Skip to content

Dataset · globis-university

Cleaned Japanese text corpus from Aozora Bunko library

globis-university/aozorabunko-clean

Public domain Japanese fiction and nonfiction literature corpus.

What's inside

The dataset contains between 10,000 and 100,000 cleaned Japanese texts collected from the Aozora Bunko online library, which publishes public domain works. The authors prepared the data in a machine learning-friendly format and published the parsing code on GitHub. The dataset is intended for text generation and classification tasks.

What you can do with it

  • Fine-tuning language models or LoRA on classic and literary Japanese language
  • Training text classification and Japanese literature style analysis models
Tasks
Text generation, Text classification
Languages
Japanese
Size
10K–100K rows
Downloads this month
1.5K
Likes
49

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("globis-university/aozorabunko-clean")