In this project, we will select a multimodal approach (RS or otherwise) and adapt it to be able to learn RS models using three types of data: item descriptions, item evaluation and user-item interaction. The proposed method will be tested on public benchmarks (possibly augmented with sensor data, if possible) and data from projects at FhP-AICOS. We will investigate if the combination of these different types of data may lead to better recommendations.
Development of a multimodal recommender system method.
Author: João Miguel Oliveira
Type: MSc thesis
Partner: FEUP – Faculdade de Engenharia da Universidade do Porto