Research On Ontology-based Query Translation In Deep Web

With the sustained growth of information on the internet, more and more Web information is no longer stored in static HTML pages but in a Web database and hidden behind a query interface form. This information constitutes Deep Web. Deep Web contains a large number of high quality information. The value of this information is much more important than the information contained in Surface Web. However, this information cannot be indexed by the traditional search engine. In order to help users get information from Deep Web quickly and accurately and improve the coverage and accuracy, many scholars are engaged in the research of Deep Web data query. The main thought of Deep Web search is integrating the information of various Web databases. This method can facilitate the unified query. At present, Deep Web data integration service has become a hot topic.Deep Web integration service involves several technical problems. It contains Entry form finding, Form pattern extraction, Pattern matching and integration, Form query translation and Query result post-treatment. This thesis mainly studies the query translation problem.Query Translation technology is an important part of Deep Web data integration. It is responsible for backfilling the query condition that filled in integrated form by users into local form. Due to the autonomy and heterogeneous of Deep Web, precise query translation cannot be realized. How to maximally guarantee the similarity of local form query and source query is now a challenging task.This thesis innovatively introduces domain ontology and common ontology into query translation. The concept and data hierarchy of Web data base are stored in domain ontology. It takes the advantages of ontology in knowledge management to guide query translation.We arrange the job as followFirst, we build domain ontology under the guidance of domain expert. In order to guarantee the correctness of ontology, we adopt manual construction and constantly update and improve the ontology in the actual use of ontology process.Then divide the entire query translation into two parts:a query analyzer and a query translator based on ontology. (1)Query analyzer is mainly responsible for generating mapping table and the model of query ability of local form. We obtain the Semantic Mapping relations by integrating form and local form in ontology, which are then stored in mapping table. We get the model of query ability by identifying the exclusivity and binding attribute. (2) Query translator consists of attribute matching, predicate mapping, query rewriting and query submition. As the congruent relationship of integrated form attributes and local form attributes is recorded in mapping table, there is no need to search ontology time after time. What we have to do is just operating the mapping table, so the efficiency of query is obviously improved. In the predicate mapping, we designed a predicate template with constraints. It can deal with various predicate templates and take different strategies toward different types of data to ensure that query translation is the smallest superset of source query. Query rewriting have the predicate template fill top-up reasonable combination according to the query ability of local form. We gain effective query of local form mainly by the dispose of exclusiveness attribute and binding attribute. Finally, submit the local query according to the generative rules of the URL and get results returned.Experiments show that the introduction of ontology can make the query translation achieve higher precision and accuracy in semantic level. The system designed in this thesis can do query translation automatically and have practical feasibility.