基于混合属性的用户检索意图识别方法

User Retrieval Intention Recognition Method Based on Mixed Attributes

针对目前知识检索方法中未能充分考虑用户的检索意图而导致的检索准确率不高的问题,本文提出了一种基于混合属性的用户检索意图识别方法,并应用于检索引擎中,用以提高其检索准确率。该方法综合考虑了用户属性和查询关键词所对应的用户检索意图,首先分别采用了决策树学习方法和朴素贝叶斯分类器进行意图识别,然后采用二元线性回归分析方法对两类检索意图进行融合,获得的检索意图用以指导检索引擎中对检索结果的排序。最后,本文以某航天企业知识库的检索为例,分别与单一属性的用户检索意图识别方法和常见的开源检索引擎进行了对比实验,实验结果表明,本文提出的方法准确率更优。

In view of the low retrieval accuracy caused by the failure to fully consider the user’s retrieval intention in the current knowledge retrieval methods,this paper proposes a user retrieval intention recognition method based on mixed attributes,which is applied to the retrieval engine to improve its retrieval accuracy.This method comprehensively considers the user retrieval intention corresponding to user attributes and query keywords.Firstly,the decision tree learning method and naive Bayesian classifier are used for intention recognition,and then the binary linear regression analysis method is used to fuse the two kinds of retrieval intentions.The obtained retrieval intentions are used to guide the ranking of retrieval results in the retrieval engine.Finally,taking the retrieval of a knowledge base of an aerospace enterprise as an example,this paper makes comparative experiments with the single attribute user retrieval intention recognition method and the common open source retrieval engine.The experimental results show that the accuracy of the method proposed in this paper is better.