Query-sensitive similarity measures for information retrieval

摘要: The application of document clustering to information retrieval has been motivated by the potential effectiveness gains postulated cluster hypothesis. hypothesis states that relevant documents tend be highly similar each other and therefore appear in same clusters. In this paper we propose an axiomatic view suggesting query (co-relevant documents) display inherent similarity is dictated itself. Because similarity, should valid for any collection. Our research describes attempt devise means which can detected. We use query-sensitive measures bias interdocument relationships toward pairs jointly possess attributes expressed a query. experimentally tested three against conventional ones do not take into account, also examined comparative measures. calculated varying numbers top-ranked six collections. results show consistent significant increase number become nearest neighbors given when are used. These suggest cluster-based system through