The roberta-base-ca-cased-sts is a Semantic Textual Similarity (STS) model for the Catalan language fine-tuned from the roberta-base-ca model, a RoBERTa base model pre-trained on a medium-size corpus collected from publicly available corpora and crawlers.
Intended uses and limitations
roberta-base-ca-cased-sts model can be used to assess the similarity between two snippets of text. The model is limited by its training dataset and may not generalize well for all use cases.
How to use
To get the correct1 model's prediction scores with values between 0.0 and 5.0, use the following code:
python
1from transformers import pipeline, AutoTokenizer
2from scipy.special import logit
34model ='projecte-aina/roberta-base-ca-cased-sts'5tokenizer = AutoTokenizer.from_pretrained(model)6pipe = pipeline('text-classification', model=model, tokenizer=tokenizer)78defprepare(sentence_pairs):9 sentence_pairs_prep =[]10for s1, s2 in sentence_pairs:11 sentence_pairs_prep.append(f"{tokenizer.cls_token}{s1}{tokenizer.sep_token}{tokenizer.sep_token}{s2}{tokenizer.sep_token}")12return sentence_pairs_prep
1314sentence_pairs =[("El llibre va caure per la finestra.","El llibre va sortir volant."),15("M'agrades.","T'estimo."),16("M'agrada el sol i la calor","A la Garrotxa plou molt.")]1718predictions = pipe(prepare(sentence_pairs), add_special_tokens=False)1920# convert back to scores to the original 0 and 5 interval21for prediction in predictions:22 prediction['score']= logit(prediction['score'])23print(predictions)
1avoid using the widget scores since they are normalized and do not reflect the original annotation values.
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
We used the STS dataset in Catalan called STS-ca for training and evaluation.
Training procedure
The model was trained with a batch size of 16 and a learning rate of 5e-5 for 5 epochs. We then selected the best checkpoint using the downstream task metric in the corresponding development set, and then evaluated it on the test set.
Evaluation
Variable and metrics
This model was finetuned maximizing the average score between the Pearson and Spearman correlations.
Evaluation results
We evaluated the roberta-base-ca-cased-sts on the STS-ca test set against standard multilingual and monolingual baselines:
Model
STS-ca (Pearson score)
roberta-base-ca-cased-sts
79.73
mBERT
74.26
XLM-RoBERTa
61.61
For more details, check the fine-tuning and evaluation scripts in the official GitHub repository.
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
If you use any of these resources (datasets or models) in your work, please cite our latest paper:
bibtex
1@inproceedings{armengol-estape-etal-2021-multilingual,
2 title = "Are Multilingual Models the Best Choice for Moderately Under-resourced Languages? {A} Comprehensive Assessment for {C}atalan",
3 author = "Armengol-Estap{\'e}, Jordi and
4 Carrino, Casimiro Pio and
5 Rodriguez-Penagos, Carlos and
6 de Gibert Bonet, Ona and
7 Armentano-Oller, Carme and
8 Gonzalez-Agirre, Aitor and
9 Melero, Maite and
10 Villegas, Marta",
11 booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
12 month = aug,
13 year = "2021",
14 address = "Online",
15 publisher = "Association for Computational Linguistics",
16 url = "https://aclanthology.org/2021.findings-acl.437",
17 doi = "10.18653/v1/2021.findings-acl.437",
18 pages = "4933--4946",
19}
Disclaimer
Click to expand
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.
