基于知识图谱的花卉病虫害知识管理方法  

Knowledge Management Method of Flower Diseases and Pests Based on Knowledge Graph

作　　者：陈明[1,2] 朱珏樟 席晓桃 CHEN Ming;ZHU Juezhang;XI Xiaotao(College of Information Technology,Shanghai Ocean University,Shanghai 201306,China;Key Laboratory of Fisheries Information,Ministry of Agriculture and Rural Affairs,Shanghai 201306,China)

机构地区：[1]上海海洋大学信息学院,上海201306 [2]农业农村部渔业信息重点实验室,上海201306

出　　处：《农业机械学报》2023年第3期291-300,共10页Transactions of the Chinese Society for Agricultural Machinery

基　　金：上海市科技创新计划项目(20dz1203800)。

摘　　要：为解决花卉病虫害领域中病虫害防治因素关系复杂、知识冗余等问题,结合知识图谱对知识组织和管理的技术,提出一种基于知识图谱的花卉病虫害知识管理方法。首先,根据文献提取包括环境在内的花卉病虫害防治要素,构建花卉病虫害本体模型并存储在RDF图中,实现对知识规范性和完整性的控制;其次,对花卉病虫害领域文本进行分析,针对分析得到的文本特点,提出融合头尾实体分离“01”标注方法、轻量级双向转换编码表示模型(A lite BERT, ALBERT)和引入词性特征的级联标注模型(CasPOSRel)的抽取框架进行三元组抽取;之后利用自定义RDF2PG映射算法,按照RDF图中的本体模型将抽取到的三元组存入Neo4j数据库中,完成对花卉病虫害知识的存储及管理。实验结果证明提出的抽取框架中标注方法、预训练模型与抽取模型相比基线方法F1值分别提升0.88、4.90、8.57个百分点,同时得到抽取结果F1值为95.07%。通过知识发现表明该知识管理方法能有效组织管理病虫害知识,帮助花卉种植人员进行更为有效的病虫害防治工作。In order to solve the problems of complex relationship of factors and mixed knowledge in the field of flower diseases and pests,combined with the knowledge organization and management technology of knowledge graph,a knowledge management method of flower diseases and pests based on knowledge graph was proposed.Firstly,according to the literatures,the flower diseases and pests control elements,including environment were extracted,the flower diseases and pests ontology model was constructed and stored in RDF to realize the control of knowledge standardization and integrity.Secondly,according to the text characteristics obtained from the analysis,the triple extraction framework was proposed which combined the“01”tagging method of head and tail entity separation,a lite bidirectional encoder representations from transformers(ALBERT)and cascade tagging model with part of speech features(Cas_(POS) Rel).Then using the custom RDF2PG mapping algorithm to complete the storage and management of flower diseases and pests knowledge.The experiments showed that the F1 value of the tagging methods,pretrained model and extraction model in proposed extraction framework was increased by 0.88,4.90 and 8.57 percentage points compared with that of baseline methods,and the F1 value of the extraction result was 95.07%.The knowledge discovery showed that the knowledge management method effectively organized and managed the knowledge of flower diseases and pests,and helped the flower growers to carry out more effective pests control work.

关 键 词：花卉 病虫害防治 知识图谱 知识抽取 知识管理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]