I wrote a speech transcriber application for automatic annotation of video files for the Annodex project [http://www.annodex.net].
1) What word error rate do you reckon is a "good" result for non-American speakers (Australian or Euro-English) in a conference talk environment
(a) talking about the weather or
(b) using a lot of technical acronyms?
2) How is it possible to add text (which is related to the speech to transcribe in order to train it better) to existing language models?
Are the texts used to train the existing language models available anywhere?
Greetings!
Chris.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I wrote a speech transcriber application for automatic annotation of video files for the Annodex project [http://www.annodex.net].
1) What word error rate do you reckon is a "good" result for non-American speakers (Australian or Euro-English) in a conference talk environment
(a) talking about the weather or
(b) using a lot of technical acronyms?
2) How is it possible to add text (which is related to the speech to transcribe in order to train it better) to existing language models?
Are the texts used to train the existing language models available anywhere?
Greetings!
Chris.