Changelog#

0.7.0#

  • Added support for aligning from manual alignments to use in training/adapting

  • Implemented Levenshtein alignment for interval overlap scoring

  • Added a workaround for a bug in mismatched MFCC and pitch features differing in frames under certain conditions in Kaldi

0.6.6#

  • Updated code in parsing phones to word pronunciations to better support transcription capabilities

  • Change build system to ignore CUDA dependencies in an effort to simplify conda-forge build system

0.6.5#

  • Changed how the silence_probability parameter of LexiconCompiler works with pronunciations that have silence probabilities, so that setting it to 0.0 will ensure that no optional silences are included

  • Changed TrainingGraphCompiler signature to require a LexiconCompiler rather than an FST/path and a word table

  • Added the functionality for adding interjection words in between each word in an alignment

0.6.0#

  • Fixed a bug in feature archives where fMLLR transforms were being ignored

0.5.1#

  • Added better error handling for creating archives for files that don’t exist

  • Added more input types for computing MFCCs

0.5.0#

  • First official release

  • Expanded functionality across all modules

0.1.0#

Initial release

On this page

This Page