A Concept Net-based semantic constraint method for query expansion

Zhichao Chen, Junmei Wang, Xiwei Yang

Abstract: Existing query expansion methods usually select the expansion terms according to the term frequency, proximity, location, etc., neglecting the semantic information related to the original query. To solve this problem, we propose a ConceptNet-based semantic constraint method for query expansion. Firstly, the ConceptNet semantic network constrains the range of candidate terms, avoiding all terms in feedback documents is selected as candidate terms. It will reduce the impact of irrelevant terms on query expansion. Secondly, BERT is used to filter out sentences related to query semantics in feedback documents and construct subsets of candidate expansion terms. Then, the two subsets of candidate terms are combined to screen out the expansion terms related to query on semantics. To verify the effectiveness of the proposed method, a series of experiments are carried out on TREC standard text retrieval data sets via applying the proposed method to pseudo-relevance feedback model called Rocchio. Experimental results show that the proposed model significantly improves the results of MAP and P@10 on four data sets.

External IDs:dblp:conf/webi/ChenWY22