Research on Juristic Sentiment Topic Model on Legal Texts

With the construction of intelligent courts, recommending cases plays an important role in judicial activities. The evaluation of legal texts makes legal texts have a kind of juristic sentiment. Given the lack of supervised information for texts in traditional topic models, JSTM (Juristic Sentiment and Topic Model) is proposed, which takes the evaluation of legal texts as the response variable by continuous transformation and uses the method of variation inference to get topic distribution. The experimental results show that the JSTM can find common and unique topics in different legal texts, and it significantly improves text recommending and dimensionality reduction.