Verification package for "What the Corpus Still Holds": scripts, outputs and dated design documents

This package contains everything behind the paper What the Corpus Still Holds: Testing Keys, Survival and Result Stability in Deposited YouTube Comment Corpora (https://doi.org/10.5281/zenodo.23805575) – scripts/ All collection and analysis scripts (Python 3), for three deposits: Chan et al. (2026), DIVERSE and the IMSyPP English YouTube Hate Speech Corpus; plus the script for Figure 1. – outputs/ The outputs reported in the paper: presence bits, counts, survival and result tables. – design/ The freeze and its addenda A1–A11, each written before the step it governs (A11 records findings made while building this package), and the H1′ freeze with its feasibility check. – data/ Video and comment IDs of DIVERSE and IMSyPP, in deposit order. IDs only. – README.md Maps every table and the figure to its script and output, and explains how to rerun each step. Every offline step was rerun on 30 October 2026 from the original deposits and gave byte-identical outputs. The online steps need your own YouTube Data API key and describe the platform on the day they are run. No comment text, author name or author ID is contained anywhere in the package. 10.5281/zenodo.23805575 Is supplement to Publication 10.5281/zenodo.21146055 Is derived from Dataset 10.5281/zenodo.10493803 Is derived from Dataset http://hdl.handle.net/11356/1454 Is derived from Dataset Contact: [e-mail]