How nurrasyid14/h works
IndoToxic 2024 analyzes Indonesian text for toxicity, mood, sentiment, subtopics, and candidate gambling promotions. The documented workflow starts with annotated comments, cleans text, produces TriModel mood/sentiment/subtopic signals, and feeds those signals plus lexicon features to a calibrated toxicity classifier. A separate prediction entry point combines TriModel predictions with a heuristic toxicity score; gambling detection is an independent pattern-based capability. The dashboard is explicitly unimplemented and disconnected from trained models. Several relationships below come from README documentation; unsampled internals are not inferred.
Main entry points: trimodel_rf.py TriModel coordinator · predict.py Text prediction · preprocessing.py Text cleaning · data_cleaned_before_encode.csv Cleaned modeling data
Components 16
Data preparation
- Annotated comments
data/raw/indotoxic2024_annotated_data-3.csv - Text cleaning
src/preprocessing.py - Cleaned modeling data
data/processed/data_cleaned_before_encode.csv
Classification
- Model training
src/train.py - TriModel coordinator
src/trimodel_rf.py - Mood classifier
src/classif/mood_rf.py - Sentiment classifier
src/classif/sentiment_rf.py - Subtopic classifier
src/classif/subtopic_rf.py - Toxicity SVM
src/classif/Toxic_or_not.py - Lexicon rules
src/classif/rule_toxicity.py
Prediction services
- Text prediction
src/predict.py - Toxicity meter
src/classif/toxicity_meter.py - Gambling detector
src/classif/gambling_ad_detector.py
Dashboard
- Dashboard app
dashboard
Other components
- Analyst
- X posts
Connections 15
- Annotated comments → Text cleaning: clean text
README.md - Text cleaning → Cleaned modeling data: writes data
README.md - Cleaned modeling data → Model training: training input
README.md - Model training → TriModel coordinator: trains heads
README.md - TriModel coordinator → Mood classifier: fits
src/trimodel_rf.py - TriModel coordinator → Sentiment classifier: fits
src/trimodel_rf.py - TriModel coordinator → Subtopic classifier: fits
src/trimodel_rf.py - TriModel coordinator → Toxicity SVM: feeds features
README.md - Lexicon rules → Toxicity SVM: supplies signals
src/classif/Toxic_or_not.py - Analyst → Text prediction: submits text
- Text prediction → TriModel coordinator: predicts heads
src/predict.py - Text prediction → Toxicity meter: scores text
src/predict.py - Analyst → Gambling detector: checks post
- X posts → Gambling detector: scans optionally
README.md - Analyst → Dashboard app: opens dashboard