Claude: how can you help me analyse open response questions on surveys? (2023, October 17).
I haven’t played around with this a lot, but I have run some of my anonymised PhD transcripts through Claude a few months ago, and was reasonably impressed with the thematic analysis it did. I think it’s worth doing your own analysis, and then passing it through a language model, almost as a sense-check (doing it in the model first will probably lead you to just go with what it says).
In this short post, Mark Carrigan shares the output from Claude in response to a prompt about analysing open-ended questions.
Huge issue about (1) reproducibility (2) data governance which might rule this out, but otherwise there are interesting possibilities here:
- Sentiment analysis.
- Topic modeling.
- Text classification.
- Keyword extraction.
- Named entity recognition.
- Summarisation.
- Comparison analysis.
- Data validation.