Nathan Dykes: Studying argumentation patterns through corpus queries

The Department of Digital Humanities and Social Studies would like to invite you to the following talk in our DH Colloquium:

Nathan Dykes: »Studying argumentation patterns through corpus queries«

Abstract

Corpus-Based Discourse Analysis typically relies on keywords or collocations to identify patterns in large bodies of text. While these approaches work well to identify broad themes and attitudes in corpora, applying them to the study of complex propositions is challenging.

In this presentation, corpus queries are proposed as an entry point to exploring statements relevant to argumentation. Using a collection of German media articles on multidrug-resistant organisms (MDRO), a range of corpus queries were developed with the CQP query language (Evert & Hardie 2011). The queries combine regular expressions and word-lists and are built to reflect different linguistic realizations of argumentative patterns (cf. Dykes et al. 2022).

Queries were developed for two types of statements: causality and reported speech. The causal queries retrieve realizations of predetermined argumentative patterns, such as those used by Dykes & Peters (2020) in their keyword study on the same corpus. For example, the negligence topos highlights a lack of hygiene measures in hospitals as a central cause for the spread of MDRO. In this study, candidate words filter the query results, providing a more targeted approach than keywords.

The quotation patterns were investigated in two ways. Firstly, a sample of matches was annotated to compare gender patterns regarding speakers and reporting verbs, similar to Taboada (2024). Secondly, the quotations themselves were clustered based on sentence similarity, grouping these complex statements.

The results suggest that corpus queries are a promising route for discourse analysts. For the retrieval of predetermined argument patterns, filtering causal queries achieves considerably higher precision than keywords alone. The exploration of quotes reveals a similar gender bias regarding speakers as in Taboada (2024), while the cluster analysis sheds light on prevalent topics and stylistic nuances shaping reported speech in this context.

References

• Dykes, N., & Peters, J. (2020). Reconstructing argumentation patterns in German newspaper articles on multidrug-resistant pathogens: a multi-measure keyword approach.
  Journal of Corpora and Discourse Studies, 3, 51-74.
• Dykes, N., Heinrich, P., & Evert, S. (2022). Retrieving Twitter argumentation with corpus queries and discourse analysis.
  Broadening the Spectrum of Corpus Linguistics: New approaches to variability and change, 105, 229.
• Evert, S., & Hardie, A. (2011, July). Twenty-first century Corpus Workbench: Updating a query architecture for the new millennium.
  In Proceedings of the Corpus Linguistics 2011 conference (pp. 1-21).
• Taboada, M. (2024). Reported speech and gender in the news: Who is quoted, how are they quoted, and why it matters.
  Discourse & Communication, 17504813241281713.

Further information and other upcoming talks in the Colloquium can be found here.