A collaborative personalized affective video retrieval system

Abstract

In this demonstration, a collaborative personalized affective video retrieval is introduced. A dataset of 155 video clips extracted from Hollywood movies were annotated by the emotion felt by participants. More than 1300 annotations from 40 participants were gathered in a database to be used for affective retrieval system. The retrieval system is able to retrieve videos based on emotional keyword query as well as arousal and valence query. The user's personal profile (gender, age, cultural background) was employed to improve the collaborative filtering in retrieval.