Code and result tables for the "Subword Pooling Makes a Difference" EACL2021 paper

Prerequisites

The source code for running the experiments is available in the probing package. This repository only contains the configuration files, the results tables and the analysis notebooks.

Running a single experiment

python $PROBING_PATH/train.py -c config/morphology.yaml --train data/morphology/gender_noun/German/train.tsv --dev data/morphology/gender_noun/German/dev.tsv

Running multiple experiments

Run all morphology experiments (6th layer only):

python $PROBING_PATH/train_many_configs.py -c config/morphology.yaml -p config/generate_morphology.py

Run all POS experiments (6th layer only):

python $PROBING_PATH/train_many_configs.py -c config/pos.yaml -p config/generate_pos.py

Cite

@inproceedings{Acs:2020,
    author    = {Judit \'Acs and \'Akos K\'ad\'ar and Andr\'as Kornai},
    title     = {Subword Pooling Makes a Difference},
    booktitle = {Proceedings of the 16th Conference of the European Chapter of the
        Association for Computational Linguistics, {EACL} 2021}
    publisher = {Association for Computational Linguistics},
    year      = {2021},
}