SentientPulse: Real-Time Brand & Social Sentiment Intelligence

🚀 Live Demo

SentientPulse is an airline-focused sentiment intelligence dashboard that analyzes customer tweets, classifies sentiment, and highlights operational issues such as delays, baggage problems, cancellations, and service complaints.

Business Value

• Tracks airline customer sentiment in near real time.
• Helps surface high-impact service issues quickly.
• Provides an executive-friendly dashboard for monitoring brand health.
• Supports issue-level insights beyond basic sentiment classification.

Dataset

This project uses the Twitter US Airline Sentiment dataset. The key columns are:

• text: tweet content.
• airline_sentiment: sentiment label.
• negativereason: optional field for negative issue analysis.

Technical Architecture

• Training script: model_trainer.py
• Inference dashboard: app.py
• Feature extraction: TF-IDF
• Classifier: Linear SVM
• Visualization: Streamlit + Plotly
• Persistence: Joblib

Project Workflow

1. Load the raw CSV dataset.
2. Remove missing rows and empty labels.
3. Normalize tweet text.
4. Train a TF-IDF + Linear SVM classification pipeline.
5. Save the pipeline as sentiment_pipeline.joblib.
6. Load the model in Streamlit.
7. Predict sentiment for new text.
8. Show live sentiment stream and airline aspect analysis.

Setup

1. Create a virtual environment

python -m venv venv

2. Activate it

# Windows
venv\Scripts\activate

# macOS/Linux
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Prepare the dataset

Place your CSV file in the project root as data.csv.

Required columns:

• text
• airline_sentiment

Optional column:

• negativereason

5. Train the model

python model_trainer.py

6. Launch the dashboard

streamlit run app.py

Notes on Production Readiness

• Missing values are dropped before training.
• Empty labels are removed.
• Text is normalized consistently in both training and inference.
• The model is saved as a reusable pipeline.
• The dashboard gracefully handles model-loading failures.

Future Enhancements

• Connect to live Twitter/X or RSS data.
• Add probability calibration.
• Replace keyword-based aspect analysis with trained ABSA.
• Add deployment with Docker and cloud hosting.

📝 License

This project is licensed under the MIT License.