Publikationen

Many IT workers suffer from stress in doing their work and there are a few counselors to help them. To cope with this, a context respectful counseling agent (CRECA) is proposed. This agent extracts emotional words from clients' utterances throughout their dialogue to detect emotion changes and provides clients such changes as a dialogue summary. If no change is detected, it replies with paraphrases of clients' utterances followed by context-respectful prompts to narrow problems. The summary responses promote the reflection of clients. This way, the counseling agent can pretend to keep recognizing clients' psychological sufferings. It behaves as if it empathized with clients and continues talking to clients without losing their trust. Keeping reflection on themselves, clients reach more problem clarification and self-awareness, which enables them to solve their problems. Since the agent provides only information from clients' sayings and the summaries focused on the change in their emotions, there occur few problems of knowledge explosion and knowledge maintenance. An experiment verifies that our new agent with summarization function more effective than the agent without summarization function (Old CRECA) and ELIZA.

To detect false information or rumors spread on Twitter on and after the Great East Japan Earthquake, a tweet credibility assessing method was proposed, based on the topic and opinion classification. The credibility is assessed by calculating the ratio of the same opinions to all opinions about a topic identified by topic models generated using Latent Dirichlet Allocation. To identify an opinion (positive or negative) about a tweet, sentiment analysis is performed using a semantic orientation dictionary. However, it is a kind of imbalanced data analysis to identify usually very few false tweets and the accuracy is a problem. The accuracy of the originally proposed method was susceptible since the sentiment opinion of most tweets was identified negative by the baseline (namely Takamura's) semantic orientation dictionary. To cope with this problem, a method for extracting sentiment orientations of words and phrases is also proposed to improve the evaluation for analyzing the credibility of tweet information. This method 1) evolutionally learns from a large amount of social data on Twitter, 2) focuses on adjective predicates, and 3) considers co-occurrences with negation expressions or multiple adjectives, between subjects and predicates, etc. The effects are proven by experiments using a large number of real tweets, in which we could detect rumor tweet much more accurately. In opposition to the baseline semantic dictionary, our method leads to succeed in imbalanced data analysis.

The paper introduces an improvement to an Educational Data Mining approach, which refrains from explicit learner modeling along with a recent refinement and evaluation. The technology models students' learning characteristics by considering real data instead of deriving their characteristics explicitly. It aims at mining course characteristics interdependencies of former students' study traces and utilizing them to optimize curricula of current students based to their performance traits revealed in their educational history. The recent refinement aims at increasing the sensitivity of the Data Mining technology by amplifying the influence of data, which shows interdependencies between the students' talents and weaknesses and weakening the influence of data from students, who perform about the same way in most courses (usually, very good or very poor in most subjects). Finally, the paper shows a validation approach by comparing the students' performance with the degree of similarity of their curriculum to the curriculum proposed by our technology.

Effective need collection is a major part of post-disaster assessment and recovery. A system for matching needs with available offers is an essential component for the recovery of communities in the aftermath of natural disasters. In the classic approach, the needs of disaster-stricken communities are collected by rescue personnel sampling a geographic grid and using emergency communication facilities. Subsequently, needs are matched to offers available at some central server and relief interventions are planned to deliver them. In this chapter, after an introduction to this approach, we explore the new notion of peer-to-peer community empowerment for some key activities of disaster management, including early warning systems, needs collection, and need-to-offer matching. Then, we present the design of a framework that leverages on the capacity for the communities to selforganize during crisis management, proposing advanced algorithms and techniques for need-offermatching. Our framework supports pre- and post-disaster use of social networks information and connectivity via an evolvable vocabulary, and supports metrics for resilience assessments and improvement.