Search papers, labs, and topics across Lattice.
This study introduces NAICS-GH, a dataset of 6,588 GitHub repositories labeled according to the North American Industry Classification System (NAICS), addressing the lack of standardized industry mapping on GitHub. The authors employed a retrieve-and-verify pipeline that integrates embeddings, FAISS retrieval, and GPT-4.1 rubric scoring to ensure high-confidence labeling, achieving 96.98% precision on a human-validated sample. This work not only facilitates empirical research on innovation geography and open-source production but also provides a robust resource for analyzing technology diffusion across sectors.
NAICS-GH reveals that over 6,500 GitHub repositories can be accurately classified by industry, unlocking new avenues for research on innovation and technology diffusion.
GitHub hosts hundreds of millions of public repositories, but the platform exposes no native mapping from repositories to standardized industry sectors. This gap limits empirical work on the geography of innovation, the industrial composition of open-source production, and the diffusion of new technologies across economic sectors. We present NAICS-GH, a publicly released corpus of 6,588 GitHub repositories drawn from source pools covering the United States, the European Union, and Australia, each labeled with a 2-digit sector from the North American Industry Classification System (NAICS 2022). Labels are produced by a retrieve-and-verify pipeline that combines BAAI/bge-large-en embeddings, FAISS retrieval, and GPT-4.1 rubric scoring. The pipeline narrows about 1.37 million source repositories to 31,178 candidate repository-sector pairs and retains 6,588 high-confidence labels with score at least 8. Re-running the retrieval pipeline end to end reproduces the candidate set to within 0.03 percent. On a 2,421-repository human-validated random sample, the released labels attain 96.98 percent precision, with Wilson 95 percent confidence interval [96.23, 97.59]. We benchmark six pretrained encoders on the released corpus; RoBERTa-large reaches 86.45 percent F1 and 86.35 percent accuracy on a held-out 20 percent test set. The dataset, Croissant metadata, pipeline code, prompts, and fine-tuned checkpoint are released under CC-BY-4.0 and MIT licenses.
