bugbakery / bugbakery/audapolis
Truecasing
Open
backend
enhancement
- Dominant language
- TypeScript
- Stars
- 1.9k
- Forks
- 62
- PR merge metrics
- No merged PRs in 30d
Description
At the moment our STT-library (`vosk`) provides output that is (a) without punctuation and (b) all-lowercase. We addressed (a) in #330, but it would be nice to also reconstruct / guess the casing. This is the NLP task of truecasing.
We need to choose a library & model for doing that.Possible libraries:
* https://github.com/daltonfury42/truecase
* https://github.com/alvations/sacremoses
Contributor guide
Assessment
This issue has not been assessed yet.