Gov_GLM:基于ChatGLM的政务助手系统  

Gov_GLM:ChatGLM-based government assistant system

作　　者：蒙醒 陈亮 王珺琳 MENG Xing;CHEN Liang;WANG Junlin(School of Automation and Electrical Engineering,Shenyang Ligong University,Shenyang 110159,China;School of Information Science and Engineering,Shenyang Ligong University,Shenyang 110159,China)

机构地区：[1]沈阳理工大学自动化与电气工程学院,辽宁沈阳110159 [2]沈阳理工大学信息科学与工程学院,辽宁沈阳110159

出　　处：《通信与信息技术》2025年第2期130-135,共6页Communication & Information Technology

摘　　要：近年来,生成式大型语言模型在人工智能领域取得了显著的技术进展。因其在自然语言处理和文本分析方面的强大能力得到包括政务在内的多个领域的广泛应用。然而,国家政策涉及复杂逻辑知识和解释性问题使用户难以解读。针对这些问题,通过使用Lora方法对ChatGLM模型进行重参数化微调,增强模型政务知识处理能力,并构建相应Prompt以优化模型问答意图理解能力。实验研究表明,问答系统的准确性和完整性均高于对比模型。通过在自制政务服务问答对话数据集上验证,Gov_GLM的BLEU分数达到75.2%,模型的准确性提高,显著降低系统问答结果的复杂度,辅助用户更好地解读政务信息。Generative large language models have made remarkable progress in the field of artificial intelligence in recent years.Be⁃cause of its powerful ability in natural language processing and text analysis,it has been widely used in many fields,including government affairs.However,in the government service work,it is difficult for users to interpret the national policy because of the complex logic knowledge and interpretation problems.To solve these problems,the Lora method is used to reparameterize and fine-tune ChatGLM mod⁃el to enhance the model′s government knowledge processing ability,and the corresponding Prompt is constructed to optimize the model′s question answering intention understanding ability.Experimental research shows that the proposed question-answering system outper⁃forms baseline models in both accuracy and completeness.By verifying on the self-made government service question and answer dia⁃logue dataset,Gov_GLM achieves 75.2%BLEU score,The model not only improves accuracy but also significantly reduces the complexi⁃ty of system responses,thereby facilitating users′comprehension of government-related information.

关 键 词：ChatGLM 政务问答 Lora方法 PROMPT BLEU 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]