PS I love you: Privacy aware sentiment classification

Hugo Alatrista-Salas, Hugo Cordero, Miguel Nunez-Del-Prado

Research output: Contribution to journalArticle in a journalpeer-review

Abstract

At first glance, one might think that people are aware of the availability of comments or posts on social networks. Therefore, one may believe that people do not share sensitive information on public social networks. Nonetheless, people's posts sometimes reveal susceptible information. These posts include mentions the use of drugs or alcohol, sexual preferences, intimate confessions and even serious medical conditions like cancer or HIV. Such privacy leaks could cost someone to get fired or even worse to be a victim of denial insurance or bad credit evaluations. In this paper, we propose a complete process to perform a privacy-preserving sentiment analysis trough Bloom filters. Our approach shows an accuracy difference between 1% and 3% less than their classic sentiment analysis task counter part while guarantying a private aware analysis.

Original languageEnglish
Pages (from-to)1507-1515
Number of pages9
JournalComputacion y Sistemas
Volume23
Issue number4
DOIs
StatePublished - 1 Jan 2019

Keywords

  • Bloom filter
  • Disclosure risk
  • Information loss
  • Privacy
  • Sentiment analysis

Fingerprint

Dive into the research topics of 'PS I love you: Privacy aware sentiment classification'. Together they form a unique fingerprint.

Cite this