Emotion Analysis using Social Media, Computer Vision, and Interactive Virtual Systems
Topics: Spatial Analysis & Modeling
,
,
Keywords: COVID-19, Social media, Emotion Analysis
Session Type: Virtual Paper
Day: Saturday
Session Start / End Time: 4/10/2021 04:40 PM (Pacific Time (US & Canada)) - 4/10/2021 05:55 PM (Pacific Time (US & Canada))
Room: Virtual 9
Authors:
Diya Li, Texas A&M University
Zhe Zhang, Department of Geography, Texas A&M University
Guoying Zhao, Inforamtion Technology and Electrical Engineering, University of Oulu, Finland
Lingli Zhu, Finnish Geodetic Institute, Remote Sensing and Photogrammetry, Finland
Nick Duffield, Department of Electrical and Computer Engineering, Texas A&M University
Shaowen Wang, Department of Geography & Geographic Information Science, University of Illinois Urbana-Champaign
Juha Hyyppä, Department of Remote Sensing and Photogrammetry, National Land Survey of Finland
,
,
,
Abstract
The latest technologies used for emotion detection mainly rely on computer vision using video recording or computer display. The current research work has been devoted purely to visually analyzing human facial expression and ignoring human's social, geographical, demographic, and spatial decision-making characteristics. In contrast, social media data (e.g., Twitter data) provide a unique opportunity to learn about people’s moods, feelings, and behaviors as they experience daily struggles in real-time. In this presentation, we will explore research intersections between traditional computer vision-based emotion detection approach and social media data mining approach to futher build research road map in emotion detection and social behaviro studies. First, we will introduce a multi-modal framework for face-based group-level emotion recognition, which encods a person’s information in a group level image. After that, we will demonstrate social media data mining approach- a CorExQ9 algorithm that integrates a Correlation Explanation (CorEx) social media learning algorithm and clinical Patient Health Questionnaire (PHQ) lexicon to detect COVID-19 related stress symptoms at a spatiotemporal scale in the United States. The algorithm overcomes the common limitations of traditional topic detection models and minimizes the ambiguity that is caused by human interventions in social media data mining. Finally, we will discuss the opportunities of integrating these two approaches in GIScience to advance spatial decision-making and social science studies.
