Placing Query Term Proximity in Search Context

Abstract

In the information retrieval system, relevance manifestation is pivotal and regularly based on document-term statistics, i.e. term frequency (tf), inverse document frequency (idf), etc.. Query Term Proximity (QTP) within matched documents is mostly under-explored for the relevance estimation in the information retrieval. In this paper, we systematically review the lineage of the notion of QTP in IR and proposed a novel framework for relevance estimation. The proposed framework is referred as Adaptive QTP based User Information Retrieval (AQtpUIR), is intended to promote the document’s relevance among all relevant retrieved ones. Here, the relevance estimation is a weighted combination of document-term (DT) statistics and query-term (QT) statistics. The notions ‘term-term query proximity’ is a simple aggregation of contextual aspects of user search in relevance estimates and query formation. Intuitively, QTP is exploited to promote the documents for balanced exploitation-exploration, and eventually navigate a search towards goals. The design analysis asserts the usability of QTP measures to balance several seeking tradeoffs, e.g. relevance, novelty, result diversity (Coverage and Topicality), and highlight various inherent challenges and issue of the proposed work.