metadata
license: cc-by-4.0
Text-Beautified-1.0
This dataset if for training small llms to fix poorly formatted text.
Info
This data is composed of 1,000 of eatch:
- Wikipedia HTML and it's plaintext counterpart
- Wikipedia MediaWiki text, and it's plaintext counterpart
- And 1,000 samples of
C4 200M Grammar Error Correction dataset(Obtained from kaggle)
Flaws
- The corrosponding wikipedia plaintext is most likely from very diffrent dates +-5 years appart