Named Entity Linking Based On Multisource Knowledge

In the age of big data, network information is boosting with an explosive speed, which makes the Internet gradually become a global shared information warehouse. At the same time, it has become more and more difficult to understand the information accurately, thus making it an urgent problem to be solved in the information age.Named entities stand for the major information in text, which is an important language unit. Therefore, the correct analysis for some ambiguous named entities is of great significance to the understanding of texts. Meanwhile, for the new named entities that doesn’t exist in the knowledge base, we should discover them in time so that new named entities can be added to the knowledge base immediately in order to expand the knowledge base. Named entity linking is the vital method to solve this problem, which links to the relevant entities in the knowledge base, realizing automatic named entity linking.Named entity linking plays a key role in solving the ambiguity and the size problem of entities in the knowledge base. The research towards entity linking includes many obstacles, such as various entity forms and complex relationship in reasoning and representation. In this paper, a novel named entity linking method based on multi-source knowledge and Ranking SVM algorithm is proposed, we make experiments on the public data sets provided by NLP&CC 2013 and NLP&CC 2014. Firstly, the query semantic extension is conducted based on the context information, at the same time the synonym dictionary is built in order to get the accurate collection of candidate entity; then we do named entity linking based on Lucene, VSM with network authentication, and multisource knowledge with Ranking SVM. Through the comparison of the three strategies, it turns out that the third method takes more semantic features and achieves outgoing preformance.