Constructing Feature Vectors for search: investigating intrinsic quality impact on search performance

Abstract: In this paper, we revisit our approach to construction of semanticlinguistic Feature Vectors (FVs) used to enhance Web search. These FVs are built based on domain semantics encoded in an ontology and augmented by relevant terminology from Web documents. The contributions of this paper are the evaluation of constructed FVs and the analysis of their impact on search performance. This completes the validation of the proposed approach concluding that the proposed metrics provide good indications of the quality of the FVs. Yet, the results suggest that the metrics need to be tailored to fit the needs of search applications.

• Type: Journal
• Year: 2010
• Where: IJWGS
• Authors: Stein L. Tomassen, Darijus Strasunskas
• Keywords: Domain Semantics, IJWGS 2010, Relevant Terminology, Semanticlinguistic Feature Vectors