-
I'm trying to transcribe greek audio and have some strange results. Thanks |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments 3 replies
-
I have a solution for you. It may be freely tested online, but it's a commercial product. |
Beta Was this translation helpful? Give feedback.
-
This is unfortunately a limitation of the trained model, and our filters may not have been very effective at excluding ASR-generated transcripts for non-English languages. I suspect incorrectly spelled transcripts in the training data have probably caused this phenomenon. I've seen similar spelling errors in Korean as well, which wouldn't make sense if the training data only contained grammatically valid text in correct spelling. Like you mentioned, integrating with an LM may improve the results; this is not directly supported, but one can extend the Lines 195 to 246 in 9e653bd You may also consider fine-tuning if you have a speech corpus in Greek, like in this example. |
Beta Was this translation helpful? Give feedback.
-
Did you find any answer to this question? I am intrerested in Greek too! |
Beta Was this translation helpful? Give feedback.
This is unfortunately a limitation of the trained model, and our filters may not have been very effective at excluding ASR-generated transcripts for non-English languages. I suspect incorrectly spelled transcripts in the training data have probably caused this phenomenon. I've seen similar spelling errors in Korean as well, which wouldn't make sense if the training data only contained grammatically valid text in correct spelling.
Like you mentioned, integrating with an LM may improve the results; this is not directly supported, but one can extend the
TokenDecoder
class to select tokens according to a language model:whisper/whisper/decoding.py
Lines 195 to 246 in 9e653bd