---
tags:
- text-classification
- toxicity
- Twitter
base_model: cardiffnlp/twitter-roberta-base-sentiment
widget:
- text: I love AutoTrain
license: mit
language:
- es
pipeline_tag: text-classification
library_name: transformers
---

# Fined-tuned roBERTa for Toxicity Classification in Spanish

This is a fine-tuned roBERTa model trained using as a base model Twitter-roBERTa base-sized for Sentiment Analysis, which was trained on ~58M tweets. The dataset for training this model is a gold standard for protest events for toxicity and incivility in Spanish.

The dataset comprises ~5M data points from three Latin American protest events: (a) protests against the coronavirus and judicial reform measures in Argentina during August 2020; (b) protests against education budget cuts in Brazil in May 2019; and (c) the social outburst in Chile stemming from protests against the underground fare hike in October 2019. We are focusing on interactions in Spanish to elaborate a gold standard for digital interactions in this language, therefore, we prioritise Argentinian and Chilean data.

- [GitHub repository](https://github.com/training-datalab/gold-standard-toxicity).
- [Dataset on Zenodo](https://zenodo.org/doi/10.5281/zenodo.12574288).
- [Reference paper](https://arxiv.org/abs/2409.09741)

**Labels: NONTOXIC and TOXIC.**

**We suggest using [bert-spanish-toxicity](https://huggingface.co/bgonzalezbustamante/bert-spanish-toxicity) or [ft-xlm-roberta-toxicity](https://huggingface.co/bgonzalezbustamante/ft-xlm-roberta-toxicity) instead of this model.**

## Validation Metrics

- Accuracy: 0.790
- Precision: 0.920
- Recall: 0.657
- F1-Score: 0.767