rohanpython9229/OpinionAI-Sentiment-DistilBERT

OpinionAI Sentiment DistilBERT

A fine-tuned DistilBERT model for 3-class sentiment classification of English product reviews.

The model classifies a review into one of three sentiment categories:

• Negative
• Neutral
• Positive

Model Details

• Base model: distilbert/distilbert-base-uncased
• Architecture: DistilBERT
• Task: Text Classification / Sentiment Analysis
• Language: English
• Number of labels: 3
• Maximum sequence length: 128 tokens
• Training epochs: 3
• Learning rate: 2e-5
• Batch size: 16
• Optimizer: AdamW (via Hugging Face Trainer)
• Weight decay: 0.01

Dataset

The model was fine-tuned on a dataset containing 1,453 product reviews.

Split	Samples
Training	1,162
Validation	145
Test	146
Total	1,453

The sentiment labels were mapped as follows:

Negative → 0
Neutral  → 1
Positive → 2

Evaluation Results

The model achieved 97.26% accuracy on the held-out test set.

Class	Precision	Recall	F1-Score	Support
Negative	0.92	1.00	0.96	48
Neutral	1.00	0.96	0.98	49
Positive	1.00	0.96	0.98	49
Overall Accuracy			0.97	146

Confusion Matrix

[[48  0  0]
 [ 2 47  0]
 [ 2  0 47]]

Rows represent the true labels and columns represent the predicted labels in the order:

Negative, Neutral, Positive

Training

The model was fine-tuned using the Hugging Face transformers library and Trainer API.

Training was performed using a Google Colab Tesla T4 GPU.

The best model checkpoint was selected based on validation accuracy.

Usage

Install the required libraries:

pip install transformers torch

Load the model and tokenizer:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_id = "rohanpython9229/OpinionAI-Sentiment-DistilBERT"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSequenceClassification.from_pretrained(model_id)

text = "This product is amazing and works perfectly!"

inputs = tokenizer(
    text,
    return_tensors="pt",
    truncation=True,
    padding=True,
    max_length=128
)

with torch.no_grad():
    outputs = model(**inputs)

prediction = torch.argmax(outputs.logits, dim=-1).item()

print(model.config.id2label[prediction])

Example output:

Positive

Limitations

• The model was trained on a relatively small dataset of 1,453 reviews.
• Performance may vary on text that differs significantly from the training data.
• The model is intended primarily for English product-review sentiment classification.
• The reported test performance should not be interpreted as general performance across all types of text.

Intended Use

This model is intended for educational, experimental, and application-development purposes involving sentiment analysis of English product reviews.

It can be used as the NLP model component of the OpinionAI Sentiment Analysis project.

License

This model is fine-tuned from distilbert/distilbert-base-uncased, which is released under the Apache License 2.0. Please refer to the base model's license and terms when using or redistributing this model