This model is based on the multilingual version of BERT, specifically google-bert/bert-base-multilingual-cased, and has been fine-tuned to classify negative emotions in Spanish tweets.
The fine-tuning was performed using a labeled corpus available from the IPN CIC repository https://www.cic.ipn.mx/~sidorov/corpus_marcado6emociones.txt, which contains Spanish tweets annotated with six emotions.
For this model, only the four negative emotions defined by Ekman were used:
0 → Anger
1 → Disgust
2 → Fear
3 → Sadness
The model is designed to support tasks in emotion detection and sentiment analysis in Spanish.