Chinese Named Entity Recognition with Integrated Channel Attention

Using an Iterated Dilated Convolutional Neural Network (IDCNN) in Chinese Named Entity Recognition (NER) helps capture local information. However, the contribution of information from different positions in a sentence to the judgment of the current position is not uniform, which needs to be considered when expanding a Dilated Convolutional Neural Network (DCNN). To address this, this paper incorporates Efficient Channel Attention (ECA) into DCNN and constructs the It-erated Dilated Convolutional Attention Network (IDCAN). The model focuses on the importance of different channel information and integrates the Linguistically-motivated Bidirectional Encoder Representation from Transformers (LERT), Bidirectional Gated Recurrent Units (BiGRU), and Conditional Random Fields (CRF) to form a Chinese NER model. The model uses LERT for word vector construction, BiGRU for capturing temporal features, IDCAN for extracting key local features, and CRF for feature decoding. On the Daily, MSRA and E-commerce Chinese datasets, the model achieved F1 scores of 96.88%, 95.15% and 90.52%, respectively, outperforming other models.