Chinese Relation Extraction of Apple Diseases and Pests Based on A Dual-Channel Neural Network.

Relation extraction is one of the basic tasks in the construction of the knowledge graph and intelligent Q&A of apple diseases and pests. In view of multiple relation categories between apple disease and pest entities and the unbalanced distribution of relation data, we propose a novel model named DC-ARE for relation extraction based on dual-channel mechanism. Specifically, the dual channels are used to enhance the semantic representation and obtain richer features of data. The performance of DC-ARE is evaluated on AppleRE, a self-made apple disease and pest relation extraction dataset containing 23 relation categories. It realizes an F1-score of 94.08% under the AppleRE dataset. Compared to BLSTM and Att-BLSTM, the F1-score increases by 4.38% and 3.25%, respectively. The F1-scores on seven small samples are all higher than comparison models. The results demonstrate that the proposed model has better performance in relation extraction of apple diseases and pests, and can solve the problem of unbalanced distribution of relation data.