Linguistic variation across Twitter and Twitter trolling

Clarke, Isobelle ORCID: 0000-0001-5541-6327 (2020). Linguistic variation across Twitter and Twitter trolling. University of Birmingham. Ph.D.

Preview

Clarke2020PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (2MB) | Preview

Abstract

Trolling is used to label a variety of behaviours, from the spread of misinformation and hyperbole to targeted abuse and malicious attacks. Despite this, little is known about how trolling varies linguistically and what its major linguistic repertoires and communicative functions are in comparison to general social media posts. Consequently, this dissertation collects two corpora of tweets – a general English Twitter corpus and a Twitter trolling corpus using other Twitter users’ accusations – and introduces and applies a new short-text version of Multi-Dimensional Analysis to each corpus, which is designed to identify aggregated dimensions of linguistic variation across them. The analysis finds that trolling tweets and general tweets only differ on the final dimension of linguistic variation, but share the following linguistic repertoires: “Informational versus Interactive”, “Personal versus Other Description”, and “Promotional versus Oppositional”. Moreover, the analysis compares trolling tweets to general Twitter’s dimensions and finds that trolling tweets and general tweets are remarkably more similar than they are different in their distribution along all dimensions. These findings counter various theories on trolling and problematise the notion that trolling can be detected automatically using grammatical variation. Overall, this dissertation provides empirical evidence on how trolling and general tweets vary linguistically.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):