Entity extraction (NER) with the same tools

See how small the step is to extracting details from a message (names, places, amounts, dates): NER labels each token with BIO tags (B-PER = beginning of a person name, I-PER = inside, O = outside). Fine-tune DistilBERT for token classification on CoNLL-2003 once, and label 30 sentences of your own in Label Studio (NER template). Scored with entity-level F1.