MYBully-EmoBERT (Manual + HITL)

Model Overview

This model is MYBully-EmoBERT, trained on manual + HITL annotations from MYBully for Emotion Detection.
It benefits from iterative HITL refinement to improve robustness.

Intended Use

  • Detecting emotions in Bahasa Malaysia tweets (e.g., anger, joy, sadness, fear).

Training Data

  • Dataset: MYBully (Bahasa Malaysia tweets).
  • Annotation: Manual + HITL
  • {'Anger': 0, 'Disgust': 1, 'Fear': 2, 'Happiness': 3, 'Neutral': 4, 'Sadness': 5, 'Surprise': 6}

Model Details

  • Base model: roberta-base-bahasa-cased
  • Fine-tuning: Multi-class classification head
  • Labels: Multiple emotions (e.g., Anger, Joy, Sadness, Fear, Neutral)

Performance

Metric Value
Accuracy 0.72
Precision 0.71
Recall 0.72
F1 0.71
Downloads last month
13
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for mohanrj/mybully-emobert-manual-hitl

Finetuned
(9)
this model

Dataset used to train mohanrj/mybully-emobert-manual-hitl

Collection including mohanrj/mybully-emobert-manual-hitl