Sign in Subscribe
Machine learning

Novel Framework for Legal Judgment Prediction Enhances Performance with Prompt Learning and Charge Keywords Fusion

Strategic Insights

— 1 min read

Researchers from Xiangtan University have developed a new framework for Legal Judgment Prediction (LJP) that addresses the limitation of existing methods, which focus on multiclass classification and single-label learning, but neglect the semantic correlation between fact descriptions and legal keyword labels. The new approach enhances the utilization of legal concept keywords through prompt learning, incorporating legal keywords into the language model to improve its ability to understand and process legal texts. The framework has been tested on the CAIL2018 dataset, showing an improvement in F1 scores ranging from 1.59% to 9.28% compared to state-of-the-art models.

Key Takeaways:

  • The existing research on LJP mainly focuses on multiclass classification methods and single-label learning, neglecting the semantic correlation between fact descriptions and legal keyword labels.
  • The novel framework for LJP introduces a method based on legal charge keywords and prompt engineering to enhance the performance for LJP.
  • The approach integrates legal keywords with fact descriptions to improve the representation capacity of case fact vectors and enhances the model's ability to understand and process legal texts.
  • A prompt template is designed to guide the reasoning process of the pre-trained language model through structured instructions, which strengthens the semantical relevance between the fact description and legal labels.
  • The experimental results on CAIL2018 datasets across different tasks show an improvement in F1 scores ranging from 1.59% to 9.28% compared to state-of-the-art models such as LADAN, NeurJudge, and CL4LJP.

Statistics:

  • The improvement in F1 scores ranges from at least 1.59% to a maximum of 9.28% compared to state-of-the-art models.
  • The F1 scores improvement demonstrates the effectiveness of the proposed method for LJP.

Sources:

  • "A Method of Legal Judgment Prediction Via Prompt Learning and Charge Keywords Fusion" (Article), Artificial Intelligence and Law, 2025.
  • Xiangtan University, School of Mathematics and Computational Science, Xiangtan 411100, Hunan, People's Republic of China.

Read more