PhD Position on Latent Multimodal Concept Analysis

Are you interested in doing cutting-edge research at the intersection of artificial intelligence, culture, and media? We are looking for a PhD candidate to investigate how to discover and analyse latent multimodal concepts found in multimedia.

The meaning of multimedia objects is influenced by the interactions between its constituent modalities. An image of a polar bear with a caption describing its hunting behaviour has a different meaning than an image of a polar bear next to an article on climate change. By modelling recurring patterns of interactions between modalities in large-scale datasets it becomes possible to discover latent concepts, thereby enabling analysis of their use and meaning. The aim of the project is to develop improved multimodal models for real world data by using new learning paradigms (e.g., contrastive, interactive, or self-supervised learning).

The position is at the Informatics Institute of the University of Amsterdam (UvA), in the Multimedia Analytics  (MultiX) lab, with close ties to the Cultural AI lab. For this position you will be supervised by dr Nanne van Noord.

We strongly encourage applications coming from a unique perspective. Tell us how your background fits with the aims of the lab.

What are you going to do

You will carry out research in the areas of Multimedia Analysis, Deep Learning, and Cultural AI in the MultiX lab. Through this research you will develop new methods for discovering and analysing latent multimodal concepts in real world data

You are expected to:

• develop Multimedia Analysis methods to discover and analyse latent multimodal concepts;
• collaborate with researchers in the MultiX and Cultural AI labs, and UvA;
• present intermediate research results at conferences and workshops, and publishing in proceedings and journals;
• pursue and compete a PhD thesis within the appointment duration of four years;
• assist in teaching activities.