metadata

annotations_creators:
  - no-annotation
language_creators:
  - crowdsourced
license:
  - cc-by-sa-4.0
task_categories:
  - text-generation
  - fill-mask
task_ids:
  - language-modeling
  - masked-language-modeling
source_datasets:
  - original
language:
  - en
pretty_name: Recursalberg

Dataset Card for Recursalberg

Waifu to catch your attention.

Dataset Details

Dataset Description

Recursalberg is a cleaned dataset of Project Gutenberg books. We downloaded all the publicly available Gutenberg books at the time and processed them.
Filtering to a total amount of tokens of ~XYZB (llama-2-7b-chat-tokenizer) / ~XYZ (RWKV Tokenizer) from primarily English language.

Curated by: KaraKaraWitch
Funded by [optional]: Recursal.ai (I work there lol)
Shared by [optional]: KaraKaraWitch
Language(s) (NLP): Primarily English
License: cc-by-sa-4.0

Dataset Sources [optional]

Source Data: gutenberg.org (see mirroring) (rclone download)

Processing

We performed the following downloading and processing steps to prepare Retenberg.

Use gutenberg's mirror rclone to download to a folder.
Index gutenberg with gutenberg_index.py (Gather all items to find html documents.)
Process all book indexes that has a html version.
- Remove Gutenberg pre html block, page numbers, html comments, table of contents
- Convert sections to markdown
- Clean new lines, standardize punctuations.
Save each html file into 1 jsonl file.

Data Keys

text (str): the book's text. converted to markdown.

Dataset Curators

KaraKaraWitch. (I typically hangout in PygmalionAI discord, sometimes EleutherAI. If something is wrong, @karakarawitch on discord.)

I'd be happy if you could spread the word and recommend this dataset for your use cases :)

Licensing Information

Complicated. Refer to gutenberg's license here,

We haven't seen any in-copyrighted books in our dataset so far so we assume it's safe for usage, unless otherwise.

Citation Information

@ONLINE{recursalberg,
  title         = {Recursalberg},
  author        = {KaraKaraWitch, recursal.ai},
  year          = {2023},
  howpublished  = {\url{TBD}},
}