Dataset from Language-Agnostic Modeling of Wikipedia Articles for Quality Assessment

By: Paramita Das , Isaac Johnson, Diego Sáez-Trumper and Pablo Aragón

This dataset was created with a modeling approach for predicting the quality of Wikipedia articles using language-agnostic features (further details are provided in the model card).

The CSV files in the features zipped folder contain the values of the language-agnostic features  of revisions through the end of 2022 for all articles in over 300 language editions of Wikipedia (enwiki.csv for English Wikipedia, dewiki.csv for German Wikipedia, frwiki for French Wikipedia, etc.). Columns are:

revision_id: Id of the revision (in the corresponding Wikipedia language edition).

page_id: Id of the page (in the corresponding Wikipedia language edition).

page_length:  Number of bytes of the revision.

num_refs:  Number of references of the revision.

num_wikilinks: Number of wikilinks of the revision.

num_categories: Number of categories of the revision.

num_media: Number of media files of the revision.

num_headings: Number of sections of the revision.

The CSV files n the scores zipped folder contain the predicted quality scores for the revisions of the above CSV files using the same folder structure. Columns are:

revision_id: Id of the revision (in the corresponding Wikipedia language edition).

page_id: Id of the page (in the corresponding Wikipedia language edition).

item_id: Id of the page in Wikidata.

pred_qual: Predicted quality score between 0 and 1.