Publication Date

12-2016

Date of Final Oral Examination (Defense)

8-20-2016

Type of Culminating Activity

Thesis

Degree Title

Master of Science in Computer Science

Department

Computer Science

Major Advisor

Maria Soledad Pera, Ph.D.

Advisor

Marissa Schmidt, MS

Advisor

Edoardo Serra, Ph.D.

Abstract

Curation is the act of selecting, organizing, and presenting content most often guided by professional or expert knowledge. While many popular applications have attempted to emulate this process by turning users into curators, we put an accent on a recommendation system which can leverage multiple data sources to accomplish the curation task. We introduce QBook, a recommender that acts as a personal docent by identifying and suggesting books tailored to the various preferences of each individual user. The goal of the designed system is to address several limitations often associated with recommenders in order to provide diverse and personalized book recommendations that can foster trust, effectiveness of the system, and improve the decision making process. QBook considers multiple perspectives, from analyzing user reviews, user historical data, and items' metadata, to considering experts' reviews and constantly evolving users' preferences, to enhance the recommendation process, as well as quality and usability of the suggestions. QBook pairs each generated suggestion with an explanation that (i) showcases why a particular book was recommended and (ii) helps users decide which items, among the ones recommended, will best suit their individual interests. Empirical studies conducted using the Amazon/LibraryThing benchmark corpus demonstrate the correctness of the proposed methodology and QBook's ability to outperform baseline and state-of-the-art methodologies for book recommendations.

Available for download on Thursday, December 20, 2018

Share

COinS