摘要
[目的/意义]大语言模型的爆火为智能问答系统带来了颠覆性变革,给图书馆参考咨询服务的智能化建设提供了创新方向。文章旨在探究大语言模型在图书馆参考咨询服务中的实际应用方案,并评估其生成答案的效果,以期为图书馆的智能化创新发展提供参考。[方法/过程]采用基于p-tuning的大语言模型微调方案提高模型的智能性,根据问答数据构建本地知识库以规范模型内容生成,并利用langchain应用框架构建咨询系统,最后设计评价指标进行主客观综合效果评估。[结果/结论]通过大语言模型微调+langchain本地知识库的联合应用方案,既能发挥模型生成内容的智能性,同时生成内容得到正确规范,生成答案BERT Score的F1值达到0.823,验证了其在参考咨询服务中的可行性,为智慧图书馆的AI革新提供创新方向。
[Purpose/significance]The explosion of large language models has brought disruptive changes to intelligent question and answer systems and provided innovative directions for the intelligent construction of library reference consulting services.This study aims to investigate the practical application scheme of Big Language Model in library reference consulting service and evaluate its effectiveness in generating answers,in order to provide reference for the development of intelligent innovation in libraries.[Method/process]A p-tuning-based fine-tuning scheme of the big language model is adopted to improve the intelligence of the model,a local knowledge base is constructed to regulate the model content generation based on the question and answer data,and a consulting system is constructed using the langchain application framework,and finally evaluation indexes are designed to evaluate the comprehensive subjective and objective effects.[Result/conclusion]The joint application scheme of Large Language Model Fine-tuning+langchain local knowledge base can bring into the intelligence of model-generated content,and at the same time the generated content is correctly standardized.The F1 value of generated answer BERT score reaches 0.823,which verifies its feasibility in reference consulting service and provides innovative directions for AI innovation in smart libraries.
出处
《情报理论与实践》
北大核心
2023年第8期96-103,共8页
Information Studies:Theory & Application
关键词
大语言模型
人工智能生成内容
智慧图书馆
参考咨询
large language models
artificial intelligence generated content
smart library
reference consulting
