metadata
language: en
tags:
- text-classification
- sentiment-analysis
- roberta
- pytorch
datasets:
- glue
- imdb
metrics:
- accuracy
- f1
model-index:
- name: sentiment-classifier
results:
- task:
type: text-classification
name: Sentiment Analysis
dataset:
name: SST-2 + IMDB
type: mixed
metrics:
- type: accuracy
value: 0.9292
- type: f1
value: 0.9413
RoBERTa Fine-Tuned for Sentiment Analysis
This model classifies English text as either Positive 😊 or Negative 😞.
Fine-tuned from roberta-base on a combination of SST-2 (Stanford Sentiment Treebank) and IMDB movie reviews.
Performance
| Metric | Score |
|---|---|
| Accuracy | 0.9292 (92.92%) |
| F1 Score | 0.9413 |
Evaluated on 20,000 held-out IMDB test samples.
How to Use
from transformers import pipeline
classifier = pipeline(
"text-classification",
model="samandar1105/sentiment-classifier"
)
result = classifier("This movie was absolutely fantastic!")
print(result)
# [{'label': 'positive', 'score': 0.998}]
Labels
| ID | Label | Meaning |
|---|---|---|
| 0 | negative | Negative sentiment |
| 1 | positive | Positive sentiment |
Training Details
| Parameter | Value |
|---|---|
| Base model | roberta-base |
| Training data | SST-2 (67K) + IMDB (25K) = 92K samples |
| Epochs | 4 |
| Batch size | 32 |
| Learning rate | 2e-5 |
| Max sequence length | 256 |
| Warmup ratio | 0.1 |
| Weight decay | 0.01 |
Limitations
- Trained on English text only
- Optimized for movie/review-style text
- Binary only (positive / negative) — no neutral class