北京信息科技大学学报（自然科学版）2025，Vol.40Issue(6)：41-48,8.DOI:10.16508/j.cnki.11-5866/n.2025.06.005
精准农业中条件感知融合知识检索方法
A condition-aware fusion knowledge retrieval method for precision agriculture
摘要
Abstract
As agricultural decision-making becomes increasingly complex,existing retrieval augmented generation(RAG)models often fail to effectively capture the impact of condition constraints on information retrieval when processing agricultural knowledge with strong condition-dependencies,resulting in retrieval results that do not match practical requirements.To address this issue,a condition-aware fusion knowledge retrieval(CAFKR)method was proposed to enhance the retrieval precision and domain adaptability of RAG systems by optimizing the retrieval process.CAFKR introduces a condition recognition framework to precisely parse condition-constraint relationships within agricultural knowledge,and combines a multi-path retrieval mechanism with an adaptive information fusion strategy to construct a precise mapping between conditions and information.These mechanisms effectively strengthen the capability to handle condition-dependencies in the agricultural domain.Experimental results indicate that CAFKR achieves a Recall@5 of 0.910 1,significantly outperforming traditional retrieval models.This approach provides a reliable foundation for the future development of RAG systems,enabling more precise and efficient knowledge retrieval in complex fields such as agriculture.关键词
精准农业/条件感知融合/多路径检索/检索增强生成Key words
precision agriculture/condition-aware fusion/multi-path retrieval/retrieval augmented generation(RAG)分类
信息技术与安全科学引用本文复制引用
赵克清,王一群,陈雯柏..精准农业中条件感知融合知识检索方法[J].北京信息科技大学学报（自然科学版）,2025,40(6):41-48,8.基金项目
科技部2030新一代人工智能重大专项(2021ZD0113603) （2021ZD0113603）
国家自然科学基金项目(62276028) （62276028）
国家自然科学基金重大研究计划(92267110) （92267110）
