Nowadays, the traditional information retrieval (IR) is inadequate for the user who requires precise results. Hence, the importance of the semantic IR arises. It is very important to move from the level of ambiguous terms to that of well-specified concepts in the indexing phase to enrich the search process. To handle the problems of semantic ambiguity of indexed terms as well as the uncertainty and imprecision inherent in the information retrieval process, a semantic indexing approach was proposed for a better document representation. It is based on indexing the associated synsets with document terms that are identified by mapping on the WordNet ontology. These synsets are defined following a term disambiguation process (WSD). The key to the proposed system is a weighting model which calculates the importance of each index item considering many factors that improve the performance of the information retrieval system. Proposed conceptual weight is based on local and global integrality, degree of re-homogenization, and degree of specificity of the concept. A corrector parameter to reduce the impact of errors in WSD process is included. The experimental evaluation of the introduced semantic IR model shows very satisfactory results compared to well-cited benchmarks.
Neji, Sameh; Jemni Ben Ayed, Leila; Chenaina, Tarek; and M. Shoeb, Abdullah
"A Novel Conceptual Weighting Model for Semantic Information Retrieval,"
Information Sciences Letters: Vol. 10
, Article 14.
Available at: https://digitalcommons.aaru.edu.jo/isl/vol10/iss1/14