A Study Of Query Expansion Model Based On External Knowledge Base

The development and improvement of knowledge graph in academia and industry has prompted more and more commercial engines to explore semantic informations and improve the retrieval performance by using knowledge graphs as the basic semantic web.Knowledge graph has been successfully applied in knowledge inference and entity search.However,the potential ability of its entities and properties for better improving search performance in query expansion remains to be further excavated.Query expansion is a technology that exploring user information needs by adding expansion terms to original query and reformulating a new query for final retrieval.Recent research has shown that query expansion based on internal resources(e.g.,pseudo relevance feedback(PRF)documents)may hurt the performance of some individual queries.In this paper,we focus on using external knowledge base that contain a large amount of semantic information and entity information to help to solve this problem.We propose a novel query expansion technique with knowledge graph(KG)based on the Markov random fields(MRF)model to enhance retrieval performance.This technique,called MRF-KG,models the joint distribution of original query terms,documents and two expanded variants,i.e.entities and properties.Global importance and local importance are considered when computing the entity distribution,the most relevant entities are then selected based on this distribution.We conduct experiments on two Web-TREC collections,WT10 G and ClueWeb12 B.Both queries and documents are annotated with Freebase entities.Experiment results demonstrate that MRF-KG outperforms traditional graph-based models,such as MRFs and LCE models,which demonstrate that:(1)Freebase has a positive effect on query expansion;(2)the joint distribution of queries,documents,entities and properties modeled by MRF-KG is helpful to get query semantic related information.