Visualizing Wahl-O-Mat & zero-shot classification

Author

Nico Blokker

Published

February 14, 2023

The Wahl-O-Mat probably is Germany’s most popular voting-advice-application. Users can compare their preferences on certain policies to those of political parties. Party-Policy relations are often visualized as bipartite network or as either of its projections.

The Wahl-O-Mat also contains justifications, in which parties argue why they support or oppose the policy in question. However, as opposed to the policies, justifications are not labeled. In this post, I first automatically label the justifications following the Codebook from the Media Frames Corpus (Card et al. 2015), only to subsequently visualize the Party-Policy-Justification network (shown below, see Blokker et al. 2022).1

Label justifications from WoM

I use Sentence-BERT (Reimers and Gurevych 2019) to automatically assign one of the categories contained in the Codebook: Each justification is compared against the category and its description (Card et al. 2015: 439). The most similar category is assigned as label (label_auto) and subsequently used as node in the network. Policies are indicated by the starting “T_” (for thesis) and black font, whereas justifications are in grey and start with “F_” (for frame). The color of the edges signals the corresponding party.

Examples Wahl-O-Mat 2021 in Germany
party thesis position justification label_auto
cdu Abschaffung des Familiennachzugs stimme nicht zu Wir wollen, dass die Zahl der Flüchtlinge, die zu uns kommen, dauerhaft niedrig bleibt. Grundsätzlich gilt, dass anerkannte Flüchtlinge einen Anspruch auf den Nachzug ihrer Kernfamilie haben. Eine Ausweitung des Familiennachzugs über die heute bestehenden Regelungen hinaus lehnen wir ab. Cultural identity
spd Abschaffung des Familiennachzugs stimme nicht zu Wer vor Verfolgung, Tod und Terror zu uns geflohen ist und als schutzberechtigt anerkannt ist, soll auch seine Familie in Sicherheit bringen dürfen. Das gebietet die Menschlichkeit und das internationale Flüchtlingsrecht gleichermaßen. Die Regelungen für den Familiennachzug von subsidiär Schutzberechtigten wollen wir daher auch wieder an die für Flüchtlinge angleichen. Dabei werden wir auch Regelungen für den Geschwisternachzug zu unbegleiteten minderjährigen Flüchtlingen schaffen. Security and defense
afd Abschaffung des Familiennachzugs stimme zu Familiennachzug für Flüchtlinge erhöht die unqualifizierte Zuwanderung nach Deutschland um ein Vielfaches und endet fast immer dauerhaft im sozialen Netz. Wir lehnen dies ab. Die Familieneinheit kann in Drittländern hergestellt werden. Cultural identity

Visualize tripartite Network

In the following example network, one can for example see that the Left party (purple) opposes – indicated by the dashed line – to abolish family reunifications (for migrants) because of reasons of ‘fairness and equality’. The AfD on the other hand supports this policy for reasons of ‘Cultural Identity’.

Tripartite Network representation

Conclusion

In this post zero-shot classification is used to quickly categorize unlabeled texts. These labels are in turn used as additional building blocks of tripartite network structures. The resulting network gives a concise overview of a) the positions parties take with regard to b) certain policies and c) with what reasoning.

References

Blokker, Nico, Tanise Ceron, André Blessing, Erenay Dayanik, Sebastian Haunss, Jonas Kuhn, Gabriella Lapesa, and Sebastian Padó. 2022. “Why Justifications of Claims Matter for Understanding Party Positions.” In Proceedings of the 2nd Workshop on Computational Linguistics for Political Text Analysis. https://old.gscl.org/media/pages/arbeitskreise/cpss/cpss-2022/workshop-proceedings-2022/254133848-1662996909/cpss-2022-proceedings.pdf.
Card, Dallas, Amber Boydstun, Justin H Gross, Philip Resnik, and Noah A Smith. 2015. “The Media Frames Corpus: Annotations of Frames Across Issues.” In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2: Short Papers), 438–44.
Longabaugh, William JR. 2012. “Combing the Hairball with BioFabric: A New Approach for Visualization of Large Networks.” BMC Bioinformatics 13 (1): 275. https://doi.org/10.1186/1471-2105-13-275.
Pedersen, Thomas Lin. 2022. Ggraph: An Implementation of Grammar of Graphics for Graphs and Networks. https://CRAN.R-project.org/package=ggraph.
Reimers, Nils, and Iryna Gurevych. 2019. “Sentence-BERT: Sentence Embeddings Using Siamese BERT-Networks.” arXiv. https://doi.org/10.48550/arXiv.1908.10084.

Footnotes

  1. To learn more about the “biofabric” layout refer to Longabaugh (2012) and the GGraph- Package (Pedersen 2022).↩︎