细粒度苹果病虫害知识图谱构建研究  

Research on Construction of Fine-Grained Knowledge Graph of Apple Diseases and Pests

作　　者：张嘉宇 郭玫 张永亮 李梅[1] 耿楠[1,2,3] 耿耀君[1] ZHANG Jiayu;GUO Mei;ZHANG Yongliang;LI Mei;GENG Nan;GENG Yaojun(College of Information Engineering,Northwest A&F University,Yangling,Shaanxi 712100,China;Key Laboratory of Agricultural Internet of Things,Ministry of Agriculture and Rural Affairs,Northwest A&F University,Yangling,Shaanxi 712100,China;Shaanxi Key Laboratory of Agricultural Information Perception and Intelligent Service,Northwest A&F University,Yangling,Shaanxi 712100,China)

机构地区：[1]西北农林科技大学信息工程学院,陕西杨凌712100 [2]西北农林科技大学农业农村部农业物联网重点实验室,陕西杨凌712100 [3]西北农林科技大学陕西省农业信息感知与智能服务重点实验室,陕西杨凌712100

出　　处：《计算机工程与应用》2023年第5期270-280,共11页Computer Engineering and Applications

基　　金：陕西省重点研发计划(2019ZDLNY07-06-01);国家重点研发计划(2020YFD1100601)。

摘　　要：鉴于现有农业知识图谱对病虫害防治相关实体、关系刻画不够细致的问题,以苹果病虫害知识图谱构建为例,研究细粒度农业知识图谱的构建方法。对苹果病虫害知识的实体类型和关系种类进行细粒度定义,共划分出19种实体类别和22种实体关系,以此为基础标注并构建了苹果病虫害知识图谱数据集AppleKG。使用APD-CA模型对苹果病虫害领域命名实体进行识别,使用ED-ARE模型对实体关系进行抽取。实验结果表明,该文模型在命名实体识别和关系抽取两项子任务中的F1值分别达到了93.08%和94.73%。使用Neo4j数据库对知识图谱进行了存储和可视化,并就细粒度苹果病虫害知识图谱可以为精准病虫害信息查询、智能辅助诊断等下游任务提供底层技术支撑进行了讨论。In view of the problem that existing agricultural knowledge graphs do not portray entities and relationships related to disease and pest control in sufficient detail, this paper takes the construction of a knowledge graph of apple diseases and pests as an example to study the construction method of fine-grained agricultural knowledge graphs. Firstly,the entity types and relationship types of apple disease and pest knowledge are defined at a fine-grained level, and a total of 19 entity categories and 22 entity relationships are classified, based on which the apple disease and pest knowledge graph dataset AppleKG is annotated and constructed. Then, the APD-CA model is used to identify named entities in the apple disease and pest field, and the ED-ARE model is used to extract the relationships between entities. The experimental results show that the F1-score of the models in this paper reaches 93.08% and 94.73% in the subtasks of named entity recognition and relationship extraction, respectively. Finally, the knowledge graph is stored and visualised using the Neo4j database, and a discussion is held on how fine-grained apple disease and pest knowledge graphs can provide the underlying technical support for downstream tasks such as accurate disease and pest information query and intelligent assisted diagnosis.

关 键 词：苹果病虫害防治 知识图谱 深度学习 循环神经网络 知识抽取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]