A Model for Named Entity Recognition in Colorectal Tumor Pathological Text
Yinfeng Tian · 2024
Pathological diagnostic reports are unstructured data in natural language. Named entity recognition technology can effectively extract key data from pathological diagnosis reports, assist doctors in understanding pathological results, and thus enhance the accuracy of diagnosis. To improve the accuracy of text recognition in Chinese colorectal cancer pathological text, a named entity recognition method is proposed, which mainly includes a long sequence text feature extraction module and a gated mechanism-based entity recognition module. To reduce the segmentation error of long sequences of Chinese text, BigBird and SoftLexicon are utilized to extract pathological text features at the character and word levels, respectively. To enhance the model's ability to capture key features of long sequential texts, a gated mechanism-based entity recognition module is proposed. Comparative experimental results show that the named entity recognition method proposed in this paper exhibits better performance in colorectal tumor pathological text information recognition, with a precision rate of 93.50%, a recall rate of 94.06%, and an F1-score of 93.76%.