Ranking of documents by measures considering conceptual dependence between terms

摘要

There is demand for a retrieval system that would support extraction of documents even without recognizing new query expressions, while offering a user interface featuring an easy way to evaluate relevance of search results. In this paper, a solution to such a need is offered. Topics of queries and documents are normally represented by search terms. Thus, terms are assumed to be associated through a concept hierarchy; measures are found that reflect, in a natural way, conceptual relations between terms through associativity between a query and a document; a method is developed to rank documents using such measures. These measures consider specificity of coverings of expressions mentioned in both the query and the document with regard for concept hierarchy. Experiments confirmed that the proposed measures were helpful in evaluating conceptual associativity, while document ranking based on these measures proved more efficient compared to traditional similarity measures.