A Revisit of Query Expansion with Different Semantic Levels

摘要

Query expansion has received extensive attention in information retrieval community. Although semantic based query expansion appears to be promising in improving retrieval performance, previous research has shown that it cannot consistently improve retrieval performance. It is a tricky problem to automatically determine whether to do query expansion for a given query. In this paper, we introduce Compact Concept Ontology (CCO) and provide users the option of exploring different semantic levels by using different CCOs. Experimental results show our approach is superior to previous work in many cases. Additionally, we integrate the proposed methods into a text-based video search system (iVSearcher), to improve the user's experience and retrieval performance significantly. To the best of our knowledge, this is the first system that integrates semantic information into video search and explores different semantic levels.