Feature: Emotional Tagging and Alert System - Sprint 1

[Tanvir-ctrl1]

Description

Added the Emotional Tagging and Alert System for Guardian Monitor — a machine learning pipeline that classifies patient nurse notes into emotional and clinical tags, combines them with a structured vitals risk model, and produces a Low/Medium/High alert level with a plain-language explanation per patient. Also includes an interactive clinical monitoring dashboard for medical staff.

Todos

• Tested and working locally
• Code follows the style guidelines of this project
• I have performed a self-review of my code
• Code changes documented
• Requested review from >= 2 devs on the team (one frontend and one backend recommended)

How to test

1. Navigate to AI Guardian/emotional_tagging_and_alert_system/
2. Run python guardian_pipeline.py — no external ML libraries required (NumPy and pandas only)
3. Pipeline will generate predictions for all 553 synthetic nurse notes
4. Open guardian_dashboard.html in any browser to view the patient alert monitoring dashboard
5. Check evaluation_report.txt for full F1 scores and alert-level performance metrics

Screenshots and/or Gifs

Dashboard displays patient alert cards sorted by severity (High → Medium → Low) with emotional/clinical tag chips, colour-coded vitals, and one-line clinical explanations.

Associated MS Planner Tasks

• Emotional Tagging and Alert System — Sprint 1

Known Issues

• Pipeline uses synthetic nurse notes only (real MIMIC-IV clinical notes require PhysioNet credentialing)
• Medium alert recall is 0.553 — improvement planned for Sprint 2 using a Random Forest vitals classifier
• scikit-learn and pyarrow unavailable in current environment — pipeline implemented from scratch in NumPy