Twitter Sentiment Analysis - NLP

This project aims to detect hate speech in tweets, identifying tweets with racist or sexist sentiments. The primary goal is to classify these hateful tweets from non-hateful ones using Natural Language Processing (NLP) techniques.

Project Overview

The dataset includes 31,962 labeled tweets provided as part of a hackathon problem on Analytics Vidhya. A tweet is labeled as '1' if it is classified as racist or sexist and '0' if it is not. The objective is to build a model to predict these labels for a given test dataset.

Hackathon Link: Practice Problem - Twitter Sentiment Analysis

Dataset Information

• Total Tweets: 31,962
• Labels:
  • 1 - Racist/Sexist
  • 0 - Not Racist/Sexist
• Format: CSV file with columns for tweet ID, label, and tweet text

Libraries Used

• pandas: For data manipulation and preprocessing
• matplotlib: For data visualization
• seaborn: For enhanced visualizations
• scikit-learn: For implementing machine learning models
• nltk: For creating applications for statistical

Algorithms Implemented

This project explores two algorithms in separate Jupyter Notebooks (.ipynb files):

1. Logistic Regression: A classic machine learning approach for binary classification.
2. LSTM (Long Short-Term Memory): A deep learning model capable of handling sequential data, particularly useful for text analysis.

Best Model Performance

• Accuracy: 95.00%

File Structure

• Twitter_Sentiment_Analysis.ipynb: Notebook implementing Logistic Regression
• LSTM_Twitter_Sentiment_Analysis.ipynb: Notebook implementing LSTM model