| --- |
| library_name: transformers |
| license: apache-2.0 |
| base_model: bert-base-uncased |
| tags: |
| - sentiment-analysis |
| - text-classification |
| - bert |
| - transformers |
| - pytorch |
| metrics: |
| - accuracy |
| - f1 |
| - precision |
| - recall |
| model-index: |
| - name: finetuned-bert-sentiment |
| results: |
| - task: |
| type: text-classification |
| name: Sentiment Analysis |
| dataset: |
| name: IMDb Movie Reviews |
| type: imdb |
| metrics: |
| - type: accuracy |
| value: 0.9225 |
| - type: f1 |
| value: 0.9238 |
| - type: precision |
| value: 0.9086 |
| - type: recall |
| value: 0.9395 |
| --- |
| |
| # 🎬 Finetuned BERT for Sentiment Analysis |
|
|
| This model is a fine-tuned version of **BERT (bert-base-uncased)** for binary sentiment classification (positive vs negative). |
|
|
| It is trained on the **IMDb movie reviews dataset**, a widely used benchmark for sentiment analysis tasks. |
|
|
| --- |
|
|
| ## 🚀 Model Performance |
|
|
| | Metric | Score | |
| |------------|--------| |
| | Accuracy | 92.25% | |
| | F1 Score | 92.38% | |
| | Precision | 90.86% | |
| | Recall | 93.95% | |
|
|
| ### Confusion Matrix Insights |
| - Strong balance between positive and negative predictions |
| - Slight tendency toward higher recall (fewer false negatives) |
| - Overall robust generalization on full test dataset (25,000 samples) |
|
|
| --- |
|
|
| ## 📌 Model Description |
|
|
| This project demonstrates fine-tuning of a pre-trained Transformer model for NLP classification tasks using the Hugging Face ecosystem. |
|
|
| Key features: |
| - Pretrained **BERT encoder** |
| - Fine-tuned for **binary sentiment classification** |
| - Implemented using **Hugging Face Transformers Trainer API** |
| - Evaluated using standard classification metrics |
|
|
| --- |
|
|
| ## 📊 Dataset |
|
|
| - **Name:** IMDb Movie Reviews Dataset |
| - **Size:** |
| - Train: 25,000 samples |
| - Test: 25,000 samples |
| - **Classes:** |
| - `0` → Negative |
| - `1` → Positive |
|
|
| The dataset is balanced across both classes. |
|
|
| --- |
|
|
| ## 🏋️ Training Procedure |
|
|
| ### Hyperparameters |
|
|
| - Learning rate: `2e-5` |
| - Batch size: `8` |
| - Epochs: `2` |
| - Optimizer: AdamW |
| - Scheduler: Linear decay |
| - Mixed precision: Enabled (FP16) |
|
|
| ### Training Details |
|
|
| - Framework: Hugging Face `Trainer` |
| - Hardware: Google Colab GPU |
| - Loss function: Cross-entropy |
|
|
| --- |
|
|
| ## 🧠 Intended Use |
|
|
| This model can be used for: |
| - Sentiment analysis on movie reviews |
| - Product review classification |
| - Social media sentiment detection |
| - NLP learning and experimentation |
|
|
| --- |
|
|
| ## ⚠️ Limitations |
|
|
| - Trained only on English text |
| - Domain-specific (movie reviews) → may not generalize perfectly to other domains |
| - Binary classification only (no neutral sentiment) |
| - May inherit biases present in training data |
|
|
| --- |
|
|
| ## 🛠️ How to Use |
|
|
| ```python |
| from transformers import pipeline |
| |
| classifier = pipeline("sentiment-analysis", model="ashwini10521/finetuned_bert") |
| |
| result = classifier("This movie was absolutely amazing!") |
| print(result) |
