Header

UZH-Logo

Maintenance Infos

Complementing Kernel Density Estimation and Topic Modelling to Visualise Political Discourse


Reveilhac, Maud; Schneider, Gerold (2022). Complementing Kernel Density Estimation and Topic Modelling to Visualise Political Discourse. In: Digital Research Data and Human Sciences DRDHum Conference 2022, Jyväskylä, Finland, 1 December 2022 - 3 December 2022. University of Jyväskylä, 12-27.

Abstract

We examine how politicians shape and convey policy issues among different communication channels and how well we can visualise issue ownership and issue framing using data-driven methods. Drawing from a political communication approach, we propose to complement two methods – Kernel density estimation and topic modelling – to visualise political discourse. Our case study is established on two main data sources: transcripts of parliamentary debates and tweets of Swiss elected politicians. We propose a two-step methodology. First, we use topic modelling to identify the main policy issues emphasised by politicians. Second, we use the topical content and political affiliation as meta-information in Kernel density estimation to calculate and display the distances between important features, the main extracted topic for each document, and political affiliation. We compare the obtained results from the transcripts of parliamentary debates and from Twitter data. Using conceptual visualisation maps enables us to qualitatively discuss whether the findings are indicative of issue ownership and stylistic markers pointing to issue framing. The proposed methodology can be applied to a variety of research questions in the realm of communication studies with an interdisciplinary and exploratory focus. It also provides a suitable way to account for polarisation processes that can be linked to affordances of communication channels and to institutional specificities of the national political context.

Abstract

We examine how politicians shape and convey policy issues among different communication channels and how well we can visualise issue ownership and issue framing using data-driven methods. Drawing from a political communication approach, we propose to complement two methods – Kernel density estimation and topic modelling – to visualise political discourse. Our case study is established on two main data sources: transcripts of parliamentary debates and tweets of Swiss elected politicians. We propose a two-step methodology. First, we use topic modelling to identify the main policy issues emphasised by politicians. Second, we use the topical content and political affiliation as meta-information in Kernel density estimation to calculate and display the distances between important features, the main extracted topic for each document, and political affiliation. We compare the obtained results from the transcripts of parliamentary debates and from Twitter data. Using conceptual visualisation maps enables us to qualitatively discuss whether the findings are indicative of issue ownership and stylistic markers pointing to issue framing. The proposed methodology can be applied to a variety of research questions in the realm of communication studies with an interdisciplinary and exploratory focus. It also provides a suitable way to account for polarisation processes that can be linked to affordances of communication channels and to institutional specificities of the national political context.

Statistics

Downloads

61 downloads since deposited on 12 Dec 2022
45 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Conference or Workshop Item (Paper), refereed, original work
Communities & Collections:06 Faculty of Arts > Institute of Computational Linguistics
06 Faculty of Arts > Department of Communication and Media Research
08 Research Priority Programs > Digital Society Initiative
06 Faculty of Arts > Zurich Center for Linguistics
08 Research Priority Programs > Digital Religion(s)
06 Faculty of Arts > Linguistic Research Infrastructure (LiRI)
Dewey Decimal Classification:410 Linguistics
Uncontrolled Keywords:topic modelling, distributional semantics, data visualisation, interdisciplinarity, qualitative interpretation
Language:English
Event End Date:3 December 2022
Deposited On:12 Dec 2022 09:29
Last Modified:28 Dec 2022 12:53
Publisher:University of Jyväskylä
OA Status:Green
Free access at:Official URL. An embargo period may apply.
Official URL:https://jyx.jyu.fi/handle/123456789/84140
  • Content: Published Version
  • Language: English