Semiotically-grounded distant viewing of diagrams: insights from two multimodal corpora

Abstract

Abstract In this article, we argue for the benefits of combining large-scale analyses of visual materials currently pursued within digital humanities with insights from multimodality research, which is an emerging discipline that studies how human communication relies on appropriate combinations of expressive resources. We show that concepts developed within the field of multimodality research provide appropriate metadata schemes for various modes of expression in large corpora and datasets. We illustrate the proposed approach using a common mode of expression, diagrams, and analyse two recent multimodal diagram corpora using statistical and computational methods. Our results suggest that multimodally-motivated metadata schemes can provide a robust foundation for computational analyses of large corpora and datasets. Even if a corpus or dataset is not designed to support full-blown analyses of multimodal communication, our results imply that multimodality theory can still be used to impose tighter analytical control over a variety of visual materials.