Research On Entity Linking Using The Extended Information From Search Engine

Named entity linking is a process of linking named entities in text to the real entity in knowledge base.This work has effectively promoted the development of machine translation,question answering system,information retrieval,information fusion,knowledge base completion and so on.In the research of named entity linking,the traditional method uses the existing knowledge base to disambiguate.However,the existing knowledge base has the shortcomings of the later update and the incomplete information of entities.It can influence the accuracy of similarity calculation which depends on the "word co-occurrence".Therefore,it is of great significance to use the external knowledge source to expand the entity information in the knowledge base.Based on the above exploration,this dissertation studies the problem of naming entity linking.The main contributions of this dissertation are as follows:(1)Whether the candidate entity is the target entity of entity mention,so we utilize the D-S evidence theory to solve the problem of uncertainty problem reasoning.We use the D-S evidence theory to fuse three features including entity name feature,context similarity feature and entity popularity feature to solve the entity disambiguation problem.And the search engine is used to extend the context similarity feature between the entity mention and the candidate entity.Experimental results show that use the D-S evidence theory and the search engine to deal with the entity linking problem is better than the contrast algorithm on precision,recall and F measure,which confirms the effectiveness of this method in entity linking.(2)In order to disambiguating entities in text synchronously and mining semantic relation between entities,we make full use of the graph model to build semantic relations between the entities.In a graph model,the graph nodes are entity mentions in text and their candidate entities.In the process of building edges between nodes,we first use a search engine to extend the context of entity mention and candidate entity,and then use the similarity calculation method to judge whether there is an edge between the entity mention and its candidate entity,then,use the method of mining indirect relations to judge whether there is an edge between candidate entities.Finally,it is proved that the effect of the named entity linking algorithm based on the graph model is better than the contrast algorithm.