I have a problem converting the binary annotation to the gold one. Using:
python3 -m prodigy ner.silver-to-gold novi202010 novi202009 batch-trained-model
there is not enough context to do proper annotation. Is there are a switch to display unsegmented text, like -U in ner.tech?
ner.silver-to-gold recipe doesn't do its own sentence segmentation – it will just stream in whatever is in the data. So it seems like the dataset already comes in segmented? If that's the case, the recipe couldn't just put the sentences back together because it's unclear what belongs together.
However, if you do know which examples you want to combine, you could write a script that does this, based on the annotations. Just make sure you translate the tokens and character offsets correctly.