Profanity filter for closed captions?

We are implementing Vosk with Jigasi to enable Closed Captions in Jitsi.
We are trying to figure out how to enable profanity (“bad language”) filter so that when closed captions shows up, it shows ******* for words from a ban list we have.
I have been looking at the various ASR levels, and that has been a deep dive, but I wonder if there isn’t an easier way to do this on the Jitsi side?
Is there any easy way to implement this just before it is shown in the Jitsi chatroom, rather than at the ASR level?
Suggestions for best approach?
Thanks!

1 Like

Speech-to-text transcription is done outside Jitsi, Jigasi mainly acts as a ‘router’ of sorts for sending the audio and then displaying the text from the json file that’s returned from the transcription service. So, if any filtering is done, my suspicion is that it would be done before the json file is sent back to Jigasi. You should check directly with the Vosk developer, he’s very responsive.

2 Likes

Its probably easier to change on Vosk server. No patch yet though.

1 Like

Please subscribe to Implement profanity filter · Issue #623 · alphacep/vosk-api · GitHub to get updates on the issue

1 Like

@Nickolay_Shmyrev subscription enabled!
Looking forward to figuring this out.
I was just told Friday that we’re also hoping to bring in a neuroscience intern for me to help mentor, to help out as well.
Will post back here when have something testable. Thanks!

1 Like

We just started onboarding today an intern that we’ll have assigned to work on this task. Will inform as we make progress.