Intersections. East European Journal of Society and Politics

Application of Natural Language Processing to the electoral manifestos of parties characterised by populist rhetoric in Central and Eastern Europe

Elena Cossu — 2025-02-17

The aim of this paper is to find patterns regarding the rhetoric employed by actors characterised by populist rhetoric in Central and Eastern Europe (CEE). To do so we examine all available electoral manifestos of all parties in CEE during the 2000–2022 period and we use Natural Language Processing with the goal of extracting meaning from a large number of texts. The analyses focus on most frequent topics using Topic Modelling, whether the manifestos express positive or negative emotions using Sentiment Analysis, and on the analysis of word frequencies. The analysis reveals a complex landscape characterized by both common themes and country-specific variations for what concerns populist rhetoric remove in the region. Populist parties consistently employ more negative sentiments in their manifestos compared to non-populist parties, regardless of being right-wing or left-wing. While they share overarching themes such as anti-EU sentiment, critiques of economic liberalism, religious and ethno-centric language, and anti-corruption narratives, populist parties skilfully tailor their rhetoric to specific national contexts and historical narratives. This adaptability allows them to resonate more effectively with local electorates. The use of negative rhetoric is a tactical choice across the populist spectrum rather than an indicator of ideological radicalism. Notably, populist parties often frame the EU as an obstacle to national sovereignty and economic prosperity while simultaneously acknowledging its benefits. The effectiveness of populist rhetoric appears to be influenced by each country’s economic and institutional context, with countries investing in productivity- and innovation-led growth providing less fertile ground for divisive populist messages.

How opposition parties politicize democracy during autocratization

Theresa Gessler — 2025-02-17

How do democratic actors rhetorically politicize their own disempowerment? Autocratization as a gradual process not only erodes democracy, but also progressively reduces the power of domestic actors to oppose this erosion. Often, incumbent governments disable the institutions meant to hold them accountable, such as parliaments. Drawing on the case of Hungary, we study how (opposition) parties rhetorically politicize democracy within the institution of parliament. As a case of rapid autocratization, Hungary saw far-reaching changes, including the transformation or abolition of many democratic institutions over the past years. New laws also restricted the rights of parliament itself, thereby narrowing the opportunities for public debate around democratic procedures. We address two questions related to this: To which extent do opposition parties politicize democracy and its procedures in a context of autocratization? And how does the way they talk about democracy differ from that of the government? The paper uses text-as-data methods – namely a dictionary of democratic principles and a word-embedding-based analysis of democracy rhetoric – to study parliamentary debates between 2010 and 2022. We find that democracy is highly salient for both the government

Analysis of the populist discourse of the Lithuanian political parties

Jogilė Ulinskaitė — 2025-02-17

This article aims to analyse the populist discourse of Lithuanian political parties over a 30-year period: 1990–2020. Since Lithuania belongs to the CEE region, the question arises whether it is witnessing a worrying rise of populism and a related backsliding of democracy. Although Lithuania is currently a stable consolidated democracy, the lack of a stable party system and clear ideological cleavages during the transition to democracy in the 1990s created a favourable environment for populism to flourish. In this article, we analyse the changes in populist discourse in Lithuania across different actors, sources and over time. This is done by applying machine learning models to identify populist content at the paragraph level through a corpus of political party manifestos, political party websites, and columns written by party members on Delfi.lt. The results show that, although elements of populist discourse are present in the texts of all Lithuanian political parties, the overall level of populist discourse has remained fairly stable over the period analysed, with a temporary increase in 2008–2009. We observe that populist discourse is more widespread in the media than in party manifestos and that non-parliamentary parties engage in this discourse more than their parliamentary counterparts.

Jobbik’s journey from radicalism to mainstream politics:

Zsófia Rakovics — 2025-02-17

Exploring the substantial influence wielded by politicians in shaping social reality and molding the public perceptions of the country may be beneficial from a social scientific point of view. Our research delves into the distinctive realm of parliamentary discourse to unravel this intricate process.

We focus on the transformative role of politicians and political parties within the public political sphere. Through an innovative approach, linguistic similarities in parliamentary speeches are harnessed to unveil the strategies of political communication, the dynamics of power dynamics, and the myriad modes of interactions.

Our research puts the spotlight on Jobbik – Movement for a Better Hungary, a notable far-right party that has navigated a decade of political success in Central and Eastern Europe. Gaining a substantial share of the vote in national elections, Jobbik’s evolution from its radical right-wing origins to a moderate conservative people’s party is the subject of scrutiny. Leveraging techniques such as natural language processing, document embedding, social network analysis, and structural topic modeling, the study dissects Jobbik’s journey from radicalism to mainstream politics.

We uncover the transformation of Jobbik through a meticulous analysis of parliamentary speeches and the use of quantitative analytical tools. We not only reveal the evolution of Jobbik’s identity but also contribute to a deeper understanding of the intricate interplay between political communication and power dynamics.

Exploring the potential and limitations of large language models as virtual respondents for social science research

Zsófia Rakovics — 2025-02-17

Social and linguistic differences encoded in various textual content available on the internet represent certain features of modern societies. For any scientific research which is interested in social differences mediated by language, the advent of large language models (LLMs) has brought new opportunities. LLMs could be used to extract information about different groups of society and utilized as data providers by acting as virtual respondents generating answers as such.

Using LLMs (GPT-variants, Llama2, and Mixtral), we generated virtual answers for politics and democracy related attitude questions of the European Social Survey (10^th wave) and statistically compared the results of the simulated responses to the real ones. We explored different prompting techniques and the effect of different types and richness of contextual information provided to the models. Our results suggest that the tested LLMs generate highly realistic answers and are good at invoking the needed patterns from limited contextual information given to them if a couple of relevant examples are provided, but struggle in a zero-shot setting.

A critical methodological analysis is inevitable when considering the potential use of data generated by LLMs for scientific research, the exploration of known biases and reflection on social reality not represented on the internet are essential.

Beyond “latent thematic structure”:

Renáta Németh — 2025-02-17

According to the textbook definition, a topic model aims to uncover the underlying topics of a corpus. Despite its widespread use across disciplines, the nature of these ‘topics’ has remained relatively underdefined. This research note attempts to fill this gap, drawing on empirical evidence to elucidate the practical application of the model. We argue that the frequency of terms within texts is influenced not only by their theme but also by factors such as genre and context, thus extending the notion of ‘latent topics’ beyond referential-semantic boundaries to include pragmatic considerations. Through case studies focusing on different genres, such as parliamentary speeches and online forums, we demonstrate the importance of pragmatics, which is often overlooked in well-known early applications that deal predominantly with formal written texts such as newspaper articles or academic papers.

Levelling up quantitative legislative studies on Central-Eastern Europe:

Miklós Sebők — 2025-02-17

The availability of ready-made textual corpora for research is crucial for social scientists, especially in the current era of rapid advancements in natural language processing (NLP) and artificial intelligence (AI) methods. Despite various useful contributions that address issues of accessibility and standardisation when it comes to such corpora, in many cases, they have limitations related to scope, geographical coverage, and time frame. This concern is particularly significant in the context of political research on Central-Eastern Europe (CEE), for which such deployment-ready databases are few and far between. In this research note, we bridge part of this gap by making available a new database: ParlText CEE. The database, prepared under the auspices of the V-Shift Momentum project at the HUN-REN Centre for Social Sciences, covers almost 1.9 million text vectors and metadata for parliamentary speeches, bills, and laws for Czechia, Hungary, Poland, and Slovakia for the period from 1990–1991 to 2022–2024. The datasets encompass relevant dates, texts, titles, and, in the case of the speech corpora, parliamentary agendas, speaker names, and parties. All data are also linked based on unique identifiers following the ParlLawSpeech standard. This paper introduces the specifics of the 1.0 release of ParlText CEE and contemplates its possible use cases.

Book Review

Tamás Varga — 2025-02-17

Social scientists, digital humanities scholars and industry professionals now regularly leverage large-scale document corpora. A large dataset of texts, while providing a wealth of information, is insufficient on its own to generate meaningful insights. It is essential to approach the dataset with well-defined research questions that guide the analytical process and ensure the relevance of the findings. Moreover, deriving meaningful answers requires the application of appropriate methodologies that are aligned with the research objectives. In addition to methodological rigor, scholars must critically assess the limitations of the dataset's validity. This involves evaluating the accuracy, reliability, and completeness of the data, as well as recognizing any inherent biases.

The book book aims to illustrate how to treat “text as data” for social science tasks and social science problems. It adopts a six-part structure, combined with several chapters and subchapters. Each part is structured around five fundamental concepts: representation, discovery, measurement, prediction, and causal inference. By doing this, it serves as a comprehensive guide for researchers, delineating the capabilities and limitations inherent in text data methodologies.

Editorial

Domonkos Sik — 2025-02-17