Edit model card


This model is a fine-tuned version of BSC-TeMU/roberta-base-bne on the catalonia_independence dataset. It achieves the following results on the evaluation set:

  • Loss: 0.9415
  • Accuracy: 0.7881

Model description

The data was collected over 12 days during February and March of 2019 from tweets posted in Barcelona, and during September of 2018 from tweets posted in the town of Terrassa, Catalonia. Each corpus is annotated with three classes: AGAINST, FAVOR and NEUTRAL, which express the stance towards the target - independence of Catalonia.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss Accuracy
No log 1.0 378 0.5534 0.7558
0.6089 2.0 756 0.5315 0.7643
0.2678 3.0 1134 0.7336 0.7816
0.0605 4.0 1512 0.8809 0.7866
0.0605 5.0 1890 0.9415 0.7881

Model in action 🚀

Fast usage with pipelines:

from transformers import pipeline

model_path = "JonatanGk/roberta-base-bne-finetuned-catalonia-independence-detector"
independence_analysis = pipeline("text-classification", model=model_path, tokenizer=model_path)

    "Junqueras, sobre la decisión judicial sobre Puigdemont: La justicia que falta en el Estado llega y llegará de Europa"

# Output:
[{'label': 'FAVOR', 'score': 0.9936726093292236}]

    "El desafío independentista queda adormecido, y eso que el Gobierno ha sido muy claro en que su propuesta para Cataluña es una agenda de reencuentro, centrada en inversiones e infraestructuras")

# Output:
[{'label': 'AGAINST', 'score': 0.7508948445320129}]

    "Desconvocada la manifestación del domingo en Barcelona en apoyo a Puigdemont"

# Output:
[{'label': 'NEUTRAL', 'score': 0.9966907501220703}]

Open In Colab

Framework versions

  • Transformers 4.11.3
  • Pytorch 1.9.0+cu111
  • Datasets 1.12.1
  • Tokenizers 0.10.3


Thx to HF.co & @lewtun for Dataset ;)

Special thx to Manuel Romero/@mrm8488 as my mentor & R.C.

Created by Jonatan Luna | LinkedIn

Downloads last month
Hosted inference API
Text Classification
This model can be loaded on the Inference API on-demand.

Dataset used to train JonatanGk/roberta-base-bne-finetuned-catalonia-independence-detector

Space using JonatanGk/roberta-base-bne-finetuned-catalonia-independence-detector

Evaluation results