News & Events

Analyzing discourse across time with time series models

by Dr Dennis Tay, Department of English, PolyU

Date                 8 October 2018
Time                 5:00pm-6:00pm
Venue               AG434
The talk will be conducted in English.

A time series is a set of consecutive measurements of a variable made at equally spaced intervals. Natural examples which occur over different time scales include daily stock prices, monthly sales figures, and annual birth/death rates. Time series analysis (TSA) is a family of methods used to i) extract the components and internal structure of time series data, and ii) develop models to predict future values in the series. While TSA is routinely used in fields like engineering and finance for process control and revenue forecasts, its potential applications to humanistic processes in fields like psychology (Jebb, Tay, Wang, & Huang, 2015) and discourse analysis (Tay, 2017, in press) are only beginning to be explored. It turns out that many real-life scenarios of discourse (e.g. therapist-client interactions over multiple sessions, or newspaper articles on a certain topic across time) can be treated as naturally occurring time series data. In this talk, my objectives are i) to introduce the basic logic and process of TSA, ii) to review and point out limitations of existing discourse analytic approaches to the relationship between discourse and time, and iii) to demonstrate the feasibility and value of applying TSA to discourse analysis. Through case studies of psychotherapy talk, newspapers, and classroom discourse, I suggest that TSA models can adequately predict the use of discourse phenomena across different contexts and time scales. Consequently, these models can be interpreted as schemas or ‘signatures’ of discourse structure, which shed new light on structural developments of discourse invisible to qualitative analysis alone. The proposed approach is still in its developmental stages and thus critical limitations will be reflected upon.

Back to top