The 1st International Workshop on Bridging the Gap Between Semantics and Multimedia Processing (SeMP 2019) workshop aims to bring together AI research in multimedia that can contribute to narrow or advance the understanding of the semantic gap that exists between multimedia signals and the concepts represented by these signals. This includes not only research that improves on traditional (symbolic or non-symbolic) AI techniques applied to multimedia, but also research that combines these techniques in novel, interesting, and experimental ways.
The workshop will provide a forum for researchers to present and discuss new ideas, work in progress, and preliminary results which can advance the state of current AI techniques and their interpretability applied to multimedia.
There is a big representational gap between the audiovisual signals that compose multimedia objects and the concepts represented by these signals. One way to narrow this gap, from the concepts side, is through symbolic AI. One can use logic to describe the multimedia content and then apply symbolic querying and reasoning to obtain relevant information from these descriptions. The problem with this purely symbolic approach is that it does not scale well.
The large volumes of multimedia data generated by current applications demand efficient methods of concept detection and annotation. This is where deep learning techniques usually come in—to narrow the gap from the signals side. Through deep learning one can learn correlations among the training dataset but overall it is not clear how these correlations map to high level semantic concepts, which brings many interpretability issues on current deep learning models. Additional challenges on both sides of the gap are posed by the diverse, multimodal, imprecise, and uncertain nature of multimedia data.
The SeMP 2019 workshop will be collocated with IEEE ISM 2019, which will take place at the Wyndham San Diego Bay Side in San Diego, CA, USA. See the IEEE ISM 2019 website for more information on the venue.