Overview
The position is part of the DFG-funded project Learning Linguistic
Inferences and Their Alternatives (PIs: Jacopo Romoli and Yulia Zinova,
Heinrich-Heine-Universität Düsseldorf). The project investigates how
language models learn linguistic inferences — including implicatures,
presuppositions, implicated presuppositions, free choice, and
distributive inferences — and whether training on one inference type
facilitates learning of others. Combining theoretical, experimental, and
computational methods, the project addresses foundational questions
about the semantics–pragmatics interface and what language models
actually learn about meaning. The project involves constructing novel
datasets, running behavioral experiments with human participants,
probing inference derivation and its ingredients in both humans and
language models as well as training and fine-tuning models using
collected data.
Broader context: LaSTing Priority Programme
This position is embedded within the DFG Priority Programme LaSTing (SPP
2556): “Robust Assessment & Safe Applicability of Language Modeling:
Foundations for a New Field of Language Science & Technology”
(https://www.lasting-spp.org) LaSTing brings together researchers across
linguistics, cognitive science, and language technology to advance our
understanding of language models for safer and more principled use,
especially in the language sciences. The programme fosters
interdisciplinary community-building, networking, and the training of a
new generation of researchers at the interface of language science and
language technology.
*Position Description*:
We invite applications for a full-time Postdoctoral Researcher (3 years)
to join the project team. The postdoc will play a central role across
the project’s work packages, contributing to the computational
components of the project as well as to the design and implementation of
online behavioral experiments, dataset construction, and data analysis.
The postdoc will also contribute to the writing and dissemination of
research outputs, and will collaborate closely with both PIs and with
Mercator Fellow Paul Marty (University of Lisbon).
Company:
Heinrich-Heine-Universität Düsseldorf
Qualifications:
Minimum Education: PhD
Specialty Areas: Computational Linguistics; Pragmatics;
Psycholinguistics; Semantics
*Required Qualifications*:
– PhD in Linguistics or a closely related field
– Expertise in computational linguistics
– Ability to engage with both the experimental and computational aspects
of the project
*Desirable Qualifications*:
– Experience with experimental methods in psycholinguistics
– Background in formal semantics and pragmatics and familiarity with
formal approaches to pragmatic inferences
– Experience working with large language models (LLMs)
– Familiarity with natural language inference (NLI) tasks and datasets
Language requirements:
Specialty Language(s): English (eng)
Educational level:
Ph. D.
