This year's baseline is an ensembled neural transition system based on the imitation learning paradigm introduced by Makarov & Clematide (2020). This is a slightly modified baseline from the 2021 shared task (Ashby et al. 2021).
-
The baseline requires Python 3.7. If your system does not use Python 3.7 by default (i.e., see
python --versionfor your default Python), create and activate either:virtualenv --python=python3.7 sigmorphon source sigmorphon/bin/activate- or a 3.7 Conda environment:
conda create --name=sigmorphon python=3.7 conda activate sigmorphon
-
Install the requirements and the library itself:
pip install --upgrade pip pip install -r requirements.txt pip install . -
Train the ensemble models for each target language and data setting subtask. Low, mixed, and high refer to the 100 The provided bash scripts are for an SGE job scheduling environment, but can be modified to use a for-loop instead. Training may take a while depending on the model settings.
qsub train_ensemble_CLUZH_low.sh qsub train_ensemble_CLUZH_mixed.sh qsub train_ensemble_CLUZH_high.sh
The baseline is made available under the Apache 2.0 license.
Makarov, P., and Clematide, S. 2020. CLUZH at SIGMORPHON 2020 shared task on multilingual grapheme-to-phoneme conversion. In Proceedings of the 17th SIGMORPHON Workshopon Computational Research in Phonetics, Phonology, and Morphology, pages 171-176.