Relation Classification in Knowledge Graph Based on Natural Language Text

Abstract

Relation classification is an important semantic processing task in natural language processing, and it is also an important task to construct knowledge graph based on natural language text. At present, the cutting-edge method in the field of natural language processing is to obtain some advanced features based on deep learning. One problem is that important features of a sentence can appear anywhere in the sentence. Another problem is that building a domain-specific knowledge map often lacks annotated data. In order to solve these problems, this paper proposes to obtain labeled corpus by distant supervision, and use bidirectional GRU to get relationship between entities. Because the corpus is Chinese so use the word vector as input. At the same time, the attention mechanism is applied to reduce the weight of the noise instance. Finally, the classification model is tested on open dataset and get a good result.