Fig. 2
From: VetTag: improving automated veterinary diagnosis coding via large-scale language modeling
Our proposed model architecture for automated diagnosis coding. Two tasks are shown: unsupervised language modeling (top) and supervised learning (bottom). The dashed red arrows represent the pretraining process on the unlabeled PSVG data, and the solid blue arrows represent the fine-tuning process on the labeled CSU data. Additional test is done on the PP data (not shown)
