农业大数据学报

• •    下一篇

基于本体的红枣领域知识智能服务研究

薛原1,冯建中1*,杨浩1,2,王剑1,高祥伟2,谢能付1,曹丹3*,白林燕4,陈彩铭1   

  1. 1.中国农业科学院农业信息研究所,北京 100081;2. 江苏海洋大学海洋技术与测绘学院,江苏 连云港 222005;3. 中国消防救援学院,北京 102201;4. 中国科学院空天信息创新研究院,北京 100094

Research on Intelligent Services for Jujube Domain Knowledge, Based on Ontology

XUE Yuan1, FENG Jianzhong1*, YANG Hao1, 2, WANG Jian1, GAO Xiangwei2, XIE Nengfu1, CAO Dan3*, BAI Linyan4, CHEN Caiming1   

  1. 1. Agricultural Information Institute of Chinese Academy of Agricultural Sciences, Beijing 100081, China; 2. School of Marine Technology and Geomatics, Jiangsu Ocean University, Lianyungang, Jiangsu 222005, China; 3. China Fire and Rescue Institute, Beijing 102201, China; 4. Aerospace Information Research Institute, Chinese Academy of SciencesBeijing 100094, China

摘要: 针对当前红枣领域知识体系碎片化、更新滞后、知识服务个性化不足、服务质量与效率低等问题实现红枣领域知识的高效组织智能服务,本文开展了基于本体与知识图谱的红枣产业领域知识智能服务研究,首先系统分析了红枣产业信息知识服务过程体系,构建了红枣领域知识服务框架,探索了红枣产业领域知识组织方式,基于领域知识本体与知识图谱开展了红枣领域知识智能服务机制与服务模式研究;进而通过对红枣产业种植经营及相关领域多源异构语料进行系统整合与处理,构建了规范化多维领域知识本体库,同时利用Bert-BiLSTM-CRF混合模型,构建了具有良好完备性、消歧容错与关联推理能力等优势的红枣产业领域知识图谱。在此基础上,基于本体和知识图谱及领域知识服务协同优化原理,研发设计了一套融合领域专用模型与大模型的红枣产业领域知识智能问答系统,并且探讨了智能服务评估的优化路径,以保障服务的高质量和可信度。本文对提升红枣领域知识智能服务的实用性与适用性、推动其产业数字化与智慧化转型升级,探索可行的技术路径提供重要的实践参考。

Abstract: To address the current issues in the red jujube domain, including fragmented knowledge systems, delayed updates, insufficient personalization of knowledge services, and low service quality and efficiency, and to meet the practical needs for efficient organization and intelligent services of red jujube domain knowledge, in this paper we carried out a study about ontology-based intelligent knowledge services, which was associated with relevant aspects, for the red jujube industry. Firstly, a fundamental framework for jujube domain knowledge services was established through systematic analysis of the knowledge service processes in jujube industry, and based on domain ontology and knowledge graphs, the intelligent service mechanism and patterns for jujube domain were studied by exploring knowledge organization approaches about jujube industry. Secondly, the ?structured, semi-structured and unstructured multi-source heterogeneous corpus information in terms of the cultivation and management of jujube industry and its related domains was systematically integrated and processed (e.g., semantic preprocessing), and then, the multi-dimensional, standardized domain knowledge and ontology bases for jujube industry were constructed, and farther, the domain knowledge graph for jujube industry, which showed good merits such as favorable completeness, well fault-tolerant and disambiguation capabilities, and strong relational reasoning competence, was built by using a BERT-BiLSTM-CRF hybrid model. Hence, to satisfy the requirements of intelligent question-and-answering (Q&A) services for jujube domain, an intelligent Q&A system, based on the domain ontology and knowledge graph and keeping to the collaborative optimization principle of domain knowledge services, was designed while employing a hybrid processing framework with the domain-specific models and (locally deployable) large language models (LLMs)‌, and moreover, the improvement approaches for evaluation of knowledge intelligence services were discussed to ensure the provision of high-quality, reliable, and satisfactory services to users and/or systems. As for all the above, the paper would provide significant practical references for enhancing the practicality and applicability of intelligent services in jujube domain, promoting the digital and intelligent transformation and upgrading of jujube industry, and thus further exploring their feasible technical pathways.