Open tuxfoo opened 1 year ago
I have a dataset of my own voice, approx 1 hour worth that I previously contributed to the common voice dataset for deepspeech.
I am considering extending this dataset to include Te Reo loan words, such as place names and common Te Reo words that English speakers use in NZ.
Would phonomes be better for these loan words, so that any voice can pronounce loan words and place names correctly?
My dataset is mostly consistent but in a few I might have using a silly voice or was using a different or buzzing microphone.
I assume that we are using transfer learning. Know of any good documentation or guides(Preferably Written) for this? https://stt.readthedocs.io/en/latest/TRANSFER_LEARNING.html
Objective
I have a dataset of my own voice, approx 1 hour worth that I previously contributed to the common voice dataset for deepspeech.
Initial Implementation Requirements
I am considering extending this dataset to include Te Reo loan words, such as place names and common Te Reo words that English speakers use in NZ.
Would phonomes be better for these loan words, so that any voice can pronounce loan words and place names correctly?
Other Considerations
My dataset is mostly consistent but in a few I might have using a silly voice or was using a different or buzzing microphone.
I assume that we are using transfer learning. Know of any good documentation or guides(Preferably Written) for this? https://stt.readthedocs.io/en/latest/TRANSFER_LEARNING.html