A method for automatically generating the emotional vectors of emoticons using weblog articles

In recent years, reputation analysis services using weblogs, message boards, and community web sites have been developed. To improve the accuracy of the reputation analysis, we have to extract emotions or reactions of writers of documents accurately. Emoticons (Emojis) are often used in weblogs, and in many cases these emoticons have the role of modalities of writers of weblog articles. That is, to estimate emotions represented by emoticons is important for reputation analysis. In this study, we propose a method to create emotional vectors of emoticons automatically using the collocation relationship between emotional words and emoticons which is derived from many weblog articles.