Document Clustering using Color representation of emotional Words

Most of the previous research on document clustering concentrated on similarity of topics discussed in documents, using the frequency of occurrence of content words. On the other hand, our document clustering method focuses on similarity of authors' perspectives or opinions that were expressed in texts. This kind of methods has not been thoroughly investigated, however, we consider that it is one of the indispensable factors for actively utilizing documents such as Web documents. In our proposed method document representation was prepared by identifying emotional expressions in texts using a predefined dictionary, and assigning a color associated by each of the expressions. The proposed method performs document clustering using the color expression of emotional words. Keyword Document Clustering, Information Extraction, Emotion Extraction, Opinion Information 1. はじめに 本稿では,Web より収集した書評を対象として,書