Twitter Sentiment Analysis using Probabilistic Modeling

Project Overview

This project aims to analyze and classify the sentiment of tweets into Positive or Negative. The project demonstrates the full pipeline of an NLP task, from raw data cleaning to building a custom probabilistic model for prediction.

Tech Stack & Libraries

• Language: Python
• Data Handling: Pandas, NumPy
• NLP Tools: NLTK (Natural Language Toolkit)
• Modeling: Custom Probabilistic Implementation
• Evaluation: Scikit-learn (Accuracy Score)

Key Features

• Custom Preprocessing Pipeline: - Removing URLs, Hashtags, and Mentions.
  • Tokenization and Stopwords removal.
  • Porter Stemming to reduce words to their root forms.
• Frequency Dictionary: Building a word-frequency mapping based on sentiment labels.
• Probabilistic Model: Implementing a Naive Bayes-like approach for sentiment classification.
• Live Prediction: A dedicated function to test the model with any custom string input.

Dataset

The model was trained using the Twitter Sentiment Analysis Dataset, which contains thousands of labeled tweets. The data is split into:

• twitter_training.csv for model building.
• twitter_validation.csv for testing and accuracy calculation.

Accuracy

The model achieved an impressive accuracy of 90.0% on the validation set.

How to Use

1. Clone the repository.
2. Ensure you have the datasets in the specified paths.
3. Run the Jupyter Notebook Sentiment Anaysis.ipynb.
4. Use the Sentiment_Analysis("Your text here") function to test custom statements.

Developed as part of an NLP & AI exploration.