Visual text analytic technologies have recently found favour in supplementing traditional manual methods of discourse analysis. One specific technique, Discursis, blends natural language processing with visualisation to support the analysis of sense-making between participants in interactions such as conversations. Discursis highlights patterns in conceptual engagement over different time scales during a multi-participant interaction, making the conceptual interactions of interlocutors visually salient. Like most current visual text analytic technologies, Discursis has primarily been used for the study of English transcripts, despite the core algorithms being theoretically generalizable to languages other than English. In this project we are developing workflows and best practice guidelines for the analysis of Cantonese Chinese transcripts using Discursis. Through this project we are identifying the challenges associated with the analysis of logographic rather than alphabetic data by contrasting the analytic output associated with three data sets processed in three different formats – (i) Chinese data (ii) English data translated using Google Translate and (iii) English data translated by human translators. Our analysis is examining differences in outputs and reveals how these different workflows can impact interpretive judgements of the discourse.
Funding / Grants
- Hong Kong Polytechnic University (2018 - 2019)