enhance pseudomization

[qwertygc]

Hello,
OpenAI has released a library for anonymising a corpus. Perhaps this could be a useful avenue for improving QualCoder’s pseudonymisation feature?

https://openai.com/index/introducing-openai-privacy-filter/

Best
Justin

[kaixxx]

Thank you for the hint. That's interesting, especially because this model can run locally. I will take a closer look if I find the time.

[ccbogel]

Yes ok. A couple of concerns and suggestions I have are:
Adding more AI modules into QC will blow out the already large size of the software. Particularly when it is packaged as an executable.
I think it could be used as an add-in option.

[qwertygc]

I have tested on another software; this does not add any extra weight to the binary compiled with pycompiler. The model is downloaded the first time it is used. It is work, but we can have things that are not anonymized.

My suggestion: Maybe it is not the function of qualcoder. Maybe it's more appropriate for a transcribe tool. That's open to debate.

[kaixxx]

I am collaborating with Prof. Yves Jeanrenaud (University of Applied Sciences Osnabrück) who is developing an AI based anonymizing tool for qualitative research: https://github.com/yjeanrenaud/yj_qualAnonymiser We might also test the new OpenAI Model. Once finished, I plan to integrate the tool in my transcription app noScribe as a plugin. We could do the same with QualCoder.