/Twitter-Sentiment-Analysis

This project applies Natural Language Processing (NLP) techniques to extract and classify the emotional tone of social media text data. We will analyze sentiment of tweets sourced from X (formerly Twitter) using machine learning.

Primary LanguageJupyter NotebookGNU General Public License v3.0GPL-3.0

🐦 Twitter-Sentiment-Analysis

This project applies Natural Language Processing (NLP) techniques to extract and classify the emotional tone of social media text data. We will analyze sentiment of tweets sourced from X (formerly Twitter) using machine learning (ML). It leverages the Sentiment140 dataset, imported via Hugging Face Datasets, as the training and testing source. Real-time tweets are retrieved using the Tweepy library with authenticated access to the Twitter API.

The goal is to evaluate the performance of various classification models including Logistic Regression, Support Vector Machine (SVM), and others to determine which performs best for sentiment classification tasks.

The reproducibility of this project will be in the form of conda 'environment.yml'.

The IMPORTANT FILE that preprocesses and evaluates models is 'nlp_sentiment_pipline.ipynb'

⚙️ Reproduce Results

In the command line execute:

  1. 'conda env create -f environment.yml'
  2. 'conda activate twitter-sentiment'

Advice: Python Kernel is easily installed with VS Code

📂 Data Source & Description

This project uses the Sentiment140 dataset hosted on Hugging Face Datasets.

Dataset Details:

  • Source: Hugging Face Datasets → stanfordnlp/sentiment140
  • Size: 1.6 million tweets
  • Labels:
    • 0: Negative sentiment
    • 2: Neutral sentiment (often excluded in binary classification tasks)
    • 4: Positive sentiment

📊 Concluding Results

Model Best Params F1-Score Accuracy
Logistic Regression {'C': 1, 'solver': 'saga'} 0.78 0.77
Multinomial Naive Bayes {'alpha': 6.0, 'class_prior': None, 'fit_prior': False} 0.76 0.76
Support Vector Machine (SVM) {'C': 0.1, 'dual': False, 'max_iter': 1000, 'penalty': 'l1'} 0.78 0.77

🔍 Sample Classified Tweets (15 Live Twitter Samples)

✅ Positive Sentiment Tweets:

  • "Welcome first meeting White House FIFA task force..."
  • "🇺🇸 Steve Witkoff officially sworn Oval Office..."
  • "RT Evening Parade longstanding tradition Marine..."
  • "Country blessed negotiator skill experience..."
  • "RT Thank President Trump inviting White House..."
  • "Great honor Prime Minister U day ago big elect..."
  • "RT Don’t need energy transition need energy ad..."
  • "RT Energy price 🇺🇸"
  • "RT Thanks hardworking partner canceling wastef..."
  • "RT Message clear DHS enforce law put safety Am..."
  • "RT 🏆 Award show like 📸 scene SmallBusinessWeek..."
  • "President Trump Vance FIFA President Gianni In..."

❌ Negative Sentiment Tweets:

  • "RT Sgt Brian Lieberman ran toward gunfire save..."
  • "RT Even know air traffic control problem isn’t..."
  • "RT Played political game expense air traffic c..."

🚀 Aspirational Improvements

  • Introduce Non-linear ML models (i.e.Generalized Additive Models (GAMs), Local Regression (LOESS), or Splines)
  • Use different Vectorization methods (i.e. Word2Vec or GloVe) in place of TF-IDF matrices

📚 Citations

  • Go, A., Bhayani, R. and Huang, L., 2009. Twitter sentiment classification using distant supervision. CS224N Project Report, Stanford, 1(2009), p.12.
  • @vprusso. twitter_python. Notes: "Tweepy APIv1 tutorial".