Research on Named Entity Recognition in Chinese Resumes Based on ResNEZHA-BiLSTM-Attention-CRF Model

To extract meaningful entity information from a substantial volume of Chinese resume data, we introduced a ResNEZHA-BiLSTM-Attention-CRF model based on multi-task pre-training. The model enhanced the accuracy and generalization ability of the ResNEZHA model by incorporating a multi-task pre-training mechanism. The model facilitated Chinese word segmentation and entity sequence extraction. Subsequently, the attention mechanism in the BiLSTM layer refines the extraction of local features. The CRF was then employed to assign labels. The model exhibited outstanding performance in the experiment, achieving accuracy, recall, and F1 values of 97.47, 97.79, and 97.73%, respectively. Research findings suggested that, compared to alternative models, the proposed model demonstrated superior performance in the task of named entity recognition in Chinese resumes.