Cascading Ranking Pipelines for Sensitivity-Aware Search

Jack McKechnie

Abstract: Search engines are designed to make information accessible. However, some information should not be accessible, such as documents concerning citizenship applications or personal information. This sensitive information is often found interspersed with other potentially useful non-sensitive information. As such, collections containing sensitive information cannot be made searchable due to the risk of revealing sensitive information. The development of search engines capable of safely searching collections containing sensitive information to provide relevant and non-sensitive information would allow previously hidden collections to be made available. This work aims to develop sensitivity-aware search engines via two-stage cascading retrieval pipelines.

External IDs:dblp:conf/ecir/McKechnie24