jmp1987's picture
Upload README.md
5fb37df verified
|
Raw
History Blame Contribute Delete
1.92 kB
---
language: de
license: cc-by-4.0
task_categories: [question-answering, text-generation]
tags: [simson, moped, ddr, automotive, german, knowledge-graph, linked-data]
pretty_name: Simson Unified Knowledge Graph
size_categories: n<1K
---
# 🧠 Simson Unified Knowledge Graph
**173 Nodes × 348 Edges** – der Klebstoff zwischen allen Simson-Datasets.
## Was das ist
Ein maschinenlesbarer Graph, der alle 6 Datasets miteinander verknüpft:
| Dataset | Status | Nodes |
|---------|--------|-------|
| racing-planet-simson-traces | Diagnose-Traces | 15 |
| simson-forum-qa-pairs | Forum-Wissen | 30 |
| simson-repair-manual | Technische Daten | 14 |
| racing-planet-product-catalog | Teilekatalog | 37 |
| simson-youtube-tutorials | Video-Tutorials | 20 |
| simson-part-problem-matrix | Diagnose-Logik | 57 |
## Edge-Typen
| Relation | Anzahl | Bedeutung |
|----------|--------|-----------|
| `similar_to` | 228 | Forum-Threads mit gleichem Topic |
| `has_cause` | 44 | Symptom → Ursache |
| `fixed_by_part` | 38 | Ursache → Produkt |
| `fixes_with` | 35 | Symptom → direktes Produkt |
| `references` | 3 | Forum → Handbuch |
| `demonstrates` | (weitere) | YouTube → Handbuch |
## Verwendung
```python
from datasets import load_dataset
import json
ds = load_dataset("jmp1987/simson-unified-knowledge-graph")
graph = json.loads(ds["train"][0]["nodes"])
edges = json.loads(ds["train"][0]["edges"])
# Finde alle Teile für ein Symptom
symptom_node = next(n for n in graph if n["label"].startswith("Motor springt"))
related = [e for e in edges if e["source"] == symptom_node["id"]]
```
## RAG-Integration
Der Knowledge Graph erlaubt:
1. **Hop-1 Retrieval**: Frage → Symptom-Node → verlinkte Produkte/Manual/Forum
2. **Hop-2 Retrieval**: Frage → Symptom → Ursache → nötige Teile → Assembly-Deps
3. **Cross-Dataset Retrieval**: Automatisches Finden relevanter Chunks aus allen Datasets