Mo7art/Stack2Graph_VD_cpp

1---2tags:3  - stackoverflow4  - qdrant5  - embeddings6  - retrieval7  - question-answering8task_categories:9  - feature-extraction10  - question-answering11language:12  - en13pretty_name: C++ StackOverflow Vector Dataset14license: cc-by-sa-4.015viewer: false16---17 18# C++ StackOverflow Vector Dataset19 20## Summary21 22This Hugging Face dataset repository contains the C++ shard of the Stack2Graph vector-database component as restorable Qdrant artifacts plus portable Parquet fallback files.23Hugging Face uses one dataset repository per programming language, so this repository is directly cloneable without an extra top-level archive wrapper.24 25The artifacts are intended for semantic and hybrid retrieval, graph entry-point finding, and retrieval-augmented generation experiments over Stack Overflow content.26 27Stack2Graph source:28[https://github.com/tha-atlas/Stack2Graph](https://github.com/tha-atlas/Stack2Graph)29 30## Hugging Face Repository Layout31 32```text33README.md34dataset_manifest.json35qdrant_snapshots/36  collections_manifest.json37  stackoverflow_c++_vector.tar.zst38  stackoverflow_c++_vector.snapshot.zst.part_00039  ...40question_metadata/41  r0_00000.parquet42chunk_records/43  r0_00000.parquet44```45 46- `README.md`: Hugging Face dataset card rendered from this template.47- `dataset_manifest.json`: language-scoped manifest listing the files in this repository.48- `qdrant_snapshots/collections_manifest.json`: language-scoped Qdrant snapshot manifest.49- `qdrant_snapshots/stackoverflow_c++_vector.*`: Qdrant restore artifacts for `stackoverflow_c++_vector`.50- `question_metadata/*.parquet`: question-level metadata used by parent-child chunk retrieval.51- `chunk_records/*.parquet`: chunk-level vector rows when parent-child indexing is enabled.52- `question_records/*.parquet`: question-level vector rows when non-chunked export is used.53 54Parent-child exports contain `question_metadata/` and `chunk_records/`. Non-chunked exports contain `question_records/` instead. A repository does not need both `chunk_records/` and `question_records/`.55 56During Hugging Face upload, the local leading language directory is removed. A local file such as `c++/chunk_records/r0_00000.parquet` appears in this repository as `chunk_records/r0_00000.parquet`.57 58## Content59 60Rows are derived from Stack Overflow questions tagged for C++.61 62The current Stack2Graph vector pipeline uses:63 64- dense vectors from `Qwen/Qwen3-Embedding-8B`65- 4096-dimensional dense embeddings with last-token pooling66- instruction-aware query embedding for retrieval67- sparse lexical vectors from `BAAI/bge-m3`68- unified question text containing title, tags, cleaned problem text, and code69- optional parent-child indexing where chunk hits collapse back to parent question IDs70 71When Qdrant snapshots are included, they are the fastest restore path. The Parquet files remain the portable fallback for rebuilding the collection.72 73## Coverage74 75This repository contains only the C++ shard. A Stack Overflow question can appear in more than one language shard when it has multiple programming-language tags.76 77The dataset is not a complete Stack Overflow mirror. Full question and answer graph context lives in the corresponding Stack2Graph knowledge graph artifacts; the vector dataset stores retrieval rows, vectors, sparse weights, and routing metadata.78 79## Recommended Use80 81Use this dataset for semantic and hybrid retrieval, Qdrant restore or ingestion tests, Stack2Graph RAG experiments, graph entry-point retrieval before QLever graph expansion, and vector database benchmarking.82 83This dataset is not intended as a standalone supervised training dataset.84 85## Restore With Stack2Graph86 87You do not need to regenerate embeddings to use this dataset.88 89Typical workflow:90 911. Clone Stack2Graph and configure `.env` with service paths and an HF token.922. Start local services:93 94```bash95docker compose up -d96```97 983. Download and restore vector artifacts:99 100```bash101python -m experiment.sources.hf --skip-kg102```103 104The loader restores Qdrant snapshots when present and falls back to Parquet ingestion when snapshots are unavailable.105 106## Manual Use107 108For manual use, inspect `dataset_manifest.json`, then either restore the Qdrant snapshot artifacts under `qdrant_snapshots/` or ingest the listed Parquet shards into a compatible vector database.109 110The target Qdrant collection name is:111 112```text113stackoverflow_c++_vector114```115 116## Quality Notes117 118- Embedding quality depends on the configured Stack2Graph export pipeline and model versions.119- Sparse text can be generated from the same unified text as dense embeddings or from a lexical variant, depending on export configuration.120- Community-generated Stack Overflow content may contain errors, outdated information, bias, or incomplete answers.121- Rebuilding from Parquet may produce operational differences if Qdrant collection settings differ from the original export.122 123## License And Attribution124 125This dataset is derived from Stack Overflow content and is distributed under `CC-BY-SA-4.0`.126Preserve required attribution and license notices when redistributing derived artifacts.127 128## Citation129 130If you use this dataset, cite the Stack2Graph project and preserve Stack Overflow attribution requirements:131 132- Kleybolte, L. A., Ventura, V., & Zarcone, A. (2026). Stack2Graph: A Structured Knowledge Representation of Stack Overflow Data for Retrieval-based Question Answering . In Proceedings of the Knowledge Graphs and Large Language Models Workshop (KG-LLM) @ LREC26 (pp. 120-133). European Language Resources Association (ELRA). https://doi.org/10.63317/583q4d3mwbdu.133