search

OpenAdaptAI / OpenAdapt

AI-First Process Automation with Large ([Language (LLMs) / Action (LAMs) / Multimodal (LMMs)] / Visual Language (VLMs)) Models
https://www.OpenAdapt.AI
MIT License
850 stars 110 forks source link

Pseudonymization (aka. Maintain Identity semantics during Scrub) #264

Open atomicrichard opened 1 year ago

atomicrichard commented 1 year ago

Feature request

When calling scrub.scrub_text, instead of just replacing with ***, we would like to retain semantic meaning. e.g. given:

Alice said hi to Bob

We want:

<PersonA> said hi to <PersonB>

and not just:

*** said hi to ***

@KrishPatel13 🙏

Motivation

I'm always frustrated when [...] so this feature would [...].

KrishPatel13 commented 1 year ago

Nice Article which describes the Pseudonymization (related Library: OpenRedact)

Link: https://medium.com/@openredact/anonymizer-a-framework-for-text-anonymization-499855f639d4 image

KrishPatel13 commented 1 year ago

Must Read: https://github.com/MLDSAI/OpenAdapt/issues/330#issuecomment-1612437801

KrishPatel13 commented 1 year ago

Tested OpenRedact/anonymizer (but the there is no API refrence, no docs) and also had issues on testing it manuallly.

Then proceeded to find another psuedonymization library, and found about Cape-Privacy

Good Choice Cape-Privacy for OpenAdapt: https://docs.capeprivacy.com/getting-started/

KrishPatel13 commented 1 month ago

Shall we close this @abrichr as completed ?

abrichr commented 1 month ago

@KrishPatel13 can you please link to where this has been implemented?