A semantic graph based approach on interest extraction from user generated texts in social media

Abstract

Micro-blogs and social networking websites have become a platform for self-expression. They reflect the thoughts, ideas and opinions of people on various subjects. Analyzing these texts to find the main topics mentioned in them is a fine method for targeted marketing. Targeted marketing involves identifying potential clients who might be interested in particular products or services and marketing them to these clients. Examples of targeted marketing include recommender systems and targeted advertising. Identifying the personal interests of users is a major factor that determines the quality of such systems. Commonly used techniques monitor online behavior of users like purchase histories, product views etc. or explicitly collect the user's interests through surveys and rating systems. However there have been only a few attempts to use user generated texts as a source for analyzing personal interests and preferences. This paper proposes a semantic graph based method to identify the likes and interests of users by analyzing their twitter feeds. It also put forward the design for a recommender system that can work along with the proposed interest extraction method. This method is purely based on the texts that a user leaves in a particular social network website or a micro blog. Unlike the other conventional methods there is no need to track the user activity on the Internet or conduct exclusive surveys and ratings to collect explicit ideas from the user.