I have checked the Arabic dataset upload and found that the deepspeech learn on the basis of sentences. I want to know instead of a single sentence in a single audio file can we feed multiple sentences in single audio file?
Yes and no. Ideally you have shorter sequences of 5-10 seconds to train. You can try longer audio, but it can be problematic.
thanks. is there a way to automatically split audio as we split para into sentences.
you can look at aeneas to help you split each audio into a seperate sentence. The tool alligns a given transcription to the timeframe it was said in a given audio. I’ve used it for english and i was happy with the results that i got. It seems there’s support for arabic (https://www.readbeyond.it/aeneas/docs/language.html) so you might want to try it out. goodluck!