Semantic Matching of Legal Cases for large Scale Judgments of China

With the development of China’s judicial cause and the technological innovation of artificial intelligence, artificial judicial intelligence has become an essential power in judicial trials. Now, more than 100 million judgments have been published online. The combination of existing literature data and natural language processing technology in artificial intelligence is of great value to judge’s trial work. In order to achieve judicial justice, judges need to refer to similar cases in which the previous judgments take effect.At the present stage, the retrieval of similar cases mainly uses keyword retrieval, legal provisions retrieval, and other methods. This kind of method is not accurate enough.From the perspective of semantic matching, this paper analyzes the judgment documents first, points out the important fact that the controversial focus is the high concentration of the facts of the case. Then, we construct a data set in the form of (query: Q, document 1: D
1
, document 2: D
2
); On this basis, this paper proposes a multi-task BCKS model. The follow-up experiments show that this method has a good effect and can achieve 91% accuracy when the data size is large enough, which provides strong technical support for the case retrieval system.