Load the preprocessed Rotten Tomatoes dataset. This version can be used as a simple matrix-like pool.

This dataset is best suited for text classification.

The training dataset contains 32712 objects. Each object is described by 22 columns of numerical, categorical and text features. The label column is not precisely specified. The dataset contains movie reviews. Every object in the dataset represents a unique review from a user and movie-specific information (such as synopsis, rating, etc.).

The validation dataset contains 8179 objects. The structure is identical to the training dataset.

Method call format


Type of return value

A two pandas.DataFrame tuple (for train and validation datasets).

Usage examples

from catboost.datasets import rotten_tomatoes

rotten_tomatoes_train, rotten_tomatoes_test = rotten_tomatoes()


The output of this example:

       id                                           synopsis rating_MPAA                                              genre  ...                 publisher        date    date_int rating_10
0   830.0  A gay New Yorker stages a marriage of convenie...           R  Art House and International | Comedy | Drama |...  ...  Las Vegas Review-Journal  2004-04-16  20040416.0       8.0
1  1161.0  Screenwriter Nimrod Antal makes an impressive ...           R  Action and Adventure | Art House and Internati...  ...                 E! Online  2005-04-22  20050422.0       6.0
2   596.0  "Arctic Tale" is an epic adventure that explor...           G                     Documentary | Special Interest  ...       New York Daily News  2007-07-27  20070727.0       6.0

[3 rows x 22 columns]