finetuned_model_emotion_detection

This model is a fine-tuned version of jhu-clsp/mmBERT-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3387
• F1 Macro: 0.5265

Model description

This model is a fine-tuned version of mmBERT for multi-label emotion classification in Spanish tweets.

The model was trained on the Spanish subset of the SemEval 2018 Task 1 Affect in Tweets dataset (Emotion Classification track). Given a tweet, the model predicts the presence or absence of eleven emotion categories:

anger anticipation disgust fear joy love optimism pessimism sadness surprise trust

The model is based on the multilingual transformer mmBERT and was fine-tuned using transfer learning with the Hugging Face Transformers library.

Intended uses & limitations

This model is intended for:

• Emotion detection in Spanish social media text.
• Sentiment and affect analysis experiments.

Limitations

• The model was trained exclusively on Twitter data and may not generalize well to other domains.
• Emotion labels are not exclusive; a tweet may express multiple emotions simultaneously.
• Performance may decrease on texts containing slang, spelling mistakes, code-switching, or cultural references not represented in the training data.

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Macro
No log	1.0	223	0.2811	0.4049
No log	2.0	446	0.2685	0.4912
0.2496	3.0	669	0.3387	0.5265

Framework versions

• Transformers 5.12.1
• Pytorch 2.11.0+cu128
• Datasets 5.0.0
• Tokenizers 0.22.2