Hideaki Takeda's Publication
- T. Fukuhara, H. Takeda and T. Nishida: Multiple-text
Summarization for Collective Knowledge Formation, in The 1999 IEEE
International Conference on Systems, Man, and Cybernetics, Tokyo (1999),
IEEE.
Multiple-text summarization method for facilitating a process of
the collective knowledge formation is proposed. There are enormous pieces of
information represented by WWW pages which contain various topics in early
stage of the community. To facilitate this process, organizing various
unordered pieces of information is needed. We facilitate this process by
multiple-text summarization. Proposed method consists of topic identification
method and context-based summarization method. Topic identification method
finds a topic including central information over the text set. We identify
partialities of topics based on skewness and kurtosis of a word frequency.
Context-based summarization method generates summaries by linking relevant
topics. Summarization in our approach is to find context based on theme and
focus of a sentence which represent central information within a sentence. We
developed a prototype system called Topic Showcase. Experimental results
demonstrate the availabilities of proposed methods in identifying topics from
each cluster and supporting users in forecasting contents of the texts.
Hideaki Takeda (National Institute of Informatics)