Automatic Contextual Storytelling in a Natural Language Corpus

Abstract

Storytelling is an ancient art and science of conveying wisdom through generations for centuries. Data-driven storytelling in the context of a natural language corpus has a huge potential for conveying fast valuable insights about the corpus for better decision making. But high dimensional unstructured nature of natural language text makes automatic extraction of stories extremely difficult. This PhD research project believes that modern storytelling is a hand in hand approach of contextual topic visualization and contextual summarization. While exploratory data visualization can provide valuable insights into the data, these insights can be used to understand and design models for producing abstract summarization. In this project, the context of a story is defined from three perspectives: a single document, a collection of multiple documents about a topic of interest and the whole corpus. In this project, exploratory data visualization is used to understand the context better and now with the achieved insights, research is focusing on abstract summarization for automatic contextual storytelling.