Hi, it seems that for the Spanish trf pretrained model, there is no 'tagger' element, so it cannot be trained. How would one, then, improve the POS tagging from that model?
Additionally, does the NER component rely on POS tagging to make decisions? Would improving the POS tagger or finetuning it have an effect on the recognition accuracy of the NER (if I chose to finetune that)?
this sounds more like a spaCy question than a Prodigy question. Could you instead ask the question on our spaCy discussions forum? You'll find the actual maintainers there who will be much better equipped to answer these kinds of questions.