Skip to content

v1.0.0

Choose a tag to compare

@github-actions github-actions released this 07 Feb 09:14
· 16 commits to main since this release

Added

Text Normalization as Grid Search Dimension

  • Systematic text normalization: Each transcription generates 2 results (raw + normalized)
  • Normalized preset: lowercase + remove punctuation + normalize spaces
  • Diff view shows both raw and normalized texts, if selected

Text normalizations

  • Configurable text transforms: ToLowerCase(), RemovePunctuation(), ExpandCommonEnglishContractions()
  • Metrics and transforms computed consistently with jiwer library

Visual Encoding

  • Symbol = Degradation type (circle = original, diamond = reverb, etc.)
  • Color = Engine (whisper = blue, nemo = purple, etc.)
  • Size = Text normalization (normalized = large, raw = small)

Changed

  • wav2vec2 engine: Output normalized to lowercase (was outputting uppercase)