Research On DataBase Schema Matching Based On Domain Knowledge

With the rapid development of Internet, the database has become the most important sources of information. Considering the different features of data domain, the users of database designed different database to place data in order to improve the efficiency. Because of these differences, the work of data integration has become harder. The crucial work of data integration is how to define the semantic map between the data schemas, named database schema matching.Firstly, the database schema and domain knowledge about database were introduced in this thesis. Secondly, a new semi-automatically approach was exploited to discover both 1:1 and complex matching with the domain knowledge. It contained four periods: pre-processing matching period, generating matching period, evaluating matching period and selecting matching period. In the first period, the new approach formalized the information about the database schema and domain knowledge. In the second period, the target attribute's searching space was divided by the type of the target, then the candidates were searched by the Searchers, lastly it generated a set of candidates with a rough similarity. In the third period, the similarity of matching between the target attribute and the candidates were calculated with the help of the domain knowledge. In the last period, the candidates were ranked by the similarity of candidate, then the candidate with the least cost was chosen as the feedback of matching, which conducted the next matching. The selecting matching period can offer a novel explanation facility that human users can interact with the system.The approach was verified by experiments, and the results were analyzed, the advantages and weaknesses of the algorithms were discussed.