Relevance in Information Retrieval

在信息检索中,“相关性”是一个关键性的概念,但是信息检索研究长久以来一直是在缺乏一个准确的相关性定义的情况下进行的。对这一概念人们做了大量研究并提出了理论框架,但这些研究成果无法很好地被加以利用。本文简要回顾了近五十年来相关性研究的概况,对研究现状进行了讨论,然后就如何将相关性研究成果与系统开发相结合提出了初步设想。Modern information retrieval studies have been conducted for more than half a century. Retrieval of documents is based on their alleged “relevance” to the user’s query. However, researchers have never reached a consensus on the definition for “relevance”. Although studies on system-oriented relevance and user-oriented relevance have been extensive, it does not seem hopeful that these insights can be incorporated into development practices of information retrieval systems. The word-counting method used by most retrieval systems has remained despite the fact that the user has been generally recognized as an active determinant in relevance judgments. Based on a brief critical overview of the literature, focusing on the two aspects mentioned above, we suggest that it is possible that the clarity of the system-oriented approach and the likely advantage of the user’s intuition are kept in an integrated system