I have labeled the document using bert.ner.manual using this command:
After labeling using bert.ner.manual can we use the normal spacy model en_vectors_web_lg to train that document or else is there any bert model in prodigy to train it ??
Hi! This use case and recipe was mostly designed for workflows that use the transformer model and tokenization directly and its main purpose is to make sure that annotations you collect manually align to the model's tokenization (which doesn't necessarily follow the linguistic concept of what's a "word").
So you wouldn't go via spaCy and just train it however you'd typically train your transformer-based model (e.g. using the transformers library). If you want to train with spaCy, there's not really an advantage in using this recipe for manual annotation.