基于无导词义消歧的语义查询扩展

Semantic Query Expansion Based on Unsupervised Word Sense Disambiguation

一个构造良好的查询是信息检索质量的基本保证,语义查询扩展技术解决了传统信息检索系统不能很好理解用户查询意图的问题,在提高检索查全率的同时保证了检索准确率.本文以查询关键字之间的语义关联为切入点,辅以隐式反馈技术获取消歧上下文,以WordNet本体库和WordNet Domains扩展库作为消歧数据源,使用基于局部上下文和基于图论的两类无导词义消歧方法进行查询关键字到本体概念的映射,最后基于概念词汇关联完成基于语义的查询扩展.综合WordNet本体库和WordNet Domains扩展库中的各项知识源对查询词义进行判定,保证了词义消歧的精度;采用无导词义消歧实现查询词义的快速判定,保证了信息检索的实时性;根据查询关键词的多寡分别提出两类消歧方法,满足了各种查询需求.

A well-formed query is a basic guarantee for the quality of information retrieval. Semantic query expansion technology solves the problem of not well understanding user's query intention in traditional information retrieval systems, and it can improve the retrieval recall while maintaining the retrieval accuracy. This article took the semantic association between query keywords as the starting point, supplemented by implicit feedback technique to get disambiguation context. Taking WordNet ontology and WordNet Domains expansion library as disambiguation data source, we provided two Word Sense Disambiguation methods, namely local context based Unsupervised Word Sense Disambiguation and graph theory based Unsupervised Word Sense Disambiguation, to mapping query keywords to ontology concepts. Finally, Semantic query expansion is done based on the concept-terminology associations. Integrating the knowledge of WordNet ontology database and WordNet Domains expanding library to determine the meaning of the query keywords ensuring the accuracy of word sense disambiguation; using Unsupervised Word Sense Dsambiguation to achieve fast query meaning determination ensuring the real-time information retrieval; providing two types of disambiguation methods based on the number of query keywords meeting the needs of all kinds of queries.