Takuya Kawada

Extracting Subjective and Objective Evaluative Expressions from the Web

There are various opinions on the Web, and analyzing them is an important task. Although many previous studies focused on analyzing subjective evaluative expressions, objective evaluative expressions which describe positive or negative facts are also informative information. In this paper, we study extraction and classification of subjective and objective evaluative expressions on Japanese Web documents using machine learning and evaluative word dictionaries.

Organizing information on the Web to support user judgments on information credibility

A vast amount of information and knowledge has been accumulated and circulated on the Web. They provide people with options regarding their daily lives and are starting to have a strong influence on governmental policies and business management. A crucial problem is that information on the Web is not necessarily credible. This paper describes an information analysis system called WISDOM, which assists users in assessing the credibility of information on the Web. WISDOM is to organize information on a given topic through the following three types of analyses: (1) extracting and contrasting opinions and important statements around the points related to the topic, (2) identifying and classifying the information sender of each page; and (3) analyzing the appearance of each page, for example, page design and writing style. Our preliminary evaluation indicates the effectiveness of WISDOM and its advantage to Google from the viewpoint of the ability of grasping the difference of information senders and opinions.

Web information analysis for open-domain decision support: system design and user evaluation

In this paper, we investigate the effectiveness of the system design of a Web information analysis for open-domain decision support. In order to make decisions, it is required to collect and compare information from various view points. In case of making decisions based on Web information, however, it is difficult to obtain diverse information from variety of sources by using current search engines. Based on this observation, we design a system for supporting open-domain decision making, which analyzes Web information. Among the major design decisions are to focus on two elements, i.e. identifying the source of information and the extraction of informative content, and to organize the two elements so that the user can quickly grasp who is saying what on the Web. The assumption behind such decisions is that information organized in such a way would facilitate proper judgments in the users decision making process. We conduct users evaluation to verify the effectiveness of our approach. In the result, it is confirmed that our system is superior to current search engine for grasping organized information from different stance of senders and supports the process of decision making, by (i) uncovering biases, (ii) showing various opinions from multiple view points, (iii) revealing information sources.

Topic relatedness in evaluative information extraction

The task of extracting opinions/evaluations related to a given topic from a large number of documents such as Web documents is crucial for developing an automatic evaluation finding system, which can handle a wide variety of topics as input. In this paper, we discuss the topic relatedness of extracted evaluation through analysis of a corpus we developed. We suggest here that the semantic relationship between the target of each extracted evaluation and a given topic helps in judging topic relatedness. In addition, we point out other factors that are beyond the analysis of topic-target relations for judging the topic relatedness of evaluation.

Summarizing evaluative information on the web for information credibility analysis

The World Wide Web comprises a wide variety of evaluative information. It consists of positive and negative opinions on innumerable topics from various perspectives, thus proving to be a useful information source for information credibility analysis. To present an informative and at-a-glance summary of any topic that a user of such an analysis system searches for, it is important to summarize many diverse evaluative expressions on the topic. In this paper, we describe a method for summarizing an extensive variety of evaluative expressions that are automatically extracted.

WISDOM: A web information analysis system

We present an information analysis system, WISDOM (Web Information Sensibly and Discreetly Ordered and Marshaled), which assists users in assessing the credibility of information on the Web from multiple viewpoints. A vast amount of information is now accumulated on the Web and becoming increasingly influential in peoples decision making. However, the Web contains a proliferation of false and misleading information. Thus, it is essential to study the methodologies and technologies that enable us to judge the credibility of information. It is for this reason that we have developed WISDOM, which organizes information on a given topic through the following three types of analysis: (1) extracting and contrasting important and controversial sentences around the points related to the topic, using semantics-oriented natural language processing (NLP) techniques; (2) identifying and classifying the authorship of each page; and (3) analyzing the appearance of each page, according to, for example, page design and writing style. WISDOM is in a state of continuous development, and in comparison to our previous demonstration at IUCS 2009, the current version of WISDOM is more advanced in the following respects. First, it extracts contrasting statements, in which statements contrasting the topic to other things, e.g., electric vehicles vs. hybrid electric vehicles. Second, it analyzes the expertise of information senders and indicates senders with high expertise with stars. Third, WISDOM analyzes a fresh 120 million Japanese Web pages; these pages are crawled and updated everyday by our crawler. WISDOM is available at http://wisdom-nict.go.jp/.