Skip to content

Siamese network for unsupervised speech representation learning

License

Notifications You must be signed in to change notification settings

bootphon/abnet3

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

74 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ABnet3

Representation learning package using side information, system for subword modeling for Zeroresource challenge.

Overview

Installation of the package

To install the ABnet3 package, you can use Anaconda, and either create a conda environment:

conda create --name abnet3 --file requirements.txt

or use a conda environment you already have with python 3 :

conda install --file requirements.txt

Once all the necessary packages are installed, simply launch:

python setup.py build && python setup.py install

If you want to work on ABnet3 and develop your own modules, instead of:

python setup.py install

you can launch:

python setup.py develop

Documentation

Tests

The package comes with a unit-tests suit. To run it, first install pytest on your Python environment:

pip install pytest
pytest test

References

[Thiolliere, R. et al. 2014] A hybrid dynamic time warping-deep neural network architecture for unsupervised acoustic modeling. In Sixteenth Annual Conference of the International Speech Communication Association

[Zeghidour, N. et al. 2016] Zeghidour, N., Synnaeve, G., Usunier, N. & Dupoux, E. (2016). Joint Learning of Speaker and Phonetic Similarities with Siamese Networks. In INTERSPEECH-2016, (pp 1295-1299)

Acknowledgments

A part of the code is inspired from the previous version in Theano of ABnet, and the examples in Pytorch