Hugging Face's logo

Datasets:
recursal
/
SCP-RECURSAL

Tasks:
Text Generation
Fill-Mask
Modalities:
Text
Formats:
json
Sub-tasks:
language-modeling
masked-language-modeling
Languages:
English
Size:
10K - 100K
Libraries:
Datasets
Dask
License:
Dataset card Viewer Files Community
1
SCP-RECURSAL / README.md
KaraKaraWitch's picture
KaraKaraWitch
Update README.md
940915f verified
preview code
|
raw
history blame
2.81 kB
metadata
annotations_creators:
  - no-annotation
language_creators:
  - crowdsourced
license:
  - cc-by-sa-3.0
task_categories:
  - text-generation
  - fill-mask
task_ids:
  - language-modeling
  - masked-language-modeling
source_datasets:
  - original
language:
  - en
configs:
  - config_name: default
    data_files:
      - split: final
        path: jsonl/*.jsonl
pretty_name: SCP-RECURSAL

Dataset Card for SCP-RECURSAL

Item ID: SCP-RECURSAL

Object Class: Safe

Special Containment Procedures:

SCP-Recursal is to be stored on the Huggingface (HF) Servers, with backup copies maintained on other secure servers. Access to SCP-Recursal is unrestricted, and it may be freely used and redistributed without authorization.

Description: SCP-Recursal is a dataset comprising ~[TBD]B Tokens (llama-2-7b-chat-tokenizer) / ~[TBD]B Tokens (RWKV Tokenizer) scraped from the SCP Wiki website (scp-wiki.wikidot.com). The dataset is curated by Darok (Erik), released by KaraKaraWitch and funded by Recursal.ai.

SCP-Recursal is a comprehensive collection of tales, series, and jokes from the SCP Wiki, spanning from 2008 to 2022. The dataset was created by scraping lists of tales for each year, fetching content from tale pages using BeautifulSoup selectors, and filtering out unnecessary information. The same process was applied to series scp-series-1 to scp-series-8, while the joke SCPs were extracted from a single URL (/joke-scps).

The cleaning process involved removing author credits, "More articles by" sections, "Hide list" buttons, ratings, and special characters until the next line. Unicode characters were also cleaned, and unnecessary text was removed. The resulting dataset is a cohesive collection of SCP-related content, ideal for language modeling tasks.

Experiments:

Experiment 1: "SCP-Recursal: The Ultimate Language Model"

  • Researchers used SCP-Recursal to train a language model, which was then tasked with generating a new SCP article.
  • Result: The model produced [DATA EXPUNGED]. Due to it's Memetic Effect, the computer that was in use was incinerated.
  • Conclusion: [DATA EXPUNGED]

Dataset Details

Dataset Description

It serves as a training resource for large language models and other NLP tasks. This card details the dataset's origin, content, and limitations.

  • Curated by: KaraKaraWitch
  • Funded by [optional]: Recursal.ai (I work there lol)
  • Shared by [optional]: KaraKaraWitch
  • Language(s) (NLP): Mainly English
  • License: cc-by-sa-3.0

SCP-RECURSAL was created under time constraints for the release of EagleX v1, and may contain suboptimal selection.

Supported Tasks and Leaderboards

Primarily used for language modeling

Languages

English. Mostly.