This is a crucial point. There’s no magic to a language model like other device Discovering models, especially deep neural networks — it’s only a Resource to incorporate ample information in the concise way that’s reusable within an out-of-sample context. With slightly retraining, BERT could be a POS-tagger thanks https://financefeeds.com/osl-group-launches-wealth-platform-for-copyright-asset-management-in-hk/