基于大型语言模型的AI招生咨询助理设计与实现  

Design and Implementation of AI Admissions Consulting Assistant Based on Large Language Models

作　　者：阮昆[1] 杨璟轩 殷旭 储雯[1] 罗婷婷 黄容[1] RUAN Kun;YANG Jingxuan;YIN Xu;CHU Wen;LUO Tingting;HUANG Rong(Office of Educational Informationization,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;School of Computer Science and Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;School of Automation,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

机构地区：[1]重庆邮电大学教育信息化办公室,重庆400065 [2]重庆邮电大学计算机科学与技术学院,重庆400065 [3]重庆邮电大学自动化学院,重庆400065

出　　处：《实验室研究与探索》2025年第2期110-116,共7页Research and Exploration In Laboratory

基　　金：2024年度重庆市高等教育考试招生研究课题(CQZSKS2024072)。

摘　　要：针对高考招生咨询业务繁忙,咨询覆盖范围有限、咨询效率不高等问题,基于检索增强生成、大型语言模型、提示词工程和检索增强生成转结构化查询语言等技术构建AI招生咨询助理,搜集学校招生信息网招生政策、常见问题、学院专业介绍等建立本地权威招生知识库,对政策咨询类问题直接在本地向量知识库检索,对数据查询类问题转化为SQL数据查询,将检索或查询结果送至大模型推理生成回复,提升提问方式的自由度以及问题回复的权威性和实时性,降低大模型幻觉,实现全天候为考生和家长提供精准化、智能化、个性化的咨询服务。在2024年高考招生咨询中,大幅度减轻学校招生咨询工作压力,有效提升招生咨询效率,促进公平获取招生信息。In response to the challenges faced by college entrance examination(CEE)admission consultation services,such as heavy workload,limited coverage,low efficiency,and the inability to provide precise,intelligent 724 consultation services,this paper proposes an AI admission Consultation Assistant.This assistant integrates advanced technologies including Retrieval-Augmented Generation(RAG),Large Language Models(LLM),Prompt Engineering,and RAG-to-SQL(converting natural language queries into SQL for database searches).By collecting policies,FAQs,and introductions of faculties and majors from university admissions websites,a local authoritative knowledge base for admissions information has been established.For admission policy queries regarding application,the system performs searches within a local vector-based knowledge base;for data retrieval queries,it accurately converts questions into SQL statements to perform database queries.The results of these searches or queries are then fed into the large language model for inference and response generation,effectively reducing hallucinations in the LLM responses and enhancing the flexibility of query formulation for data retrieval questions,as well as the authority and timeliness of the answers provided.This paper achieves round-the-clock provision of precise,intelligent,and personalized consultation services for examinees and parents.During the CEE consultations in 2024,it cumulatively responded to over 30,000 admission consultation queries,significantly alleviating the pressure on school admisson consultation work,improving consultation efficiency,and promoting fairness in accessing admissions information.

关 键 词：大型语言模型 检索增强生成技术 提示词工程 招生咨询 

分 类 号：G434[文化科学—教育学] TP18[文化科学—教育技术学] TP391.1[自动化与计算机技术—控制理论与控制工程]