Abstract

A fundamental right of the users of computer applications is that they can know the privacy policies (PP) that such applications establish. It is particularly relevant that they know about the treatment that they accept regarding the use of their data. However, these PP are very extensive and written in administrative-legal and commercial language, which makes them difficult to read and understand. The aim of this paper is to automatically summarize the PPs of five social network applications (Facebook, Twitter, TikTok, Snapchat and Instagram) in spanish, through extractive and abstractive techniques. For this purpose, three representation approaches from Natural Language Processing are used, these are: Graph Analysis, TF-IDF and Gensim. Fifteen summaries were automatically generated and evaluated in order to measure the readability and relevance, by an expert in law, based on 20 questions prepared by a study of the University of Austin, Texas (Zaeem et al., 2018). Finally, based on a classification of each privacy policy according to different risk factors, the Gensim method is found to be the most suitable for the representation and summarization of the PP's. The PP of Snapchat is also identified as the application that best meets these risk factors.