Discovering Entity Relationship And Semantic Annotations Base On Wikipedia Encyclopedia Knowledge Resources

The continuous development of Internet Technology promotes the development of new generation of Internet publishing carrier. As the result, Wikipedia encyclopedia also arises at the historic moment. For Wikipedia encyclopedia, Internet users pay close attention to it by its spirit freedom, open and sharing of Wikipedia encyclopedia. Today, Wikipedia encyclopedia has become one of the most dynamic, influential, and widely spread effect network publication. The linked relation between the data provides a possible of development of semantic web application service. However, Wikipedia encyclopedia editors do not provide semantic annotations when they edit the entities. This leads to the lack of relationship between Wikipedia encyclopedia’s data. Entity linking aims to find suitable entity for keywords in the knowledge base, and establishes links with the key words, so that the degree of relations between the knowledge base data can be strengthen. Adding entity linking in the knowledge base can transform no-structure or low-structure text into a high-structure of the data. As a result, it greatly enhances the readability of the knowledge base entries.Since Chinese Wikipedia knowledge base, such as Baidu Wikipedia, severely lacks the relations of encyclopedia entities, the entity linking methods based on Wikipedia perform low efficiency in Chinese knowledge base. So in this paper, according to the characteristics of the knowledge structure in Chinese knowledge base, five eigenvalues are defined to describe the degree of correlation between entities from different features. Then, we propose an approach to automatically discover the missing entity linking, and establish reliable links on Chinese encyclopedia. By identify entity mentions in the given infobox and text, a candidate matching table is built for each entity. After that, a logistic regression model is used to evaluate the contribution of the five eigenvalues(weight). So, the best matching can be found from the Candidate set, and entity linking relations as well as corresponding semantic annotation can be established.To assess the effectiveness of the proposed method, experiments are conducted on Baidu Wikipedia. Experimental results show that the five features can achieve good effect and our method also can efficiently find missing entity links. Furthermore the accuracy and recall rate are obviously superior to other link methods. Hence, a good effect of the semantic annotation can be obtained.