Capturing Cross-Cultural Political Schemas: A Computational Analysis of Northern and Southern European Parliamentary Debates.
- Investigated whether the political schemas held by supporters of opposing political affiliations are, and to what extent, similar across various Southern and Northern European countries.
- The study used corpora of parliamentary debates and employs distributional semantics, word embeddings, sentiment analysis, and machine learning to answer the research question.
- Corpora source: http://hdl.handle.net/11356/1910
- Corpora were retrieved using this HuggingFace repository made by the Center za jezikovne vire in tehnologije Univerze v Ljubljani: https://huggingface.co/datasets/cjvt/ParlaMint3
One of the features that distinguishes political ideologies is how some concepts are differently represented across opposing affiliations. Depending on the underlying ideology, a concept can be associated with different terms and viewed positively or negatively. Political schemas are the responsible mental framework that categorizes political information and are shared among followers of an ideology. Previous studies have used computational methods to examine opposing ideological schemas in affiliated texts. For instance, by learning patterns of co-occurrence, word embeddings trained on affiliated texts have illustrated how ideology shapes word associations. Similarly, sentiment analysis has revealed highly divergent words across opposing political camps. However, previous studies have only focused on political biases from one country. This study employed word embeddings and sentiment analysis to provide a cross-cultural investigation of political schemas through linguistic divergences. Spanish, Portuguese, Danish, and Swedish parliamentary debates from the ParlaMint-en 4.1 corpora were used. In addition to exploring word associations, data from embeddings and sentiment analysis were also used in an ideology classification task. Both extrinsic and intrinsic evaluations revealed that ideological alignment and cultural proximity did not guarantee identical schemas. However, left- and right-wing affiliations exhibit similar patterns of linguistic divergences for certain words and topics across these countries. Similar to previous research, neutral words and words referring to (inter)national crises had higher contextual semantic similarity compared to terms indicating ideological differences. Words denoting issues such as social justice and identity had higher divergences in sentiment and usage frequency across opposing affiliations.