Pre-query Processing For Semantic-oriented Web Search

The Internet has become the main carrier of information, and its size is still rapidly growing. Faced with a flood of Internet information, search engine has become a major tool for people to retrieve needed information on the Internet efficiently.Currently, most of the existing commercial search engines are based on keyword matching. With the development of search engine, the quality of search results are improved a lot than the quality in the past, but there are still several key problems:1) it’s difficult for a user to find the right keywords to describe his search intent, because the user don’t know about the information needed; 2) keyword matching doesn’t support semantic understanding, and the problem is particularly evident when the query contains ambiguous words or synonyms; 3) Some researches on query logs found that queries are too short, resulting in a search engine is difficult to understand the user search intent. In recent years, there have been a lot of researches to address these issues. From the perspective of the search process, these studies can be put into three categories:pre-query processing, such as query expansion and query reformulation; query processing, such as optimizing retrieval model and establishing the conceptual index; post-query processing, such as re-ranking the search results. However, these studies fragmented, lack of integrity solutions, the effect has yet to be further improved.In this regard, the author analyzes the existing problems of the search engines. With the semantic-oriented query language is designed by us, term-oriented web search is extended to semantic-oriented web search. The quality of search engine is optimized during pre-query processing. Thus, improve the effectiveness and quality of web search. On the specific work and achievements include:Semantic-oriented query language is designed, which supports semantic interaction between user and search engine. The core syntax of language is Concept Binding (CB), which can disambiguate query terms. On the basis of CB, three kinds of semantic queries are derived, namely conceptual query, relational query and analogy query. The language can help users express their search intentions; on the other hand, search engine can understand users more easily.Proposed a concept mapping algorithm based on ontology for understanding each CB in query, which is the first step of our pre-query processing. For each CB, firstly, get the candidate set of concepts corresponding to ambiguous word by glossary table of ontology. Then calculate semantic relatedness between concepts based on ontology, and an evaluation function f is constructed based on semantic relatedness. Finally, the concept that makes f to reach the maximum value should be returned. Experiments show that the algorithm has a high accuracy rate.Proposed a query expansion method based on ontology and query logs, which is the second step of our pre-query processing. Firstly, a candidate set of concepts is constructed through ontology. Then exploiting the relatedness among queries in the same query session, the score of each candidate is calculated. Finally, candidates are sorted in descending order according to the score, and the top-N candidates are chosen. Experiments show that our method is effective.