This release corresponds to the version of the Code2Doc dataset curation pipeline
described in the accompanying research paper:

“Code2Doc: A Curated Dataset for High-Quality Code Documentation.”

Purpose

This release provides a frozen, reproducible implementation of the full dataset
curation pipeline, including:

• Repository-level data extraction
• Multi-stage heuristic filtering
• Documentation quality scoring
• Exact and near-duplicate removal
• Heuristic detection of potentially AI-generated documentation

All thresholds, heuristics, and design choices correspond exactly to those reported
in the paper.

Reproducibility

• All filtering thresholds are centralized and configurable
• Deduplication is deterministic
• No post-release changes should be assumed for this version

This release is archived via Zenodo and intended for use in reproducible research.
Future development will occur in subsequent versions.