README.md

metadata

license: apache-2.0
language: ca
tags:
  - catalan
datasets:
  - catalonia_independence
metrics:
  - accuracy
model-index:
  - name: roberta-base-ca-finetuned-mnli
    results:
      - task:
          name: Text Classification
          type: text-classification
        dataset:
          name: catalonia_independence
          type: catalonia_independence
          args: catalan
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.7611940298507462
widget:
  - text: >-
      Puigdemont, a l'estat espanyol: Quatre anys després, ens hem guanyat el
      dret a dir prou
  - text: >-
      Llarena demana la detenció de Comín i Ponsatí aprofitant que són a Itàlia
      amb Puigdemont
  - text: >-
      Assegura l'expert que en un 46% els catalans s'inclouen dins del que es
      denomina com el doble sentiment identitari. És a dir, se senten tant
      catalans com espanyols. 1 de cada cinc, en canvi, té un sentiment
      excloent, només se senten catalans, i un 4% sol espanyol.

roberta-base-ca-finetuned-catalonia-independence-detector

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

• Loss: 0.6065
• Accuracy: 0.7612

Training and evaluation data

This dataset contains two corpora in Spanish and Catalan that consist of annotated Twitter messages for automatic stance detection. 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	377	0.6311	0.7453
0.7393	2.0	754	0.6065	0.7612
0.5019	3.0	1131	0.6340	0.7547
0.3837	4.0	1508	0.6777	0.7597
0.3837	5.0	1885	0.7232	0.7582

Model in action 🚀

Fast usage with pipelines:

from transformers import pipeline

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

independence_analysis(
    "Assegura l'expert que en un 46% els catalans s'inclouen dins del que es denomina com el doble sentiment identitari. És a dir, se senten tant catalans com espanyols. 1 de cada cinc, en canvi, té un sentiment excloent, només se senten catalans, i un 4% sol espanyol."
    )
# Output:
[{'label': 'AGAINST', 'score': 0.7457581758499146}]

independence_analysis(
    "Llarena demana la detenció de Comín i Ponsatí aprofitant que són a Itàlia amb Puigdemont"
    )
# Output:
[{'label': 'NEUTRAL', 'score': 0.7436802983283997}] 
    
independence_analysis(
    "Puigdemont, a l'estat espanyol: Quatre anys després, ens hem guanyat el dret a dir prou"
    )
# Output:
[{'label': 'FAVOR', 'score': 0.9040119647979736}]

Framework versions

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

Citation

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