Human-media interaction

Multimedia is a rich field for studying interactions at different scales: how different modalities of the content (i.e., video, audio and language use) interact with each other to provide rich storytelling; how the content in the movies and videos convey message to the viewers, and how it influences the viewers' perception; and how viewers and consumers interact with the media content: proximally (e.g., movie ratings, box office ratings), perceptually (affect, user reaction, perceived sentiment) and long-term (e.g., influence of movie on portrayal on certain stereotypes)

Media is said to also mirror the society we live in, but the reverse is often equally true. Research in different forms of media processing have been around for a while, but they have usually only focused on the low and mid-level tasks, such as indexing, summarization, without making an attempt to close the loop with any impact metrics on the audience. With the advent of big data processing, however, the field of computational media research is now gaining steam, and significant efforts are being put to study the affective content, societal impact and trends in media data. We believe that this workshop comes at a critical time when interest in human-centric media analytics is on the rise.


The first Media Analytics for Societal Trends workshop was organized in 2018 at San Diego in conjunction with the International Conference of Multimedia and Expo (ICME). The workshop brought together engineers and computer scientists from academia and industry with social scientists and media experts. The workshop attracted a large audience and showcased cutting edge research on different aspects of media analytics with an attempt to understand its impact on different societal patterns. The speakers and panels addressed the open challenges and future directions of computational media intelligence.

Workshop details

Abstract submission

The workshop will include presentations for submitted abstracts, invited talks and panel discussions. Submitted abstracts can be upto 2 pages long but authors are also encouraged to submit more details if available at the time of submission.

Workshop program

  • Workshop date : Oct. 14, 2019
  • Venue : Suzhou, China
  • Time : 2:00 - 5:00 pm

Time Topic
1400 - 1420 Introduction
1420 - 1455 Invited Talk 1
1455- 1515 Workshop Paper 1
1515 - 1545 Coffee Break
1545 - 1615 Invited Talk 2
1615 - 1635 Workshop paper 2
1635 - 1655 Workshop paper 3
1655 - 1700 Closing comments

Workshop papers: abstracts
  • “Victim or Perpetrator Analysis of Violent Character Portrayals from Movie Scripts” , Victor Martinez et al., University of Southern California
  • “Multi-queries based Attention for Facial Emotion Recognition”, Wan Ding et al., Tianjin Normal University
  • "RWF-2000: A Large Video Database for Violence Detection", Ming Cheng et al., Duke Kunshan University

Invited talks
  • "Understanding emotions and sentiment in multimedia", Mohammad Soleymani
  • "Multimodal Behavioral Analytics: What Hand Movemenets Reveal about Domain Expertise", Sharon Oviatt

Workshop co-organizers

Program Committee Members