Conference Programme Details
The conference will run from Thursday 20 November, 9:30 CET, until Friday 21 November, 17:30 CET.
Day One
| Time | Thursday 20 November 2025 |
| 09:00 – 09:30 | Registration |
| 09:30 – 09:45 | Welcome remarks |
| 09:45 – 10:45 | Keynote by Alessandro Lenci Chair: TBD Beyond prediction: What LLMs miss about meaning and why AbstractLarge Language Models (LLMs) have achieved remarkable fluency, generating text that often feels indistinguishable from human writing. Yet beneath this surface competence lies a profound question: do these systems truly capture meaning? This talk explores the conceptual and cognitive limits of current LLMs, focusing on the distinction between statistical prediction and semantic representation. In particular, I will ask why pattern recognition alone cannot yield genuine semantic understanding. |
| 10:45 – 11:15 | Coffee break |
11:15 – 11:35 | Section 1: Human-LLM comprehension Chair: Gábor Prószéky Natalia Moskvina, Raquel Montero, Masaya Yoshida, Ferdy Hubers, Paolo Morosi, Walid Irhaymi, Jin Yan, Elena Pagliarini, Fritz Günther and Evelina Leivada – Language comprehension in LLMs and humans across languages |
| 11:35 – 11:55 | Zétény Bujka, András Lukács, Péter Vedres and Anna Babarczy – Do Large Language Models possess a theory of mind? A comparative evaluation using the Strange Stories paradigm |
| 11:55 – 12:15 | Zoltán Bánréti and László Hunyadi – Challenging AI: How does an Artificial Intelligence learn an artificial language? |
| 12:15 – 12:35 | Christian Lang, Marco Gierke and Ngoc Duyen Tanja Tu – Orthographic diversity in Large Language Models – A case study of foreign word spelling in German |
| 12:35 – 12:55 | Lili Tamás, Mariann Lengyel and Noémi Ligeti-Nagy – Language Models achieve human-level sarcasm detection |
| 12:55 – 14:00 | Lunch |
| 14:00 – 15:00 | Keynote by Erhard Hinrichs Chair: Veronika Lipp The added value of LLMs for lexicography and for lexical semantics AbstractWith the availability of deep learning methods, LLMs, and generative AI, the question has been posed whether dictionaries — and lexicographic resources more generally –can be created by purely automatic means. This hypothesis has been identified as “the end of lexicography” by, among others, Gilles-Maurice de Shriver and David Joffe. On the basis of three use cases from digital lexicography, I want to examine this hypothesis and draw some more general conclusions about the added value of LLMs and generative AI for lexicography and for lexical semantics. |
15:00 – 15:20 | Section 2: Idioms, metaphors and terminology Chair: Marko Tadić Vasile Păiș, Maria Mitrofan, Verginica Barbu Mititelu and Dan Tufiș – Large Language Models as multiword expressions annotators |
| 15:20 – 15:40 | Irene Russo and Paola Vernillo – Seeing the unsaid: Visualizing English idioms with text-to-image generation |
| 15:40 – 16:00 | Gábor Simon, Tímea Borbála Bajzát, Natabara Máté Gyöngyössy, Péter Gergő Molnár, Noémi Prótár and Balázs Indig – Large Language Models in metaphor identification: The case of presuicidal interactions |
| 16:00 – 16:20 | Alexey Matyushin – LLMs as tools for drafting ad hoc pharmaceutical glossaries |
| 16:20 – 16:50 | Coffee break |
16:50 – 17:10 | Section 3: Language games and power plays Chair: TBD Ágoston Tóth – An LLM-motivated theory of language |
| 17:10 – 17:30 | Dániel Golden – Large Language Models and the philosophy of language games |
| 17:30- 17:50 | Silje Susanne Alvestad, Nele Poldvere, Asbjørn Følstad and Petter Bae Brandtzæg – Fakespeak in the age of Large Language Models: A comparative study of persuasion in AI-generated and human-written propaganda narratives |
| 17:50 – 18:10 | Katerina Zoi and Dimitrios Mysiloglou – Reconstructing the past: How LLMs reflect or adapt to Papadiamantis’ cultural worldview |
| 18:10 – 18:15 | Closing remarks |
| 18:30 – | Conference dinner |
Day Two
| Time | Friday 21 November 2025 |
| 09:30 – 10:30 | Keynote by András Kornai Chair: Dávid Márk Nemeskey The linguistic power of LLMs AbstractOpinion on the power of LLMs is on a broad spectrum. At the high end we find the view that these models are so good that we no longer need to deal with messy humans and can do all sorts of exciting linguistics by inspecting LLMs. At the low end we find the view that LLMs are stochastic parrots that cannot possibly have any bearing on how natural language works in humans. In this talk we approach the matter from the perspective of formal language theory, and conclude that there is nothing in natural language that stands in the way of treating LLMs as full and faithful models of human linguistic competence. |
| 10:30 – 11:00 | Coffee Break |
11:00 – 11:20 | Section 4: Chat, translate, evaluate Chair: Dan Tufiș Noémi Prótár and Dávid Márk Nemeskey – Bridging the gap between qualitative and quantitative – How linguistic analysis can help automatic text simplification evaluation |
| 11:20 – 11:40 | Cecilia Domingo, Paul Piwek, Svetlana Stoyanchev and Michel Wermelinger – Reference processing in pair-programming dialogue |
| 11:40 – 12:00 | Bálint Levente Mórász and László János Laki – The impact of example-selection metrics on LLM-based machine translation |
| 12:00 – 12:20 | Roberto Jiménez de la Torre and Carlos Á. Iglesias – A modular LLM-enhanced agent-based system for the generation and evaluation of journalistic interview questions |
| 12:20 – 12:40 | Tommaso Sgrizzi, Asya Zanollo and Cristiano Chesi – Syntactic maps or surface hacks? Testing restructuring verb order and clitic placement in LLMs |
| 12:40 – 13:40 | Lunch |
13:40 – 14:00 | Section 5: LLMs in practice Chair: Tamás Váradi Botond Szemes and Kata Dobás – Digital literary memory in Central-East-Europe. Analysing Wikipedia with LLMs |
| 14:00 – 14:20 | Réka Dodé, Gábor Madarász, Mátyás Osváth, Kristóf Varga and Enikő Héja – Opportunities and challenges in classifying Hungarian scientific texts by field of science |
| 14:20 – 14:40 | Kata Ágnes Szűcs, Noémi Vadász, Zsolt Záros, Emese Varga and Zoltán Szatucsek – Integrating Large Language Models in structural data processing in Hungarian civil registers |
| 14:40 – 15:00 | Boglárka Vermeki – Modelling language proficiency with Puli-BERT-Large: A case study on CEFR classification in Hungarian learner texts |
| 15:00 – 15:20 | Zijian Győző Yang, Ágnes Bánfi, Réka Dodé, Gergő Ferenczi, Flóra Földesi, Enikő Héja, Mariann Lengyel, Gábor Madarász, Mátyás Osváth, Bence Sárossy, Kristóf Varga and Noémi Ligeti-Nagy – Toward Hungarian-centric language understanding: Hungarian-adapted PULI Large Language Models |
| 15:20 – 15:50 | Coffee Break |
| 15:50 – 16:45 | Panel: Do LLMs “understand” language? Moderator: Csaba Pléh (Central European University) Panellists: Erhard Hinrichs (University of Tübingen), András Kornai (HUN-REN SZTAKI, BME), Alessandro Lenci (University of Pisa) and Tamás Váradi (ELTE Research Centre for Linguistics) |
| 16:45 – 17:00 | Closing remarks, rewards |