In our next LangSci talk, Steffen Eger, Heisenberg Group Leader of the Natural Language Learning Group (NLLG) at University of Mannheim, will give a talk on „Syntactic language change in English and German„.

The hybrid talk will take place in building A 2.2, room 2.02. and on MS Teams January 25th at 16:15!

In our next LangSci colloquium, Seana Coulson, Professor of Cognitive Science, University of California, San Diego will give a talk on „Distributional Semantics and Embodied Meaning: What do Large Language Models have to say?„.

The talk will take place as a hybrid event in building A2.2, room 1.20.2 and via MS Teams next Thursday at 16:15!

In our next LangSci colloquium, Christian Bentz Assistant Professor at the Department of General Linguistics, University of Tübingen will give a talk on „Language complexity in shallow and deep time„.

The talk will take place as a hybrid event in building A2.2, room 1.20.2 and via MS Teams next Thursday at 16:15!

Last week Elke Teich was invited to give a talk on “Language use in science – convergence, diversity, dynamicity“ at the Symposium on Theoretical Foundations for Interdisciplinarity (November 30 – December 1) which took place at the Helsinki Collegium for Advanced Studies. You can find the abstract below:

Language plays a crucial role in scientific processes as it fulfills the core epistemic functions of denoting and interpreting. As science developed to become an established socio-cultural domain of activity from the 17th century onwards, a specific linguistic code emerged for scientific communication that is characterized by abstractness, technicality, and informational density. While optimizing communication among experts, the specific linguistic properties of scientific language (e.g. complex nominal expressions, implicit logical relations, high lexical density) can hinder communication with non-specialist audiences, learners, and even across disciplines.

Against this background, in my talk I ask to what extent and in which regards scientific language is conducive or inimical to interdisciplinarity. Specifically, I will show results from selected studies highlighting three fundamental aspects of scientific language: convergence, i.e.  the tacit agreement to use specific linguistic features rather than others; diversity, i.e. variation in language use according to discipline; and dynamicity, i.e. continuous adaptation and change over time. The studies are empirical using samples of scientific English from contemporary and historical usage, spanning from the mid-17th century to the 21st century. The data were collected from scientific journals, including the Transactions and Proceedings of the Royal Society of London, and compiled into linguistic corpora according to best practices (meta-data, linguistic annotation; Fischer et al., 2020). For analysis, we applied computational language models (statistical models, embeddings), combined with selected exploratory methods (e.g. clustering, classification, (relative) entropy) for evaluating and interpreting model outputs (Bizzoni et al., 2020, Teich et al., 2021). In terms of linguistic theory, we combine a functional approach (Halliday & Martin, 1993; Matthiessen, 2019) with a formal theory of communication based on Shannon’s (1948) information theory, recently widely applied as a linking theory between language and cognition (Crocker et al., 2015; https://sfb1102.uni-saarland.de/).

Last October C7 has hosted three scholars from the University of Tirana and the Polytechnic of Tirana to collaborate on the Saarbrücken Treebank of Albanian Fiction (STAF). The project aims to develop a computational resource for Albanian, following the Universal Dependency guidelines to manually annotate a collection of sentence for dependency relations, parts of speech and morpho-syntactic features.

Congratulations to Marian Marchal and Iuliia Zaitova for winning the Small Grant Scheme 2023! The two PhD students convinced reviewers and submission board of their proposed research projects. Marian Marchal receives support for a research visit at the University of Edinburgh with the project „Disentangling relation and content prediction using eyetracking-while-reading“. Iuliia Zaitova’s project, submitted in cooperation with Wei Xue, investigates „Validating the effects of additional visual information on spoken perception of mutual intelligibility“.

In our next LangSci talk of the semester, Sharid Loáiciga, Associate Senior Lecturer in Computational Linguistics and Coordinating Director of the Centre for linguistic theory and studies in probability (CLASP) at the University of Gothenburg, will give a talk on „Keeping reference in pre-trained language models and multimodal corpora„.

The hybrid talk will take place in building A 2.2, room 2.02. and on MS Teams November 16th at 16:15!

In our first LangSci talk of the semester, Karin Harbusch, professor at the Institut für Computervisualistik at the University of Koblenz, will give a talk on „Two new German corpora and their use: (1) A German Parallel Clausal Coordinate Ellipsis Corpus, and (2) A Treebank of Leichte Sprache texts„.

The hybrid talk will take place in building A 2.2, room 2.02. and on MS Teams November 9th at 16:15!

„Limits and benefits of information-theoretic perspectives in spoken communication“
Interspeech 2023 Satellite Workshop, Dublin, Aug 19, 2023

Organized by Bistra Andreeva, Omnia Ibrahim, Bernd Möbius, Ivan Yuen (all Saarland University), and James Whang (Seoul National University)

Information-theoretic factors have been identified and used to account for fine-grained phonetic variations in speech, as rational speakers and listeners can adapt speech to be efficient in transmitting information. As conversation unfolds, the distribution and density of information changes over time. It is therefore important to consider to what extent, and how, interlocutors track and predict these informational updates, which in turn guide linguistic and phonetic encoding. The workshop brought together researchers examining the limits and benefits of information-theoretic perspectives in dynamic spoken communication contexts, in both human-human and human-computer interaction.

Abstracts of presentations are available at the workshop website at https://lbit2023.lst.uni-saarland.de/

We are looking forward to recieving your proprosals by October 15, 2023. Detailed information can be found here.