Scholarlink

This work presents a question-answering (QA) system combining a knowledge graph (KG) and a large language model (LLM) to boost response accuracy and efficiency in answering tourism-associated queries. The system builds a structured KG from trusted data sources as a foundation for precise retrieval of factual questions and natural and context-aware responses generated by GPT. In experiments, the integration of KG with GPT proves to increase accuracy to 91% on simple and 86% on complex questions compared to the case with GPT as a solitary system at the levels of 80% and 74%, respectively. Besides, response times decrease with the system reaching 1.1 seconds on simple questions and 2.4 seconds on complicated ones, better than GPT alone (1.4 seconds and 3.0 seconds, respectively). These results illustrate the benefits of combining KG with LLMs to produce more precise, uniform, and efficient QA performance.

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

Enhancing QA Systems with Knowledge Graphs and LLMS: A Case Study in Vietnamese

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • A. Nguyen Thi Thuy,
  • Tieu Phung Mai Suong,
  • Chien D. C. Ta,
  • Nguyen Hong Vu,
  • Nguyen Phuoc Dai

摘要

This work presents a question-answering (QA) system combining a knowledge graph (KG) and a large language model (LLM) to boost response accuracy and efficiency in answering tourism-associated queries. The system builds a structured KG from trusted data sources as a foundation for precise retrieval of factual questions and natural and context-aware responses generated by GPT. In experiments, the integration of KG with GPT proves to increase accuracy to 91% on simple and 86% on complex questions compared to the case with GPT as a solitary system at the levels of 80% and 74%, respectively. Besides, response times decrease with the system reaching 1.1 seconds on simple questions and 2.4 seconds on complicated ones, better than GPT alone (1.4 seconds and 3.0 seconds, respectively). These results illustrate the benefits of combining KG with LLMs to produce more precise, uniform, and efficient QA performance.