Dual Scaling in Data Mining from Text Databases
The availability of multimedia text document information has disseminated text mining among researchers. Text documents, integrate numerical and linguistic data, making text mining interesting and challenging. We propose text mining based on a fuzzy quantification model and fuzzy thesaurus. In text mining, we focus on: 1) Sentences included in Japanese text that are broken down into words. 2) Fuzzy thesaurus for finding words matching keywords in text. 3) Fuzzy multivariate analysis to analyze semantic meaning in predefined case studies. We use a fuzzy thesaurus to translate words using Chinese and Japanese characters into keywords. This speeds up processing without requiring a dictionary to separate words. Fuzzy multivariate analysis is used to analyze such processed data and to extract latent mutual related structures in text data, i.e., to extract otherwise obscured knowledge. We apply dual scaling to mining library and Web page text information, and propose integrating the result in Kansei engineering for possible application in sales, marketing, and production.