Emotion Analysis as a Means of Categorizing Content 


Emotion and Mood are an inseparable part of everyday life. However, typical
internet search and discovery are focused mainly on topical and semantic
aspects of web content. The talk will present MoodBase a unique web portal
which delivers aggregated content based on the reaction it elicits rather
than the content per se. MoodBase segments and displays content by emotional
categories (e.g., funny, amazing, weird or romantic) as opposed to common
topical categorization classes (e.g., news, sports, finance, etc.).  Moodbase
collects content available on the web, focusing on communities and social
media sites that have user generated comments and discussions. The user
reactions are then analyzed to infer how people reacted to the subject, what
emotions where raised, and basically, what mood did the content put the
readers in. I will describe the research done and the methods used to detect
emotion in texts focusing on the analysis of social media and user generated
commentary. I will also review related work, present several interesting
insights and pose new challenges.