Storied navigation : toward media collection-based storytelling

Abstract

Life is filled with stories. Modern technologies enable us to document and share life events with various kinds of media, such as photos, videos, etc. But people still find it time-consuming to select and arrange media fragments to create coherent and engaging narratives. This thesis proposes a novel storytelling system called Storied Navigation, which lets users assemble a sequence of video clips based on their roles in telling a story, rather than solely by explicit start and end times. Storied Navigation uses textual annotations expressed in unconstrained natural language, using parsing and Commonsense reasoning to deduce possible connections between the narrative intent of the storyteller, and descriptions of events and characters in the video. It helps users increase their familiarity with a documentary video corpus. It helps them develop story threads by prompting them with recommendations of alternatives as well as possible continuations for each selected video clip. We view it as a promising first step towards transforming today's fragmented media production experience into an enjoyable, integrated storytelling activity.