Research on Named Entity Recognition Technique for Researchers in Power domain Based on RoBERTa-CNN-BiLSTM-CRF

The traditional named entity recognition method has many problems, which makes it unsatisfactory for application in specific knowledge domains. To improve the processing efficiency of named entity recognition for researchers in the power domain, this paper designs a named entity recognition technique for researchers in the power domain that combines RoBERTa-Convolutional Neural Network (CNN)Bidirectional Long and Short-term Memory network (BiLSTM)Conditional Random Field (CRF). The technical framework helps to intelligently process the information of scientific researchers in the power domain while providing a natural language processing foundation for data visualization methods such as knowledge graph establishment and portrait construction. Future work will explore the framework’s applicability in more segmented domains and other specialized term recognition tasks