A research platform that detects fear-mongering patterns in video transcripts and correlates them with biometric stress signals (heart rate, HRV, EDA) from wearables like Fitbit, Apple Health, or Google Fit.
- Multi-source transcript ingestion: YouTube URLs, manual text input, SRT/VTT subtitles, or MP4 transcription via Whisper.
- ML-powered fear detection: NLP pipeline combining emotion classification, zero-shot NLI, and lexical cue analysis.
- Flexible text segmentation: Character-based, sentence-based, or hybrid segmentation with adjustable parameters.
- Biometric data integration: Fitbit API, Apple Health exports, Google Fit API, or custom CSV uploads.
- Correlation analytics: Time-aligned analysis of fear scores and physiological responses.
- Interactive visualization: Dual-axis charts, distribution analytics, and downloadable datasets.
fear-monger-detector/
│
├── demos/ # Demo scripts and Jupyter notebooks
├── docs/ # Architecture diagrams and documentation
├── src/
│ ├── backend/
│ │ ├── fear_monger_processor/ # Fear detection NLP pipeline
│ │ │ ├── model.py # Model loading and caching
│ │ │ ├── inference.py # Batch prediction
│ │ │ ├── transcript.py # YouTube transcript fetching
│ │ │ └── utils.py # Segmentation and analysis
│ │ ├── fitbit_app/ # Fitbit ingestion and alignment
│ │ │ ├── fitbit_client.py # OAuth2 + API calls
│ │ │ ├── fitbit_utils.py # Data processing
│ │ │ └── aligner.py # Time-based correlation
│ │ └── ted_talks_app/ # TED transcript database
│ │ └── data_loader.py # CSV loader with metadata
│ └── frontend/
│ └── correlation_engine/ # Streamlit application
│ ├── app.py # Main UI
│ ├── config.py # Configuration constants
│ └── styles.css # Custom styling
├── assets/ # Images and media for documentation
├── requirements.txt # Python dependencies
├── .env # Environment variable template
└── README.md
- Backend: FastAPI + Celery for async processing
- Storage: Postgres/TimescaleDB + MinIO (object storage)
- ML/NLP: Falconsai/fear_mongering_detection
- Frontend: Streamlit (MVP) or Next.js (scalable)
- APIs: Fitbit Web API, YouTube Transcript API
- Visualization: Plotly, Matplotlib
Pipeline: Video Transcript → NLP Fear Score → Alignment with Biometric Data → Analytics → Visualization
| Fear Analysis Stream | Heart Rate Stream (Fitbit) | Merged Correlation |
|---|---|---|
| Transcript text / SRT/VTT | Raw heart rate readings | Align by playback window |
| Segment text → fear_score | Trim to selected time window | Merge on relative timestamp |
| Normalize fear_score (0 → 1) | Normalize heart rate (0 → 1) | Plotly chart: fear vs heart |
| Summary metrics & CSV | Chart & CSV | Correlation stats & CSV |
- Fear intensity timeline with highlighted transcript segments.
- Overlay of heart rate/HRV against fear spikes.
- Lagged cross-correlation plots to see if fear spikes precede stress responses.
- Python 3.10+
- Docker & docker-compose (for running services)
- Fitbit/Google API credentials (if using live connectors)
# Clone repository
git clone https://github.com/torontoai-hub/fear-monger-detector.git
cd fear-monger-detector
# Create virtual environment
python -m venv .venv
source .venv/bin/activate # Linux/macOS
# .venv\Scripts\activate # Windows
# Install dependencies
pip install -r requirements.txt
# Setup environment variables
cp .env.example .env
# Edit .env with your Fitbit credentials# .env file
FITBIT_CLIENT_ID=your_client_id_here
FITBIT_CLIENT_SECRET=your_client_secret_here
FITBIT_REDIRECT_URI=http://localhost:8080/callbackThis guide demonstrates how to analyze a video transcript to detect fear-mongering and correlate the results with biometric stress data, such as heart rate from a Fitbit device.
1️⃣ Launch the Application
- Navigate to the app:
cd src/frontend/correlation_engine
streamlit run app.pyThe app will open in your browser at http://localhost:8501
2️⃣ Run Fear Analysis
Option A: YouTube Video
- Paste a YouTube URL into the input field
- The transcript will be fetched automatically
- Preview appears in the "Transcript Preview" expander
Option B: Manual Input
- Paste transcript text into the text area
- Text is immediately ready for analysis
Option C: TED Talks Database
- Open the TED Talks sidebar expander
- Sort by title, views, date, or duration
- Navigate pages and select a talk
- Click to analyze
- Segment Mode: Characters | Sentences | Both
- Max Characters: 200-600 (default: 400)
- Max Sentencess: 1-10 (default: 5)
- Fear Threshold: 0.0-1.0 (default: 0.7)
- Scores above threshold = high risk
- Smoothing Window: 1-10 segments (default: 3)
- Reduces noise in timeline
- Type: Line | Bar | Area chart
- Hover Length: 20-500 characters (default: 30)
- Total paragraphs analyzed
- Average fear score vs. threshold
- Peak score and minimum score
- High-risk segment count and percentage
- 🔴 High Risk: Average score ≥ threshold
- 🟡 Moderate: 0.5 ≤ Average score < threshold
- 🟢 Low Risk: Average score < 0.5
- Distribution pie chart (Low / Medium / High)
- Interactive timeline chart (Line / Bar / Area)
- Paragraph-level analysis table
- CSV export with all scores and timestamps CSV export with all scores and timestamps
Prerequisites
- Fitbit account with intraday heart rate data enabled
- Completed OAuth2 authentication (first-time setup)
- Heart rate data recorded on the target date
Steps
- Navigate to the Fitbit Heart Rate Data section
- Select the date using the date picker
- Click the Load Data button
- Wait for the API response (≈5–10 seconds)
- View the heart rate chart for the full day
Troubleshooting
-
"No data available"
- Ensure you wore your Fitbit device that day
- Check that heart rate tracking was enabled
- Verify OAuth2 token hasn't expired
-
"Invalid credentials"
- Re-run OAuth2 flow
- Check
.envfile for correct client ID/secret
Setup Playback Window
- Navigate to Fear vs. Heart Rate Analysis
- Configure the time window:
- Start Time: HH:MM when you started watching
- Duration: Minutes of video playback (1–30)
- Click Load Fitbit vs Heart Data
What Happens Next 🔄 Alignment Process
- Fetch Fitbit data for the specified window (1-min resolution)
- Map transcript segments to relative timestamps
- Interpolate fear scores to match heart rate frequency
- Normalize both streams (0 → 1) for comparison
- Merge datasets on aligned timestamps
📊 Output
- Dual-axis chart: Fear score (line) + heart rate (bars)
- Summary metrics:
- Average fear score
- Average heart rate (bpm)
- Number of aligned data points
- Aligned data table: Preview first 20 rows
- CSV download: Full merged dataset
Use Cases
- Identify if fear spikes correlate with heart rate increases
- Detect delayed physiological responses (lag analysis)
- Compare emotional content with objective stress markers
- Export data for custom statistical analysis in R/Python
✅ Transcript Quality
- Use complete transcripts with proper punctuation
- Verify timestamps are accurate
- Remove non-verbal annotations like
[music]or[applause]
✅ Fitbit Integration
- Ensure intraday API access is enabled
- Wear the device continuously during playback
- Sync Fitbit before fetching data
✅ Analysis Parameters
- Start with the default threshold (0.7) and adjust as needed
- Use smoothing for noisy transcripts
- Compare multiple videos to establish baseline patterns
✅ Correlation Interpretation
- Correlation ≠ causation
- Account for individual baseline differences
- Consider confounding factors (e.g., caffeine, stress, environment)
Research Applications
- Media Studies: Quantify fear-mongering in news coverage
- Psychology: Study emotional responses to persuasive content
- Public Health: Analyze stress responses to health messaging
- Education: Teach critical media literacy with objective metrics
Personal Use
- Self-awareness: Track emotional responses to media consumption
- Content Curation: Identify high-stress content to avoid
- Mindfulness: Correlate media habits with physiological stress
Code Quality (Priority: High)
- Refactor
app.pyinto modular components - Add comprehensive unit tests (
pytest) - Implement type hints throughout
- Setup CI/CD pipeline (GitHub Actions)
- Add logging and error tracking (Sentry)
Features (Priority: Medium)
- Multi-video batch analysis
- Apple Health integration
- Google Fit API support
- Export to PDF reports
- Real-time streaming analysis
- Multi-language support (non-English transcripts)
ML Improvements (Priority: Medium)
- Fine-tune model on domain-specific data
- Add explainability (LIME/SHAP)
- Emotion classification (fear, anger, sadness, joy)
- Stance detection (pro/anti topics)
Infrastructure (Priority: Low)
- Migrate to FastAPI backend
- Setup PostgreSQL + TimescaleDB
- Add Redis caching layer
- Implement Celery task queue
- Deploy with Docker Compose
- Production-ready Next.js frontend
This project is for research, educational, and demonstration purposes only. It is not a medical device and does not provide medical advice. Correlation does not imply causation. Results should not be used to make health-related decisions. Individual physiological responses vary significantly
- Use responsibly and transparently
- Respect privacy when analyzing content
- Consider potential biases in training data
- Do not use for surveillance or manipulation
MIT License. See LICENSE for details.
Pull requests are welcome! For major changes, please open an issue first to discuss what you’d like to change.