README.md

metadata

dataset_info:
  features:
    - name: short_book_title
      dtype: string
    - name: publication_date
      dtype: int32
    - name: url
      dtype: string
    - name: text
      dtype: string
  splits:
    - name: train
      num_examples: 28602
    - name: validation
      num_examples: 50
    - name: test
      num_examples: 100
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*.jsonl
      - split: validation
        path: data/validation-*.jsonl
      - split: test
        path: data/test-*.jsonl

PG19 JSONL

This dataset is a JSONL conversion of the Hugging Face dataset emozilla/pg19. The source dataset is a parquet version of pg19.

Dataset Structure

The dataset has three splits:

• train: 28,602 rows in 23 JSONL shards
• validation: 50 rows in 1 JSONL shard
• test: 100 rows in 1 JSONL shard

Each line is a JSON object with the same fields as the source parquet dataset:

• short_book_title: short title for the book
• publication_date: publication year
• url: Project Gutenberg URL
• text: book text

Source

Converted from emozilla/pg19. For upstream dataset-building details, curation notes, and licensing context, see the source dataset card and the original pg19 dataset.