/soft-attributes

This dataset consists of sets of movie titles, with each set annotated with a single English soft attribute (such as 'confusing' or 'romantic') and a reference movie. For each set, a rater has placed the movies into three sets: more, equally, and less <attribute> than the reference movie. From these sets, approximately 250,000 pairwise preferences can be derived for 60 distinct soft attributes studied.

SoftAttributes: Relative movie attribute dataset for soft attributes

COPYRIGHT NOTICE

This is the work of Krisztian Balog, Filip Radlinski and Alexandros Karatzoglou from Google LLC, made available under the Creative Commons Attribution 4.0 License. A full copy of the license can be found at https://creativecommons.org/licenses/by/4.0/

DATA COLLECTION METHODOLOGY

The dataset consists of sets of movie titles, with each set annotated with a single English soft attribute (subjective descriptive property, such as 'confusing' or 'romantic') and a reference movie. For each set, a crowd worker has placed the movies into three sets: more, equally, and less than the reference movie. There are 5,991 such sets, from which one can infer approximately 250,000 pairwise preferences over movies for the 60 distinct soft attributes studied.

A full description of the data, methodology, and analyses as well as user instructions can be found in the associated research paper, which can be cited as:

@inproceedings{balog-et-al-2021-soft-attributes,
  title = {On Interpretation and Measurement of Soft Attributes for Recommendation},
  author = {Krisztian Balog and Filip Radlinski and Alexandros Karatzoglou},
  booktitle = {Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval ({SIGIR})},
  year = 2021
}

EXPLANATION OF DATA FILES

The dataset is provided a CSV format in the file soft-attributes.csv.

Each record consists of a categorization of up to 10 movie titles by a single rater with regards to a single soft attribute relative to an anchor movie. It has the following fields:

  • rater_id: A unique integer id for each rater from 1 to 100.
  • reference_title: The title of the movie with respect to which the ratings are made.
  • soft_attribute: The soft attribute with respect to which the ratings are made.
  • less_than: List of movie titles judged less <soft_attribute> than <reference_title>
  • about_as: List of movie titles judged about the same <soft_attribute> as <reference_title>
  • more_than: List of movie titles judged more <soft_attribute> than <reference_title>

Further details about the data are provided in Data-Description.pdf