Named Entity Recognition in the Education Domain Based on BERT-BiLSTM-CRF - Using Data Structures as an Example
Xinwen Cheng · 2023
In the field of education, due to the use of traditional methods for named entity identification, there are some cases where the entity feature information extraction accuracy is relatively low, and the identification efficiency is relatively low. This paper presents a named entity identification research approach based on the BERT-BiLSTM-CRF model. This method first uses the pre-training model of BERT (Bidirectional encoder representations from transformers) to obtain the word vector of the input sequence semantic; Then the trained word vector is input into the bidirectional long and short-term memory network Model to obtain contextual features; Finally, according to the conditional random fields of the annotation rules and sequence decoding ability to output the sequence annotation results of the maximum probability, In this way, the named entity recognition model of education field is constructed. This paper also uses adversarial training to improve the model accuracy, recall rate and F1 value.