dialectautoencoder

Performs dialect analysis using autoencoders. Generates output files that can be ported to GabMap or visualized using SplitsTree.

Please consider citing the following paper if you use the code:

Taraka Rama and Çağrı Çöltekin. LSTM Autoencoders for Dialect Analysis. Proceedings of Third Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial3), COLING 2016, Osaka, Japan, 2016. http://www.aclweb.org/anthology/W/W16/W16-48.pdf#page=37

Requirements:

• Need Keras (https://keras.io/) with Tensorflow (https://www.tensorflow.org/) as a backend for running the code.

Running the program:

• The main program is in src folder as ae.py
• Run the program as python3 src/ae.py data/Germany40.utf8
• The program generates a nexus file (.nex extension) and a vectors file in the data folder

Contact:

In case of any questions, contact [email protected]