Abstract:

According to the actual needs of the task of “legal judgment prediction” in the field of intelligent judicial services, the research ideas and implementation ways were discussed, and the overall framework and the specific process of this task were introduced.Based on the massive real cases obtained by China Judgments Online and the evaluation dataset of CAIL2018, the categories were sorted out.The format of the experimental dataset was standardized.And the prediction dataset of legal judgment prediction based on legal judgment documents was built.For the judgment prediction model,the high-quality sentences by using the method of decision elements extraction were extracted.Then refer to the judge’s judgment ideas, the whole task of legal judgment prediction was transform into three subtasks, namely the law articles prediction, the charge prediction, and the penalty prediction.Meanwhile, construct the prediction models based on the judgment elements respectively.The experimental results show that the proposed methods achieves excellent results on the criminal law judgment prediction dataset.

Key words: legal judgment prediction, decision elements extraction, law articles prediction, charge prediction, penalty prediction