sentiment_roberta_taglish

Sentiment RoBERTa Taglish

A fine-tuned RoBERTa model for sentiment analysis on Taglish (Tagalog + English) Shopee comments. Developed as part of a project for DOST-STII-IRAD by an APC College group.

---

Model Details

• Base model: dost-asti/RoBERTa-tl-sentiment-analysis
• Tokenizer: AutoTokenizer.from_pretrained("dost-asti/RoBERTa-tl-sentiment-analysis")
• Task: Sequence classification (Sentiment analysis)
• Labels:
  • 0 = Negative
  • 1 = Neutral
  • 2 = Positive
• Dataset used: letijo03/sentiment-analysis-taglish-shopee-comment (train split)
• Framework: PyTorch

---

Training & Evaluation

Evaluation Metrics

Test Set Performance:

• Accuracy: 0.8630
• Macro F1 Score: 0.6939
• Weighted Average F1 Score: 0.8686

---

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification

# Load tokenizer and model from Hugging Face Hub
tokenizer = AutoTokenizer.from_pretrained("ldlazaro/sentiment_roberta_taglish")
model = AutoModelForSequenceClassification.from_pretrained("ldlazaro/sentiment_roberta_taglish")

# Example inference
text = "Ang ganda ng araw na ito!"
inputs = tokenizer(text, return_tensors="pt")
pred = model(**inputs).logits.argmax(-1)
print(pred.item())