fetch_toxicity#

skrub.datasets.fetch_toxicity(data_home=None)[source]#

Fetch the toxicity dataset (classification) available at skrub-data/skrub-data-files

This is a balanced binary classification use-case, where the single table consists in only two columns:

  • text: the text of the comment

  • is_toxic: whether or not the comment is toxic

Size on disk: 220KB.

Parameters:
data_home: str or path, default=None

The directory where to download and unzip the files.

Returns:
bunchsklearn.utils.Bunch

A dictionary-like object with the following keys:

  • toxicity : pd.DataFrame, the dataframe. Shape: (1000, 2)

  • X : pd.DataFrame, features, i.e. the dataframe without the target labels. Shape: (1000, 1)

  • y : pd.DataFrame, target labels. Shape: (1000, 1)

  • metadata : a dictionary containing the name, description, source and target

  • path : str, the path to the toxicity CSV file