A Multilingual Semantic Search Assistant to Facilitate Information Retrieval

This paper describes a multilingual semantic search assistant (MSSA) that could help users find proper keywords in FAO fulltext-based search engines. The MSSA implements four independent functions each associated with the AGROVOC multilingual thesaurus to assist users in choosing better keywords: a Venn diagram-based Boolean search interface, an animated AGROVOC concept browser, cross-language query expansion for five official FAO languages, and domain-specific synonym expansion. A usability study of cross-language support and domain-specific synonym expansion was performed through a comparative evaluation of precision and recall in the WAICENT Information Finder. The cross-language support functionality yielded a significant increase in recall without any decrease in precision when compared to a simple search in. Domain-specific synonym expansion, on the other hand, made no significant difference in precision and recall. We suggest some ways to improve the performance of MSSA. Keyword: AGROVOC, multilingual semantic search, search assistant, multilingual, semantic, information retrieval, Food and Agriculture Organization of the United Nation, WAICENT Information Finder, Web Services, Boolean search, topic browser, cross-language query expansion, synonym expansion