poltextlab_admin

On 13 November 2025, Martin Bánóczy delivered a lecture at the HUN-REN Centre in Budapest as part of the “HUN-REN Cloud Meetup”. His presentation, titled Optimising the Fine-Tuning of Large Language Models, addressed key challenges in computational social science, with a focus on improving the efficiency of text classification tasks. He outlined the limitations of manual coding and presented machine-learning-based solutions developed within the Babel Machine project. Bánóczy discussed three major classification tasks: the Comparative Agendas Project, its extension using an enhanced codebook for media analysis, and a further expansion incorporating media-specific coding schemes. He examined factors affecting model performance and highlighted obstacles that complicate multilingual fine-tuning processes.

A new study by Miklós Sebők, Áron Buzogány, Julia Fleischer, Theresa Gessler, Anna Takács, Sean M. Theriault and Ákos Holányi, titled “Crisis-exploitation or fear-mongering? A research agenda for the comparative study of policy crises and illiberal policy frames”, has been published in Journal of European Public Policy. The paper examines how legislative politicians in Austria, Germany, Hungary and the United States use policy crises (migration and COVID-19) to advance illiberal policy frames (IPFs). Their key finding is that the use of these illiberal frames does not closely track objective crisis metrics (such as asylum seeker numbers or COVID casualties), suggesting that the narrative strategy is one of sustained fear-mongering rather than opportunistic crisis-exploitation. The authors operationalise IPFs via a novel codebook and large-language-model text-analysis across four countries and two issues, offering a systematic research agenda for studying illiberal policy-framing. The full study is available here: https://www.tandfonline.com/doi/full/10.1080/13501763.2025.2583176  Sebők, M., Buzogány, Á., Fleischer, J., Gessler, T., Takács, A., Theriault, S. M. & Holányi, Á. (2025) Crisis-Exploitation or Fear-Mongering? A Research Agenda for the Comparative Study of Policy Crises and Illiberal Policy Frames, Journal of European Public Policy. DOI: 10.1080/13501763.2025.2583176.

On 11 November 2025, Miklós Sebők delivered a lecture at University of Kentucky as part of the 2025 Fall Seminar Series. His presentation, titled Using AI Assistants in Comparative Research: The Case of the Comparative Agendas Project and the Babel Machine, introduced new AI-driven tools developed by poltextLAB to support large-scale political text classification. Sebők outlined a fine-tuning agent that enables researchers to adapt multilingual transformer models through natural-language instructions in Slack, automating validation, data strategy selection, and GPU-based training processes. He also presented RobotAssistant, a multi-model comparison framework that deploys classification tasks simultaneously across systems such as Claude, GPT, and DeepSeek. Together, these innovations aim to democratise gold-standard machine coding and reduce technical barriers in comparative political research.  Further details of the event can be found here: https://martin.uky.edu/events/seminar-miklos-sebok

On 10 November 2025, Miklós Sebők delivered a workshop titled Generative AI for Researchers — Tips and Tricks at the University of Kentucky’s Martin School of Public Policy and Administration as part of the 2025 Fall Seminar Series. The session introduced participants to the concepts, tools, and research applications of generative artificial intelligence (GenAI), exploring both introductory and advanced topics. Sebők discussed the ethical and responsible use of GenAI, demonstrated QuantiCheck — a custom GPT designed for peer review — and presented prompt-based methods for literature searches, classification, and regression. He also compared local and API-based models, highlighting how GenAI can be integrated with R and Python, and showcased the role of AI agents in automating data collection and classification. Further details of the event can be found here: https://martin.uky.edu/events/workshop-miklos-sebok

Miklós Sebők delivered a presentation at the University of Texas in Austin as part of a research visit programme, during which he introduced two newly developed poltextLAB dashboards: a New York Times analysis dashboard and a Congressional Speech Dashboard. Dr. Sebők highlighted how these tools integrate AI-based CAP sentiment measurement and additional coding schemes to support the systematic study of media narratives and legislative discourse. Using illustrative examples, he demonstrated how the dashboards enable comparative tracking of political topics, sentiment shifts, and rhetorical strategies in real time.

On 7 November 2025, Miklós Sebők delivered a presentation at the University of Texas at Austin during the panel Building the Dataset, organised as part of the broader event “A Celebration of The Policy Agendas Project and Bryan Jones.” In his contribution, Sebők highlighted the AI-assisted methodological and conceptual innovations required to construct robust, longitudinal policy datasets. He discussed challenges in ensuring coding consistency, integrating heterogeneous political information, and adapting established frameworks to new digital and multilingual sources. Sebők underscored the unique Importance of the Policy Agendas codebook developed by Bryan D. Jones and Frank Baumgartner. His presentation showed examples of how the collaborative, cross-national data project strengthens the analytical foundations of the Policy Agendas Project and enables more precise comparative research on policy change.

Ádám Kerényi chaired a panel on Competitiveness and Economic Development at the “2nd Network for European Studies Conference”. The event took place on 6 November, and was hosted by National University of Public Service. The session featured presentations by the following speakers: Enikő Győri (Member of the European Parliament), Csaba Fási and Petra Szűcs (National University of Public Service), Gabriella Németh (University of Szeged), Krisztián Manzinger and Ákos Kántor (Károli Gáspár University of the Reformed Church in Hungary).

On 29 October 2025, Miklós Sebők held a workshop for the University of Tampa’s faculty and computer science students titled Generative AI for Research (with some computer science applications). The session explored how generative artificial intelligence tools, including large language models, can enhance academic research workflows across the social sciences and computer science. Dr Sebők demonstrated practical examples of AI-assisted data collection, text analysis, and literature synthesis, showing how these tools can accelerate empirical and theoretical investigations.