Multi-Feature Fusion Method for Chinese Pesticide Named Entity Recognition

Chinese pesticide named-entity recognition (NER) aims to identify named entities related properties from unstructured information texts. In view of the characteristics massive, fragmented, professional, and complex semantic relationships data, a deep learning method based on multi-feature fusion was applied improve accuracy NER. this study, data set is manually annotated by begin inside outside (BIO) sequence annotation scheme. Bi-directional long short-term memory (BiLSTM) iterated dilated convolutional neural networks (IDCNN) combined with conditional random field (CRF) form model BiLSTM-IDCNN-CRF, it implement in sets. IDCNN introduced enhance representation ability local feature capture text. BiLSTM network are obtain long-distance dependence relationship context features different granularity Finally, CRF used labeling task. According experiment results, rate, recall F1 score BiLSTM-IDCNN-CRF were 78.59%, 68.71%, 73.32%, respectively, which significantly better than other compared models. Experiments show that can effectively extract text helpful constructing knowledge graph intelligent question-answering.