Would like to understand whether prodigy tool supports the following,
TSV file (as input).
Active Learning on NER side.
Can a CoreNLP NER output (not necessarily to be one derived from prodigy) be fed to the tool and rework on the NER outputs (for example – can a tag ‘ORG’ be changed to ‘NON_ORG’ wherever applicable).
One way you can load TSV is to use the CSV loader and pass a tab as its delimiter.
You can perform active learning using the ner.teach recipe. It updates a model in the loop, and based on your annotations, Prodigy will decide the best next question to annotate.
Yes that should be possible. Once you have the CoreNLP NER output, you need to convert them into a format that Prodigy accepts and run the appropriate NER recipe (ner.manual for "standard NER" and ner.teach for an active-learning powered one). You can customize the tags by passing arguments to the --labels parameter.