Datasets:

tommasobonomo
/

ITAGutenberg

dataset_info:
  features:
    - name: gutenberg_id
      dtype: int64
    - name: title
      dtype: string
    - name: text
      dtype: string
    - name: tokenized_length
      dtype: int64
    - name: metadata
      struct:
        - name: authors
          sequence: string
        - name: bookshelves
          sequence: string
        - name: encoding
          dtype: string
        - name: languages
          sequence: string
        - name: subjects
          sequence: string
        - name: summaries
          sequence: 'null'
        - name: url
          dtype: string
  splits:
    - name: train
      num_bytes: 480844883
      num_examples: 1084
  download_size: 295576071
  dataset_size: 480844883
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
language:
  - it
tags:
  - project_gutenberg
  - text
pretty_name: ITAGutenberg

ITAGutenberg

A collection of all books written in Italian that appear on Project Gutenberg, meant for pretraining of Large Language Models. We collected the plaintext version of each book and lightly processed it to remove licensing text that is usually included in the original text of Project Gutenberg.

🚀 Quickstart

Simply download the dataset as you would with any other Hugging Face dataset:

from datasets import load_dataset

dataset = load_dataset("tommasobonomo/ITAGutenberg")

📊 Data Schema

The dataset is organized with the following schema:

gutenberg_id (str) — Project Gutenberg key identifying the book.
title (str) — Title of the book.
text (str) — Full text of the book from Project Gutenberg.
tokenized_length (int) — Length in tokens of the book. Computed through the cl100k_base encoding from openai/tiktoken.
metadata (dict) — Additional contextual information about the book, including:
- authors (list[str]) — Name of the book’s author(s).
- bookshelves (list[str]) — Name of bookshelves that include this book, as per Project Gutenberg.
- encoding (str) — Name of the encoding of the original book source.
- languages (list[str]) — Languages that appear in a specific book.
- summaries (list[str]) — Summaries collected from Project Gutenberg for a specific book.
- url (str) — URL from which the text of the book was downloaded.