An Approach for Sentiment Based Product-Feature Diversification of User Generated Reviews

Information shared online by the web users, is increasingly becoming a good source for others to learn from it. In many cases, the shared information is reflecting users’ experiences and opinions about events or use of certain products. The volume of this shared information is huge. It is humanly very time consuming to read all this information and make an informed decision. The challenge is to analyze shared contents automatically, find dimensions of discussions, associated opinions and summarize them so that the user can have an extensive overview of the information for decision making. The research work presented in this paper addresses this challenge of information diversification by using probabilistic topic modeling and opinion extraction techniques. The proposed method automatically extracts dimensions of a particular discussion and combines it with the opinions presented in the discussion for information diversification purpose. Experiments on the real-world dataset indicates that our method is able to extract dimensions of a discussion and successfully associate it with the opinions expressed against these dimensions. The method presents users with both