A NOVEL THEORY OF SUPPORT IN SOCIAL MEDIA DISCOURSE (WITH ARTIFICIAL INTELLIGENCE AND LINGUISTICS ANALYSIS)

This thesis aims to inform the way that people are directly affected by various issues and conditions and how they can support each other on social media. It explores their utilisation of novel salient high-frequency and diverse device-enabled discourse categories of purpose and content. The predominate analysis is from computing and artificial intelligence. It is with an aspiration for the novel methods and novel theory to find its place in corpus linguistics—subtle linguistic analysis guides the exploratory research. An investigation of sophisticated patterns of support alludes to an architecture for social media discourse. The primary patterns can include device-enabled discourse categories of purpose and content. My thesis calls this vose and nent. Advice with stance-taking offers an entry into the large-scale analysis. It forms part of a multitude of online topics and discourses from diabetes advice to raising charity money. The thesis focuses mainly on domain-specific targets of diabetes, blood, and pumps from many others ranging from targets of usernames, child, school, greetings, places, time, to years. The thesis proposes a novel theory of support constructs in social media discourse. The thesis makes a methodological contribution that seeks to combine Artificial Intelligence (AI) computational analyses with corpus linguistics (CL) and qualitative SFL analysis. It is carried out on a big large-scale dataset of 218,068, anonymised Facebook Diabetes UK posts and 16,137 anonymous diabetes-related users of the platform. AI with Latent Dirichlet Allocation (LDA) topic-modelling probabilistic modelling, automated content analysis, an annotation for entity recognition and Discourse Analysis (DA) is utilised with consideration of its limitations but to analyse the corpora for potential patterns. An adapted anonymisation process is used on the data to meet the ongoing challenges of online ethical research requirements. People living with diabetes employ high-frequency patterns of relevant device-enabled categories of purpose and content with for example linguistic forms of advice with stance-taking, diabetes, blood, pumps, to humour/sarcasm or questioning or raising charity funds in their support of themselves and each other amongst other interactional ways. They may not respond directly to each other in consecutive posts but to many other similar posts with similar targets. The focus on particularly advice with stance-taking and domain-specific targets helps to place the novel device-enabled discourse categories in context. These can be in a broader context of power and solidarity, demonstrating social relations concerning risk and trust. Hence, the uncertainty and the variation of effect displayed when sharing information for support. The support trends from a probabilistic model of social media support are confirmed with log-likelihood, precision measures and a multi-method approach. The implications of the new theory are aimed at healthcare communicators to work with organisations. To help their social media users support each other by understanding a peer-focused view of chronic illness support. Corpus linguistics may benefit from the use of combined AI and DA approaches to anonymised large-scale online data. This thesis also offers preliminary work for support-bots to be programmed to utilise the language patterns to support people who need them automatically. The bots may be able to have conversations instantaneously with many people but to do so in natural ways.

Grantor dc:publisher

Oxford Brookes University

Year dc:date

2020

Author dc:creator

• Solomon, Bazil Stanley

Contributors dc:contributor

• Crook, Nigel
• Lischinsky, Alon
• Boness, Kenneth

Statement dc:rights

• All rights reserved

Language dc:language

en

DOI dc:identifier

https://doi.org/10.24384/4xky-qv73

OAI identifier oai:identifier

tle:242e2a1f-8ac2-433c-8123-737e97ca838c:d6bd9758-527a-46cd-bfe2-c433766e8fca:1