest categories. Using a learning algorithm derived for this representation, our scheme adapts quickly to significant changes in user interest, and is also able to learn exceptions to interest categories.

The emergence of, what is now called, ‘emotional intellige nce’ has revealed yet another aspect of human intelligence. Emot ions were shown to have a major impact on many of our everyday tasks, including decision-making, planning, communication, and behavior. Researchers have recently acknowledged this major role that emotions play, and thus we see a variety of models bei ng presented on simulating emotions in agents. However, emotion is not a simple process, it is often linked with many other pr ocesses, one of which is learning. As it has long been emphasized throug h psychology literature, memory and experience help shape and build the dynamic nature of the emotional process. In this pa per, we introduce PETEEI (a PET with Evolving Emotional Intelligence). PETEEI is a general model for simulating emotions in agents, with a particular emphasis on incorporating vario us learning mechanisms so that it can produce emotions accor ding to its own experience. Furthermore, it was modeled to recogniz e and cope with the various mood and emotional changes of its owne r. We have implemented PETEEI using fuzzy logic. An evaluation involving twenty-one subjects indicated that simulating the dynamic emotional process through learning provides a significantly more believable agent.