Judicial Case Screening Based on LDA

Abstract

Under the background of judicial responsibility system, making similar judgments according to similar cases is vital for front-line judges to solve complicated problems such as non-standard use of law and inconsistency of judicial ruling standards. In this paper, a method is proposed for judicial cases based on the LDA topic model. The case, penalty and legal provisions were set. Gibbs Sampling algorithm was employed to estimate the probability distribution of topics on the implicit topic set in a text and calculate the similarity between texts by cosine similarity. The quality of screening was used as a final evaluation indicator. The verification of massive experiments shows that the case screening method based on LDA and cosine similarity has a satisfactory effect.