This was previously done manually in a few places, the new extractor
engine will apply this to any text-containing node.