Automatic Impression Indexing based on Appraisal Dictionary from Tweet

Abstract

With the development of social media, interest in text mining has been growing steadily during the past few years. Although these research studies have been conducted on product development and marketing, no research has been conducted to explore cultures that create new value not found in existing products. Therefore, we propose a method of extracting topics related to specic product domains and indexing impressions from tweets containing a wide range of descriptions of cultures and values. Our method is based on the appraisal dictionary as well as the structures of impressions for the target domain to become clear. A result of our method has been applied to the lawn culture. We extracted 13 ground topics and the impressions related to each of them. Moreover one of them was an unexpected concept. This result indicates the usefulness of the method in exploring a new concept of value related to a specic field.