Kaldi

Kaldi is a toolkit for speech recognition, intended for use by speech recognition researchers and professionals. Find the code repository at http://github.com/kaldi-asr/kaldi.

• About the Kaldi project
• Other Kaldi-related resources (and how to get help)
• Downloading and installing Kaldi
• Versions of Kaldi
• Software required to install and run Kaldi
• Legal stuff
• Kaldi tutorial
• Kaldi for Dummies tutorial
• Examples included with Kaldi
• Frequently Asked Questions
• Glossary of terms
• Data preparation
• The build process (how Kaldi is compiled)
• The Kaldi coding style
• History of the Kaldi project
• The Kaldi Matrix library
• External matrix libraries
• The CUDA Matrix library
• Kaldi I/O mechanisms
• Kaldi I/O from a command-line perspective.
• Kaldi logging and error-reporting
• Parsing command-line options
• Other Kaldi utilities
• Clustering mechanisms in Kaldi
• HMM topology and transition modeling
• Decision tree internals
• How decision trees are used in Kaldi
• Decoding graph construction in Kaldi
• Decoding-graph creation recipe (test time)
• Decoding-graph creation recipe (training time)
• Support for grammars and graphs with on-the-fly parts.
• Finite State Transducer algorithms
• Decoders used in the Kaldi toolkit
• Lattices in Kaldi
• Acoustic modeling code
• Feature extraction
• Feature and model-space transforms in Kaldi
• Deep Neural Networks in Kaldi
  • Karel's DNN implementation
  • Dan's DNN implementation
  • The "nnet3" setup
  • 'Chain' models
• Online decoding in Kaldi
• Keyword Search in Kaldi
• Parallelization in Kaldi
• Kaldi Tools