Chinese named entity recognition of apple diseases and pests based on iterative dilated convolution
Yue Zhang, Pan Pu, Lvwen Huang, Bo Qian, Yong Liu · 2023
In order to solve the problems of Chinese named entity recognition in the field of apple diseases and pests, including the difficulty of identifying long entities and polysemous words, we proposed a Chinese named entity recognition model based on iterative dilated convolution. Specifically, the Bert pre-trained model was used as the embedding layer to enhance the semantic representation of the text by integrating the text position information. The iterative dilated convolutional neural network (IDCNN) was used to extract text context features and output all possible sequences labels to the conditional random field (CRF) to get the optimal sequence label. To verify the validity of the model, we constructed a corpus containing 6 entity categories of apple diseases and pests, and experiments were performed on the corpus. The experimental results showed that the precision, recall and F1 score of the proposed model are 93.29%, 93.68% and 93.49%, respectively, which are superior to the other three conventional models. It provides a new idea for the recognition of complex named entities in the field of agriculture.