Workshop on Multimedia Analytics for Societal Trends (MAST)


A workshop at ICME 2018

The creation, storage and organization, distribution, and consumption of multimedia content have evolved rapidly in the past decade. Large media industries, such as film or news industries have witnessed huge changes in every step, starting from how new content is conceived to how it is eventually shared with the viewers. This has led to rapid democratization of the field inviting content-makers from all spheres, and an unprecedented growth in the consumption of multimedia content via both modern and traditional mediums. This widespread reach of media has extended beyond movies and ads to internet-based plat- forms that share user-generated images and videos. While automated analysis is indispensable for traditional multimedia areas i.e. navigating, indexing and organizing diverse and vast media databases, more recently, an emerging trend in this area has been to improve and facilitate per- sonal and social activities, insight generation, and interaction experience. Research effort has been directed towards developing computational tools and methodologies for systematic study of trends and biases in commercially produced media forms, such as movies. Yet another emerging area involves studying the impact of such content on the end users. One of the major research challenges in this area is that at the core of reliable analytics lie reli- able algorithms. These algorithms must be robust under a diverse set of synthesized yet seemingly realistic background conditions. Depending on the type of media, these conditions could manifest themselves in the audio or video channels and could even vary within the duration of the content, thereby making it challenging to apply off-the-shelf techniques from other domains. Analysis of such content necessitates the design and training of customized algorithms that seek to exploit specific properties of or additional structure in the data. Infact, for most vision or audio related tasks, produced media data proves to be one of the most difficult benchmarks. This issue is further compounded by absence of any large in-domain datasets with reliable annotations. As a result, research in this field often requires a mix of clever data mining techniques and approaches from semi-supervised or transfer learning. Finally, this research area is also becoming exceedingly multi-disciplinary requiring skills from a variety of fields including engineering, film studies, psychology and social sciences. Thus the main purpose of this workshop is to facilitate conversation between different groups of researchers and provide a platform where they can share progress and updates in recent research on media analytics for societal trends. We sincerely hope that submitted manuscripts will help develop this new and important area of research, enabling discussion on its potential challenges and shedding light on some areas for future work.


The workshop will offer a timely dissemination of research updates to benefit the academic and industry researchers and professionals working in the fields ranging from multimedia computing, multimodal signal processing to social science. To this end, we solicit original research papers related (but not limited) to the topics listed below:

1. Media analytics and methodologies (e.g. automated discovery of rich analytics related to gender, profession, ethnicity, personality, stereotypes, topics of discussion/conversation, and analysis of their relationships and dynamics)

2. Impact prediction and analysis (e.g. popularity, virality, memorability, commercial success and influence of media content on society)

3. Methodologies and analytics for analyzing less-studied (computationally) media, such as, advertisements and animated movies.

4. Affect and sentiment analysis from media (e.g. emotional appeal, persuasiveness, emotion perception and communication)

5. Large-scale data collection, benchmarking and challenges

6. Evaluation protocols and metrics for methods analyzing societal trends.

Sumbission Instructions

Papers will be evaluated based on their novelty, presentation, contributions and relevance to the workshop topic. The papers must be written in English and describe original unpublished work. Extensions on previously published work must show significant additional work in order to be considered for publication. Reviewers will make an initial determination on the suitability and scope of all submissions.

Important Dates

Paper submission due: March 19, 2018

Decision notification: April 23, 2018

Camera ready version due: May 11, 2018

Organizers and Contact Information

Naveen Kumar (primary contact) Sony Interactive Entertainment America,

Tanaya Guha Assistant Professor, Indian Institute of Technology, Kanpur

Krishna Somandepalli PhD student, University of Southern California

Shri Narayanan Professor, University of Southern California