tahamueed23/sentiment_roberta_english_finetuned

🚀 Model Overview

• Base Model: siebert/sentiment-roberta-large-english
• Task: Sentiment Classification
• Languages: English
• Dataset Source: Google Sheets CSV (custom dataset)
• Columns Used: Sentence, Language, Sentiment
• Filtered: Only English rows were kept

---

📊 Training Results (by Epoch)

Epoch	Training Loss	Validation Loss	Accuracy	Precision	Recall	F1
1	0.627600	0.484212	0.875636	0.874333	0.875636	0.872211
2	0.250100	0.419673	0.905031	0.901430	0.905031	0.902535
3	0.229800	0.467646	0.906727	0.903297	0.906727	0.904490

---

🧪 Final Evaluation Metrics

The model shows strong performance with ~90%+ across all core metrics (accuracy, precision, recall, F1).

⚠️ A warning from scikit-learn indicated some classes had no predicted samples. This usually happens when the dataset is imbalanced. Class-balancing or weighted loss can improve minority class scores.

---

🧠 Intended Use

This model is suitable for:

• Product review sentiment
• Student Feedback
• Customer feedback analysis
• Social media sentiment
• General English text polarity detection

---

🚫 Limitations

• Trained only on English sentences — performance on Urdu / Roman Urdu is not guaranteed.
• Dataset may be imbalanced.
• Not designed for multi-label sentiment or non-general texts (e.g., sarcasm).

---

🧩 How to Use

Pipeline Usage

from transformers import pipeline

model_name = "tahamueed23/sentiment_roberta_english_finetuned"

classifier = pipeline("text-classification", model=model_name, tokenizer=model_name)

classifier("I really love this product!")