Skip to main content
Kent Academic Repository

Discursive Signatures: A Computational Method for Mapping How Meaning Varies Across Communities

Hill, Mark J., Miller, Vincent, Moreira, Tiago (2026) Discursive Signatures: A Computational Method for Mapping How Meaning Varies Across Communities. In: Anthology of Computers and the Humanities Volume 6. Association for Computers and the Humanities, Austin, Texas, USA (doi:10.63744/dVXq85XJAhwJ) (KAR id:117674)

Abstract

This paper presents a computational method for studying how the meaning of cultural artefacts varies across online communities. Rather than treating topic model outputs as thematic labels, we propose using full topic-proportion vectors as discursive signatures that capture meaning at both the level of individual expression and the community contexts that shape it. By aggregating these signatures at the community level, validating groupings with PERMANOVA, applying hierarchical clustering, and projecting the resulting space via PCA, we produce a topology of discourse: a spatial representation of how meaning-making varies systematically across contexts. We demonstrate the method through a case study of mental health technology discourse on Reddit (289,039 posts across 445 subreddits), showing how the same technologies are conceptualised differently depending on community context. The method is designed to be transferable to any domain where researchers seek to understand how technologies, practices, or concepts are variably constituted across online spaces.

Item Type: Conference proceeding
DOI/Identification number: 10.63744/dVXq85XJAhwJ
Uncontrolled keywords: computational discourse analysis, cultural analytics, social media, community-level modelling, discourse topology, mental health technology
Subjects: H Social Sciences
Q Science > QA Mathematics (inc Computing science) > QA 76 Software, computer programming,
Institutional Unit: Schools > School of Social Sciences > Criminology, Philanthropy, Social Policy, Social Work, Sociology
Former Institutional Unit:
There are no former institutional units.
Funders: University of Kent (https://ror.org/00xkeyj56)
Depositing User: Vince Miller
Date Deposited: 10 Nov 2026 15:20 UTC
Last Modified: 11 Nov 2026 11:44 UTC
Resource URI: https://kar.kent.ac.uk/id/eprint/117674 (The current URI for this page, for reference purposes)

University of Kent Author Information

  • Depositors only (login required):

Total unique views of this page since July 2020. For more details click on the image.