Profanity and Offensive Language Filter On/Off | Voters

Profanity and Offensive Language Filter On/Off

Kyle Lancaster

Give users the option to turn on a "profanity filter" which would not transcribe profanity, or display profanity as "----", or at the very least not auto-predict profanity including the "n-word" or variations of the "n-word" when it was not said. Thanks!

November 3, 2021

Canny AI

Merged in a post:

remove course language

Jay

Nothing is more unprofessional that using profanity in a presentation. In addition to removing filler words, it would be nice to remove expletives.

June 6, 2024

Gabe Michalski

Merged in a post:

Profanity and Offensive Language Filtering

Matt Browning

We used Descript for automated transcription and video subtitles of a recorded live streaming event to add subtitles to the on-demand version of the event. Today, the event producer passed along a message from the end customer regarding offensive language in their transcript. Please see the attached screen shots, where Descript, as the customer put it, shockingly transcribed "never jumped out at me" as "n-word" jumped out at me. To see such racial bias surface even in the phrases programmed into AI for speech recognition, to weight n-word over never in the phrase that continues as "...jumped out at me" should have all of us deeply concerned. So I am reaching out to request swift improvement in the speech recognition algorithms, or offensive language filtering. I would like it if I could provide a list of words to NEVER be used in my transcripts, ever!

April 12, 2024

DubG

This is crazy, but understandable as the AI is a LLM trained on data entered from the idiot humans that exist here on earth that have used that word in so many different confusing, yet all offensive contexts. Here's my question: is it possible to fine tune a LLM with another layer of data, to "find" every offensive word such as the 4-lettered like fuXX, shXX, damX, cuXX, etc, or other offensive words and tune it to automatically "replace" them with other words like "Fuzzy navel" "shizzat muckimmings" "cannopolopogus" ... ya know... like a five year old, so something we can all understand, as a world, united, yet full of idiots? Wouldn't that be fairly easy considering any text editor can do that with a few intuitive commands?

DubG

I just noticed these complaints were from 2020. Whoa.... PR low on the radar here?

Kyle Lancaster

We had a similar issue with the "N-word" being wrongly detected in our content. Could there be a "profanity filter" users can turn on to remove certain words form being predicted, or if they are said just put in "----" in the transcription? Cheers!

Jon Libbey

This same thing happened to me. There needs to be a way to stop it from happening