🐦 Twitter Sentiment Analysis – NLP Portfolio Project

Author: Utkarsh Tekriwal
Toolset: Python | Hugging Face Transformers | Gradio | Pandas

---

🔍 Project Overview

Turn raw Twitter data into real‑time sentiment insights. This project demonstrates how to ingest tweet text, preprocess it, run it through a fine‑tuned RoBERTa model, and expose the results via an interactive Gradio web interface.

---

📁 Input & Data

• Sample CSV (data/sample_tweets.csv)
  • Column: TweetText
  • Contains example tweets for quick testing
• CSV Upload
  • Any user‑provided CSV with a TweetText column can be analyzed live

---

🛠️ Built With

• Python 3.8+
• Transformers (cardiffnlp/twitter-roberta-base-sentiment) for sentiment classification
• Gradio for web UI
• Pandas for data handling
• Matplotlib for simple bar charts

---

⚙️ How It Works

1. Load & Clean
   • Remove URLs, mentions, and emojis
2. Tokenize
   • Use the RoBERTa tokenizer to prepare inputs
3. Predict
   • Run through the fine‑tuned RoBERTa model
   • Apply softmax to get probabilities for Positive, Neutral, Negative
4. Visualize
   • Render sentiment distribution bar chart
   • Show JSON summary and top‑label with confidence
5. Interact
   • Users upload their own CSV via Gradio
   • Results update instantly

---

📊 Key Metrics & Insights

Metric	Description
✅ Total Tweets	Number of tweets processed
👍 Positives	Count & percentage of positive tweets
🤝 Neutrals	Count & percentage of neutral tweets
👎 Negatives	Count & percentage of negative tweets
📈 Distribution Chart	Bar chart showing sentiment breakdown
🔍 Top Example	Highest confidence prediction displayed

---

🖼️ App Preview

---

📂 Folder Structure

Twitter-Sentiment-Analysis/
├── assets/                    # App screenshots
│   └── app_screenshot.png
├── data/                      # Sample input CSV
│   └── sample_tweets.csv
├── src/                       # Application code
│   ├── sentiment_model.py     # Load & run the RoBERTa model
│   └── gradio_app.py          # Gradio UI entrypoint
├── notebooks/                 # Exploratory analysis (optional)
│   └── Twitter_Sentiment_Analysis.ipynb
├── requirements.txt           # Project dependencies
├── README.md                  # (You are here)
└── LICENSE                    # MIT License

---

🚀 Getting Started

1. Clone the repo

   git clone https://github.com/utkarsh-tekriwal/Twitter-Sentiment-Analysis.git
   cd Twitter-Sentiment-Analysis

2. Install dependencies

   pip install -r requirements.txt

3. Run the app

   python src/gradio_app.py

   • Opens http://localhost:7860
   • Upload your own CSV and explore sentiment distributions

---

💡 Why This Project Matters

• Real‑World Relevance: Brands and analysts rely on sentiment to gauge public opinion.
• State‑of‑the‑Art Model: Leveraging a Transformer fine‑tuned on Twitter data ensures high accuracy.
• No‑Code UI: Gradio makes deployment and sharing effortless.
• Extensible Design: Swap in any Hugging Face model, add new visualizations, or integrate with other dashboards.

---

📜 License

This project is licensed under the MIT License.