Research On Question Correlation And Answer Ranking Based In Question Answering Community

In the Q & A community, user is not only the consumer of information, but also the creator. Large numbers of users created vast data, the content generated by user often appear difference between question and answer, meanwhile, some junk information also bring obstacle to the information that meet user need. Hence, it is challenging to filter and rank the question and answers. The goal of this paper is to improve the effectiveness to search information for users and share the knowledge in the Q & A community.This paper explores the ranking of questions and answers in the Q & A community. Because of the difference of manners between recommend questions and candidate answers, this paper didn't use the same methods to deal with the recommend questions, candidate answers and best answer. Recommend questions supply the relevant questions for users in order to help him to find the answer faster. This paper discusses the common method for question relevance and proposes 4 efficient text features, and experimented with them in the Q & A community. This method proved the relevance of the recommend questions, and proposes the ranking idea and proved the feasibility to collect sentence in the Q & A community. In the ranking research, the latent correlation between question and answers is considered, and analogism method is used to compare the relation between new Q & A relations with the existed accurate relation. Thus the most similar answer is taken as the best answer. In the analogism method, plenty of Q & A pairs are collected to guarantee high quality answers, based on which auxiliary set is built to evaluate the same kind of relation between the new Q & A pairs and the auxiliary set along with a regression model.The experimental results showed that, it is effective to distinguish the relevance of questions using binary methods. Relevant questions after distinguished can be candidate questions. The evaluation result showed, the analogism method performs better in all of aspects. In the Q & A community with high noise, it is helpful to analogy the existed Q & A relations.