中文社区问答系统中问题检索技术研究

发布时间：2018-01-12 02:35

本文关键词：中文社区问答系统中问题检索技术研究　出处：《北京理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着互联网技术的迅猛发展,网络中的信息呈现指数级增长,网络的日益普及,逐渐改变着人们获取信息的主要方式。近年来社区问答系统迅速增长,对人类的信息获取方式产生了革命性影响。而在网络信息急剧增加的同时,超负荷的巨大信息量也使人们产生了阅读困扰。因此,如何在海量信息中快捷、准确地获取感兴趣的内容,成为人们面临的一个新的难题。目前,社区问答系统已获得广泛应用,其中相似问题检索可将相似问题的答案推荐给用户,从而避免了用户重复提交问题,也方便用户更加快速地获取问题的答案。本文针对社区问答系统问句相似度计算问题,提出了一种改进的TFIDF算法。首先,按照用户的查询意图对问句进行分类,然后根据特征词在类别中的分布对权值进行调整;其次,将问句的主题词归入特征项进行TFIDF计算。实验结果表明,与传统TFIDF算法和参考改进算法相比,该算法明显提高了检索性能。
[Abstract]:With the rapid development of Internet technology, the information in the network is increasing exponentially. With the increasing popularity of the network, the main way to obtain information has been gradually changed. In recent years, the community question answering system has grown rapidly. It has a revolutionary impact on the way of human information acquisition. While the information on the network is increasing rapidly, the huge amount of information overload also causes people to read puzzles. Therefore, how to quickly in the mass of information. It has become a new difficult problem for people to get the content of interest accurately. At present, the community question answering system has been widely used, in which similarity question retrieval can recommend the answer to similar question to the user. In order to avoid the user to submit the question repeatedly, but also to facilitate the user to obtain the answer more quickly. This paper aims at the question similarity calculation problem of community question answering system. An improved TFIDF algorithm is proposed. Firstly, the question sentences are classified according to the user's query intention, and then the weights are adjusted according to the distribution of the feature words in the category. Secondly, the topic words of the question are classified into the feature items for TFIDF calculation. The experimental results show that the algorithm improves the retrieval performance obviously compared with the traditional TFIDF algorithm and the reference improved algorithm.
【学位授予单位】：北京理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.3

【相似文献】