CIVIL CASE JUDGMENT PREDICTION METHOD BASED ON MULTI-TASK LEARNING

Aiming at the problem of multiple combinations of laws and regulations for civil case prediction, this paper proposes a civil case judgment prediction method based on multi-task learning. It used a variety of strategies such as CNN model fusion and threshold setting, and used the dependence between legal disputes and legal articles to realize the joint judgment prediction of legal disputes and legal articles in civil cases. Based on the civil cases of China Judgment Online, a dataset of 100,000 civil cases was constructed, and multiple sets of experiments were performed on the dataset. Experimental results show that compared with traditional prediction models, this method is more reasonable and effective for the prediction task with dependency.