/accented-speech-recognition

Assessment of current speech recognition platforms on accented English speech, and re-training of Mozilla Deep Speech with foreign accents.

Primary LanguagePython

accented-speech-recognition

This project has two phases:

  1. Assessing Amazon Transcribe, Google Speech-to-text, and Mozilla Deep Speech on accented English speech. We use the JiWER library to compute WER score, and the Speech Accent Archive dataset.
  2. Re-training of Mozilla Deep Speech with foreign accents using Mozilla's training API.