图书情报知识 ›› 2024, Vol. 41 ›› Issue (1): 6-11.doi: 10.13366/j.dik.2024.01.006

• 观点视角 • 上一篇    下一篇

数智赋能信息资源管理新路径: 指令工程的概念、内涵和发展

陆伟1, 2 ,汪磊1, 2, 程齐凯1, 2, 刘家伟1, 2, 黄永1, 2   

  1. 1. 武汉大学信息管理学院,武汉,430072;
    2. 武汉大学信息检索与知识挖掘研究所,武汉,430072
  • 出版日期:2024-01-10 发布日期:2024-03-04
  • 通讯作者: 程齐凯(ORCID:0000-0003-3904-8901),博士,副教授,研究方向:文本挖掘、信息检索,Email:chengqikai@whu.edu.cn。
  • 作者简介:陆伟(ORCID:0000-0002-0929-7416),博士,教授,研究方向:信息检索、AI 治理、人机协同,Email:weilu@whu.edu.cn;汪磊(ORCID:0009-0005-5077-4607),硕士研究生,研究方向:信息抽取、文本挖掘,Email: wanglei@whu.edu.cn;刘家伟(ORCID:0000-0002-2774-1509),博士研究生,研究方向:信息检索、信息安全,Email:laujames2017@whu.edu.cn;黄永(ORCID:0000-0001-5953-6908),博士,副教授,研究方向:文本挖掘、科学计量,Email:yonghuang1991@whu.edu.cn。
  • 基金资助:
    本文系国家自然科学基金重点项目“数智赋能的科技信息资源与知识管理理论变革”(72234005)和国家自然科学基金面上项目“基于机器阅读理解的科学命题文本论证逻辑识别”(72174157)的研究成果之一。

A New Approach for Information Resources Management Empowered by Data Intelligence: The Concept, Connotation and Development of Instruction Engineering

LU Wei1, 2 , WANG Lei1, 2, CHENG Qikai1, 2, LIU Jiawei1, 2, HUANG Yong1, 2   

  1. 1.School of Information Management, Wuhan University, Wuhan, 430072;
    2. Information Retrieval and Knowledge Mining Laboratory, Wuhan University, Wuhan, 430072
  • Online:2024-01-10 Published:2024-03-04
  • Contact: Correspondence should be addressed to CHENG Qikai, Email: chengqikai@whu.edu.cn, ORCID: 0000-0003-3904-8901
  • Supported by:
    This is an outcome of the Key Project "Data and Intelligence Empowered Theoretic Change of Scientific Information Resource and Knowledge Management Theory"(72234005)and the project "Argumentation Logic Recognition of Scientific Proposition Text based on Machine Reading Comprehension"(72174157), both supported by National Natural Science Foundation of China.

摘要: 新一轮科技革命和产业变革方兴未艾,大数据、人工智能等系列数智技术对信息资源管理学科产生了深远影响。在大模型背景下,指令工程通过高质量、体系化、流程化的指令设计引导模型生成结果,是高效发挥大模型能力的重要途径,可以用于解决学科相关重要问题。本文首先介绍了指令工程的概念,然后详细梳理了指令的构成要素、设计模式以及指令工程的特点和意义,并探讨了指令工程赋能信息资源管理的建设路径。未来,指令工程的研究和发展还需要关注通用及领域指令工程建设、指令工程标准化、知识产权保护、安全性和体系化测试评估等问题,以期能够在各行业复杂的应用场景中更好地发挥指令的效能。

关键词: 指令工程, 大模型, 信息资源管理, 数智赋能

Abstract: The emerging wave of scientific revolution and industrial transformation is burgeoning, with technologies such as big data and artificial intelligence exerting a profound impact on the information resources management discipline. Under the background of large language models(LLMs), instruction engineering guides the generation of model results through high-quality, systematic and procedural instruction design, which is an important approach to efficiently leverage the capabilities of LLMs and can be applied to resolve discipline-related issues. This article firstly introduces the concept of instruction engineering, and then thoroughly analyzes the constituent elements and design patterns of instructions, as well as the features and significance of instruction engineering. We also discuss the construction path of instruction engineering empowered information resources management. In the future, research and development in instruction engineering should focus on the construction of both general and domain-specific instruction engineering, standardization, intellectual property protection, security and systematic testing and assessment, aiming to enable instructions to effectively demonstrate their efficacy in complex application scenarios across various industries.

Key words: Instruction engineering, Large language model, Information resources management, Data intelligence empowerment