Exploiting Semantics for Personalized Story Creation

Abstract

The task of creating albums or multimedia output from consumer content is becoming increasingly difficult as the amount of content grows. This work presents a system for using semantic information to automate the process of selecting and combining digital assets into summary presentations or storylines, as well as determining triggers for when to generate such content. The system obtains semantic information from a variety of sources, including the capture metadata, image and video understanding algorithms, user profiles and third party ontologies; all such semantic information is stored in a triple store. Prolog-based rules leverage the triple store to provide a knowledgebase for determining when to create particular types of output and how to select assets for such output. This knowledgebase greatly simplifies the task of creating consumer-grade multimedia content.