StaResGRU-CNN with CMedLMs: A stacked residual GRU-CNN with pre-trained biomedical language models for predictive intelligence

As a task requiring strong professional experience as supports, predictive biomedical intelligence cannot be separated from the support of large amount external domain knowledge. By using transfer learning to obtain sufficient prior massive text data, it is essential promote performance specific downstream and decision-making models. This an efficient convenient method, but has not been fully developed for Chinese Natural Language Processing (NLP) in field. study proposes Stacked Residual Gated Recurrent Unit-Convolutional Neural Networks (StaResGRU-CNN) combined with pre-trained language models (PLMs) text-based tasks. Exploring related paradigms NLP based on expert knowledge comparing some English We have identified key issues that or those present difficulties application field biomedicine. Therefore, we also propose series bioMedical Models (CMedLMs) detailed evaluations learning, are introduced improve tasks solve serve medical system better. Additionally, free-form Electronic Medical Record (EMR)-based Disease Diagnosis Prediction proposed, which used evaluation analyzed together Clinical Named Entity Recognition, Biomedical Text Classification Our experiments prove introduction significantly improves their different granularity. And our proposed model achieved competitive these • Provided CMedLMs, Med Pre-trained (PLM) Designed GRU-CNN PLMs predictiontype Detailed NER, classification & prediction text. The first comprehensive work extensive evaluations.