Preserving Privacy of User Reviews in Recommender Systems
Keywords
Loading...
Authors
Issue Date
2022-09-27
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In this interdisciplinary thesis we combine the fields of natural language processing, recommender
systems, and privacy. People tend to disclose information about themselves in user reviews, which
does not benefit other users for deciding whether or not to consume an item. Such information does
harm the privacy of the user and an accumulation of texts written by the user has the potential to
de-anonymize a user. In this work, we have found that protecting the privacy of users by removing
sensitive mentions from reviews has no substantial effect on the performance of a standard contentbased
recommender model, both for the rating prediction task, and for top-N recommendation.
These findings lay the foundation of using user generated texts in a responsible manner, while
retaining effectively identical usability.
Description
Citation
Faculty
Faculteit der Sociale Wetenschappen