Evaluation of Automated Storytelling Procedures

• Activity Number: 304
• Type: Topic Contributed
• Date/Time: Tuesday, August 11, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Defense and National Security
• Author(s): John Rigsby* and Daniel Barbará
• Companies: Naval Surface Warfare Center and George Mason University
• Keywords: automated storytelling ; literature-based discovery ; text mining ; evaluation ; knowledge discovery ; topic detection and tracking

Abstract

Automated storytelling attempts to create a chain of documents linking one article to another one, telling a coherent and cohesive story to explain the events that connect the two article end points. In our world of ever more numerous sources of information, including scientific publications, news articles, web resources, emails, blogs, tweets, etc., automated storytelling mitigates information overload by presenting readers with the clarified chain of information most pertinent to their needs.

In this paper, multiple storytelling algorithms are quantitatively evaluated using methods based on several distance methods: cosine distance, Soergel distance, and Kullback-Leibler divergence.