Chinese Named Entity Recognition Based on Gated Graph Neural Network

Most Chinese Named Entity Recognition (CNER) models based on deep learning are implemented based on long short-term memory networks (LSTM) and conditional random fields (CRF). The serialized structure of LSTM is easily affected by word ambiguity and lack of word boundary information. In this regard, we propose a Chinese named entity recognition model based on a gated graph neural network (GGNN).We use the BERT model to generate pre-training encoding vectors of characters, and introduce global nodes to capture the global information in the sentence. Finally, we exploits multiple interactions between the characters in the graph structure, all matching words, and the entire sentence to solve the problem of word ambiguity.The comparative experimental results on the three CNER datasets show that the GGNN model has a better effect on named entity recognition.