融合预训练语言模型的知识图谱在政务问答系统中的应用研究

打开文本图片集

摘要：该文针对当前政务问答系统面临的复杂语境理解、政策法规解释等问题，探讨了如何将预训练语言模型与知识图谱进行有效融合，以实现更加精准、全面和个性化的政务信息问答服务，构建了政务问答系统框架，利用知识图谱和大模型工具验证了该方法在提高问答准确率、增强上下文理解能力方面的显著优势。

关键词：知识图谱；自然语言处理；预训练语言模型；三元组；知识库

doi：10.3969/J.ISSN.1672-7274.2024.09.063

中图分类号：TP 3                 文献标志码：A            文章编码：1672-7274（2024）09-0-03

Research on the Application of Knowledge Graph Integrated with Pre-trained Language Models in Government Question-answering Systems

ZHANG Chaoyang， SHEN Jianhui， YE Weirong

（Zhejiang Public Information Industry Co.， LTD.， Hangzhou 310000， China）

Abstract： Aiming at the problems of complex context understanding and interpretation of policies and regulations faced by the current government question answering system， this paper discusses how to effectively integrate pre-trained language models and knowledge graphs， so as to realize more accurate， comprehensive and personalized government information question answering service. The framework of government question answering system is constructed， and the significant advantages of this method in improving the accuracy of question answering and enhancing the context understanding ability are verified by using knowledge graph and large model tools.

Keywords： knowledge graph; natural language processing; pre-trained language model; triple; knowledge base

0   引言

政务问答系统的核心在于如何更好地建模语言、理解和输出文本信息，本文以政务服务垂直领域在线咨询问答场景为例，探索预训练语言模型与知识图谱的融合应用。（剩余3992字）