Discursis: A Computational Methodology for the Analysis of Communication Data

Discursis is a new communication analytic technology that allows an analyst to visually examine and assess text based communication data. Discursis can be used to analyse conversations, web forums, social media, training scenarios, and many other individual and multi-party communication situations. Discursis automatically processes transcribed text to build a statistically grounded language model, and uses this language model to highlight participant interactions around specific topics over the time-course of the communication. Discursis can assist analysts in understanding the structure, information content, and inter-speaker relationships that are present within input data. In addition to rich descriptive visual outputs, Discursis also provides quantitative measures of key metrics, such as topic introduction; topic consistency; and topic novelty. In this talk I will showcase recent findings from the application of Discursis to a variety of datasets, including doctor/patient consultations, television broadcast interviews, audience panel discussion programs, and conversations with persons with language impairments.