Manufacturing companies are concerned about ordinary customers™f opinions of their products. This paper presents an affect-based text mining system for Japanese that can aid in the analysis of customers™f reviews of commercial products. From free-text product reviews, the system creates an Adjective-Relation Database containing relations extracted between adjective and noun phrases. Using a pre-existing lexicon of general affect words together with scaled Positive/Negative evaluation, the system provides a quantified evaluation of modified-noun phrases appearing in the product survey. In addition, using correspondence analysis techniques, an Affect Map reflecting the preferences of target customers is created displaying the relations between user profile data and positive and negative affect-laden words. In a practical experiment, we used the results of such correspondence analysis between high-frequency adjective phrases and customer profiles on a questionnaire-based survey of opinions on a brand-name handbag, to further modify our affect lexicon, improving correspondence results.