OnCoCo Analysis System¶
✅ Status: Implemented
The OnCoCo analysis system is fully implemented and ready for production use.
Overview¶
OnCoCo (Online Counseling Conversations) classifies counseling conversations at sentence level with a 68‑category system (XLM‑RoBERTa Large, bilingual DE/EN). The analysis is available in the UI and includes live progress, distributions, transition heatmaps, and statistical pillar comparisons.
Documentation¶
| Document | Description | Status |
|---|---|---|
| Concept | Implementation and architecture overview | ✅ Updated |
Features¶
- 68‑class categorization (40 counselor, 28 client)
- Live analysis with Socket.IO progress updates
- Label distributions (filter by pillar, level, role)
- Transition matrices as heatmaps + top transitions
- Pillar comparison (matrix comparison metrics)
- KIA data sync directly from the UI
- Resume/force‑resume for stuck analyses
UI Entry Points¶
- Overview:
/oncoco - New analysis:
/oncoco/config - Results:
/oncoco/results/:id - Model & labels:
/oncoco/info
Relevant Files¶
app/
├── routes/oncoco/ # Info, Pillars, Analyses, Results, Matrix, Debug
├── services/oncoco/
│ ├── oncoco_service.py
│ └── oncoco_labels.py
└── services/judge/kia_sync_service.py # KIA data sync
llars-frontend/src/components/OnCoCo/
├── OnCoCoOverview.vue
├── OnCoCoConfig.vue
├── OnCoCoResults.vue
├── OnCoCoInfo.vue
├── TransitionHeatmap.vue
└── MatrixComparisonMetrics.vue