This deposit is the research-data-and-methods companion to a study of diacritic reduction (šišana) and restoration (dešišavanje) in computer-mediated Serbo-Croatian (the sr/hr/bs macro-language). It documents how the research was conducted and provides reproducible results. Main findings: diacritic stripping is a discrete writing mode predicted by register and keyboard environment rather than low literacy; the restoration task has a 0.49% dictionary-plus-unigram ceiling with error concentrated in about 400 forms; and the apparent cross-language accuracy gap is an artifact of dictionary coverage, not language — the three varieties converge under a controlled protocol. The accompanying open-source engine, ioDiacritics (MIT), is a small, deterministic, fully on-device restorer (no model, no network, precision-first). The deposit includes the English and Serbo-Croatian companion documents, two supplementary-data documents (resolution-table statistics, task inventory, reliability passports, code-switching collision inventories, sociolinguistic measurements), figure data and captions, and data-provenance/licensing notes. All numbers are reproducible from the pinned source commit; bundled dictionaries carry their own source licenses (see NOTICE).
Ilya V. Osipov (Sun,) studied this question.