This is a Python implementation for Nonparametric Bayesian Double Articulation Analyzer (NPB-DAA). The NPB-DAA can directly acquire language and acoustic models from observed continuous speech signals.

This generative model is called hiererichel Dirichlet process hidden language model (HDP-HLM), which is obtained by extending the hierarchical Dirichlet process hidden semi-Markov model (HDP-HSMM) proposed by Johnson et al. An inference procedure for the HDP-HLM is derived using the blocked Gibbs sampler originally proposed for the HDP-HSMM.

• Ubuntu 16.04 LTS
• Python 3.6.5
• Numpy 1.14.2
• Scipy 1.0.1
• Scikit-learn 0.19.1
• Matplotlib 2.2.2
• Joblib 0.11
• Cython 0.28.2
• tqdm 4.23.4
• pybasicbayes 0.2.2
• pyhsmm 0.1.6

1. Install GNU compiler collection to use Cython.

$ sudo apt install gcc

1. Install the necessary libraries for installation.

$ pip install numpy future six
$ pip install cython

1. Install pybasicbayes.

$ git clone https://github.com/mattjj/pybasicbayes
$ cd pybasicbayes
$ python setup.py install

1. Install pyhsmm.

$ git clone https://github.com/RyoOzaki/pyhsmm
$ cd pyhsmm
$ python setup.py install

The repository of pyhsmm was forked and updated by Ryo Ozaki. If you want to install pyhsmm of master repository, please go to https://github.com/mattjj/pyhsmm But, the master repository's codes include some bugs in cython codes.

1. Install pyhlm (this).

$ git clone https://github.com/RyoOzaki/npbdaa npbdaa
$ cd npbdaa
$ python setup.py install

There is a sample source of NPB-DAA in "sample" directory. Please run the "unroll_default_config" before run "pyhlm_sample", and you can change the hyperparameters using the config file "hypparams/defaults.config".

$ cd sample
$ python unroll_default_config.py
$ python pyhlm_sample.py
$ python summary_and_plot.py

• Taniguchi, Tadahiro, Shogo Nagasaka, and Ryo Nakashima. Nonparametric Bayesian double articulation analyzer for direct language acquisition from continuous speech signals, 2015.

• Matthew J. Johnson and Alan S. Willsky. Bayesian Nonparametric Hidden Semi-Markov Models. Journal of Machine Learning Research (JMLR), 14:673–701, 2013.

Tadahiro Taniguch, Ryo Nakashima, Nagasaka Shogo, Tada Yuki, Kaede Hayashi, and Ryo Ozaki.

• MIT
  • see LICENSE