Automatic Chinese Financial Knowledge Graph Constructing Framework
摘要：As rapid growth of complexity of finance, building a Chinese Financial Knowledge Graph（CFKG） from scratch become more and more important. The key of construction of CFKG is the argument relationship extracting method. There are many researches focusing on extracting argument relationship from text based on open domain methods or closed domain methods. However, closed domain methods rely on vast tagged data source which is difficult to acquire and most open domain methods have the problem of high false extracting rate on financial data. So we propose a financial domain open domain extractor which not only adopt the advantages of syntax and semantic dependence but also adopt the advantages of Siamese Network for classification. Experimental results show that our solution outperforms the existing best method DSNFs in terms of precision and recall measure and our improved Siamese Network can reduce the number of false relations. In the end, we build an experimental graph from news data, in which we can explore personnel flow, implicit relationship, etc.