Raconteur : intelligent assistance for conversational storytelling with media libraries

Abstract

People who are not professional storytellers sometimes have difficulty putting together a coherent and engaging story, even when it is about their own experiences. However, consider putting the same person in a conversation with a sympathetic, interested and questioning listener, suddenly the story comes alive. There's something about the situation of being in a conversation that encourages people to stay on topic, make coherent points, and make the story interesting for a listener. Raconteur is a system for conversational storytelling between a storyteller and a viewer. It provides intelligent assistance in illustrating a life story with photos and videos from a personal media library. Raconteur performs natural language processing on a text chat between two users and recommends appropriate media items from the annotated library, each file with one or a few sentences in unrestricted English. A large commonsense knowledge base and a novel commonsense inference technique are used to understand event relations and determine narration similarity using concept vector computation that goes beyond keyword matching or word co-occurrence based techniques. Furthermore, by identifying larger scale story patterns such as problem and resolution or expectation violation, it assists users in continuing the chatted story coherently. A small user study shows that people find Raconteur's suggestions helpful in real-time storytelling and its interaction design engaging to explore stories together.