This is the is the Japanese segment of FineWeb which has been preprocessed to streamline the task of selecting educational content. More information is presented in https://github.com/Lone-Wolfgang/Distill-FineWeb2/tree/main.
The main file is preprocessed/jpn_Jpan.parquet. This DataFrame has the columns:
- domain: The domain from which webpages were collected.
- count: The number of pages that were collected from the domain.
- tld: The top-level domain from the domain column.
The intermediate files align with the raw data from the Japanese segment of FineWeb.
In the directory, intermediate/jpn_Jpan/stripped, the DataFrames have the columns:
- url: The URL from which text was collected.
- domain: The domain of the URL.
- id: The ID from the FineWeb corpus.
- date: The date when the page was collected by CommonCrawl.
In the directory, intermediate/jpn_Jpan/grouped, the DataFrames have the columns:
- domain: The domain from which webpages were collected.
- count: The number of pages which were collected from the domain.