Purely entity-based semantic search for information retrieval

Over the past decade, Knowledge bases (KB) have been increasingly used in almost all information retrieval tasks. To improve the ad hoc document retrieval task, KB have often been utilized to complete and enrich the representation of queries and documents. Although many approaches have used KB for such purpose, understanding how effectively leverage entity-based representation in conjunction with term-based representation still needs to be resolved. In this thesis, we propose a Purely Entity-based Semantic Search Approach for Information Retrieval (PESS4IR). We explore its strengths and weaknesses to know how it would be effectively leverageable alongside any other model. The approach includes (i) its own entity linking method, Entity Linking for Document Text (EL4DT), which is designed to be appropriate for document text. Moreover, the approach includes (ii) an inverted indexing method for the indexing task. For document retrieval and ranking, (iii) an appropriate ranking method is designed to take advantage of all the strengths of the approach. We report the findings on the performance of our approach tested by queries annotated by two entity linking tools, REL and DBpedia Spotlight. The experiments are performed on the standard TREC 2004 Robust collection and MSMARCO collections. By using the REL method on Robust collection, for queries whose all terms are annotated and whose average annotation scores are greater than or equal to 0.75, our approach achieves the maximum nDCG@5 score (1.000). Thus, using our approach with any document retrieval method would be an added value, unless that method achieves the maximum nDCG@5 score for those highly annotated queries.