About the project
verba is an open, unified corpus of Ukrainian proverbs, adages, and sayings, compiled from five digitized collections spanning 1841 to the present.
48 787 записів; кожен — із сучасним написанням і тематичними мітками, із пошуком за словом, темою, джерелом і за змістом.
What this is
For decades, Ukrainian paremiology remained scattered across individual printed collections with varying spelling conventions and no shared machine-readable format. verba unifies these collections into one corpus, making it searchable, researchable, and reusable.
Each entry includes: the original text (preserving the source's historical spelling), a modern spelling, 1–3 themes from a fixed set of 27, the source and reference, and a variant group (linking probable dialectal equivalents).
Sources
How it was made
Historical editions were digitized; OCR (tesseract) was applied to scans, while Nomys material was extracted from a critical edition. Duplicates were soft-merged, and probable dialectal variants were linked rather than removed. Texts are presented as close to the sources as possible.
Honest limitations: the Nomys collection (1864) represents the best possible OCR (≈75–80% character accuracy), so occasional recognition errors occur. Modern spellings and themes were AI-generated and are approximate (modern spelling ≈95% acceptable, themes ≈85%). The primary theme tag is most reliable.
In the data itself, typography is reduced to plain ASCII (straight quotes, apostrophe, hyphen) for easy programmatic processing; Ukrainian typography (« », em dash, ') is applied only for display — on the website and cards.
How to use
- Web app — search, filters by theme and source, semantic search "by meaning", Browse mode, and saved proverbs.
- REST API (
/api/v1) — JSON · JSONL · XML · CSV · TSV; see the API documentation. - Download — the full corpus in every format, along with a data card and Croissant metadata, in GitHub releases and as a dataset on Hugging Face.
License and citation
Curation and enrichment (corpus structure, modern spellings, themes, cleaned annotations, variant groups) — CC BY 4.0. Historical texts (1841–1909) — public domain. Modern collections (Bobkova; Mlodzynskyi, 2009) remain under publisher rights; included for research and education, with attribution via the sources field.
How to cite:
Yemelianov, Dmytro (2026). verba — Ukrainian Proverbs Corpus (v1.0.2). https://verbacorpus.org · ORCID 0009-0002-9244-7426
Support the project
verba is a non-commercial, open-data project. The site and bot are free; donations fund its growth: native-speaker audio for the proverbs, new sources, and dialectal variants. Thank you for helping keep the living word alive 🌾