Research and Design on Cognitive Computing Framework for Predicting Judicial Decisions

摘要

This paper aims to provide a cognitive computing framework to meet the challenges of semantic understanding, knowledge learning and judicial reasoning in the Chinese legal domain. In our framework, legal factors are first represented in a formal way; secondly, legal factors are extracted, and concepts and their relations are augmented with a combination of rule-based and deep learning methods; thirdly, a predication model is generated and trained to make judicial decisions. When a fact description is brought into the model, the probability of judicial decisions will be given automatically. Two elementary results are obtained: I. Our method can effectively predict the decisions for divorce cases with different expression styles, and offers better performance than traditional methods like Support Vector Machine (SVM); II. Our machine learning predicting results can be easily understood by general public as applied induction rules are given.