Specialized pre-training of smaller language models can improve their performance on healthcare-related text processing tasks compared to general-purpose language models.
