Research on Named Entity Recognition in Inorganic Chemistry Based on Bert-BiLSTM-CRF
Wei Jiang, Qiang Xu, Hui Ding · 2022
Aiming at the problems of fuzzy entity boundary and poor recognition effect in named entity recognition in the field of inorganic chemistry, this paper adopts the BERT-BiLSTM-CRF model. First, the BERT pre-training model is used for semantic extraction to generate word vectors representing contextual semantic information, and then the word vectors are input into the BiLSTM model for training to obtain contextual features, and finally, the CRF model is used to obtain a reasonable sequence with the maximum probability. The entity recognition effect of the model used has the highest F1 value of 0.9429 on the test corpus. The experimental results show that the BERT-BiLSTM-CRF model can effectively identify named entities in the field of inorganic chemistry, and the recognition effect is better than other traditional models.