WORKSHOP DESCRIPTIONS
Referring expression choice in grounded contexts: Linguistic, cognitive, and computational aspects
Louise McNally and Gemma Boleda
Website: https://www.upf.edu/web/glif/esslli2026-workshop
When we refer to entities and events in our environment, particularly (but not only) when visual information is present, we have choices. Depending on what has been said before, who or what else is in a scene, and the characteristics of what we want to refer to, we might say (among other options) the person running, the runner, the woman in the red shirt, the one with the glasses, or them over there. While some factors influencing RE choice have been amply studied, others have received less attention (e.g. choices involving taxonomic granularity, cross-classifiability of referents, the scenes referents appear in). This workshop aims to gain further insight into the range of this variation; to highlight its theoretical and practical relevance; and to promote synergies between researchers at the interfaces of linguistics, cognitive science, and computation who have studied different aspects of referring expression choice in grounded contexts.
Natural Language Meets Logic and Machine Learning (the NALOMA workshop)
Lasha Abzianidze and Hitomi Yanaka
Website: https://naloma.github.io
There has been an ever-growing interest in tasks targeting Natural Language Understanding and Reasoning. Although deep learning models have seemingly achieved human-like performance in many such tasks, it has also been repeatedly shown that they lack the precision, generalization power, reasoning capabilities, and explainability found in more traditional, symbolic approaches. Thus, current research has started employing hybrid methods, combining the strengths of each tradition and mitigating its weaknesses. This workshop would like to promote this research direction and foster fruitful dialogue between the two disciplines by bringing together researchers working on hybrid methods in any subfield of Natural Language Understanding and Reasoning.
NALOMA began by focusing on bridging the gap between machine learning and natural logic, but has since broadened to include works integrating symbolic methods and machine learning. Even so, combining deep learning with natural logic remains a highly promising direction. The remarkable capabilities of LLMs, which directly operate on natural language expressions, offer a distinct advantage to natural logic — a family of logics with formulas resembling natural language expressions.
Peter Sutton and Louise McNally
Website: https://sites.google.com/view/pac2026/
Polysemous expressions have multiple closely related senses, and there is a substantial amount of semantic and computational linguistic research on polysemy found in different categories of expressions, such as adjectives, nouns and verbs. However, the extent to which there is interaction between researchers working on polysemy in different categories of expressions is limited.
This workshop will focus on approaches to polysemy across different grammatical categories so as to promote engagement between different strands of polysemy research. The workshop will have talks by two invited speakers, and also contributed talks. We will welcome submissions on the modelling the meaning of polysemous expressions, via theoretical, experimental, corpus-based, and computational methodologies.
Semantics and compositionality for expressiveness and complexity
Tomáš Jakl and Dan Marsden
Website: https://tomas.jakl.one/events/esslli-2026-workshop
The aim of the workshop is to bring together researchers who apply semantic approaches in the study of expressiveness and complexity problems in logic. Typical tools of the former type come from category theory, algebra, and topology, whereas expressiveness and complexity concern algorithm-oriented fields of theoretical computer science, such as automata theory, descriptive complexity, finite model theory, and constraint satisfaction.
Examples of such applications include the algebraic approach to the CSP (Constraint Satisfaction Problem), which was key in resolving the famous Feder–Vardi dichotomy; the Game Comonad research programme, which uses category theory in finite model theory; and the use of categorical and topological tools in the study of formal languages and automata theory. We hope the workshop provides a platform to foster interactions and collaborations for this growing community.
Conceptual structure of attitudes: Language and cognition
Natasha Korotkova and Salvador Mascarenhas
Website: https://natasha-korotkova.github.io/attitudes2026.html
Propositional attitudes, linguistically manifested through predicates like “believe”, “intend”, “know” or “want”, constitute one of the central topics in linguistics and philosophy, and they are the foundation of belief-desire-intention psychology. Propositional attitudes allow us to express key aspects of our mental lives for external use in communication and internal use in reasoning. They have been studied by linguists and philosophers, focusing on mathematically precise formulations of the conditions for truthful attitude ascriptions; and by psychologists, focusing on experimentally testing the behavioral consequences of attitudes. These traditions show considerable overlap and complementarity, yet their interaction over the past twenty years has been limited. In this interdisciplinary workshop, we bridge this gap by bringing together researchers from these different fields whose work bears on propositional attitudes. Our chief goal is to assess the state of the art on propositional attitudes, and to identify areas for interdisciplinary collaborations.
Human Label Variation in Discourse and Pragmatic Phenomena
Massimo Poesio and Manfred Stede
Website: https://hlv-dp.github.io/
Over the past 10-15 years, the NLP community has given steadily-increasing attention to the idea of regarding annotation disagreement not as a nuisance but as a potential asset that can be exploited for understanding the respective task better. In practice, this means that multiple annotations are not just averaged (and the individual annotations then thrown away) but taken as a spectrum that constitutes a "complex ground truth", for instance in the form of probability distributions over labels attached to items. While some of the pioneering work on appreciating human label variation (HLV) came from research in discourse tasks such as anaphoric reference and discourse structure, the great majority of the recent work has concentrated on much less subjective tasks, so that the realms of pragmatics and discourse have seen relatively little progress. This is on the one hand surprising - pragmatics and discourse are highly prone to interesting variation in human judgement - but may well be an effect of the relative complexity of the annotation tasks, which make treatments of HLV difficult. In this workshop, we aim to bring together experts on HLV and on discourse phenomena, and to identify ways forward: What are examples of successful approaches to-date; what are ideas to tackle phenomena that are hitherto understudied for HLV; are there commonalities between different pragmatic/discourse annotation tasks that invite similar treatment for capturing and exploiting HLV?