Analogy_for_inflection

This is the code and data we used for experiments in the following paper:

Ling Liu and Mans Hulden. 2020. Analogy models for neural word inflection. COLING 2020.

Installation

This repository was tested on Python 3.6.10.

All dependencies can be installed as follows:

pip install -r requirements.txt

Reconstruct partial paradigms from shared task data

python src/get_paradigms.py

Train the models and make predictions

bash src/runAll.sh <data_reformatting_method>

<data_reformatting_method> can be # change here for other data reformatting methods: 1src, 1src1crosstable, 1src2crosstable, 2src, or leave1outMSDchunk.

Details on training parameters for different data_organization_method can be found in Appendix A of the paper.

Data hallucination

We used data hallucination implementation provided by the SIGMORPHON 2020 shared task 0, available here: example/sigmorphon2020-shared-tasks/augment.sh. We conducted the data hallucination on the training data created with the 1src data reformatting method.

Evaluation

python src/evaluate.py <language> <data_organization_method>