基于期望与K次方差的信息检索质量评估模型的研究

A NEW MEASURE FOR EVALUATION OF INFORMATION RETRIEVAL QUALITY

查全率和查准率是评估信息检索系统检索质量的两个基本标准 .长期以来 ,基于这两个标准 ,存在着多种评价方法 .但是 ,这些方法基本上是对查全率和查准率做简单的处理 ,仅反映检索的平均性能 ,而对检索稳定性没有分析 ,并且缺乏一套科学的、系统的评估体系 .针对这种情况 ,借鉴概率学中的期望和方差的思想 ,用数学语言严格定义了查全期望、查准期望 ,K次查全方差和 K次查准方差等概念 .在这些概念的基础上 ,给出了信息检索质量评估准则 .与其它模型相比 ,该模型能从检索的平均质量和检索的稳定性两方面反映检索系统的性能 ,因此 。

Recall and precision are the basic criteria for evaluation of retrieval quality in information retrieval systems. Much effort and research has been done to solve the problem of the evaluation. Many methods have been proposed for evaluation of retrieval quality. However, they only reflect the average performance of retrieval but lack the analysis of stability of retrieval. In this paper, a new measure for evaluation of information retrieval quality is brought forward. The measure is based on some concepts that are called recall expectation, recall K variance, precision expectation, and precision K variance. Compared with other evaluation models of retrieval quality, this measure can reflect both the average performance and the stabilization of retrieval. Therefore, this model is better for evaluating the retrieval quality of information retrieval systems.