A Unified Framework for Opinion Retrieval

The popularity of Web 2.0 has promoted the Web to be a valuable source for accessing opinions. Unfortunately, due to the large number of user generated content, it is difficult to access and utilize the opinion resource efficiently. Developing an opinion retrieval system is a promising way to overcome the problem of overloaded opinion information. In this paper, we propose a unified framework for opinion retrieval, which is based on generative model and opinion mining technologies. Within this framework, relevance, quality and temporal dimension information of user generated content are incorporated in a unified language model, and on the top of which an opinion summary is proposed. We have developed an opinion retrieval system, (ORS), to retrieve opinions in customer review domain. Our evaluation on a real-world data set shows that ORS can effectively retrieve, summarize and visualize customer opinions.