Abstract
Entity Linking refers to the task of linking entity mentions in the given text with their referent entities in a knowledge base, which is a key technology of knowledge base expansion. However, the performance of traditional Chinese entity linking methods are affected by the incomplete Chinese knowledge base. Also they rarely use the semantic relevance between entities. Therefore, we propose a Chinese collective entity linking method based on the consistency of the topic, which considers both the content similarity and topic relevance of the co-occurrence entities, and propose a method for calculating the topic consistency of entities. This method implements batch links for multiple ambiguous entities that appear in the same text, and reduces the reliance on the local knowledge base by using the combination of the local knowledge base and the external knowledge base. Experimental results show that our method performs well over the traditional methods. And it is potentially effective.
Highlights
Nowadays, we are in such an era of information explosion
In our method we model each entity mention in the given text and its candidate entity as distinct nodes in a graph, and model topic consistency by links between nodes
We propose a Chinese collective entity linking algorithm based on topic consistency and propose a method for calculating the topic consistency of the entity
Summary
How to effectively obtain valuable information from massive network data is a hot research problem in the field of Internet information extraction technology. The main problems are the diversity and ambiguity of natural language. The diversity of natural language refers to that the same meaning can be expressed in many different forms. This is due to the elasticity of expression of natural language. As for the ambiguity of the natural language, it means that the same word, phrase, or sentence have different meanings in different contexts. Because of the existence of ambiguity and ambiguity, it is difficult for the users to retrieve the relevant information of the target entity quickly and accurately when searching the information on the network [1]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have