Opinion mining is a natural language processing technique which extracts subjective information from natural language text. To estimate an opinion about a query in large data collection, an opinion retrieval system that retrieves subjective and relevant information about the query can be useful. We present an opinion retrieval system that retrieves subjective and query-relevant tweets from Twitter, which is a useful source of obtaining real-time opinions. Our system outperforms previous opinion retrieval systems, and it further provides subjective information about Twitter authors and hashtags to describe their subjective tendencies.

In this paper, we propose a new probabilistic GLR parsing method that can solve the problems of conventional methods. Our proposed Conditional Action Model uses Surface Phrasal Types (SPTs) encoding the functional word sequences of the sub-trees for describing structural characteristics of the partial parse. And, the proposed GLR model outperforms the previous methods by about 6~8%.