基于搜索日志的查询推荐算法研究

发布时间：2019-01-07 20:27

【摘要】：种类众多、数量庞大的信息每天都会发布到网络中，网络成为海量信息资源的载体，人们已经越来越习惯于通过网络寻求答案、文章、产品、人物和地点等诸多事物。搜索引擎正是实现这一过程的一种基本应用，而在丰富多彩的网络信息中，如何找到用户真正需要的、有用的信息，时刻困扰着人们，为了解决这一实际而有意义的问题，引出了本文的研究重点—查询推荐。查询推荐通俗解释也就是：为了满足搜索引擎用户对提高其搜索意图及定位准确性的迫切需求，结合用户所提交的检索词和用户后续行为，，采用一定的策略和算法为其推荐合适的检索词的过程。可以说在查询推荐在提高搜索引擎可用性方面，扮演着很重要的角色。在本文中方法的设计和实现过程中，熟悉并掌握了搜索引擎的工作原理以及体系架构，同时以自然语言处理技术和Web信息挖掘技术为理论指导，以搜索日志为数据分析对象，重点研究搜索引擎个性化服务功能中的查询推荐技术。本文所采用的方法中分为两阶段，为保证在线推荐的实时性，设计了离线处理阶段，实现数据预处理、构建用户查询—点击二部图、二部图聚类等操作。第二阶段是在线推荐阶段，实现基于用户日志的查询推荐。经过实验表明，本文所提出的方法不仅保证了查全率，而且提高了查准率。
[Abstract]:There are many kinds of information, a large number of information will be published to the network every day, the network has become the carrier of massive information resources, people have become more and more accustomed to the Internet to find answers, articles, products, people and places and many other things. Search engine is just a basic application to realize this process, and in the rich and colorful network information, how to find the useful information that the user really needs, always puzzles people, in order to solve this practical and meaningful problem, The research focus of this paper is query recommendation. In order to meet the urgent need of search engine users to improve their search intention and positioning accuracy, the search recommendation is explained by the user, which combines the keywords submitted by the user and the follow-up behavior of the user. The process of recommending appropriate search words for them with certain strategies and algorithms. It can be said that query recommendation plays an important role in improving the usability of search engines. In the process of design and implementation of the method in this paper, we are familiar with and master the working principle and architecture of search engine. At the same time, we take the natural language processing technology and Web information mining technology as the theoretical guidance, and take the search log as the data analysis object. Focus on the search engine personalized services in the function of query recommendation technology. The method used in this paper is divided into two stages. In order to ensure the real-time performance of online recommendation, off-line processing stage is designed, data preprocessing is realized, user query-click bipartite graph, bipartite graph clustering and so on are constructed. The second stage is online recommendation phase, which realizes query recommendation based on user log. The experimental results show that the proposed method not only guarantees the recall rate, but also improves the precision rate.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3

【参考文献】