A Query Expansion Algorithm Based on Phrases Semantic Similarity

Abstract

During the indexing process of traditional search engine, web pages become a list of terms, but single term cannot represent the rich content of web pages, which makes information retrieval methods mainly based on terms matching often result in depressed precision. This paper proposes a novel query expansion technique that has phrases as its expansion unit. Phrases typically have a higher information content and a smaller degree of ambiguity than their constituent words, and therefore represent the concepts expressed in text more accurately than single terms. This method extracts key phrases from original results, and calculates the semantic similarity between the query phrase and each phrase extracted using the semantic similarity algorithm based on WordNet, and then expands the query with the most similar phrases to search again. Experimental results show that the proposed algorithm can provide more precision than the traditional query expansion methods.