Multimedia Recommender Systems

This tutorial introduces multimedia recommender systems (MMRS), in particular, recommender systems that leverage multimedia content to recommend different media types. In contrast to the still most frequently adopted collaborative filtering approaches, we focus on content-based MMRS and on hybrids of collaborative filtering and content-based filtering. The target recommendation domains of the tutorial are movies, music and images. We present state-of-the-art approaches for multimedia feature extraction (text, audio, visual), including deep learning methods, and recommendation approaches tailored to the multimedia domain. Furthermore, by introducing common evaluation techniques, pointing to publicly available datasets specific to the multimedia domain, and discussing the grand challenges in MMRS research, this tutorial provides the audience with a profound introduction to MMRS and an inspiration to conduct further research.