LitStory-Dataset-v1.0.0

# LitStory Dataset v1.0.0 [](https://doi.org/10.5281/zenodo.XXXXXXX) [](https://creativecommons.org/licenses/by/4.0/) > Companion dataset for the manuscript: > Sheng N., Zhang B., Han Y. **Fine-Grained Narrative Unit Recombination for Coherent Literary Text-to-Animation Storyboard Generation.** Submitted to *The Visual Computer*, 2026. --- ## What's in this archive This archive contains the **LitStory benchmark dataset** — 2,500 literary excerpts from 99 source works across four genres, professionally annotated with: - Story atom decompositions (four-type taxonomy) - Temporal relationship graphs - Shot-level storyboard sequences (47,600 total frame annotations) - Emotion labels (28-tag controlled vocabulary) - Visual style labels (20-style controlled vocabulary) Inter-annotator agreement: **Fleiss' κ = 0.82** (substantial agreement). ## Contents (55 files / 57 MB) ``` LitStory-Dataset/ ├── README.md ← This file ├── LICENSE ← CC-BY-4.0 ├── metadata.xlsx ← Excel workbook (8 sheets, ~1.8 MB) │ ├── texts.jsonl ← 2,500 literary excerpts (~5 MB) ├── annotations.jsonl ← Full annotations (~50 MB) │ ├── splits/ ← 5-fold cross-validation │ ├── README.md │ ├── fold_1.json │ ├── fold_2.json │ ├── fold_3.json │ ├── fold_4.json │ └── fold_5.json │ ├── storyboards/ ← Representative storyboards (40 frames) │ ├── LS_CL_00005/ ← Classic Literature sample (8 frames) │ ├── LS_FA_00660/ ← Fantasy Fiction sample (8 frames) │ ├── LS_CH_01253/ ← Children's Stories sample (8 frames) │ ├── LS_HI_02013/ ← Historical Narratives sample (8 frames) │ └── LS_CL_00102/ ← Additional Classic sample (8 frames) │ └── examples/ ← Preview images ├── dataset_overview.png ← Statistical overview figure └── example_LS_*.png ← Full storyboard previews (5 images) ``` ## How to use with the LitStory code 1. Download the LitStory code repository: ```bash git clone https://github.com/a18907052542-lang/Achieving-Narrative-Coherence-from-Literary-Texts.git cd Achieving-Narrative-Coherence-from-Literary-Texts ``` 2. Extract this dataset archive into the code repository's `data/` directory: ```bash unzip LitStory-Dataset-v1.0.0.zip mv LitStory-Dataset/* data/ ``` 3. Verify: ```bash python -c " import json n = sum(1 for _ in open('data/texts.jsonl', encoding='utf-8')) print(f'Loaded {n} excerpts (expected: 2500)') assert n == 2500 " ``` ## Dataset statistics | Aspect | Value | |---|---| | Excerpts | 2,500 | | Source works | 99 | | Genres | 4 (Classic Lit / Fantasy / Children's / Historical) | | Genre distribution | 650 / 600 / 625 / 625 | | Original languages | 12 | | Avg word count | 351.9 ± 84.7 | | Avg atoms per excerpt | 19.6 ± 3.8 | | Total frame annotations | ~47,600 | | Annotators | 3 (≥5 years industry experience each) | | Fleiss' κ | 0.82 | | Adjudication rate | 12.2% | ## License The annotations are released under **CC-BY-4.0**. The source texts are treated in two tiers: - **Public-domain works** (e.g., *Pride and Prejudice*, *Dream of the Red Chamber*): full text included with `content_type = "public_domain_passage"`. - **Copyrighted modern works** (e.g., *Harry Potter*, *The Little Prince*): released as `content_type = "academic_locator"` with only chapter/section references and opening previews of <15 words. Researchers must obtain full text from licensed editions. See the `LICENSE` file for full terms. ## Citation If you use this dataset, please cite: ```bibtex @article{sheng2026litstory, title = {Fine-Grained Narrative Unit Recombination for Coherent Literary Text-to-Animation Storyboard Generation}, author = {Sheng, Nan and Zhang, Bo and Han, Yeteng}, journal = {The Visual Computer}, year = {2026}, note = {Companion dataset: \url{https://doi.org/10.5281/zenodo.XXXXXXX}} } ``` ## Contact **Bo Zhang** (corresponding author) School of Fine Arts and Design, Huainan Normal University Email: [e-mail]