A Content-Based Recommendation Framework for Judicial Cases.

Zichen Guo,Tieke He,Zemin Qin,Zicong Xie,Jia Liu

DOI: https://doi.org/10.1007/978-981-15-0118-0_7

2019-01-01

Abstract:Under the background of the Judicial Reform of China, big data of judicial cases are widely used to solve the problem of judicial research. Similarity analysis of judicial cases is the basis of wisdom judicature. In view of the necessity of getting rid of the ineffective information and extracting useful rules and conditions from the descriptive document, the analysis of Chinese judicial cases with a certain format is a big challenge. Hence, we propose a method that focuses on producing recommendations that are based on the content of judicial cases. Considering the particularity of Chinese language, we use “jieba” text segmentation to preprocess the cases. In view of the lack of labels of user interest and behavior, the proposed method considers the content information via adopting TF-IDF combined with LDA topic model, as opposed to the traditional methods such as CF (Collaborative Filtering Recommendations). Users are recommended to compute cosine similarity of cases in the same topic. In the experiments, we evaluate the performance of the proposed model on a given dataset of nearly 200,000 judicial cases. The experimental result reveals when the number of topics is around 80, the proposed method gets the best performance.