One useful approach to autocoding could be to conduct named entity recognition, count occurrences across the corpus, and then prompt the user somehow to decide which entities ought to be coded. Try this out at least; I find myself coding language that's in the corpus which doesn't feel like a valuable use of time.
The corpus anonymize command is pretty close to this functionality.
One useful approach to autocoding could be to conduct named entity recognition, count occurrences across the corpus, and then prompt the user somehow to decide which entities ought to be coded. Try this out at least; I find myself coding language that's in the corpus which doesn't feel like a valuable use of time.
The
corpus anonymizecommand is pretty close to this functionality.