Leistungsanrechnung für Studierende

Bezüglich der Anforderungen für die Bestätigung der Aktiven Teilnahme wenden Sie sich bitte an die Organisatorin / den Organisator des LAK.

Archiv

Andrea Bruera

Polysemous neurons and grandmother cells: language models as tools to study fine-grained semantics in the brain

In this talk I will try to show how we can use language models and brain imaging to address traditional questions in the field of semantics and linguistics, like polysemy or the meaning of proper names. I will go through some experimental results from my Ph.D. and current postdoc that go in this direction. Along the way I will provide brief introductions to the different neuroscientific techniques used (EEG, fMRI), as well as to language models and some theoretical grounding. I hope this will provide a common ground for a shared discussion, taking into consideration multiple points of view on the value, as well as the downsides, of cutting-edge approaches to meaning and language.

Mathias Barthel

The Psycholinguistics of Conversation: Fusion Experimental Control and Unscripted Interaction

Turn-taking in conversation requires speakers to process incoming speech while simultaneously planning their own utterances. This talk presents research that investigates how interlocutors manage these concurrent cognitive demands. Using a multi-method approach combining controlled experiments and analyses of natural conversation, it demonstrates that speakers commonly begin planning their responses while still listening to their interlocutor, leading to partial overlap of comprehension and production processes. Laboratory evidence targeting different levels of linguistic processing reveals that comprehending incoming speech can be shallow due to high cognitive load during phases of overlapping planning. Pupillometric data from natural conversation shows that speakers nonetheless systematically engage in early response planning despite these processing costs, at least at turn transitions in which they are (co-)selected as a next speaker. This planning-in-overlap strategy appears to serve multiple conversational goals: it enables shorter gaps between turns, helps to avoid unwanted implications of delayed responses, and secures speaking opportunities in multi-party settings. The combination of these findings suggests that speakers prioritise timely turn transitions over processing efficiency and depth, supporting a planning-prioritised model of turn-taking in conversation. The results demonstrate how experimental and observational methods can complement each other to advance our understanding of the cognitive mechanisms underlying smooth conversation.

Manfred Krifka

Modelling Discourse Moves and Countermoves in Commitment Spaces

Recent work in formal semantics and pragmatics has developed models of how the update of the common ground in conversation is negotiated. One particularly popular framework is the Table Model of Farkas & Bruce (2010) that assumes an automaton next to the common ground with transition rules and the ability to store or delete propositions. I will discuss shortcomings of this framework and propose as an alternative Commitment Spaces (CS, Krifka 2015) that can be understood as traditional common grounds with possible continuations. Starting from the proposal of Krifka (2022), I will first show how reactions like (1) can be modelled. For example, in (1a) Speaker A adds a commitment to the truth of a proposition p to the CS followed by a disjunction of an acceptance of p or an objection against accepting p by B. The denying reaction B: No would lead to non-acceptance of p into the common ground, but the commitment of A to the truth of p would remain.

1.a) A: Tom lives in \Borneo. B: Yes. / Okay. / No.

b) A: Does Tom live in /Borneo?

B: Yes. / No. / *Okay.

c) A: Does Tom live in /Borneo or in \Sumatra?

B: Tom lives in BORneo. / *Yes. / *No. / *Okay.

d) A: Does Tom live in /Borneo or \not?

B: He does. / He doesn’t. ?Yes. / ?No. / *Okay.

I will then discuss options for the modelling of other discourse moves: declarative questions (2a), assertions disjoined with questions (2b), question tags (2c), and high and low negation questions (2d,e).

2.a) A: Tom lives in /Borneo?

b) A: Tom lives in \Borneo, or does he live in /Sumatra?

c) A: Tom lives in \Borneo, doesn’t he?

d) A: Doesn’t Tom live in /Borneo?

e) A: Does Tom /not live in Borneo?

I will furthermore explore options to model certain countermoves: incredulity questions (3a), deaccented doubting questions (3b) and metalinguistic negations of the type (3c) that address the locutionary act.

3.a) A: Tom lives in Borneo.

B: Tom lives in \Bor/neo?

b) A: Tom lebt in Borneo.

B: Wie, Tom lebt in Borneo? (German)

c) A: Tom lives in BorNEo.

B: Tom doesn’t live in BorNEo, he lives in BORneo.

I will finally touch on what it would take to model retractions as in (4) (cf. Bussière-Caraes et al. 2022).

4) A: Tom lives in Borneo. B: Interesting. Does he like it there? A: Well — let me take this back. He lives in Sumatra.

Bussière-Caraes, Lwenn, Incurvati, Luca, Sbardolini, Giorgio, & Schlöder, Julian. 2022. Nevermind: on retraction as a speech act. In Dan Zeman & M Hincu (eds.) Retraction Matters. Springer.

Farkas, Donka, & Kim Bruce. 2010. On reacting to assertions and polar questions. Journal of Semantics 27(1).

Krifka, Manfred 2015. Bias in Commitment Space Semantics: Declarative questions, negated questions, and question tags. SALT 25.

Krifka, Manfred 2022. Adjacency pairs in common ground update: Assertions, questions, greetings, offers, commands. SemDial 26.

UNIVERSITÄT ZU KÖLN

Philosophische FakultätInstitut für deutsche Sprache und Literatur I

Linguistischer Arbeitskreis

Programm Wintersemester 2025-26

Leistungsanrechnung für Studierende

Archiv