Scholarlink

This study addresses the challenges of developing question-answering systems (QAS) tailored for the Kazakh legal domain. We explore modern NLP models, including GPT-3.5, Llama-2, and Llama-3, assessing their effectiveness in processing legislative texts and question-answer datasets. Leveraging structured data from two major Kazakh legal platforms, we implemented and evaluated methodologies for data scraping, processing, and model fine-tuning. The findings reveal the superior performance of the Llama-3 model in terms of accuracy, F1 score, and adaptability to the Kazakh language, despite its low-resource nature. Our results demonstrate the feasibility of advanced QAS for legal texts, paving the way for improved public access to legal information.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Developing Advanced Question-Answering Models for Legal Kazakh Texts: A Comparative Study of Modern Approaches

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Diana Rakhimova,
  • Vladislav Karyukin,
  • Aidana Karibayeva,
  • Asia Sarsenbayeva

摘要

This study addresses the challenges of developing question-answering systems (QAS) tailored for the Kazakh legal domain. We explore modern NLP models, including GPT-3.5, Llama-2, and Llama-3, assessing their effectiveness in processing legislative texts and question-answer datasets. Leveraging structured data from two major Kazakh legal platforms, we implemented and evaluated methodologies for data scraping, processing, and model fine-tuning. The findings reveal the superior performance of the Llama-3 model in terms of accuracy, F1 score, and adaptability to the Kazakh language, despite its low-resource nature. Our results demonstrate the feasibility of advanced QAS for legal texts, paving the way for improved public access to legal information.