Named Entity Recognition of Benefit Enterprise Policy Based on RoBERTa_wwm_ext-BiLSTM-CRF

Abstract

To promote the development and innovation of enterprises, government departments at all levels have issued numerous enterprise benefiting policies. Named entity recognition technology can be used to extract key domain information in policy texts to help companies quickly find matching policies. Due to the problems of blurred entity boundary and polysemy of words, named entity recognition in the policy text of benefiting enterprises. In this paper, the deep learning model RoBERTa_wwm_ext-BiLSTM-CRF is applied, and the pre-training model RoBERTa_wwm_ext is applied to the field of enterprise benefiting policy for the first time. The semantic information of policy words is obtained through the RoBERTa_wwm_ext model, BiLSTM is used to extract the semantic features of the context, and finally, use CRF to sequence optimize output. The current mainstream entity recognition models are compared through experiments, and the results show that the model is better than the comparison model in terms of accuracy, recall and F1, and the F1 value reaches 86.4%. Problems such as word polysemy can be well represented.