HyRa: An Effective Hybrid Ranking Model

Abstract

Information retrieval (IR) systems are concerned of the processing of the textual content of the document, the query, and issues of system effectiveness. IR models are means to integrate many sources of evidence on documents to achieve an effective retrieval system. Unlike traditional parameters such as word frequency in documents, semantic information retrieval is an environment that requires the application of semantic techniques. These techniques calculate a degree of query-document relevance to assign the highest rankings to the documents which are semantically closer to the query. To meet the user’s information needs and to optimize the performance of IR system, an innovative model was proposed to calculate the relevance of query-document pairs. The proposed model is a hybrid one that based on a query likelihood language model and a conceptual weighting model. It exploits the semantic similarity between document and query concepts and some document/collection statistics. The conducted experiments on the proposed model show improvement of the standard performance measures over the benchmark competitors. A statistical significance test is conducted to exclude the performance random variation of the compared models. Paired t-tests show that the proposed model is highly significant improves the most important measures.