Intelligent Processing of Judicial Documents Based on Deep Learning

Legal judgement prediction is a critical application in the law and AI field, and has received great attention in recent years. In this paper, we focus on the recommendations of applicable law articles for a given criminal case. This is a multi-label problem in nature, since it is common for a sequence of criminal activities to violate the law in many aspects. We compare two different strategies for identifying the articles. The first one employs only the Lawformer, which is a Longformer-based model designed specifically for the legal domain, and the second one will combine the applications of the Lawformer with a traditional technique for keyword selection. The experimental results indicate that integrating the deep learning and the traditional methods provide a better result.