Eurong2/storydream-illustrated-stories

StoryDream — Illustrated Graded Stories (KO / EN)

11,680 illustrated story pages across 50 story/language combinations, each page pairing an image with its story text. Every story is available at three reading levels and split into three narrative parts.

Structure

{language}/{story_id}/level_{level}/{part_type}/page_{page_num}.jpg
metadata.csv

• —language — ko or en
• —story_id — original story id (25 stories per language)
• —level — 1, 2, 3 (reading difficulty; 1 is easiest)
• —part_type — intro, body, conclusion
• —page_num — 1-based page number, restarting within each part

Fields in metadata.csv

Loading

from datasets import load_dataset

ds = load_dataset("imagefolder", data_dir="path/to/repo", split="train")
print(ds[0]["image"], ds[0]["text"])

Known caveats

Read these before training on the pairs.

Korean text versions. The source spreadsheet contained two different Korean texts for every (story, level, part) slot, both pointing at the same image. The two are not paraphrases — they narrate different beats of the story, so at most one of them matches what is actually drawn. This release keeps the second version (text_version = v2) for all 2,602 Korean pages that have one. The image–text alignment on the Korean split should therefore be treated as unverified.

145 Korean pages have no text. These slots existed only in the discarded first version, so the image is included with an empty text and text_version = none.

34 pages reuse the previous page's image. The source data had no image for these pages, so the preceding page's image was carried forward. They are marked with borrowed_image = true. Five of them borrowed across a part boundary (the blank page was page 1 of its part), so their image belongs to the end of the previous part.

English split. All 8,933 English rows were unique with no version conflict; text_version is empty for them.

Provenance

Texts are level-adapted retellings of classic public-domain stories (e.g. Alice's Adventures in Wonderland, The Frog Prince); illustrations were generated for this dataset.