Page 1 of 1

Attach a dictionary to subtitle-generation

PostPosted: Mon Nov 13, 2023 8:25 am
by acebone
Hi

I am trying to use the AI-function to generate subtitles. My language is danish.

The AI will "invent" words, that does not exist.

An example could be "Socialdemokraterne" (the socialdemocrats) - which the AI thinks is "socialtumkraterne". Now a baby may say it like that, but the word does not exist at all.

Another example would be "børnefamilier" (families with children), which AI interprets as "børnepamiljer"

It's almost every second or third sentence, that contains these nonsense words - not a small part of them is downright unrecognizable.

I have no knowledge of AI, but I would suggest that the functionality consult a dictionary somewhere in the process - either when training the AI, or as a last check before outputting.

Re: Attach a dictionary to subtitle-generation

PostPosted: Fri Nov 17, 2023 9:42 pm
by Jim Simon
Have you tried specifically selecting Danish under Project Settings?

Re: Attach a dictionary to subtitle-generation

PostPosted: Sat Nov 18, 2023 6:31 pm
by Charles Bennett
The AI would have to have access to or learn a lot of dictionaries to cover all the languages. ;)

Re: Attach a dictionary to subtitle-generation

PostPosted: Thu Nov 23, 2023 8:48 pm
by acebone
Jim Simon wrote:Have you tried specifically selecting Danish under Project Settings?


Yes (of course)


Charles Bennett wrote:The AI would have to have access to or learn a lot of dictionaries to cover all the languages. ;)


Yup - that's not a problem with current technology. Think of smart phones, that do spell checks and text prediction.