A comprehensive framework for legal dispute analysis integrating prompt engineering and multi-dimensional knowledge...

Area of Science:

• Artificial Intelligence
• Legal Informatics
• Natural Language Processing

Background:

• Current Large Language Models (LLMs) struggle with complex legal concepts, consistent reasoning, and accurate source citation.
• Intelligent legal assistance systems require robust legal dispute analysis capabilities.

Purpose of the Study:

• To develop a framework enhancing LLM performance in legal dispute analysis.
• To address limitations in legal concept understanding, reasoning consistency, and source citation for LLMs.

Main Methods:

• A framework combining hierarchical prompt engineering (task definition, knowledge background, reasoning guidance) and multi-dimensional knowledge graphs (ontology, representation, instance layers).
• Four legal concept retrieval methods: direct code matching, semantic vector similarity, ontology path reasoning, and professional terminology matching.
• Systematic testing on 500 samples from six benchmark datasets.

Main Results:

• Significant performance improvements in mainstream LLMs for legal dispute analysis.
• F1 score increased from 0.356 to 0.714.
• BLEU-4 reached 0.451, ROUGE-L F1 improved from 0.34 to 0.71.
• Legal professional content quality scores increased by 18-20 points.

Conclusions:

• The proposed framework effectively enhances LLM capabilities for legal dispute analysis.
• This approach contributes to the advancement of intelligent legal assistance systems.
• The framework offers a technical solution for more accurate and reliable legal AI.