Japanese Web Novels with 23 Million Records

23,212,809 records of Japanese web novels, containing over 80.8 billion characters and 55.4 billion tokens. The dataset was personally collected by OmniAICreator and last updated in September 2025.

Use Cases

• Fine-tune Japanese language models on the 55.4 billion tokens of text for text generation tasks.
• Train text classification models using the Japanese language content of the 23 million web novel records.
• Analyze linguistic patterns and character usage across the 80.8 billion characters of Japanese text.

Strengths

• Contains 23,212,809 records, providing a substantial corpus for NLP tasks.
• Includes 55.4 billion tokens, offering extensive text for language model training.
• Focuses on Japanese web novels, providing a specific domain of text data.

Limitations

• Column structure and specific features per record are unknown, limiting structured analysis.
• Access is restricted to machine-learning purposes and requires a detailed request, which may delay use.
• The source and collection methodology for the 23 million web novels are not detailed, raising questions about data provenance.

Provenance

Source

Personally collected by OmniAICreator from Japanese web novels.

Collection Method

Personal collection; specific methodology not detailed.

Freshness

Last updated on 2025-09-22.