Collective Happiness

The increasing adoption of social media has provided a unique opportunity to quantitatively characterize human behavior at a broader scale. Status updates from Twitter, in particular, have been aggregated for large scale sentiment analysis. While the methodologies are diverse in their applications and findings, they often focus on textual analysis and ignore significant media data, such as images.

In this work, we use geolocated images to determine patterns of happiness, and to sense underlying societal events and community characteristics. Comparing to NLP techniques which are often very tightly coupled with specific language, image content based systems allows cross-cultural sentiment analysis.