search

alphacep / vosk-api

Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node
Apache License 2.0
7.35k stars 1.04k forks source link

Adding new words to Filipino model #1573

Closed moodpanda closed 1 month ago

moodpanda commented 1 month ago

just wanna ask if is it possible to use my dataset it is formatted like this audio, transcription in csv file.

thanks for answering

juliustuliao commented 1 month ago

up same question

nshmyrev commented 1 month ago

This question is too vague for me to answer. If you need help you need to provide the details. What problem are you tryign to solve, what data do you have and so on.

moodpanda commented 1 month ago

I want the model to recognize new words or sentences using my own dataset. Currently, my dataset is formatted with each entry containing the path to an audio file and its corresponding transcription. I am new to Kaldi and unsure of how to properly format my data. inorder to perform model adaptation

my dataset example: audio_file_path, transcription data/chunk_001.wav, hello, world

nshmyrev commented 1 month ago

Vosk models are adapted with just text, not the audio + text.

What is the language of your dataset. What models did you try? What is the current accuracy of the model.

moodpanda commented 1 month ago

my dataset language is filipino and I'm trying to use model adaptation vosk-model-tl-ph-generic-0.6 to add new words or sentence on the model vocabulary

nshmyrev commented 1 month ago

@moodpanda this model is precompiled and we can not modify it. You have to contact Fed directly for update, https://github.com/feddybear/flipside_ph only he can do it.

moodpanda commented 1 month ago

@moodpanda this model is precompiled and we can not modify it. You have to contact Fed directly for update, https://github.com/feddybear/flipside_ph only he can do it.

got it thankyou very much I will try to email him if he still active thankyou

nshmyrev commented 1 month ago

He certainly can help you. Best.