Personalized Search Methods Based On Multiple Semantic Relationships

In recent years,with the rapid development and application of Internet technology,online digital content has shown an explosive growth trend.When users use traditional search systems to search for information,they often get a lot of irrelevant search results,and these results are often far from the user's most genuine search intentions.In the information retrieval,the query words input by the user during the query are relatively few,and the information matching is likely to produce ambiguity,resulting in incomplete and inaccurate search results.In order to reduce the information burden,the demand for personalized search is increasing,and how to accurately return documents related to a single user's query intention has become an important research topic.In this situation,many researchers have developed an interest in social tags.They use the tags tagged by users on network resources to learn the user's behavior and interests and achieve personalized query expansion.Adopting social tags can improve the quality of search,but the actual tag system is often sparse,and the tags themselves are not standardized and random.The traditional social tag-based personalized query expansion search results are often not ideal.In terms of the choice of extended words,since the tags and the marked resources themselves may be literally far from each other,and are not an accurate description of the resources.Therefore,it is not efficient to perform query expansion with tags.In order to solve these problems,the main contributions of this paper are as follows:(1)This paper uses the tag-topic model method to build the user interest model.The model is based on words to learn implicit topics.In this method,the choice of extended words does not depend only on terms matching,but on the relevance of the topics.(2)In this paper,we extend the previous study by considering multiple relationships between tags,documents,and words extracted from the document as multiple relationships.We use the relationship between the learning tags of the word embedded model(WE),between words and between tags and words.Build three association diagrams based on the similarity between labels and words,and integrate all the relationships into a multi-semantic query expansion framework for personalized search.(3)This paper further proposes a personalized query expansion method,which uses the multi-semantic relationship of social tags to select the extended words.Experimental results based on large-scale real socialized tag datasets show that the proposed method outperforms traditional non-personalized search and other traditional personalized query expansion methods based on socialized tagging system.