BART-Ca fine-tuned on the CaSum dataset for summarization
Table of Contents
<details> <summary>Click to expand</summary>
- Model description
- Intended uses and limitations
- How to use
- Limitations and bias
- Training
- Evaluation
- Additional information
Model description
The BART-ca model has been fine-tuned on summarization with the CaSum dataset that has been created along with the model. We also evaluate on an out-of-distribution dataset, VilaSum.
The model has been fine-tuned on news articles and is expected to work best with that type of text.
Intended uses and limitations
You can use this model for text summarization.
How to use
Here is how to use this model with the pipeline API:
from transformers import pipeline
summarizer = pipeline("summarization", model="projecte-aina/bart-base-ca-casum")
ARTICLE = """"El projecte AINA generarà els recursos digitals i lingüístics necessaris per facilitar el desenvolupament d’aplicacions basades en la intel·ligència artificial i les tecnologies de la llengua, com ara els assistents de veu, els traductors automàtics o els agents conversacionals en català. L’objectiu últim és que la ciutadania pugui participar en català en el món digital al mateix nivell que els parlants d’una llengua global, com ara l’anglès, i evitar així l’extinció digital de la llengua. El primer recurs generat és el corpus del català per entrenar els algoritmes d’intel·ligència artificial (IA), el més gran creat fins al moment, amb 1.770 milions de metadades associades a paraules. El proper pas serà generar els models de la llengua, models de la parla i models de traducció utilitzant xarxes neuronals multicapa, perquè les empreses que creen aplicacions basades en intel·ligència artificial (IA), com ara assistents de veu, traductors automàtics, agents conversacionals, etc., puguin fer-ho fàcilment en català."""
print(summarizer(ARTICLE, max_length=130, min_length=30, do_sample=False))
>>> [{'summary_text': 'El projecte AINA generarà els recursos digitals i lingüístics necessaris per al desenvolupament d’aplicacions basades en la intel·ligència artificial en català’'}]
Limitations and bias
At the time of submission, no measures have been taken to estimate the bias embedded in the model. However, we are well aware that our models may be biased since the corpora have been collected using crawling techniques on multiple web sources. We intend to conduct research in these areas in the future, and if completed, this model card will be updated.
Training
Training data
As training data, we used the CaSum dataset extracted from a newswire corpus crawled from the Catalan News Agency.
Training procedure
Tokenization
The training corpus has been tokenized using a byte version of Byte-Pair Encoding (BPE) with a vocabulary size of 51,200 tokens.
Hyperparameters
The fine-tuning hyperparameters were taken from the fairseq's Fine-tuning BART on CNN-Dailymail summarization task example.
Evaluation
Variable and metrics
We use Rouge-1 and Rouge-L for evaluation on two different test sets: the CaSum test set and an out of distribution test set, VilaSum.
Evaluation results
Below the evaluation results on the summarization task compared with the multilingual mBERT and the Catalan NASCA with two different testsets: CaSum and VilaSum.
Test set | Model | Rouge-1 | Rouge-L |
---|---|---|---|
CaSum | BART-Ca | 41.39 | 36.14 |
NASCA | 24.42 | 19.89 | |
mBART | 43.95 | 38.11 | |
VilaSum | BART-Ca | 35.04 | 29.70 |
NASCA | 23.18 | 19.09 | |
mBART | 33.17 | 27.52 |
Additional information
Author
Text Mining Unit (TeMU) at the Barcelona Supercomputing Center (bsc-temu@bsc.es)
Contact information
For further information, send an email to aina@bsc.es
Copyright
Copyright (c) 2022 Text Mining Unit at Barcelona Supercomputing Center
Licensing information
Funding
This work was funded by MT4All CEF project and the [Departament de la Vicepresidència i de Polítiques Digitals i Territori de la Generalitat de Catalunya](https://politiquesdigitals.gencat.cat/ca/inici/index.html#googtrans(ca|en) within the framework of Projecte AINA.
Citation information
If you use any of these resources (datasets or models) in your work, please cite our latest preprint:
@misc{degibert2022sequencetosequence,
title={Sequence-to-Sequence Resources for Catalan},
author={Ona de Gibert and Ksenia Kharitonova and Blanca Calvo Figueras and Jordi Armengol-Estapé and Maite Melero},
year={2022},
eprint={2202.06871},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
Disclaimer
<details> <summary>Click to expand</summary>
The models published in this repository are intended for a generalist purpose and are available to third parties. These models may have bias and/or any other undesirable distortions.
When third parties, deploy or provide systems and/or services to other parties using any of these models (or using systems based on these models) or become users of the models, they should note that it is their responsibility to mitigate the risks arising from their use and, in any event, to comply with applicable regulations, including regulations regarding the use of Artificial Intelligence.
In no event shall the owner and creator of the models (BSC – Barcelona Supercomputing Center) be liable for any results arising from the use made by third parties of these models.