Research On Domain Ontology-Based Web Entity Event Extraction

With the fast development of Internet technology, Web has been a huge information source whose data is still growing. Because Web pages have some characteristics such as diversity, dynamics and no structural, it is difficult to get information people really interested in from the Web. How to extract event information people concerned about comprehensively and accurately from the massive Web pages is the currently hot research issue, which would provide support for market intelligence analysis, electronic commerce and public opinion analysis. Oriented to large-scale Web data, Web information extraction technology provides a way for people to get valid information, by converting unstructured or semi-structured data into structured data.Ontology-based Web information extraction is a method of combining ontology and information extraction technology. Making full use of the ontology-based description for specific domain combined with Web information extraction technology has showed great advantage to improve the accuracy of information extraction. How to construct domain ontology and give full play to the role of ontology for Web information extraction is a problem to be solved. Under the background of market intelligence domain, we launches a related work about the problem of Web entity event information extraction based on domain ontology. The main contributions are summarized as follows:(1)Based on the analysis of the existed ontology construction methods, we proposes a method of ontology construction suitable for specific domain that could effectively reduce the participation of domain experts and improve construction efficiency. Under the guidance of the method, the paper constructs an ontology of market intelligence by drawing on knowledge of e-commerce sites and reusing existed ontology. For the changing relationship over time between entities, this paper presents a dynamic entity-relationship model.(2)This paper improves the ontology-based Web entity-event extraction framework, taking advantage of the rich ontology concepts,examples and relationships. Considering of the characteristics of the event structure in the ontology, the paper uses the strategy of classification during the event extraction. Firstly, the sentences are classified according to event category. Then, the events are extracted in accordance with extraction template combined extraction rules. Experimental results show the feasibility and effectiveness of the event extraction. This paper presents an improved DAG-SVMs classification method for multiple class. Experiments show that the method has good classification accuracy and classification efficiency, at the same time, it obtains higher recall and precision when compared with the general classification algorithm.