Séminaires

Experimental evidence for a constructionist approach to grammar

Adele Goldberg (Princeton U)

Lectures will explain the power of usage-based constructionist approaches to language to capture the commonalities, relationships and interactions among words, idioms, and more abstract syntactic patterns. Lecture #1 will focus on lexical semantics and conventional metaphors. Lecture #2 will address the issue of productivity and Lecture #3 will focus on island constraints.

Crowdsourcing for linguistics

Titus von der Malsburg (U Stuttgart)

The use of crowdsourcing methods in linguistics and related fields has remarkably increased in recent years.  However, these methods often pose technical and ethical challenges (e.g., Cuskley & Sulik, 2022; Eyal et al., 2021; Fort, Adda, and Cohen, 2011; Stewart, Chandler, and Paolacci, 2017), and researchers are usually not given sufficient training to navigate them.  This compact course, spanning three 90-minute sessions, aims to fill this gap by offering a comprehensive overview of the relevant aspects of these methods to beginners and helping current users enhance their understanding.

Topics include: Technical implementation of experiments, self-hosting of experiments and cloud-hosting solutions, recruitment and compensation, ethical and data protection issues, vulnerable communities, experimental designs and paradigms, best-practices for minimization of attrition and maximization of data quality, data  management, data exclusion, and analysis, role of online experiments/tasks in the larger research workflow (preregistration, reproducibility, etc.).

Language models, comparison with human behavior

Benoît Crabbé (Université Paris Cité)

This class introduces current Large Language Models such as GPT-4 and chatGPT and asks the question: what can we learn on language from these models ?
First we introduce current language modelling methods at large with an historical perspective.
Second we introduce notions of information theory and illustrate how information theory can be used to link model predictions to human behavioral or physiological measures.
Assuming that language is essentially a communication system, we ask to which extent humans are optimal communicators and to which extent Large Language Models are such optimal communicators.
As humans are not optimal communicators but rather communicators subject to physiological constraints, we discuss a family of methods that can be used to express these constraints in language models.

Key Bibliography:

• Michael Hahn, Richard Futrell, Roger Levy, Edward Gibson (2022), A resource-rational model of human processing of recursive linguistic structure, Proceedings of the National Academy of Sciences, 119(43).
• Richard Futrell, Edward Gibson, Roger P. Levy (2020) Lossy-Context Surprisal: An Information-Theoretic Model of Memory Effects in Sentence Processing, Cognitive Science, 44(3).
• Vaswani, Ashish, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin (2017). "Attention is all you need." Advances in neural information processing systems 30.

Distributional methods for Morphology

Olivier Bonami (LLF, U Paris Cité)

In the last two decades, considerable progress has been made in our understanding of morphological systems thanks to the application of quantitative methods to large datasets. These have questioned received wisdom on the organization of morphology by uncovering types of patterns in the data that are beyond the reach of manual examination of small samples. Most of the relevant studies however have focused on matters of form, for lack of an operational way of studying the syntactic and semantic import of morphological contrasts at scale. Things have started to change in the mid-2010s, with seminal studies such as Marelli & Baroni (2015) showing that distributional word vectors could be used as good enough proxies for the content of words for purposes of studying the syntactic and semantic import of morphological relations.

This lecture series will be dedicated to reporting on this general line of research, showcasing important results in four areas:

• The featural organization of inflectional paradigms
• Competition between morphological processes, in both inflection and derivation
• The nature of the inflection-derivation opposition
• The organization of the word formation system
• Along the way I will also discuss some of the methodological challenges raised by the use of distributional methods in the study of morphology.

Crossing frontiers for language acquisition

Chiara Mazzocini (LPL, AMU) & Jonathan Ginzburg (LLF, U Paris Cité)

The course will survey recent attempts at characterising the early communicative abilities of children and how they emerge. After a brief introduction into how dialogue can be formally described, we will consider in turn: (i) laughter—-which emerges around 4 months and which turns out to exemplify complex cognitive abilities children possess many months before they can utter words; children’s laughter emerges piecemeal, increasing in the functions it can achieve and differing radically from the laughter of their carers, (ii) The one word stage and 2-3 word stage—the one word stage is interesting because it exemplifies a wide range of expressive possibilities children manifest by combining words and context—context-based compositionality; quite gradually children add to this initial grammar multi-word constructions (ii) The timeline of adult non-sentential constructions—-we will discuss the emergence of the main non-sentential constructions of adult grammar and how they emerge piecemeal, contradicting the view of ellipsis as driven by a generalised mechanism of phonological omission.

Key Bibliography:

• Mazzocconi M and J Ginzburg  2022`A longitudinal characterisation of typical laughter development in mother-child interaction from 12 to 36 months: formal features and reciprocal responsiveness'. The Journal of Non-Verbal Behavior, pp.1-36.
• Mazzocconi, C, & J Ginzburg,  2023” Growing up laughing: Laughables and pragmatic functions between 12 and 36 months". Journal of Pragmatics, 212, 117-145
• Ginzburg J, C Mazzocconi, and Y Tian 2020 `Laughter as Language'. Glossa: A Journal of General Linguistics, 5(1), 104, 1--51.
• Ginzburg J and D Kolliakou  2018 `Divergently Seeking Clarification: the emergence of clarification interaction'. Topics in Cognitive Science (2018) 10(2): 335-366
• Sara Moradlou, Xiaobei Zheng, Ye Tian, and Jonathan Ginzburg 2021 `Wh-Questions are understood before polar-questions: Evidence from English, German, and Chinese'. In Journal of Child Language 28:1-27.

Introduction to Super Linguistics

Pritty Patel Grosz & Patrick Georges Grosz (Ling department, U Oslo)

Super Linguistics (using ‘super’ in its Latinate meaning ‘beyond’) applies methodology from – and inspired by – formal linguistics to diverse non-standard objects beyond human language. This course focuses on three diverse case studies from this novel research sub-field: (i) the semantics of interactive human gestures (ii) chimpanzee and bonobo gestural meanings and (iii) the relationship between facial expressions and emojis. We connect these newly established objects of study to more traditional objects of study, such as imperative semantics and the semantics-pragmatics of discourse management.