Twitter-Sentiment-Analysis-using-CNN-LSTM-and-Sentiment140-dataset

Overview

This project implements a comprehensive sentiment analysis pipeline for Twitter data using the Sentiment140 dataset. The objective is to classify tweets as positive or negative using both classic machine learning and deep learning (CNN-LSTM) models.

Workflow

1. Data Loading

   • Download Sentiment140 and load into a pandas DataFrame

2. Preprocessing

   • Drop unused columns
   • Demojize tweets
   • Expand contractions and slangs
   • Clean text (lowercase, remove punctuation, etc.)
   • Tokenize, remove stopwords, lemmatize
   • Spelling correction
   • Hashtag normalization

3. Vectorization & Embedding

   • TF-IDF for classic ML
   • Tokenizer + padded sequences for LSTM/CNN-LSTM

4. Model Training

   • Train classic ML models for baseline
   • Build and train a CNN-LSTM model with Keras

5. Evaluation

   • Achieve 85% train / 80% test accuracy
   • Evaluate with classification metrics

6. Live Prediction

   • Load trained model and tokenizer
   • Preprocess and predict sentiment for user-typed sentences

---

Usage

1. Clone the repository and install dependencies
2. Run preprocessing scripts to generate vectorized data
3. Train the model using provided training scripts
4. Use the live prediction script to classify new tweets

---

Results

Metric	Value
Train Accuracy	85%
Test Accuracy	80%
Model	CNN-LSTM
Dataset	Sentiment140

---

Live Demo

You can input new sentences and get instant sentiment predictions using the trained model.

---

Acknowledgements

• Sentiment140 Dataset
• TensorFlow
• Keras
• scikit-learn

---