real time ASR for multilingual languages same time?
#1
by
DreamyP
- opened
Can we use this for multilingual ASR? like in a single audio if there r say 2 languages and we can differentiate it n it translates into both languages simultaniously?
The LID model will predict a single language (or single distribution over languages) per audio input - for multi-language prediction, we could batch our audio into smaller chunks and classify the language for each chunk. We'll then know what language is spoken in each chunk, so we can segment the audio where it changes languages, and pass each segment to the correct ASR model trained on that specific language