论文部分内容阅读
运用查询扩展中的局部反馈技术和伪文档反馈技术,提出一种面向微博的查询扩展方法。将候选词分为3个层级进行考察,分别为主题-词语层、文档-词语层和词语-词语层,对应3个层次提出权重计算方法和相似度计算方法。最后,通过实验对方法进行分析比较,实验结果显示,综合考虑主题-词语权重和文档-词语权重得到的扩展词更能满足用户的需求。
By using the local feedback technology and the pseudo-document feedback technology in query expansion, a query expansion method for Weibo is proposed. The candidate words are divided into three levels to investigate, respectively, the theme - word layer, document - word layer and word - word layer, corresponding to three levels proposed weight calculation method and similarity calculation method. Finally, through the experiment, the method is compared and analyzed. The experimental results show that the extended words that take into account the theme - word weight and document - word weight can better meet the needs of users.