A framework for opinion question answering

Question answering is a useful task to help people seek the knowledge of what they want to know. Previous study mainly focuses on factoid question answering, which serves the needs to answer factual questions. Due to rapidly increasing scale of user generated contents on the Web, people are more interested in opinion questions that can reflect others' opinions. In this paper, we propose a framework for opinion question answering by combining opinion mining with traditional question answering methods. Besides, we use question-answer opinion patterns to extract and rank candidate answers from text snippets. Experimental results on TREC Blog08 dataset reveal the potential of our framework.