Defense (PhD, Masters, Pre)
PhD DefenseToward a Fairer Information Retrieval System |
|
||
Monday, June 15, 2020, 01:00pm - 03:00pm |
|||
Speaker: Ruoyuan Gao
Location : Remote via Webex
Committee:
Prof. Chirag Shah (Chair)
Prof. Yongfeng Zhang
Prof. Geral de Melo
Dr. Fernando Diaz (External Member, Microsoft Research Montréal)
Event Type: PhD Defense
Abstract: With the increasing popularity and social influence of information retrieval (IR) systems, various studies have raised concerns on the presence of bias in IR and the social responsibilities of IR systems. This dissertation explored ways to investigate, understand and evaluate fairness-aware IR. This dissertation makes the following contributions: (1) An investigation of existing bias presented in search engine results. Several topical diversity fairness ranking strategies were explored to understand the relationship between relevance and fairness in search results. The experimental results showed that different fairness ranking strategies result in distinct utility scores and may perform differently with distinct datasets. (2) A statistical framework to analyze the relationship between data and fairness algorithms. A series of use cases were presented to demonstrate how this framework could be applied to provide insights to fairness optimization problems. (3) A novel evaluation metric situated in fairness-aware ranking and an effective ranking algorithm that jointly optimized fairness and utility. The experiments demonstrated that the proposed metric and algorithm were able to capture multiple aspects of the ranking and improve fairness while closely approximating the ideal utility.
: