VietJobs-37K: A Vietnamese Multi-Label Occupation Classification Dataset

VietJobs-37K is a Vietnamese-dominant multi-label occupation-classification dataset containing 37,274 job postings under a 60-leaf, 12-macro-group occupation taxonomy based on Vietnam Decision No. 34/2020/QD-TTg. The release includes an author-constructed descriptive crosswalk to ISCO-08; independent expert validation of that crosswalk is not claimed. This privacy-corrected release includes fixed train/dev/test splits, taxonomy files, a 1,000-record manually reviewed Gold-1000 evaluation anchor, a 500-record blind double-annotation subset artifact, validation summaries, dataset-side verification and audit scripts, README, DATASHEET, and machine-verifiable manifests. Record-local company aliases and direct contact identifiers are masked in model-facing text while record IDs, label sets, split membership, and order are preserved. The package supports peer review and reproducibility of the accompanying article published in IEEE Access. No separate public source-code repository is provided; the required verification and audit scripts are bundled with this release. Redistribution and downstream use remain subject to source-platform terms and applicable Vietnamese personal-data regulations.