Chinese named entity recognition for apple diseases and pests based on character augmentation

• A homemade corpus for apple diseases and pests is constructed, which contains 21 entity categories in total. • An original Chinese named entity recognition model for apple diseases and pests is proposed, which improves the recognition effect by incorporating dictionaries and similar words into the character representation. • We propose two strategies to integrate similar words into a neural network. • Our proposed model is highly competitive with the state-of-the-art models in terms of recognition performance and efficiency and has a certain generalization. Aiming at the problems of Chinese named entity recognition in the field of apple diseases and pests, including various entities categories, entities with aliases or abbreviations, and the difficulty of identifying rare entities, we propose a novel Chinese named entity recognition model APD-CA based on character augmentation. Specifically, we incorporate dictionaries and similar words into the character-based BiLSTM-CRF model to augment character representation. To verify the validity of the model, experiments were performed on ApdCNER, a manually derived Chinese apple disease and pest corpus containing 21 entity categories. The experimental results show that the precision, recall, and F1-score of the APD-CA model based on ApdCNER are 92.29%, 91.99%, and 92.14%, respectively, which are improved compared with those for the baseline model and four other state-of-the-art models. The improvement verifies that the proposed model in this paper has performance advantages in named entity recognition in the field of apple diseases and pests. Other experimental results also prove that this model has efficiency advantages and certain generalization advantages.