一种基于语言模型的微博检索技术

摘要

InMicroblogging service has attracted billions of online users and emerged to be a dominant web medium for sharing and spreading instant information. With so many valuable information produced on this platform, it is meaningful to research on microblogs. Information retrieval as a fundamental role of data mining attracts lots of studies on microblog. Compared with traditional web pages and articles, microblog has its own unique characteristics, including sparse, dynamic, noisy, short length and etc. In order to meet the information need of users , we propose to query expansion method based on dynamic pseudo relevance feedback ,which combine language model and the characteristics of time, and the precision is improved to 53.9%.%步入互联网时代，微博吸引了成千上万的用户并且以迅猛的速度发展成为主流的网络媒体。人们通过微博服务传播与分享即时信息，在这个平台上产生了大量宝贵的信息，因此基于微博的研究具有十分重要的意义。与传统的网页和文章相比，微博文本具有自身独特的特性，包括文本稀疏性、动态性、噪声和长度短等。基于满足用户从海量微博中获取信息的目的，本文在语言模型的基础上结合微博文本的时间特性，提出一种基于动态伪相关反馈模型的查询扩展方法，使准确率提高至53.9%。