krupaj/my-bert-imdb2

BERT Fine-Tuned for IMDB Sentiment Classification

Model Description

This model is fine-tuned from bert-base-uncased to classify movie reviews as Positive or Negative.

It uses the IMDB movie review dataset and is designed to predict the sentiment expressed in a review.

Model Details:

• Base Model: bert-base-uncased

• Task: Text Classification / Sentiment Analysis

• Dataset: Stanford IMDB Dataset

• Number of Classes: 2

• Labels: Negative (0), Positive (1)

• Framework: PyTorch, Hugging Face Transformers

This model can be used for:

• Classifying movie reviews by sentiment.

• Practicing transfer learning and BERT fine-tuning.

• Exploring NLP text classification tasks.

How to Use:

Try My BERT Sentiment Analysis Model 🤗

Step 1: Install the required libraries

from transformers import pipeline

classifier = pipeline( "text-classification", model="krupaj/my-bert-imdb2" )

review = input("Enter a movie review: ") result = classifier(review)[0]

Convert model labels into readable sentiment names

label_mapping = { "LABEL_0": "NEGATIVE", "LABEL_1": "POSITIVE" }

sentiment = label_mapping.get(result["label"], result["label"])

print("\nMovie Review:", review) print("Predicted Sentiment:", sentiment) print("Confidence Score:", round(result["score"] * 100, 2), "%")

Step 3: Enter your own movie review

# Try any movie review you like!

review = "This movie was absolutely fantastic!"

# Predict the sentiment
result = classifier(review)

# Display the result
print("Movie Review:", review)
print("Prediction:", result[0]["label"])
print("Confidence Score:", round(result[0]["score"] * 100, 2), "%")

Example output

Movie Review: This movie was absolutely fantastic!
Prediction: LABEL_1 (LABEL_0 → NEGATIVE / LABEL_1 → POSITIVE)
Confidence Score: 98.00%

Training Details

The model was fine-tuned using the IMDB dataset with the Hugging Face Transformers library.

Training configuration:

Epochs: 3

Learning Rate: 2e-5

Training Batch Size: 8

Maximum Sequence Length: 256 tokens

Optimizer: Hugging Face Trainer default for the installed configuration

The training setup used a training dataset, a validation dataset, and a separate test dataset.

Limitations:

• Performance may vary on reviews that differ from the IMDB training data.

• Sarcasm, mixed sentiment, and complex language may be challenging.

• Prediction scores are model confidence scores, not guaranteed probabilities of correctness.

• This model is intended for experimentation and educational use.