大语言模型方法在情报实践中的应用  

Application of Large Language Model Methods in Scientific and Tech⁃ nical Intelligence Practice

作　　者：化柏林[1] 王英泽 HUA Bolin;WANG Yingze(Department of Information Management,Peking University,Beijing 100871)

机构地区：[1]北京大学信息管理系,北京100871

出　　处：《科技情报研究》2025年第1期53-64,共12页Scientific Information Research

基　　金：国家社会科学基金重大项目“大数据驱动的科技文献语义评价体系研究”(编号:21&ZD329)。

摘　　要：[目的/意义]大语言模型凭借着对大规模数据集处理的较强能力和在各项自然语言处理任务中的超群表现,在各行业中皆有出色发挥,而情报领域以文本数据为主要信息源,所以大语言模型非常适用于情报工作,并为情报实践带来了新一轮变革浪潮。[方法/过程]从文本的低维稠密向量表示、大规模预训练模型、微调与提示学习、高质量大规模训练数据、人类对齐技术5个维度讨论了大语言模型的优势。[结果/讨论]大语言模型在情报识别、情报跟踪、情报比评、情报预测等情报任务中皆有广泛应用,且带来了显而易见的优化提升或范式改变。[Purpose/significance]With the strong ability to process large-scale datasets and outstanding performance in various natural language processing tasks,large language models(LLMs)have excelled across multiple industries.Since scientific and technical intelligence primarily relies on textual data,LLMs are naturally well-suited for this field,ushering in a new wave of transformative changes.[Method/process]This article discusses the advantages of LLMs from five perspectives:low-dimensional dense vector representations of text,large-scale pre-trained models,fine-tuning and prompt learning,high-quality large-scale training data,and human alignment techniques.[Result/conclusion]LLMs have extensive applications in tasks such as intelligence identification,intelligence tracking,intelligence evaluation,and intelligence prediction,resulting in significant optimization improvements or paradigm shifts.

关 键 词：大语言模型 情报学 情报方法 情报实践 深度学习 文本信息 

分 类 号：G350.7[文化科学—情报学] TP18[自动化与计算机技术—控制理论与控制工程]