Odysseus

UCD's ambitious Horizons initiative has heralded the introduction of full modularisation through a credit-based undergraduate programme. This in turn introduces a significant "choice problem" for new students. Students will be faced with a significant information overload problem when it comes to picking the right courses (electives) to strengthen or complement their core studies. These choices may have a significant impact on their lives going forward. Poor choices may limit their degree options or may lead to examination failures. This project seeks to apply automatic recommendation technologies with a view to developing a intelligent assistant that is capable of guide student's through this selection problem. The system will explore the use of content-based and collaborative approaches to recommendation and focus on different modes of feedback and preference elicitation that may be suited to this particular task and domain.

This project represents a challenging recommendation task involving the varied preferences and experiences of many different users with diverse objectives and learning styles. At the same time there are many sources of recommendation knowledge that might be brought to bear on the problem, from content-rich course descriptions to ratings-style student performance measures. A significant part of this project will focus on investigating the potential for these different sources of knowledge to play a guiding role in course recommendation. In addition, we will further focus on the different opportunities that exist to elicit feedback from individual users as well as the opportunities that exist to explain the reason behind a particular recommendation. Recommendation explanation is a particular important research topic in the field and the advantages of justifying or explaining particular recommended courses to students should be clear.

The successful candidate will focus on the development of a basic recommendation engine that is capable of integrating with the existing Horizons database (course information) to provide a form of content-based recommendation in the first instance. The project team will also seek to gain access to additional recommendation knowledge (such as historical student choices and possibly anonymised examination results for use in a collaborative recommendation layer.

In addition, the project will introduce the student to best practice in evaluation methodologies through a combination of artificial- and live-user trials with a view to understanding the potential benefits of the resulting recommender system.

Relevance of Project to the Host Laboratories:

Computer Science and Informatics at UCD hosts one of the leading research groups in the area of personalization and recommendation comprising a significant collection of academic staff and students bringing together expertise from artificial intelligence, information retrieval, user modelling and user interface design. In addition, the new UCD Horizon's initiative provides a unique opportunity to explore the use of this technology in this important domain of choice.

Supervisors:

Dr. Lorraine McGinty & Prof. Barry Smyth (AIC, UCD)

Keywords:

Recommender systems, personalization, e-learning.