Finding a way of demarcating between science and pseudoscience is a long-standing philosophical problem. Instead of looking for a theoretical distinction between these two abstract entities, our proposal is to look at the problem from an empirical perspective and try to identify characteristic features that would distinguish instances of pseudoscience from established science. In this regard, digital humanities provide invaluable tools. We have taken a journal publishing research in astrology and are looking at its citation patterns, publication structure, as well as various syntactic, stylometric and semantic properties to try and identify specific indicators that might be related to some of the traditional criteria of demarcation.

The submission presents the Collection of Multilingual Parallel Sequences (CoMParS) as a structured basis for evaluating LLMs that aim to model cross-linguistic functional-semantic phenomena. CoMParS is developed within the long-term project Grammar of German in a European perspective (GDE) at the Leibniz Institute for the German Language (IDS) in Mannheim. GDE seeks to construct a multi-contrastive grammar of German against the background of other European languages, using cognitively motivated functional domains (Givón 1984; Zifonun 2017; Frajzyngier 2021) as tertia comparationis linked to traditional part-of-speech categories.

The paper outlines the research question (form-driven and distribution-based constraints on LLMs used for cross-linguistic research), presents a functionally-oriented resource (CoMParS), and sketches ways in which CoMParS may assist in improving the described state of affairs.

Diplomatic discourse is characterised by a rigorous pragmatic granularity, in which subtle lexical variations encode consequential shifts in legal authority, geopolitical alignment, and institutional commitment. These high-stakes nuances are frequently obscured by the semantic smoothing inherent in general-purpose Large Language Models (LLMs), which prioritize broad linguistic patterns over specialized institutional registers (McGillivray et al, 2020) (Bender et al., 2020).

This paper introduces DiploBERT, a domain-specific transformer encoder developed through Domain-Adaptive Pre-Training (DAPT) on a curated 5.5GB corpus of diplomatic and institutional archives (Gururangan, S., et al., 2020). By modeling the formulaic hedging, strategic ambiguity, and modal precision characteristic of diplomatic prose, DiploBERT transforms patterns that general models perceive as statistical noise into analytically meaningful signals.

This work provides an infrastructural resource for the Digital Humanities, bridging computational language modeling and diplomatic hermeneutics, enabling scholars to move beyond reductive keyword searches toward a context-aware distant reading. In doing so, it offers a new lens for analysing the datafication and transformation of international communications.

This paper links all individuals who spoke in the German Bundestag between 1949 and 2021 to Wikidata, creating an enriched, longitudinal dataset that connects parliamentary speech transcripts with structured biographical metadata. We evaluate the coverage, composition, and potential biases of the retrieved properties and statements, with particular attention to gender, professional background, historical legacies, and transnational dimensions such as place of birth, languages, and foreign awards. The results demonstrate both the analytical potential of combining GermaParl with Wikidata and the importance of critically assessing uneven metadata coverage in open knowledge graphs.

The history of Latin in Polish lands spans from the threshold of the second millennium to at least the end of the 18th century (Axer 2004). Despite significant scholarship, a synthetic approach to the historically variable functions and forms of Latin in Polish territories remains absent. The project DiaLexPoL aims to create a lexical database of Neo-Latin (ca. 1550–1800) building upon the existing Dictionary of Medieval Latin in Poland (SŁŚ) as a diachronic baseline.

We present two experiments evaluating LLM-assisted sense assignment as a core method for compiling this resource (cf. Tiberius et al. 2024; Lugli 2025), focusing on high-frequency, polysemous sociopolitical terms such as CIVITAS, CONSILIUM, and IMPERIUM, whose referents changed substantially across the medieval–early modern transition (McGillivray & Nowak 2025). In Experiment 1, the LLM reassigns dictionary quotations to existing sense definitions; in Experiment 2, it processes novel concordance lines from a Neo-Latin corpus. We vary models (GPT-4o, Claude, Llama, Qwen, Mistral), prompting strategies, and definition languages, measuring accuracy against expert gold standards and inter-annotator agreement. Cases where no existing sense fits are analyzed for genuine semantic innovation in the Neo-Latin period.

We discuss the results within the post-editing lexicography framework (Lew 2024; Jakubíček & Rundell 2023) and reflect on the well-documented risks of LLM adoption, including irreproducibility and loss of evidence-based workflow. Importantly, enthusiastic adoption of LLMs as a way of coping with funding cuts may prove counterproductive if the efficiency gains we demonstrate are used to justify eliminating the very expert labour they currently augment.

Positioned in the area of corpus linguistics, translation and multifactorial statistics, this study aims to explore to what extent and why repeatedly used reporting verbs signalling direct or indirect speech were preserved, i.e. when the most frequent Polish equivalent was consistently used, or avoided in translation, i.e. when other reporting verbs, non-reporting verb equivalents, multi-word items or omissions were used. Press releases represent a highly conventionalized genre situated at the interface of institutional communication and journalistic discourse (Jacobs 1999; Catenaccio 2008; Perrin 2013), and EU press releases are produced within a specific organizational setting (European Commission 2003). As a research material, we used carefully selected text samples of a RAPID corpus obtained from the ELRC-SHARE repository (Piperidis et al. 2018): our English-Polish parallel sub-corpus included 2,133 bilingual segments. Each segment was carefully annotated with respect to nine potential predictors, including orthographic, syntactic, semantic and textual ones, that might have impacted repetition shift in translation, which was our dependant binary variable. The study was conducted in two stages. First, we identified translation patterns in the study corpus, and, second, we conducted mixed-effects logistic regression modelling (Kabacoff 2015; Winter 2019) to identify the predictors that impact either the preservation or avoidance of the English original’s reporting verbs repetitions in the Polish translations. The analysis conducted in the R environment (R Core Team 2025) revealed that repetition avoidance is linked with indirect and mixed reported speech and with syntactic changes in translation. Random effects revealed that beyond these fixed predictors repetition shifts are significantly influenced by verb- and text-specific factors. Both fixed and random effects taken together explain 41 observed variability in the data. The insights from regression modelling offer a clear multifactorial picture of translators’ decision-making when handling repetitions in the EU press releases translated in institutional settings from English into Polish.