Derivational Graph Auto-encoder (DGA)

This repository contains the code and data for the ACL paper A Graph Auto-encoder Model of Derivational Morphology. The paper introduces the Derivational Graph Auto-encoder (DGA), a model that learns embeddings capturing information about the compatibility of affixes and stems in derivation.

Dependencies

The code requires Python>=3.5, torch, torch_geometric, pickle, numpy, pandas, scipy, and sklearn.

Data

You can find the Derivational Graphs (DGs) for the nine subreddits in data/graphs. The derivational embeddings trained on the DGs are located in src/model/embeddings.

Experiments

To replicate the experiments from the paper, run run_models.sh in src/model.

Citation

If you use the code or data in this repository, please cite the following paper:

@inproceedings{hofmann2020dga,
    title = {A Graph Auto-encoder Model of Derivational Morphology},
    author = {Hofmann, Valentin and Sch{\"u}tze, Hinrich and Pierrehumbert, Janet},
    booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
    year = {2020}
}