Democracy in context: using a distributional semantic model to study differences in the usage of democracy across languages and countries

Abstract

Cross-cultural survey research rests upon the assumption that if survey features are kept constant, data will remain comparable across languages, cultures and countries. Yet translating concepts across languages, cultures and political contexts is complicated by linguistic, cultural, normative or institutional discrepancies. Such discrepancies are particularly relevant for complex political concepts such as democracy, where the literature on political support has revealed significant cross-cultural differences in people’s attitudes toward democracy. Recognizing that language, culture and other socio-political variables affect survey results has often been equated with giving up on comparative research and many survey researchers have consequently chosen to simply ignore the issue of comparability and measurement equivalence across languages, cultures and countries. This paper contributes to the debate, using a distributional semantic lexicon, which is a statistical model measuring co-occurrence statistics in large text data. The method is motivated by structuralist meaning theory, stating that words with similar meanings tend to occur in similar contexts, and that contexts shape and define the meanings of words. Compared to other methodological approaches aimed at identifying and measuring cross-cultural discrepancies, this approach enables us to systematically analyze how the concept of democracy is used in its natural habitat. Collecting geo-tagged language data from news and social online source documents this paper descriptively explores varieties in meanings of democracy across a substantial number of languages and countries, and maps ways in which democracy is used among online populations and regions worldwide.

Citizens, Elites, and the Legitimacy of Global Governance

“…However much care is put into translating standardized surveys, respondents may nevertheless interpret survey questions differently across contexts. Sometimes, linguistic terms with the exact same semantic meaning as the original survey text do not exist; other times, such words exist, but are interpreted differently by respondents because of cultural and political context (Dahlberg et al 2020). Next to efforts to achieve greater commensurability of survey questions, researchers should recognize that contextual knowledge is necessary in order to more fully interpret recorded quantitative patterns.…”

A systematic literature review of how and whether social media data can complement traditional survey data to study public opinion

“…SMD can also be useful for validating survey measures. For instance, the study by Dahlberg et al [20] investigated the meanings of democracy in a cross-country perspective to better understand differences in the usage of the term "democracy" across languages and countries. The authors' findings aimed to inform survey measurements about the different conceptualisations of democracy, notably by highlighting translations and language equivalence issues in survey items.…”

A Distributional Semantic Online Lexicon for Linguistic Explorations of Societies

“…Although the lexicon can be used to study almost any given word (provided that it exists in the available corpora), this section will illustrate how the lexicon can be used and applied in comparative social science research. In a recent article published in Zeitschrift für Vergleichende Politikwissenschaft, the lexicon is utilized for studying one of the key concepts in political science, in this case democracy (see Dahlberg et al, 2020). The term democracy has been shown to carry various meanings across different institutional and cultural settings that differentiate from the normative democratic definitions typically used by political scientists (Welzel, 2013).…”

“…Throughout this process, the 3,420 original neighbor terms were reduced to 808 separate "democracy topics." More detailed information about the translation and clustering processes is available in Dahlberg et al (2020).…”