Publications

In this paper, we present examples of current research foci and results of analyses of the Collaborative Research Centre subproject “Translation as Rational Communication” that focuses on the specific linguistic properties of interpreted and translated texts distinguishing them from original productions. We describe the creation and annotation of EPIC-UdS, a multilingual corpus of simultaneous interpreting for English, German and Spanish. We give an overview of the corpus variants and explore various applications of the corpus. Building on the ‘translationese’ hypothesis from translation studies, we investigate whether simultaneous interpreted language resembles written translated language with regard to specific features or whether it carries ‘interpretese’ features as a result of a unique language transfer process so that traces of these features tend to occur in all simultaneously interpreted texts and distinguish them from other texts. For instance, with regard to the simplification hypothesis put forward by various translation scholars, we can observe in interpreted texts that there is a tendency towards syntactic simplification. Another analysis shows that interpreted language is characterised by a particular use of discourse particles. EPIC-UdS contains rich metadata and fine-grained linguistic annotations tailored for diverse applications across a broad range of linguistic subfields. This paper provides the first overview in German on the EPIC-UdS corpus with the aim of bringing together results from individual studies, such as Bizzoni/Teich 2019; Karakanta/Vela/Teich 2018; Lapshinova-Koltunski et al. 2021a; Lapshinova-Koltunski/Przybyl/Bizzoni 2021b; Lapshinova-Koltunski/Pollkläsener/Przybyl 2022; Przybyl/Teich 2021; Pollkläsener 2021; Przybyl et al. 2022a, 2022b, and giving a concise and informative summary of applications and impacts of the project.

• 10.15203-99106-165-6-16     (0.49MB)
• https://www.uibk.ac.at/iup/buecher/9783991061656.html
• https://doi.org/10.15203/99106-165-6-16

@incollection{Menzel_2025_EPIC,
title = {EPIC-UdS – ein mehrsprachiges Korpus als Grundlage f{\"u}r die korpusbasierte Dolmetsch- und {\"U}bersetzungswissenschaf},
author = {Katrin Menzel and Heike Przybyl and Ekaterina Lapshinova-Koltunski},
editor = {Astrid Schmidhofer and Mar{\'i}a {\'A}ngeles Recio Ariza},
url = {https://www.uibk.ac.at/iup/buecher/9783991061656.html},
doi = {https://doi.org/10.15203/99106-165-6-16},
year = {2025},
date = {2025},
booktitle = {Zukunftsperspektiven in der Translationswissenschaft Ausgew{\"a}hlte Beitr{\"a}ge der Translata IV},
isbn = {978-3-99106-165-6},
pages = {323-349},
publisher = {innsbruck university press},
address = {Innsbruck},
abstract = {In this paper, we present examples of current research foci and results of analyses of the Collaborative Research Centre subproject “Translation as Rational Communication” that focuses on the specific linguistic properties of interpreted and translated texts distinguishing them from original productions. We describe the creation and annotation of EPIC-UdS, a multilingual corpus of simultaneous interpreting for English, German and Spanish. We give an overview of the corpus variants and explore various applications of the corpus. Building on the ‘translationese’ hypothesis from translation studies, we investigate whether simultaneous interpreted language resembles written translated language with regard to specific features or whether it carries ‘interpretese’ features as a result of a unique language transfer process so that traces of these features tend to occur in all simultaneously interpreted texts and distinguish them from other texts. For instance, with regard to the simplification hypothesis put forward by various translation scholars, we can observe in interpreted texts that there is a tendency towards syntactic simplification. Another analysis shows that interpreted language is characterised by a particular use of discourse particles. EPIC-UdS contains rich metadata and fine-grained linguistic annotations tailored for diverse applications across a broad range of linguistic subfields. This paper provides the first overview in German on the EPIC-UdS corpus with the aim of bringing together results from individual studies, such as Bizzoni/Teich 2019; Karakanta/Vela/Teich 2018; Lapshinova-Koltunski et al. 2021a; Lapshinova-Koltunski/Przybyl/Bizzoni 2021b; Lapshinova-Koltunski/Pollkl{\"a}sener/Przybyl 2022; Przybyl/Teich 2021; Pollkl{\"a}sener 2021; Przybyl et al. 2022a, 2022b, and giving a concise and informative summary of applications and impacts of the project.},
pubstate = {published},
type = {incollection}
}

Copy BibTeX to Clipboard

Contrastive focus accent in spoken language indicates that alternatives to the focused element are relevant for interpretation. The sentence “Could I have some TEA, please?”, with contrastive accent on tea, is probably the response to an offer of several alternative beverages. Research shows that contrastive focus accent helps listeners remember such alternatives. We investigated the time-course of and mechanisms behind the effects of contrastive focus accent on memory with a variant of the subsequent memory effect paradigm with eventrelated potentials (ERPs). The ERP time-locked to a critical word was more positive-going if participants remembered two earlier mentioned alternatives than just one, but only if the critical word had been contrastively accented. This effect further was only observed when the critical word itself was remembered. These findings suggest that contrastive focus marking triggers a reinstatement of the preceding sentence context (retrieval practice) by which these elements are prioritised in memory

• Spaleck.etal.2025     (1.67MB)
• https://doi.org/10.1080/23273798.2025.2503906
• https://doi.org/10.1080/23273798.2025.2503906

@article{Spaleck.etal.2025,
title = {Contrastive focus accent retroactively modulates memory for focus alternatives: evidence from event-related potentials},
author = {Katharina Spalek and Regine Bader and Sandra Glaser and Gerrit H{\"o}ltje and Axel Mecklinger},
url = {https://doi.org/10.1080/23273798.2025.2503906},
doi = {https://doi.org/10.1080/23273798.2025.2503906},
year = {2025},
date = {2025},
journal = {Language, Cognition and Neuroscience},
pages = {1-18},
publisher = {Routledge},
abstract = {Contrastive focus accent in spoken language indicates that alternatives to the focused element are relevant for interpretation. The sentence “Could I have some TEA, please?”, with contrastive accent on tea, is probably the response to an offer of several alternative beverages. Research shows that contrastive focus accent helps listeners remember such alternatives. We investigated the time-course of and mechanisms behind the effects of contrastive focus accent on memory with a variant of the subsequent memory effect paradigm with eventrelated potentials (ERPs). The ERP time-locked to a critical word was more positive-going if participants remembered two earlier mentioned alternatives than just one, but only if the critical word had been contrastively accented. This effect further was only observed when the critical word itself was remembered. These findings suggest that contrastive focus marking triggers a reinstatement of the preceding sentence context (retrieval practice) by which these elements are prioritised in memory},
pubstate = {published},
type = {article}
}

Copy BibTeX to Clipboard

@article{scholman_etal_2025_disconaija,
title = {DiscoNaija: A discourse-annotated parallel Nigerian Pidgin-English corpus},
author = {Merel Scholman and Marian Marchal and AriaRay Brown and Vera Demberg},
year = {2025},
date = {2025},
journal = {Language Resources and Evaluation},
pubstate = {forthcoming},
type = {article}
}

Copy BibTeX to Clipboard

This study analyzes the attention patterns of fine-tuned encoder-only models based on the BERT architecture (BERT-based models) towards two distinct types of Multiword Expressions (MWEs): idioms and microsyntactic units (MSUs). Idioms present challenges in semantic non-compositionality, whereas MSUs demonstrate unconventional syntactic behavior that does not conform to standard grammatical categorizations. We aim to understand whether fine-tuning BERT-based models on specific tasks influences their attention to MWEs, and how this attention differs between semantic and syntactic tasks. We examine attention scores to MWEs in both pre-trained and fine-tuned BERT-based models. We utilize monolingual models and datasets in six Indo-European languages —  English, German, Dutch, Polish, Russian, and Ukrainian. Our results show that fine-tuning significantly influences how models allocate attention to MWEs. Specifically, models fine-tuned on semantic tasks tend to distribute attention to idiomatic expressions more evenly across layers. Models fine-tuned on syntactic tasks show an increase in attention to MSUs in the lower layers, corresponding with syntactic processing requirements.

• 2025.findings-naacl.228     (0.31MB)
• https://aclanthology.org/2025.findings-naacl.228/

@inproceedings{zaitova-etal-2025-attention,
title = {Attention on Multiword Expressions: A Multilingual Study of BERT-based Models with Regard to Idiomaticity and Microsyntax},
author = {Iuliia Zaitova and Vitalii Hirak and Badr M. Abdullah and Dietrich Klakow and Bernd M{\"o}bius and Tania Avgustinova},
editor = {Luis Chiruzzo and Alan Ritter and Lu Wang},
url = {https://aclanthology.org/2025.findings-naacl.228/},
year = {2025},
date = {2025},
booktitle = {Findings of the Association for Computational Linguistics: NAACL 2025},
isbn = {979-8-89176-195-7},
pages = {4083-4092},
publisher = {Association for Computational Linguistics},
address = {Albuquerque, New Mexico},
abstract = {This study analyzes the attention patterns of fine-tuned encoder-only models based on the BERT architecture (BERT-based models) towards two distinct types of Multiword Expressions (MWEs): idioms and microsyntactic units (MSUs). Idioms present challenges in semantic non-compositionality, whereas MSUs demonstrate unconventional syntactic behavior that does not conform to standard grammatical categorizations. We aim to understand whether fine-tuning BERT-based models on specific tasks influences their attention to MWEs, and how this attention differs between semantic and syntactic tasks. We examine attention scores to MWEs in both pre-trained and fine-tuned BERT-based models. We utilize monolingual models and datasets in six Indo-European languages —  English, German, Dutch, Polish, Russian, and Ukrainian. Our results show that fine-tuning significantly influences how models allocate attention to MWEs. Specifically, models fine-tuned on semantic tasks tend to distribute attention to idiomatic expressions more evenly across layers. Models fine-tuned on syntactic tasks show an increase in attention to MSUs in the lower layers, corresponding with syntactic processing requirements.},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard

This study investigates the relation between the surprisal (or unpredictability) of linguistic items and anticipatory voicing assimilation in Bulgarian obstruents. Using a corpus of speech read by 140 Bulgarian speakers and wordlevel language models, we calculated unigram surprisal for word forms ending in obstruents followed by a word-initial obstruent of the opposite underlying [±voice] specification. Percentage of voicing was computed for 9,712 word-final obstruents. Linear mixed models were used to determine the effect of surprisal on the percentage of voicing in assimilating obstruents. The results confirm that Bulgarian obstruents do indeed in general assimilate to the voicing of a following obstruent: voiceless obstruents become voiced before voiced ones, while voiced obstruents are devoiced before voiceless ones. Crucially, however, surprisal had a significant effect on the percentage of voicing found in assimilating obstruents: in words with higher surprisal values, we found significantly lower degrees of voicing in voiceless obstruents before voiced ones, as well as significantly less devoicing of voiced obstruents before voiceless ones. This shows that assimilation is stronger in lowsurprisal words, while in high-surprisal words speakers attempt to maintain the underlying [±voice] specification of an obstruent to a higher degree. Our findings add to a growing body of research that demonstrates that processes once thought of as entirely categorical in fact exhibit gradient variation in fine phonetic detail, which is attributable to speakers’ awareness of statistical patterns in language use and their response to the predictability of linguistic items in maintaining a balance between phonetic encoding and information density.

• sabev_etal_essv2025     (2.58MB)
• https://www.essv.de/paper.php?id=1249

@inproceedings{sabev_etal_essv2025,
title = {The effects of lexical frequency on anticipatory voice assimilation in Bulgarian obstruents},
author = {Mitko Sabev and Bistra Andreeva and Bernd M{\"o}bius and Ivan Yuen and Omnia Ibrahim},
editor = {Sven Grawunder},
url = {https://www.essv.de/paper.php?id=1249},
year = {2025},
date = {2025},
booktitle = {Studientexte zur Sprachkommunikation: Elektronische Sprachsignalverarbeitung 2025},
isbn = {978-3-95908-803-9},
issn = {0940-6832},
pages = {163-169},
publisher = {TUDpress},
address = {Dresden},
abstract = {This study investigates the relation between the surprisal (or unpredictability) of linguistic items and anticipatory voicing assimilation in Bulgarian obstruents. Using a corpus of speech read by 140 Bulgarian speakers and wordlevel language models, we calculated unigram surprisal for word forms ending in obstruents followed by a word-initial obstruent of the opposite underlying [±voice] specification. Percentage of voicing was computed for 9,712 word-final obstruents. Linear mixed models were used to determine the effect of surprisal on the percentage of voicing in assimilating obstruents. The results confirm that Bulgarian obstruents do indeed in general assimilate to the voicing of a following obstruent: voiceless obstruents become voiced before voiced ones, while voiced obstruents are devoiced before voiceless ones. Crucially, however, surprisal had a significant effect on the percentage of voicing found in assimilating obstruents: in words with higher surprisal values, we found significantly lower degrees of voicing in voiceless obstruents before voiced ones, as well as significantly less devoicing of voiced obstruents before voiceless ones. This shows that assimilation is stronger in lowsurprisal words, while in high-surprisal words speakers attempt to maintain the underlying [±voice] specification of an obstruent to a higher degree. Our findings add to a growing body of research that demonstrates that processes once thought of as entirely categorical in fact exhibit gradient variation in fine phonetic detail, which is attributable to speakers’ awareness of statistical patterns in language use and their response to the predictability of linguistic items in maintaining a balance between phonetic encoding and information density.},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard

Настоящото изследване разглежда връзката между честотността (предсказуемостта) на езиковите единици и регресивната асимилация на обструентите по признака звучност в българския език. Въз основа на данните от речеви корпус и езикови модели на ниво словоформа изчисляваме стойността на изненадата на думи с краесловни съгласни /t/, /ʃ/ и /x/ в състава на формообразуващи и словообразуващи морфеми. Чрез смесен линеен модел (LMM) анализираме как предсказуемостта, звучността на следващия обструент и типът морфема влияят върху реализираната звучност на съгласните. Резултатите показват, че езиковата предсказуемост влияе върху прецизността на артикулацията, като модулира степента на регресивната асимилация в зависимост от морфологичния контекст. Установяваме, че обструентите в по-малко предсказуемите думи се произнасят с по-прецизна артикулация, отколкото обструентите в по-предсказуемите думи, особено в състава на словообразуващи морфеми. Това подчертава комплексната динамика между информационната плътност, морфологията и артикулационния процес, което може да даде основа за по-нататъшни изследвания в областта на фонетиката и фонологията.

---

This study investigates the relationship between the frequency (or predictability) of linguistic units and the anticipatory voice assimilation of obstruents in Bulgarian. Using a speech corpus and word-level language models, we calculate the surprisal of word forms ending in the consonants /t/, /ʃ/, and /x/ in inflectional and lexical morphemes. We then employ a linear mixed model (LMM) to analyse how surprisal, the voicing of the following obstruent, and morpheme type affect the voicing realised in the examined consonants. The findings demonstrate that linguistic predictability affects articulatory precision by modulating the degree of anticipatory assimilation, with different effect sizes for different morpheme types. More specifically, obstruents in less predictable words are articulated with greater precision than those in more predictable words, especially within lexical morphemes. This reveals a complex interaction between information density, morphology and articulation, providing avenues for further research in phonetics and phonology.

• 0_0_25_BISTRA ANDREEVA, MITKO SABEV_357-369_BG     (0.68MB)
• https://www.balgarskiezik.eu/p-2025/0_0_25_BISTRA%20ANDREEVA,%20MITKO%20SABEV_357-369_BG.pdf

@article{AndreevaSabev2025,
title = {Ефектът На Лексикалната Честотност И Типа Морфема Върху Регресивната Асимилация На Българските Обструенти По Признака Звучност (The Effects Of Lexical Frequency And Morpheme Type On Anticipatory Voice Assimilation In Bulgarian Obstruents)},
author = {Bistra Andreeva and Mitko Sabev},
url = {https://www.balgarskiezik.eu/p-2025/0_0_25_BISTRA%20ANDREEVA,%20MITKO%20SABEV_357-369_BG.pdf},
year = {2025},
date = {2025},
journal = {БЪЛГАРСК ЕЗИК, ПРИЛОЖЕНИЕ / BULGARIAN LANGUAGE, SUPPLEMENT},
pages = {357–369},
volume = {72},
abstract = {Настоящото изследване разглежда връзката между честотността (предсказуемостта) на езиковите единици и регресивната асимилация на обструентите по признака звучност в българския език. Въз основа на данните от речеви корпус и езикови модели на ниво словоформа изчисляваме стойността на изненадата на думи с краесловни съгласни /t/, /ʃ/ и /x/ в състава на формообразуващи и словообразуващи морфеми. Чрез смесен линеен модел (LMM) анализираме как предсказуемостта, звучността на следващия обструент и типът морфема влияят върху реализираната звучност на съгласните. Резултатите показват, че езиковата предсказуемост влияе върху прецизността на артикулацията, като модулира степента на регресивната асимилация в зависимост от морфологичния контекст. Установяваме, че обструентите в по-малко предсказуемите думи се произнасят с по-прецизна артикулация, отколкото обструентите в по-предсказуемите думи, особено в състава на словообразуващи морфеми. Това подчертава комплексната динамика между информационната плътност, морфологията и артикулационния процес, което може да даде основа за по-нататъшни изследвания в областта на фонетиката и фонологията.

---

This study investigates the relationship between the frequency (or predictability) of linguistic units and the anticipatory voice assimilation of obstruents in Bulgarian. Using a speech corpus and word-level language models, we calculate the surprisal of word forms ending in the consonants /t/, /ʃ/, and /x/ in inflectional and lexical morphemes. We then employ a linear mixed model (LMM) to analyse how surprisal, the voicing of the following obstruent, and morpheme type affect the voicing realised in the examined consonants. The findings demonstrate that linguistic predictability affects articulatory precision by modulating the degree of anticipatory assimilation, with different effect sizes for different morpheme types. More specifically, obstruents in less predictable words are articulated with greater precision than those in more predictable words, especially within lexical morphemes. This reveals a complex interaction between information density, morphology and articulation, providing avenues for further research in phonetics and phonology.},
pubstate = {published},
type = {article}
}

Copy BibTeX to Clipboard

We present a general analysis of the lexical and grammatical differences between Brazilian and European Portuguese by applying entropy measures, including Kullback-Leibler divergence and word order entropy, across various linguistic levels. Using a parallel corpus of BP and EP sentences translated from English, we quantified these differences and identified characteristic phenomena underlying the divergences between the two varieties. The highest divergence was observed at the lexical level due to word pairs unique to each variety but also related to grammatical distinctions. Furthermore, the analysis of parts-of-speech (POS), dependency relations, and POS tri-grams provided information concerning distinctive grammatical constructions. Finally, the word order entropy analysis revealed that while most of the syntactic features analysed showed similar patterns across BP and EP, specific word order preferences were still apparent.

• 2025.vardial-1.2     (0.3MB)
• https://aclanthology.org/2025.vardial-1.2/

@inproceedings{alves-2025-information,
title = {Information Theory and Linguistic Variation: A Study of Brazilian and European Portuguese},
author = {Diego Alves},
editor = {Yves Scherrer and Tommi Jauhiainen and Nikola Ljubešić and Preslav Nakov and Jorg Tiedemann and Marcos Zampieri},
url = {https://aclanthology.org/2025.vardial-1.2/},
year = {2025},
date = {2025},
booktitle = {Proceedings of the 12th Workshop on NLP for Similar Languages, Varieties and Dialects},
pages = {9-19},
publisher = {Association for Computational Linguistics},
address = {Abu Dhabi, UAE},
abstract = {We present a general analysis of the lexical and grammatical differences between Brazilian and European Portuguese by applying entropy measures, including Kullback-Leibler divergence and word order entropy, across various linguistic levels. Using a parallel corpus of BP and EP sentences translated from English, we quantified these differences and identified characteristic phenomena underlying the divergences between the two varieties. The highest divergence was observed at the lexical level due to word pairs unique to each variety but also related to grammatical distinctions. Furthermore, the analysis of parts-of-speech (POS), dependency relations, and POS tri-grams provided information concerning distinctive grammatical constructions. Finally, the word order entropy analysis revealed that while most of the syntactic features analysed showed similar patterns across BP and EP, specific word order preferences were still apparent.},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard

@article{Delogu-etal-2025,
title = {On the biphasic nature of the N400-P600 complex underlying language comprehension},
author = {Francesca Delogu and Christoph Aurnhammer and Harm Brouwer and Matthew W. Crocker},
year = {2025},
date = {2025},
journal = {Brain & Cognition},
pubstate = {forthcoming},
type = {article}
}

Copy BibTeX to Clipboard

Interpreting implicit discourse relations involves complex reasoning, requiring the integration of semantic cues with background knowledge, as overt connectives like “because” or “then” are absent. These relations often allow multiple interpretations, best represented as distributions. In this study, we compare two established methods that crowdsource implicit discourse relation annotation by connective insertion: a free-choice approach, which allows annotators to select any suitable connective, and a forced-choice approach, which asks them to select among a set of predefined options. Specifically, we re-annotate the whole DiscoGeM 1.0 corpus – initially annotated with the free-choice method – using the forced-choice approach. The free-choice approach allows for flexible and intuitive insertion of various connectives, which are context-dependent. Comparison among over 130,000 annotations, however, shows that the free-choice strategy produces less diverse annotations, often converging on common labels. Analysis of the results reveals the interplay between task design and the annotators’ abilities to interpret and produce discourse relations.

• 2025.comedi-1.2     (0.28MB)
• https://aclanthology.org/2025.comedi-1.2/

@inproceedings{yung-demberg-2025-crowdsourcing ,
title = {On Crowdsourcing Task Design for Discourse Relation Annotation},
author = {Frances Pik Yu Yung and Vera Demberg},
editor = {Michael Roth and Dominik Schlechtweg},
url = {https://aclanthology.org/2025.comedi-1.2/},
year = {2025},
date = {2025},
booktitle = {Proceedings of Context and Meaning: Navigating Disagreements in NLP Annotation},
pages = {12-19},
publisher = {International Committee on Computational Linguistics},
address = {Abu Dhabi, UAE},
abstract = {Interpreting implicit discourse relations involves complex reasoning, requiring the integration of semantic cues with background knowledge, as overt connectives like “because” or “then” are absent. These relations often allow multiple interpretations, best represented as distributions. In this study, we compare two established methods that crowdsource implicit discourse relation annotation by connective insertion: a free-choice approach, which allows annotators to select any suitable connective, and a forced-choice approach, which asks them to select among a set of predefined options. Specifically, we re-annotate the whole DiscoGeM 1.0 corpus - initially annotated with the free-choice method - using the forced-choice approach. The free-choice approach allows for flexible and intuitive insertion of various connectives, which are context-dependent. Comparison among over 130,000 annotations, however, shows that the free-choice strategy produces less diverse annotations, often converging on common labels. Analysis of the results reveals the interplay between task design and the annotators’ abilities to interpret and produce discourse relations.

},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard

Recent studies suggest that language users generate and maintain multiple predictions in parallel, especially in tasks that explicitly instruct participants to generate predictions. Here, we investigated the possibility of parallel gradedness of linguistic predictions in a simple reading task, using a new measure—imbalance—that captures the probabilistic difference between multiple sentence completions. We focus on prenominal gender-marked articles in German to obtain prediction-specific effects. Native speakers of German read predictable or unpredictable gender-marked nouns that were preceded by prediction-consistent or -inconsistent prenominal articles. Sentence frames either biased expectations more strongly toward the most likely continuation of the sentence, or balanced expectations between the first and second most likely continuation. The results showed reading facilitation for gender-marked articles when sentences were more biased but slowing when sentences were more balanced, irrespective of article predictability. We conclude that readers issue multiple prenominal predictions and weigh them according to their likelihood, providing evidence for parallel gradedness of prenominal predictions. The results are discussed in light of theoretical models on prediction and rational sentence processing.

• eScholarship UC item 7g30m0th     (1.74MB)
• https://escholarship.org/uc/item/7g30m0th
• https://doi.org/10.5070/G6011.1636

@article{HäuserBorovsky2025,
title = {Got it right up front? Further evidence for parallel graded prediction during prenominal article processing in a self-paced reading study},
author = {Katja H{\"a}user and Arielle Borovsky},
url = {https://escholarship.org/uc/item/7g30m0th},
doi = {https://doi.org/10.5070/G6011.1636},
year = {2025},
date = {2025},
journal = {Glossa Psycholinguistics},
volume = {4},
number = {1},
abstract = {Recent studies suggest that language users generate and maintain multiple predictions in parallel, especially in tasks that explicitly instruct participants to generate predictions. Here, we investigated the possibility of parallel gradedness of linguistic predictions in a simple reading task, using a new measure—imbalance—that captures the probabilistic difference between multiple sentence completions. We focus on prenominal gender-marked articles in German to obtain prediction-specific effects. Native speakers of German read predictable or unpredictable gender-marked nouns that were preceded by prediction-consistent or -inconsistent prenominal articles. Sentence frames either biased expectations more strongly toward the most likely continuation of the sentence, or balanced expectations between the first and second most likely continuation. The results showed reading facilitation for gender-marked articles when sentences were more biased but slowing when sentences were more balanced, irrespective of article predictability. We conclude that readers issue multiple prenominal predictions and weigh them according to their likelihood, providing evidence for parallel gradedness of prenominal predictions. The results are discussed in light of theoretical models on prediction and rational sentence processing.},
pubstate = {published},
type = {article}
}

Copy BibTeX to Clipboard

The diversity in readers’ cognitive abilities, including working memory capacity and prior knowledge, necessitates texts that align with individual comprehension levels. We address the challenge of rewriting text to match readers’ unique needs, approximating readers to specific grade levels. Unlike prior approaches that rely on fine-tuned models and large training datasets, our method leverages in-context learning (ICL), making it effective in data-sparse scenarios. By precisely controlling linguistic features such as syntactic depth, our approach delivers tailored rewrites aligned with specific grade levels. We demonstrate state-of-the-art performance in generating grade-specific adaptations, highlighting the potential of ICL-based methods to enhance text accessibility and inclusivity.

• Thillainathan2025Controllable     (0.39MB)
• https://ai4ed.cc/workshops/aaai2025

@inproceedings{Thillainathan2025Controllable,
title = {Controllable Text Adaptation Using In-context Learning with Linguistic Features},
author = {Sarubi Thillainathan and Alexander Koller},
url = {https://ai4ed.cc/workshops/aaai2025},
year = {2025},
date = {2025},
booktitle = {AAAI2025 AI for Education - Tools, Opportunities, and Risks in the Generative AI Era},
abstract = {The diversity in readers’ cognitive abilities, including working memory capacity and prior knowledge, necessitates texts that align with individual comprehension levels. We address the challenge of rewriting text to match readers’ unique needs, approximating readers to specific grade levels. Unlike prior approaches that rely on fine-tuned models and large training datasets, our method leverages in-context learning (ICL), making it effective in data-sparse scenarios. By precisely controlling linguistic features such as syntactic depth, our approach delivers tailored rewrites aligned with specific grade levels. We demonstrate state-of-the-art performance in generating grade-specific adaptations, highlighting the potential of ICL-based methods to enhance text accessibility and inclusivity.},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard

Language models are able to handle compositionality and, to some extent, noncompositional phenomena such as semantic idiosyncrasy, a feature most prominent in the case of idioms. This work introduces the MultiCoPIE corpus that includes potentially idiomatic expressions in Catalan, Italian, and Russian, extending the language coverage of PIE corpus data. The new corpus provides additional linguistic features of idioms, such as their semantic compositionality, part-of-speech of idiom head as well as their corresponding idiomatic expressions in English. With this new resource at hand, we first fine-tune an XLM-RoBERTa model to classify figurative and literal usage of potentially idiomatic expressions in English. We then study cross-lingual transfer to the languages represented in the MultiCoPIE corpus, evaluating the model’s ability to generalize an idiom-related task to languages not seen during fine-tuning. We show the effect of ‘cross-lingual lexical overlap’: the performance of the model, fine-tuned on English idiomatic expressions and tested on the MultiCoPIE languages, increases significantly when classifying ‘shared idioms’— idiomatic expressions that have direct counterparts in English with similar form and meaning. While this observation raises questions about the generalizability of cross-lingual learning, the results from experiments on PIEs demonstrate strong evidence of effective cross-lingual transfer, even when accounting for idioms similar across languages.

• MultiCoPIE for cross-lingual PIE disambiguation     (0.25MB)
• https://multiword.org/mwe2025/

@inproceedings{Sentsova-etal-2025,
title = {MultiCoPIE: A Multilingual Corpus of Potentially Idiomatic Expressions for Cross-lingual PIE Disambiguation},
author = {Uliana Sentsova and Debora Ciminari and Josef van Genabith and Cristina Espa{\~n}a-Bonet},
url = {https://multiword.org/mwe2025/},
year = {2025},
date = {2025},
booktitle = {21st Workshop on Multiword Expressions (MWE 2025) @NAACL2025},
address = {Albuquerque, New Mexico, U.S.A.},
abstract = {Language models are able to handle compositionality and, to some extent, noncompositional phenomena such as semantic idiosyncrasy, a feature most prominent in the case of idioms. This work introduces the MultiCoPIE corpus that includes potentially idiomatic expressions in Catalan, Italian, and Russian, extending the language coverage of PIE corpus data. The new corpus provides additional linguistic features of idioms, such as their semantic compositionality, part-of-speech of idiom head as well as their corresponding idiomatic expressions in English. With this new resource at hand, we first fine-tune an XLM-RoBERTa model to classify figurative and literal usage of potentially idiomatic expressions in English. We then study cross-lingual transfer to the languages represented in the MultiCoPIE corpus, evaluating the model’s ability to generalize an idiom-related task to languages not seen during fine-tuning. We show the effect of ‘cross-lingual lexical overlap’: the performance of the model, fine-tuned on English idiomatic expressions and tested on the MultiCoPIE languages, increases significantly when classifying ‘shared idioms’— idiomatic expressions that have direct counterparts in English with similar form and meaning. While this observation raises questions about the generalizability of cross-lingual learning, the results from experiments on PIEs demonstrate strong evidence of effective cross-lingual transfer, even when accounting for idioms similar across languages.},
pubstate = {forthcoming},
type = {inproceedings}
}

Copy BibTeX to Clipboard

• 2025_nodalida_1_2     (0.35MB)
• https://dspace.ut.ee/items/ef26bd7f-e708-41b3-b5c8-84cf8057ab71

@inproceedings{Alves-2025,
title = {Diachronic Analysis of Phrasal Verbs in English Scientific Writing},
author = {Diego Alves},
url = {https://dspace.ut.ee/items/ef26bd7f-e708-41b3-b5c8-84cf8057ab71},
year = {2025},
date = {2025},
booktitle = {Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025)},
publisher = {University of Tartu Library},
address = {Tallinn, Estonia},
abstract = {

},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard

• 2025_nodalida_1_34     (0.5MB)
• https://dspace.ut.ee/items/26e26504-9379-42cf-8f85-361a04dcd114

@inproceedings{Kunilovskaya-etal-2025,
title = {Predictability of Microsyntactic Units across Slavic Languages: A translation-based Study},
author = {Maria Kunilovskaya and Iuliia Zaitova and Wei Xue and Irina Stenger and Tania Avgustinova},
url = {https://dspace.ut.ee/items/26e26504-9379-42cf-8f85-361a04dcd114},
year = {2025},
date = {2025},
booktitle = {Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025)},
publisher = {University of Tartu Library},
address = {Tallinn, Estonia},
abstract = {

},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard

• Lemke     (0.58MB)
• https://journals.linguisticsociety.org/proceedings/index.php/ELM/article/view/5836
• https://doi.org/10.3765/elm.3.5836

@proceeding{Lemke_2025,
title = {Investigating fragment usage with a gamified utterance selection task},
author = {Tyll Robin Lemke},
url = {https://journals.linguisticsociety.org/proceedings/index.php/ELM/article/view/5836},
doi = {https://doi.org/10.3765/elm.3.5836},
year = {2025},
date = {2025},
booktitle = {Experiments in Linguistic Meaning},
pages = {447-459},
abstract = {

},
pubstate = {published},
type = {proceeding}
}

Copy BibTeX to Clipboard

Recent studies suggest that language users generate and maintain multiple predictions in parallel, especially in tasks that explicitly instruct participants to generate predictions. Here, we investigated the possibility of parallel gradedness of linguistic predictions in a simple reading task, using a new measure—imbalance—that captures the probabilistic difference between multiple sentence completions. We focus on prenominal gender-marked articles in German to obtain prediction-specific effects. Native speakers of German read predictable or unpredictable gender-marked nouns that were preceded by prediction-consistent or -inconsistent prenominal articles. Sentence frames either biased expectations more strongly toward the most likely continuation of the sentence, or balanced expectations between the first and second most likely continuation. The results showed reading facilitation for gender-marked articles when sentences were more biased but slowing when sentences were more balanced, irrespective of article predictability. We conclude that readers issue multiple prenominal predictions and weigh them according to their likelihood, providing evidence for parallel gradedness of prenominal predictions. The results are discussed in light of theoretical models on prediction and rational sentence processing.

• eScholarship UC item 7g30m0th     (1.74MB)
• https://escholarship.org/uc/item/7g30m0th
• https://doi.org/10.5070/G6011.1636

@article{haeuser-borovsky-2025,
title = {Got it right up front? Further evidence for parallel graded prediction during prenominal article processing in a self-paced reading study},
author = {Katja H{\"a}user and Arielle Borovsky},
url = {https://escholarship.org/uc/item/7g30m0th},
doi = {https://doi.org/10.5070/G6011.1636},
year = {2025},
date = {2025},
journal = {Glossa Psycholinguistics},
volume = {4},
number = {1},
abstract = {Recent studies suggest that language users generate and maintain multiple predictions in parallel, especially in tasks that explicitly instruct participants to generate predictions. Here, we investigated the possibility of parallel gradedness of linguistic predictions in a simple reading task, using a new measure—imbalance—that captures the probabilistic difference between multiple sentence completions. We focus on prenominal gender-marked articles in German to obtain prediction-specific effects. Native speakers of German read predictable or unpredictable gender-marked nouns that were preceded by prediction-consistent or -inconsistent prenominal articles. Sentence frames either biased expectations more strongly toward the most likely continuation of the sentence, or balanced expectations between the first and second most likely continuation. The results showed reading facilitation for gender-marked articles when sentences were more biased but slowing when sentences were more balanced, irrespective of article predictability. We conclude that readers issue multiple prenominal predictions and weigh them according to their likelihood, providing evidence for parallel gradedness of prenominal predictions. The results are discussed in light of theoretical models on prediction and rational sentence processing.},
pubstate = {published},
type = {article}
}

Copy BibTeX to Clipboard

• 67922b75422ca     (0.54MB)
• https://openhumanitiesdata.metajnl.com/articles/10.5334/johd.285
• https://doi.org/10.5334/johd.285

@article{Talamo-2025,
title = {Introducing STAF: The Saarbr{\"u}cken Treebank of Albanian Fiction},
author = {Luigi Talamo},
url = {https://openhumanitiesdata.metajnl.com/articles/10.5334/johd.285},
doi = {https://doi.org/10.5334/johd.285},
year = {2025},
date = {2025},
journal = {Journal of Open Humanities Data},
pages = {1–6},
volume = {11},
number = {3},
abstract = {

},
pubstate = {published},
type = {article}
}

Copy BibTeX to Clipboard

Information status — the newness or givenness of referents in discourse — is known to affect the production of language at many different levels. At the morphosyntactic level, information status gives rise to special words orders, elisions, and other phenomena that challenge the notion that morphosyntax can be considered independent of discourse context. Though there are many language-specific corpora annotated for information status and its related phenomena, coreference and anaphora resolution, what is not available at present is a cross-lingually consistently annotated corpus or annotation scheme that would allow for comparativestudy of these phenomena across many diverse languages. In this paper we present our work to build such a resource. We are annotating a parsed, parallel corpus of prose in many languages for information status and coreference resolution, so that like-for-like cross-lingual comparisons can be made at the intersection of discourse and syntax. Our corpus can and will be used both for corpus analysis and for model training.

• 2024.tlt-1.7     (0.17MB)
• https://aclanthology.org/2024.tlt-1.7/

@inproceedings{dyer-etal-2024-multilingual,
title = {A Multilingual Parallel Corpus for Coreference Resolution and Information Status in the Literary Domain},
author = {Andrew Dyer andRuveyda Betul Bahceci and Maryam Rajestari and Andreas Rouvalis and Aarushi Singhal and Yuliya Stodolinska and Syahidah Asma Umniyati and Helena Rodrigues Menezes de Oliveira Vaz},
editor = {Daniel Dakota and Sarah Jablotschkin and Sandra K{\"u}bler and Heike Zinsmeister},
url = {https://aclanthology.org/2024.tlt-1.7/},
year = {2024},
date = {2024},
booktitle = {Proceedings of the 22nd Workshop on Treebanks and Linguistic Theories (TLT 2024)},
pages = {55-64},
publisher = {Association for Computational Linguistics},
address = {Hamburg, Germany},
abstract = {Information status — the newness or givenness of referents in discourse — is known to affect the production of language at many different levels. At the morphosyntactic level, information status gives rise to special words orders, elisions, and other phenomena that challenge the notion that morphosyntax can be considered independent of discourse context. Though there are many language-specific corpora annotated for information status and its related phenomena, coreference and anaphora resolution, what is not available at present is a cross-lingually consistently annotated corpus or annotation scheme that would allow for comparativestudy of these phenomena across many diverse languages. In this paper we present our work to build such a resource. We are annotating a parsed, parallel corpus of prose in many languages for information status and coreference resolution, so that like-for-like cross-lingual comparisons can be made at the intersection of discourse and syntax. Our corpus can and will be used both for corpus analysis and for model training.},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard

The aim of this paper is to empirically test the claim that subordinate clauses tend to preserve conservative features in language change. To this end, the diachronic behavior of two well-understood and frequently adduced features of grammar, namely null subject pronouns and order of subject, object and verb, is analyzed for main and adverbial clauses in a balanced corpus of 45 IndoEuropean languages. This study combines qualitative and quantitative analysis by drawing on individual descriptive grammars and parallel corpora respectively. Additionally, diachronic change is modeled using phylogenetic comparative methods. The data suggest that adverbial clauses can in some cases develop asymmetries with respect to their independent counterparts, either through innovation or through preservation of conservative features, possibly due to a communicative need to distinguish clause types by means of grammar. However, the general tendency is for adverbial clauses to change much in the same way as main clauses. This finding contradicts previous claims and calls for a reassessment of studies on the diachronic nature of distinct clause types.

• 02-Talamo-ok     (0.54MB)
• https://www.italian-journal-linguistics.com/current-issue/
• https://doi.org/10.26346/1120-2726-225

@article{Talamo-etal-2024,
title = {A quantitative approach to clause type and syntactic change in two Indo-European corpora},
author = {Luigi Talamo and Annemarie Verkerk andIker Salaberri},
url = {https://www.italian-journal-linguistics.com/current-issue/},
doi = {https://doi.org/10.26346/1120-2726-225},
year = {2024},
date = {2024},
journal = {Italian Journal of Linguistics},
pages = {53-82},
volume = {36},
number = {2},
abstract = {The aim of this paper is to empirically test the claim that subordinate clauses tend to preserve conservative features in language change. To this end, the diachronic behavior of two well-understood and frequently adduced features of grammar, namely null subject pronouns and order of subject, object and verb, is analyzed for main and adverbial clauses in a balanced corpus of 45 IndoEuropean languages. This study combines qualitative and quantitative analysis by drawing on individual descriptive grammars and parallel corpora respectively. Additionally, diachronic change is modeled using phylogenetic comparative methods. The data suggest that adverbial clauses can in some cases develop asymmetries with respect to their independent counterparts, either through innovation or through preservation of conservative features, possibly due to a communicative need to distinguish clause types by means of grammar. However, the general tendency is for adverbial clauses to change much in the same way as main clauses. This finding contradicts previous claims and calls for a reassessment of studies on the diachronic nature of distinct clause types.},
pubstate = {published},
type = {article}
}

Copy BibTeX to Clipboard

This paper presents a study on the usage of noun+noun compounds and verbal complement structures in 18th century scientific articles in the Royal Society Corpus (RSC) comparing translated to non-translated English texts. Departing from the hypothesis that the translations will conform stronger to traditional patterns of the English language, the analysis shows that these historical translations and non-translated texts are similarly marked by the ongoing reorganisation of the noun phrase, but translations
contain more innovative complementation patterns. Additionally, a surprisal analysis shows that the analysed patterns tend to occur in more predictable and conventionalised contexts in non-translated texts than in translation.

 

• 18_Noun_noun_compounds_and_ver     (0.56MB)
• https://sites.google.com/view/tt2024/schedule-and-proceedings

@inproceedings{Menzel2024Noun,
title = {Noun + noun Compounds and Verbal Complements as Non-normalised Features in Late Modern English Scientific Translations},
author = {Katrin Menzel},
url = {https://sites.google.com/view/tt2024/schedule-and-proceedings},
year = {2024},
date = {2024-12-26},
booktitle = {Proceedings of 7th Translation in Transition Conference},
address = {Batumi: Shota Rustaveli State University},
abstract = {This paper presents a study on the usage of noun+noun compounds and verbal complement structures in 18th century scientific articles in the Royal Society Corpus (RSC) comparing translated to non-translated English texts. Departing from the hypothesis that the translations will conform stronger to traditional patterns of the English language, the analysis shows that these historical translations and non-translated texts are similarly marked by the ongoing reorganisation of the noun phrase, but translations
contain more innovative complementation patterns. Additionally, a surprisal analysis shows that the analysed patterns tend to occur in more predictable and conventionalised contexts in non-translated texts than in translation.},
pubstate = {published},
type = {inproceedings}
}

Copy BibTeX to Clipboard