Model Description
- Model type: Multi-class classifier on top of a transformer
- Language(s) (NLP): Spanish varieties: Argentinian (ar), Chilean (cl), Mexican (mx), Spanish (es), and the rest (mix)
- License: GPL-3.0
- Finetuned from: XLM-RoBERTa large
- Preprocessing and tokenisation: the same as XLM-RoBERTa
We provide models for a 3-class (es, mx, mix), a 4-class (cl, es, mx, mix) and a 5-class problem (ar, cl, es, mx, mix). For each case, models with 3 different seeds and the versions with one and two splits of the training documents are included. See the documentation of docTransformer for more detailed information.
Model Sources
- Repository: https://github.com/CEREAL-es/CEREAL
- Paper: Elote, Choclo and Mazorca: on the Varieties of Spanish (NAACL 2024)
- Data: Find the corpora at Zenodo [https://zenodo.org/records/11390829]
Use
Use the CEREAL classification models with docTransformer.
Example Usage
Use these models for evaluation, classification or explanation using integrated gradients:
Slurm
Evaluation (gold label available)
srun --ntasks 1 --gpus-per-task 1 python -u docClassifier.py --task evaluation -f trainedModel -o C4_cereal2splits_seed1.bin -b2 --sentence_batch_size 2 --split_documents True --test_dataset data/multivariant3all.test --plotConfusionFileName modelSplit2Seed3test.png
Classification (gold label unavailable)
srun --ntasks 1 --gpus-per-task 1 python -u docClassifier.py --task classification -f trainedModel -o C4_cereal2splits_seed1.bin -b1 --sentence_batch_size 2 --split_documents True --test_dataset ../es/es_meta_part_1.jsonl.unk
Explanation
srun --ntasks 1 --gpus-per-task 1 python -u docClassifier.py --task explanation -t data/testExample.mx -f trainedModel -o C4_cereal1split_seed1.bin -b1 --split_documents False --xai_threshold_percentile 90
Citation
BibTeX:
@inproceedings{espana-bonet-barron-cedeno-2024-elote,
title = "Elote, Choclo and Mazorca: on the Varieties of {S}panish",
author = "Espa{\~n}a-Bonet, Cristina and
Barr{\'o}n-Cede{\~n}o, Alberto",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.naacl-long.204",
pages = "3689--3711"
}
APA:
España-Bonet, Cristina and Barrón-Cedeño, Alberto. (2024, June). Elote, Choclo and Mazorca: on the Varieties of Spanish. In Proceedings of the 2024 Annual Conference of the North American Chapter of the Association for Computational Linguistics: NAACL 2024 (pp. 3689-3711).