CoolFace
Datasetpublic

VItaldob/viciebski-pralietaryj-yiddish

Viciebski Pralietaryj — Yiddish blocks Blocks of newspaper text set in Yiddish (Hebrew script), cut from scans of Viciebski pralietaryj («Віцебскі пралетарый»), a newspaper published in Vitebsk, Byelorussian SSR (Belarus), in 1930, 1931 and 1933. 2684 images from 78 pages across 78 issues. No transcriptions — this is a raw corpus for OCR/HTR work, not a labelled set. Belarusian blocks are present. The Yiddish pages ran as inserts inside a Belarusian newspaper, and blocks were… See the full description on the dataset page: https://huggingface.co/datasets/VItaldob/viciebski-pralietaryj-yiddish.

sourceHugging Faceotherupdated 16d agoView on Hugging Face
0likes471downloads
Dataset Card

Viciebski Pralietaryj — Yiddish blocks

Blocks of newspaper text set in Yiddish (Hebrew script), cut from scans of Viciebski pralietaryj («Віцебскі пралетарый»), a newspaper published in Vitebsk, Byelorussian SSR (Belarus), in 1930, 1931 and 1933.

2684 images from 78 pages across 78 issues. No transcriptions — this is a raw corpus for OCR/HTR work, not a labelled set.

[image]

Belarusian blocks are present. The Yiddish pages ran as inserts inside a Belarusian newspaper, and blocks were selected a page at a time rather than one by one, so mixed pages contributed their Cyrillic columns, mastheads and ads as well. Confirmed cases include full Belarusian columns on 1933-181 p2 and 1933-084 p3; more went unchecked. Filter on flagged if you need Yiddish only.

Note that flagged: true marks blocks the detector was sure about, but flagged: false does not mean Belarusian. Most often it is a headline too short to judge — the middle image above is one — and it is also where the Belarusian blocks sit. Read it as "script unknown", not as a language label.

Contents

data/ holds the images and metadata.jsonl:

fieldmeaning
file_nameimage file in data/
issueissue id, YEAR-NUMBER
year1930, 1931 or 1933
pagepage of the issue
blockblock index on the page
kindheader, column or ad
flaggedblock detected as Hebrew script
page_shareshare of flagged blocks on that page
width, heightpixels
python
from datasets import load_dataset
ds = load_dataset("VItaldob/viciebski-pralietaryj-yiddish", split="train")

Caveats

  • —Selection is automatic and was not proofread against the scans.
  • —Block types come from the segmentation step, not from a human.
  • —Whole Belarusian pages are not included, but individual Belarusian blocks are.

Rights

Issues from the 1930s are in the public domain. The scans come from a publicly available publication and are not the compiler's own; no rights are claimed over them, and none are added by the segmentation or the metadata.

Tagged license: other / public-domain rather than cc0-1.0: CC0 is a waiver of rights, and there are none left here to waive.

Developed as part of the Zhnivo initiative — a project building genealogical databases from Belarusian archival records.

Compiled 2026-09-09.