Research And Implementation On Method Of Entity Linking Baseed On Wikipedia

With the expansion of network resources,entity-linking are becoming more important in the field of Natural Language Processing.There are a lot of real-world entity information in internet data.In order to put these knowledge into better use,people build many professional knowledge base to manage these information. This requires more relevant knowledge about these entities for knowledge integration.At the same time,because of the ambiguity of knowledge entries,the expansion process cannot be carried out smoothly.Entity-linking technology is an important way to solve this problem.It can make entity mention in text data automatically correspond to entity concept in real world,and achieve entity disambiguation.Entity-linking as a method of knowledge filling and information extraction is extremely important in the field of Internet application and scientific research.The main content of entity-linking research in paper can be divided into the following sections: The first part is about finding candidate entities based on Wikipedia knowledge base.we use Wikipedia to construct mapping database which we will use in above process.In the meantime we come up with a method to generate candidate set.The second part explores the method of entity disambiguation.We propose to carry out feature engineering from multiple perspectives and use a model to expand features to further explore new features.As for the selection of the disambiguation model,a supervised machine learning model is used to carry out the entity disambiguation process.In the process of model analysis,some improvement strategies are put forward to optimize the effect and efficiency of the model,and to perfect the process of entity linking.The third part mainly focus on the design and implementation of a system based on the Wikipedia data link entity and above research.The system implements a series of sub tasks included in the entity-linking process.Then the advantages and disadvantages of the existing method and the method of this paper on the entity link are compared with the experiment.