A Big Data Approach to the Study of Speech and Co-Speech Gesture

Dr Peter Uhrig from the FAU Erlangen-Nürnberg, Germany, was the speaker on 26 February 2020.

Dr Uhrig explores how digital methods have been mainstream in linguistic research and the study of co-speech gesture for a long time. The computer helps researchers with their tasks of annotating and interpreting data, but the actual analysis is often done manually and thus highly labour-intensive.

In the first part of his talk Dr Uhrig presents methods to speed up manual analyses of spoken and multimodal data in various ways by using automatic annotation on various levels of analysis, including text, pronunciation, and gesture.

The second part discusses how and to what extent we can further automate linguistic research methodologies with so-called big data methods. Based on automatic annotation, error rates and bias in the data are assessed manually. Once we have an appropriate understanding of our dataset, we can derive insights from the data at a scale that is beyond what can be managed manually. The lecture includes a brief demonstraion of the research infrastructure available in the project as well as small case studies to illustrate the methodology.