speech tools

These files consitute a set of files that can be used to run the analyses for cross-linguistic analysis of large speech model embeddings.

We currently support functionality for the following models and corpora

wav2vec 2.0 HuBERT (TODO: WavLM)

Hindi Commonvoice (13) AI for Bharat () Vaani ()

English Librispeech Wall Street Journal Corpus

Korean Seoul Corpus

We (plan to) support the following functionality:

• data cleaning

• feature extraction

• model fine-tuning

• evaluation

  • ABX

  • 2AFC