基于大语言模型的科技项目查新方法研究

打开文本图片集

关键词：大语言模型；科技项目查新；信息检索；检索策略；语义分析中图分类号：G251.5 文献标识码：A

Research on Novelty Retrieval Methods for Scientific and Technological Projects Based on Large Language Models

Abstract LargeLanguage Models （LLMs）canasist professionals inconducting noveltyretrieval for scientific and technological projects，therebyimproving both theeficiencyandqualityoftheretrieval process.This paper explores the potential ofLLMs insupporting novelty retrieval and summarizes four application paradigms： using LLMs to analyze novelty points and construct search queries;employing LLMs to expand and identify novelty retrieval search terms; leveragingLLMs tocompare technical similarities and diferences across the literature;and directlyapplying LLMs to retrieve closely related patent documents.Drawing onpractical novelty retrieval cases，the study validates the feasibility and advancement of LLM-assisted approaches and further explainsaspects of their working mechanisms.Finally，the paper proposes recommendations fordevelopingdomain-specific largemodels fornovelty retrieval，including： enhancing thecross-disciplinaryadaptabilityofLLMs tobuild more versatilemodels;improving theircapacityto interpretthe deeper meanings of literature;increasing model interpretabilityand transparency；and establishing mechanisms for continuous learning and updating.

KeyWords Large Language Models; noveltyretrieval for scientificand technologicalprojects; informationretrieval; retrieval strategy; semantic analysis

1引言

目前，传统的科技项目查新主要依赖查新人员的专业知识和经验开展，查新人员通过查新委托书技术要点描述、项目申报书或结题报告等文件资料理解查新技术要点，期间要不断与委托人和领域专家沟通，尝试查新关键词的组配并查找文献。（剩余8708字）