This model was created to support experiments for evaluating phonetic transcription
with the Buckeye corpus as part of
https://github.com/ginic/multipa.
This is a version of facebook/wav2vec2-large-xlsr-53 fine tuned on a specific subset of the Buckeye corpus.
For details about specific model parameters, please view the config.json here or
training scripts in the scripts/buckeye_experiments folder of the GitHub repository.
These experiments keep the total amount of data equal to half the training data with the gender split 50/50, but further exclude certain speakers completely using the --speaker_restriction argument. This allows us to restrict speakers included in training data in any way. For the purposes of these experiments, we are focussed on the age demogrpahic of the user.
For reference, the speakers and their demographics included in the training data are as follows where the speaker age range 'y' means under 30 and 'o' means over 40:
