Apple disease and insect pest named entity recognition method based on dynamic dictionary features and CBAM

The invention relates to an apple disease and insect pest named entity recognition method based on dynamic dictionary features and CBAM, and the method comprises the steps: dynamically obtaining dictionary information in an embedded layer through employing a channel attention network based on a bidirectional long-short term memory-conditional random field model of characters, and fusing the four-corner vector information of the characters, so as to improve the representation capability of the characters; in a sequence coding layer, in order to compensate for the BiLSTM feature extraction capability, a space attention network is introduced, and a parallel connection space attention module is newly added; and the CRF layer constrains the label sequence and obtains a globally optimal label sequence. 本申请涉及一种基于动态词典特征和CBAM的苹果病虫害命名实体识别方法，包括：基于字的双向长短时记忆-条件随机场模型，在嵌入层利用通道注意力网络动态获取词典信息，同时融合字的四角向量信息，以提高字的表征能力；在序列编码层，为弥补BiLSTM特征提取能力，引入空间注意网络，新增并行连接的空间注意力模块；CRF层约束标签序列并获得一个全局最优的标注序列。