Research On Domain Ontology Construction Method Based On Web Meaningful Table Extraction

As an extension of the current Web, the Semantic Web has become a hot research field, while the ontology which can provide semantic to the Web is the key to the realization of the Semantic Web. However, ontology construction technology is not mature as yet. Currently ontologies are constructed mainly by domain experts and ontology engineers manually, which is not only time-consuming but also error prone. How to use existing data sources automatically or semi-automatically construct high-quality ontologies has aroused the interest of many researchers. Tables which contain a lot of structured data are widespread and easily accessible on the Web, so they can be used as a good data source to build the domain ontology. In this thesis, we study the problem of how to use Web meaningful tables to build the domain ontologies.There are two main issues on current related researches:on one hand, the technologies of extracting Web tables cannot meet the requirements of building high-quality domain ontologies, and most of the existing approaches cover the narrow range and only consider the tables encoded by table tags. Also, other approaches extract the tables only based on the visual features, which result in the low precision; On the other hand, the existing approaches for building ontologies from Web tables always need the support of the external knowledge bases, and the approaches often cannot extract the hierarchies of classes and properties and the instances of ontologies simultaneously.In this thesis, a method of constructing domain ontologies based on Web meaningful table extraction is proposed. Firstly, this thesis gives the definition of Web meaningful table and summarizes the coding types of them. Then, we propose a hybrid Web meaningful table extraction algorithm considering both the DOM structural features and positional relation between nodes rendered in the Web browser, which extends the scope of the existing Web meaningful table extraction techniques, and makes non-table-coded Web meaningful tables can be extracted. And the thesis proposes an algorithm using the distribution characteristics of cells’ data type to determine the table dimensions and to identify the table-header position. Then, the thesis gives a set of mapping rules from Web meaningful table to the ontology description language OWL (Web Ontology Language). We have developed a prototype system to verify the effectiveness of Web meaningful table extraction algorithm and ontology construction method. Finally, we draw some conclusions and give several future research directions.