🎬 IMDb Sentiment Analysis using LSTM

A Deep Learning-based Sentiment Analysis application that predicts whether a movie review is Positive or Negative using an LSTM (Long Short-Term Memory) neural network trained on the IMDb 50K Movie Reviews dataset.

🚀 Live Demo

🔗 https://sentiment-analysis-lstm-p6opvqwd68r8li5zuiappmb.streamlit.app/

---

📌 Features

• Predict sentiment of movie reviews
• Deep Learning model using LSTM
• Text preprocessing pipeline
• Interactive Streamlit UI
• Confidence score visualization
• Real-time inference

---

🛠️ Tech Stack

Machine Learning & NLP

• Python
• TensorFlow / Keras
• NLTK
• NumPy
• Pandas
• Scikit-learn

Frontend

• Streamlit

Model Architecture

• Embedding Layer
• LSTM Layer
• Dense Layers
• Sigmoid Output Layer

---

📂 Dataset

Dataset used:

📊 IMDb Dataset of 50K Movie Reviews

• 50,000 movie reviews
• Binary sentiment labels
• Positive / Negative classification

Source: https://www.kaggle.com/datasets/lakshmi25npathi/imdb-dataset-of-50k-movie-reviews

---

🔄 Data Preprocessing

The following preprocessing steps were applied:

• Convert text to lowercase
• Remove HTML tags
• Remove URLs
• Remove punctuation
• Remove stopwords
• Lemmatization
• Tokenization
• Sequence Padding

🧠 Model Architecture

Input Review
      ↓
Tokenizer
      ↓
Padding (maxlen = 200)
      ↓
Embedding Layer
      ↓
LSTM (128 Units)
      ↓
Dense Layer
      ↓
Dropout
      ↓
Output Layer (Sigmoid)
      ↓
Positive / Negative

📊 Results

Accuracy | ~85-90%

📁 Project Structure

Sentiment-Analysis-LSTM/
│
├── app.py
├── sentiment_lstm.keras
├── tokenizer.pkl
├── requirements.txt
├── README.md
├── Sentiment.ipynb

📸 Application Preview

---

🎯 Future Improvements

• Bidirectional LSTM
• GRU Comparison
• Attention Mechanism
• BERT Integration
• Explainable AI (SHAP/LIME)
• Docker Deployment
• REST API using FastAPI
• React Frontend