---
license: cc-by-4.0
pretty_name: "The Missing Link: Knowledge Graph-Guided Discovery of Novel Prompt Compositions"
tags:
- knowledge-graph
- hypergraph
- prompt-engineering
- machine-learning
- owl
- rdf
size_categories:
- n<1K
---

# P-KG - From "The Missing Link: Knowledge Graph-Guided Discovery of Novel Prompt Compositions (A.S. Kumar et al., 2026)"

This repository contains the knowledge graph and hypergraph utilised in the work "The Missing Link: Knowledge Graph-Guided Discovery of Novel Prompt Compositions (A.S. Kumar et al., 2026)".

## Licensing
This data is released under the **Creative Commons Attribution 4.0 International (CC-BY-4.0)** license. You are free to share, copy, and adapt this dataset for any purpose, including commercially, provided you give appropriate credit.

## Data Schema & Ontologies
The dataset uses a shared vocabulary defined in `ontology.ttl`. The core classes include:
- `Technique`: Prompt engineering techniques (e.g., Chain-of-Thought, ReAct).
- `AlgorithmicComponent`: Components containing execution, search, or retrieval logic.
- `PromptComponent`: Components defining prompt structures (examples, constraints).
- `DataFlow`: Components handling loop, feedback, or decomposition flows.
- `Task`: Evaluation datasets and tasks (e.g., GSM8K, HumanEval).
- `CognitiveCapability`: Intermediate capability nodes mapping Techniques to Tasks.
- `BenchmarkResult`: Specific model outputs on benchmarks.

---

## Directory Structure

```
/
├── README.md             # This documentation file
├── metadata.json         # Dataset metadata, versioning, and summary statistics
├── ontology.ttl          # Shared OWL ontology defining classes and properties
├── flat_graph/           # Flat graph representation (binary relations)
│   ├── nodes.csv         # Flat graph node table
│   ├── edges.csv         # Flat graph edge table
│   ├── graph.json        # Unified hierarchical JSON graph
│   ├── graph.ttl         # RDF Turtle triples representation
│   └── node_features.npy # NumPy embedding matrix [num_nodes, 384] for nodes
└── hypergraph/           # Hypergraph representation (set-based techniques)
    ├── nodes.csv         # Hypergraph node table (excludes Techniques)
    ├── hyperedges.csv    # Hyperedges (Techniques) table
    ├── hyperedge_members.csv # Join table linking hyperedges to member nodes
    ├── hypergraph.json   # Unified hierarchical JSON hypergraph
    ├── hypergraph.ttl    # RDF Turtle reified hypergraph representation
    └── node_features.npy # NumPy embedding matrix [num_nodes, 384] for nodes
```

---

## Dataset Statistics

### Flat Graph
- **Nodes**: 242
- **Edges**: 779
- **Node Embeddings**: `flat_graph/node_features.npy` has shape `[242, 384]` (384-dimensional SentenceTransformer embeddings generated using the `all-MiniLM-L6-v2` model).

### Hypergraph
- **Nodes (non-Technique members)**: 151
- **Hyperedges (Technique entities)**: 91
- **Node Embeddings**: `hypergraph/node_features.npy` has shape `[151, 384]`.