Judgment Prediction Based on Tensor Decomposition With Optimized Neural Networks

Area of Science:

• Artificial Intelligence
• Legal Technology
• Smart Justice

Background:

• Traditional judgment prediction methods struggle with complex case data and lack fine-grained prediction capabilities.
• Existing models require extensive legal expertise and manual labeling, limiting scalability.
• Accurate extraction of crucial information from legal documents remains a challenge.

Purpose of the Study:

• To propose an advanced judgment prediction method utilizing tensor decomposition and optimized neural networks.
• To improve the accuracy and efficiency of artificial intelligence in legal case analysis.
• To address the limitations of traditional feature models and classification algorithms in legal judgment prediction.

Main Methods:

• Representing legal cases as normalized tensors using OTenr.
• Decomposing tensors into core tensors via a guidance tensor with GTend.
• Optimizing the guidance tensor with RnEla, incorporating Bi-LSTM and Elastic-Net regression for similarity correlation.

Main Results:

• The proposed method, combining OTenr, GTend, and RnEla, demonstrated superior accuracy in predicting legal judgments.
• Core tensors effectively captured structural and elemental case information, enhancing prediction.
• Experimental results on a real legal case dataset confirmed the method's effectiveness over prior approaches.

Conclusions:

• The tensor decomposition approach with optimized neural networks offers a significant advancement in AI-driven legal judgment prediction.
• This method provides a more robust and accurate way to analyze and predict outcomes of legal cases.
• The findings highlight the potential of advanced AI techniques in the smart justice domain.