Named entity recognition in the food field based on BERT and Adversarial training

Aiming at extracting effective entity information from unstructured corpus in the food field, a named entity recognition (NER) method based on BERT (Bidirectional Encoder Representations from Transformers) and Adversarial training is proposed. The task of NER requires both the identification of entity boundaries and entity types. In order to improve the precision of identifying entity boundaries, we use the BERT word embedding method to enhance the feature extraction ability of input information. To optimize the NER task, adversarial training is introduced, which not only use the shared information obtained from task training of Chinese word segmentation (CWS) and NER, but also prevent the private information of CWS task from generating noise. The experiment is based on the corpus of two categories which are Chinese food safety cases and People's Daily news, respectively. Among them, the Chinese food safety cases data set is used to train the NER task, and People's Daily news data set is used to train the CWS task. We use adversarial training to improve the precision of the NER task for entity recognition (including person, location, organization, food and toxic substance). The Precision rate, Recall rate and F1 score are 95.46%, 89.50% and 92.38% respectively. Experimental results show that this method has a high precision rate for Chinese NER task where the boundary of a specific domain is indistinct.