Hugging Face's logo

Datasets:
Gsk068
/
JP-TH_Literary_Translation_URL_Alignment_Index

Modalities:
Text
Formats:
csv
Size:
< 1K
Libraries:
Datasets
pandas
Polars
Dataset card Data Studio Files
xet
 Community
JP-TH_Literary_Translation_URL_Alignment_Index / README.md
Gsk068's picture
Gsk068
Upload README.md
ba6463c verified
preview code
|
Raw
History Blame Contribute Delete
14.2 kB

JP–TH Literary Translation URL Alignment Index

This release provides a copyright-conscious metadata index and reproducibility package for a Japanese–Thai literary translation dataset associated with the study Context-Aware Prompting for Japanese–Thai Literary Translation in a Low-Resource Setting.

Overview

The release is designed to support reproducible academic research on Japanese–Thai literary machine translation, context-aware prompting, prompt engineering, and low-resource NLP. Due to copyright constraints, this public release provides URL alignments, synopsis/source-summary URLs, official train/test split definitions, prompt templates, and non-reconstructive verification metadata. It does not redistribute the underlying Japanese source text, Thai translation text, synopsis text, machine translation outputs, or post-edited outputs.

Release type

This release should be understood as a URL-alignment index and reproducibility package, not as a raw Japanese–Thai parallel corpus.

The release intentionally excludes:

  • Japanese source text
  • Thai translated text
  • Synopsis or plot-summary text
  • Machine translation outputs
  • Post-edited translations
  • Metric outputs containing translated text
  • Local/private source filenames

Instead, it provides metadata and reproducibility materials, including:

  • Work-level identifiers and titles
  • Japanese source URLs
  • Thai translation URLs
  • Synopsis/source-summary URLs used by the context-aware methods
  • Official work-level train/test split
  • Prompt templates used in the experiments
  • Character counts and SHA-256 hashes for verification
  • Dataset construction notebook/scripts
  • Sanity-check summaries

Repository contents 

jp_th_url_alignment_release/
├── README.md
├── data/
│   ├── manifest.csv
│   ├── work_metadata.csv
│   ├── synopsis_manifest.csv
│   └── splits/
│       ├── work_splits.csv
│       └── test_work_ids.csv
├── prompts/
│   ├── README.md
│   ├── prompt_manifest.csv
│   ├── prompt_templates.json
│   ├── post_edit_basic.txt
│   ├── post_edit_context.txt
│   ├── context_aware_translation.txt
│   ├── context_aware_translation_with_dictionary.txt
│   └── context_aware_translation_with_revised_dictionary.txt
├── checks/
│   ├── release_summary.json
│   ├── missing_jp_urls.csv
│   ├── missing_th_urls.csv
│   ├── duplicate_url_pairs.csv
│   ├── works_missing_synopsis_metadata.csv
│   └── synopsis_without_chapter_data.csv

The exact folder structure may vary slightly by release version, but the files above represent the intended public release components.

Data schema

data/manifest.csv

Chapter-level URL-alignment manifest. Each row represents one aligned Japanese source URL and Thai translation URL.

Column Description
work_id Numeric identifier of the literary work.
work_title Title of the work used for metadata identification.
row_id Unique row identifier generated for this release.
chapter_index_in_file Chapter/episode order within the internal work-level source file used during dataset construction. This is provided only as an ordering index, not as a public source filename.
jp_url URL of the Japanese source chapter/episode.
th_url URL of the corresponding Thai translation chapter/episode.
jp_domain Domain of the Japanese source URL.
th_domain Domain of the Thai translation URL.
split Official split assignment: train or test.
access_date Date on which the URL/text was accessed during dataset construction.
jp_char_count Character count of the Japanese source text used internally. The text itself is not released.
th_char_count Character count of the Thai translation text used internally. The text itself is not released.
jp_sha256 SHA-256 hash of the internally used Japanese source text for verification only.
th_sha256 SHA-256 hash of the internally used Thai translation text for verification only.

data/work_metadata.csv

Work-level metadata, including official split assignment and synopsis/source-summary URL metadata.

Column Description
work_id Numeric identifier of the literary work.
work_title Title of the work used for metadata identification.
split Official work-level split assignment: train or test.
row_count Number of chapter/episode rows included for the work in manifest.csv.
jp_synopsis_url URL associated with the Japanese synopsis or source-summary information, when available.
th_synopsis_url URL associated with the Thai synopsis, novel page, or source-summary information, when available.
jp_synopsis_domain Domain of the Japanese synopsis/source-summary URL.
th_synopsis_domain Domain of the Thai synopsis/source-summary URL.
has_synopsis_metadata Indicates whether synopsis/source-summary URL metadata is available for the work.
jp_synopsis_char_count Character count of the Japanese synopsis/source-summary text used internally, when available. The text itself is not released.
th_synopsis_char_count Character count of the Thai synopsis/source-summary text used internally, when available. The text itself is not released.
jp_synopsis_sha256 SHA-256 hash of the internally used Japanese synopsis/source-summary text for verification only.
th_synopsis_sha256 SHA-256 hash of the internally used Thai synopsis/source-summary text for verification only.

data/synopsis_manifest.csv

Work-level synopsis/source-summary URL manifest. This file documents the synopsis-related URL inputs used by the context-aware methods.

Column Description
work_id Numeric identifier of the literary work.
work_title Title of the work used for metadata identification.
jp_synopsis_url URL associated with the Japanese synopsis or source-summary information.
th_synopsis_url URL associated with the Thai synopsis, novel page, or source-summary information.
jp_synopsis_domain Domain of the Japanese synopsis/source-summary URL.
th_synopsis_domain Domain of the Thai synopsis/source-summary URL.
split Official work-level split assignment: train or test.
access_date Date on which the synopsis/source-summary URL/text was accessed during dataset construction.
jp_synopsis_char_count Character count of the Japanese synopsis/source-summary text used internally, when available. The text itself is not released.
th_synopsis_char_count Character count of the Thai synopsis/source-summary text used internally, when available. The text itself is not released.
jp_synopsis_sha256 SHA-256 hash of the internally used Japanese synopsis/source-summary text for verification only.
th_synopsis_sha256 SHA-256 hash of the internally used Thai synopsis/source-summary text for verification only.

data/splits/work_splits.csv

Official work-level split file.

Column Description
work_id Numeric identifier of the literary work.
work_title Title of the work used for metadata identification.
split Official split assignment: train or test.

data/splits/test_work_ids.csv

List of works reserved for the test set.

Column Description
work_id Numeric identifier of the test-set work.
work_title Title of the test-set work used for metadata identification.

prompts/prompt_manifest.csv

Metadata for the released prompt templates.

Column Description
prompt_id Short identifier for the prompt template.
file_name Name of the corresponding .txt prompt template file.
prompt_name Human-readable prompt name.
purpose Intended role of the prompt in the experiment.
input_placeholders Placeholder variables used by the prompt template.
expected_output_format Expected response format from the language model.
source_pages_in_uploaded_pdf Page reference in the internal prompt list document used during release preparation.

Split policy

The dataset uses a work-level split. All chapters or rows from the same work are assigned to the same split to reduce leakage of content, style, terminology, and character names between training and evaluation.

The official test set contains the following work IDs:

10215, 10896, 13035, 13141, 13624, 13878, 14587, 15354, 3981, 4052,
4500, 5423, 5460, 6534, 6710, 6937, 7252, 7441, 9277, 9548

All other works are assigned to the training set.

Context-aware translation metadata

The context-aware methods used plot summaries or synopses as part of the prompt input. For reproducibility, this release includes the work-level URLs from which synopsis/source-summary information was obtained or associated.

The actual synopsis or plot-summary text is not redistributed.

Prompt templates

The prompts/ directory contains the prompt templates used in the experiments, including:

  • Basic post-editing prompt
  • Context-aware post-editing prompt
  • Context-aware translation prompt
  • Context-aware translation prompt with reference dictionary and named entity extraction
  • Revised dictionary-based context-aware translation prompt with stricter named entity extraction rules

The prompt templates contain placeholders only and do not include copyrighted Japanese or Thai text.

Intended use

This release is intended for academic research and educational purposes only, including:

  • Low-resource machine translation research
  • Japanese–Thai literary translation studies
  • Context-aware prompting research
  • Prompt engineering research
  • Reproducibility studies in NLP
  • Educational demonstrations of dataset documentation and translation-resource construction

This release is not intended for:

  • Commercial redistribution of copyrighted literary content
  • Unauthorized reproduction or republication of Japanese source text, Thai translation text, or synopsis text
  • Automated mass collection or scraping that violates the terms of the linked source websites
  • Use that infringes copyright or other intellectual property rights
  • Misrepresentation of this release as a complete open parallel corpus

User responsibility and copyright notice

This release provides metadata, URLs, prompt templates, split definitions, and non-reconstructive verification metadata only.

Users are solely responsible for ensuring that their access to, collection of, reconstruction of, or use of the linked source materials complies with applicable laws, copyright regulations, and the terms of the original websites.

The authors do not own, redistribute, or grant rights to the underlying Japanese source texts, Thai translations, or synopsis texts. The presence of a URL in this release does not imply permission to copy, redistribute, or commercially use the linked content.

The authors are not responsible for misuse of this release or for any use of linked third-party content that violates applicable laws, website terms, or intellectual property rights.

License

The metadata, prompt templates, split definitions, scripts, and documentation included in this release are licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0).

This license permits reuse, adaptation, and redistribution of the released materials for non-commercial purposes, provided appropriate credit is given.

This license applies only to the materials included in this release. It does not apply to the underlying Japanese source texts, Thai translations, synopsis texts, or any other third-party content linked by URL. The authors do not own, redistribute, or grant reuse rights to those linked third-party materials.

Limitations

  • This release does not include the raw Japanese source text, Thai translation text, synopsis text, machine translation outputs, or post-edited outputs.
  • Some URLs may become unavailable, modified, moved, or removed after the access date.
  • The release does not guarantee that users can reconstruct the exact internally used text if the source websites change.
  • Character counts and hashes are provided only for verification against the internally used text version.
  • The work-level split reduces leakage across chapters from the same work, but genre conventions, recurring tropes, source-platform style, or translator style may still be shared across works.
  • Synopsis/source-summary URLs are provided for reproducibility of the context-aware setup, but the synopsis text itself is not included.
  • This release does not provide permission to redistribute the linked literary works, translations, or synopses.

Citation

If you use this dataset index, prompt templates, split definitions, or associated reproducibility materials, please cite the associated paper:

@INPROCEEDINGS{11597007,
  author={Sikkhamarn, Korawit and Vittayakorn, Sirion},
  booktitle={2026 23rd International Joint Conference on Computer Science and Software Engineering (JCSSE)}, 
  title={Context-Aware Prompting for Japanese--Thai Literary Translation in a Low-Resource Setting}, 
  year={2026},
  volume={},
  number={},
  pages={576--581},
  keywords={Modeling;Translation;Context-aware services;Machining;Large language models;Machine translation;Printing;Dictionaries;Conferences;Measurement;Japanese--Thai Literary Translation;Large Language Models;Prompt Engineering;Low-Resource Natural Language Processing;Parallel Corpus},
  doi={10.1109/JCSSE68839.2026.11597007}
}

Version information

  • Created date: 2026-07-09
  • Access date: 2026-07-09
  • Verification metadata included: True
  • Release status: draft/publication-preparation version

Contact

For questions about this dataset release, please contact:

Korawit Sikkhamarn
Master of Science in Information Technology
King Mongkut’s Institute of Technology Ladkrabang, Bangkok, Thailand
67076002@kmitl.ac.th