农业文本语义理解技术综述

随着互联网和人工智能技术的发展，农业知识智能化服务逐渐承担起为农业生产管理提供有效技术指导的作用。本文对农业文本语义理解中的关键技术及应用进行综述。首先按照自然语言处理中基于规则、机器学习和深度学习的语义处理方法介绍其在农业领域应用的进展；然后阐述了针对农业知识特性的语义分析方法，涵盖农业文本分析主要过程的储存、表达、计算，包括农业知识图谱的知识抽取、融合、表示、推理，TF-IDF、Word2Vec、BERT等农业文本表示模型与CNN、RNN、Attention等分类模型；阐述了可用于分词、向量化表达等的通用语料库和农业领域常用语料库；从农业智能问答、农业语义检索、农业智能管理决策方面阐述语义理解在农业领域中的应用；最后从农业语料库标准化构建、语义理解模型复杂度、多模态语义处理、多区域多语言语义理解等方面对农业文本的语义理解研究趋势进行了展望。

Review of Semantic Analysis Techniques of Agricultural Texts

With the development of Internet and artificial intelligence technology, agricultural knowledge intelligent services have gradually assumed the role of providing effective technical guidance for agricultural production management, especially during the epidemic. The key technologies and applications in the semantic understanding of agricultural knowledge service texts were reviewed. Firstly, its progress in agriculture was introduced according to the semantic processing methods based on rules, machine learning and deep learning in natural language processing. Then, the semantic analysis method for the characteristics of agricultural knowledge was introduced, covering the storage, expression and calculation of the main process of agricultural text analysis, including knowledge extraction, knowledge fusion, knowledge representation and knowledge inference of agricultural knowledge graph. The representation model of agricultural text such as TF-IDF, Word2Vec and BERT and classification models such as CNN, RNN and Attention were presented. Then the common corpus was described. The application of semantic understanding in agriculture from the aspects of agricultural intelligent question answering, agricultural semantic retrieval and agricultural intelligent management decision as well were introduced. Finally, the research trend of agricultural text semantic understanding was prospected from the aspects of standardization construction of agricultural corpus, complexity of semantic understanding model, multi-modal semantic processing, multi-region and multi-language semantic understanding.