How prateeks-py/emotion-detection works
This repository is a reproducible emotion-dataset bundle with an adjacent baseline classifier and interactive local coaching script. The main workflow builds synthetic records (optionally mixed with public multilingual records), writes CSV/SQLite outputs and a manifest, then can train a classifier for use by the coach. The README documents dataset provenance, mixed-mode generation, and validation. The graph also includes the coach because it is a runnable product capability in the tree; its local Ollama dependency is shown separately. Edges reflect the builder, trainer, and coach excerpts, plus README-documented workflows.
Main entry points: build_dataset.py Dataset builder · emotional_coach.py Conditional-trajectory coach · train_emotion_model.py Classifier training · emotion_model.joblib Trained classifier
Components 13
Dataset generation
- Dataset builder
build_dataset.py
Emotion modeling
- Classifier training
train_emotion_model.py - Trained classifier
emotion_model.joblib - Evaluation report
evaluation_report.txt
Interactive coaching
- Conditional-trajectory coach
emotional_coach.py
Dataset artifacts
- SQLite schema
schema.sql - Emotion CSV
emotion_dataset.csv - Emotion database
emotion_dataset.sqlite - Provenance manifest
manifest.json
Other components
- Dataset maintainer
- Hugging Face dataset
- Chat user
- Local Ollama service
Connections 13
- Dataset maintainer → Dataset builder: runs build
README.md - Dataset builder → Emotion CSV: writes rows
build_dataset.py - Dataset builder → SQLite schema: reads schema
build_dataset.py - Dataset builder → Emotion database: writes records
build_dataset.py - Dataset builder → Provenance manifest: writes provenance
build_dataset.py - Dataset builder → Hugging Face dataset: loads public rows
build_dataset.py - Classifier training → Emotion CSV: reads records
train_emotion_model.py - Classifier training → Trained classifier: saves classifier
train_emotion_model.py - Classifier training → Evaluation report: writes metrics
train_emotion_model.py - Chat user → Conditional-trajectory coach: enters messages
emotional_coach.py - Conditional-trajectory coach → Trained classifier: loads and predicts
emotional_coach.py - Conditional-trajectory coach → Emotion database: queries examples
emotional_coach.py - Conditional-trajectory coach → Local Ollama service: sends prompt
emotional_coach.py