With ALA, we offer a service for the automated affiliation of a language, based on standardized data from the Lexibank repository for CLDF wordlists. Language Affiliation is hereby understood as the task by which one tries to identify to which language family a given language belongs.

If you use our model or workflow, please cite our paper as follows:

Blum, Frederic and Herbold, Steffen and List, Johann-Mattis (forthcoming): From Isolates to Families: Using Neural Networks for Automated Language Affiliation. In: Proceedings of the Association of Computational Linguistics 2025. 1-12. https://doi.org/10.48550/arXiv.2502.11688

In order to run all the code, you can install the packages from the requirements.txt file (the command python here refers to the Python version that you use on your computer, on some systems, this may result in the command python3, but we use python as a generic placeholder here). You should make sure to use a fresh virtual environment before installing any packages.

python -m pip install -r requirements.txt
python -m pip install .

In order to download the data necessary for the experiments, you can run the Makefile, by opening a terminal in the examples folder and installing the additional packages:

cd examples
python -m pip install -r requirements.txt

You can now download the data and create the SQLite databases.

make download
make prepare

To run the experiments, you should open a terminal in the examples folder. Here, we have prepared shell scripts that call the Python commands for convenience.

sh runs-bl.sh
sh runs-comparison-nn.sh
sh runs-experiments.sh