Robust Chinese Named Entity Recognition Based on Fusion Graph Embedding

Named entity recognition is an important basic task in the field of natural language processing. The current mainstream named methods are mainly based on deep neural network model. vulnerability itself leads to a significant decline accuracy when there adversarial text text. In order improve robustness under conditions, this paper proposes Chinese model fusion graph embedding. Firstly, encodes and represents phonetic glyph information input through learning integrates above-multimodal knowledge into model, thus enhancing Secondly, we use Bi-LSTM further obtain context Finally, conditional random used decode label entities. experimental results OntoNotes4.0, MSRA, Weibo, Resume datasets show that F1 values increased by 3.76%, 3.93%, 4.16%, 6.49%, respectively, presence text, which verifies effectiveness