Geo distance-based event detection in social media

Abstract

In the past two decades, the increase in the digital technology created an interest within the people to capture the daily activities they come across. The popularity of the social media sites enabled sharing of these experiences leading to large number of multimedia content available on the web. Especially, large numbers of photo collections are made into the social media sites, making the task of retrieving photos difficult. The retrieval task of the social photo collections has been made easy by identifying them as events, known as social event detection SED. The proposed work describes the SED task, by utilising the various sources of metadata coupled with photos on an image community website Flickr. The photos are grouped into an event-oriented way by utilising the photo ID, title, tags, description, date, time and geo-location. Especially, three different measures of great circle distance GCD have been used on the latitude and longitude values of the photos for event detection. The proposed methodology grouped the photos into respective events of sports, technical and protest events simultaneously.