The Second Workshop on Bridges and Gaps between Formal and Computational Linguistics
| 08:30-08:50 | Welcome |
| 08:50-09:10 | On the relative impact of categorical and semantic information on the induction of self-embedding structures (Antoine Venant, Yutaka Suzuki) [abstract] |
| 09:10-09:30 | Modelling Expectation-based and Memory-based Predictors of Human Reading Times with Syntax-guided Attention (Lukas Mielczarek, Timothée Bernard, Laura Kallmeyer, Katharina Spalek, Benoit Crabbé) [abstract] |
| 09:30-10:30 | Keynote talk by Anna Rogers Title: Studying Generalization in the Age of Contamination Abstract: In the age of Large Language Models, we can no longer be sure that the test data was not observed in training. This talk discusses the main approaches to studying generalization, and presents a new framework for working with controlled test-train splits across linguistically annotated data at scale. |
| 10:30-11:00 | coffee break |
| 11:00-11:20 | Plural Interpretive Biases: A Comparison Between Human Language Processing and Language Models (Jia Ren) [abstract] |
| 11:20-11:40 | Towards Developmentally Motivated Curriculum Learning for Language Models (Arzu Burcu Güven, Rob van der Goot, Anna Rogers) [abstract] |
| 11:40-12:00 | Natural Language Inference with CCG Parser and Automated Theorem Prover for DTS (Asa Tomita, Mai Matsubara, Hinari Daido, Daisuke Bekki) [abstract] |
| 12:00-12:20 | Modal Subordination in Dependent Type Semantics (Aoi Iimura, Teruyuki Mizuno, Daisuke Bekki) [abstract] |
| 12:20-14:00 | lunch break |
| 14:00-14:20 | Coordination of Theoretical and Computational Linguistics (Adam Przepiórkowski, Agnieszka Patejuk) [abstract] |
| 14:20-15:20 | Keynote talk by Kees van Deemter Title: Classifying Hallucinations in Data-Text NLG: Avoiding the Pitfalls Abstract: Algorithms that produce textual output can sometimes “hallucinate”, producing texts that express information that differs from what is required. In this presentation, I will talk about hallucination in Data-Text NLG, focusing on situations in which the task of the algorithm is to express a known body of information both fully and accurately. Various attempts have been made to clarify the notion of hallucination, and to distinguish between different types of hallucinations that can occur in the above-mentioned situations.I will examine some of these classifications and ask: (1) Are the existing classifications well defined? (2) How feasible in practice is it to apply these classifications to concrete cases of Data-Text NLG? (This is joint work with Eduardo Calo and Albert Gatt, both at Utrecht University.) (3) How useful are the distinctions that these classifications make, for example for determining the seriousness of a hallucination, or for redesigning the NLG algorithm so as to avoid hallucinations? And finally, if time permits (4) What does our investigation tell us about hallucinations in other NLG situations, for instance in Question-Answering? |
| 15:30-16:00 | coffee break |
| 16:00-16:20 | An instructive implementation of semantic parsing and reasoning using Lexical Functional Grammar (Mark-Matthias Zymla, Kascha Kruschwitz, Paul Zodl) [abstract] |
| 16:20-16:40 | Exploring Gaps in the APS: Direct Minimal Pair Analysis in LLM Syntactic Assessments (Timothy Pistotti, Jason Brown, Michael J. Witbrock) [abstract] |
| 16:40-17:00 | Evaluating The Impact of Stimulus Quality in Investigations of LLM Language Performance (Timothy Pistotti, Jason Brown, Michael J. Witbrock) [abstract] |
| 17:00-17:20 | Syntax-Guided Parameter Efficient Fine-Tuning: Integrating Formal Grammatical Constraints into Language Models (Prasanth Yadla; remote) [abstract] |