Selfie examinations: Applying computer vision, hashtag scraping and sentiment analysis to finding and interpreting selfies

In this preliminary investigation of selfies (self portraits in social media), we assess methods for identifying selfies and examine selfies as a form of emotional expression. We compare the accuracy of hashtags and computer vision in discriminating selfies form other images, using human ratings as a reference. Customized software was used to scrape (acquire from social media feeds) 2700 probable selfies and randomly select a sample of 100 images. To describe the emotional attributes of selfies we classified photos using customized sentiment analysis software and qualitatively examined photos. Although the majority of the selfies are upbeat, approximately 20% contain negative emotion words, with some reflecting isolation, disengagement and despair. Our examination shows promise for automatic detection of selfies using a blend of metadata, sentiment analysis and computer vision. The observation of selfies associated with despair and disengagement suggests opportunities to naturalistically assess and immediately address psychosocial needs.