Annotation of Children's Oral Narrations: Modeling Emergent Narrative Skills for Computational Applications

Abstract

We present an annotation method for developing a model of children’s comprehension that differentiates between their recall for the objective content of a story and inferred content. We apply the annotation method to a corpus of retellings, in which children retell the same story on three successive days. Our results indicate differences over time: on Day three, children have a more evenly distributed recall of events throughout the story, and include significantly more inferences. The results suggest a cognitive bootstrapping effect. We discuss the potential for application to diagnostic assessment of children’s narrative skills and tutorial applications.