Sample documents for testing RAG (Retrieval-Augmented Generation) pipelines — chunking strategies, embedding quality, and retrieval evaluation.
Each file is an excerpt of the same academic text (The Role of Semantic Web Technologies in Smart Environments, a PhD thesis by Faisal Razzak, Politecnico di Torino, 2013), truncated at different sizes to exercise chunking behaviour at scale.
| File | Approx. size |
|---|---|
sample_50k.md |
50 KB |
sample_100k.md |
100 KB |
sample_250k.md |
250 KB |
sample_500k.md |
500 KB |
Complete peer-reviewed papers converted from PDF to Markdown, with all sections, tables, algorithms, and references preserved.
| File | Source | Description |
|---|---|---|
intelligent_energy_optimization_smart_home.md |
IEEE Trans. Smart Grid, Vol. 3, No. 4, Dec. 2012 | Energy optimization for smart homes using Domotic Effects and Boolean SAT; includes 3 data tables and 2 algorithms |
spamming_internet_of_things.md |
Procedia Computer Science, Vol. 10, 2012 | IoT spam via QR code spoofing and ECDSA-based countermeasures; includes 2 experimental results tables |
- Chunking benchmarks — compare fixed-size, sentence, paragraph, and semantic chunking on real prose
- RAG pipeline smoke tests — load a known document and assert expected retrieval results
- Embedding model evaluation — measure retrieval quality across different embedding models and chunk sizes
- Latency / throughput profiling — stress-test indexing and query latency against realistic document sizes
The original thesis is publicly available. These excerpts are provided for non-commercial research and tooling purposes only.