We evaluate cognitive impairment (CI) classification from transcripts of speech in English, Slovene, and Korean. We compare zero-shot large language models (LLMs) used as direct classifiers under three input settings—transcript-only, linguistic-features-only, and combined—with supervised tabular approaches trained under a leave-one-out protocol. The tabular models operate on engineered linguistic features, transcript embeddings, and early or late fusion of both modalities. Across languages, zero-shot LLMs provide competitive no-training baselines, but supervised tabular models generally perform better, particularly when engineered linguistic features are included and combined with embeddings. Few-shot experiments focusing on embeddings indicate that the value of limited supervision is language-dependent, with some languages benefiting substantially from additional labelled examples while others remain constrained without richer feature representations. Overall, the results suggest that, in small-data CI detection, structured linguistic signals and simple fusion-based classifiers remain strong and reliable signals.

We introduce the EEC-SL dataset, an adaptation of the Equity Evaluation Corpus from English to Slovenian. Based on 11 sentence templates, the dataset contains 8,640 sentences, including pairs of minimally-distant sentences, varying with regard to one of two variables: gender (female or male), and ethnicity (Slovenian or not-Slovenian). In order to validate our selection of personal names, we create a localised version of the Implicit Association Test for ethnic bias, in which participants show a significant implicit bias favouring Slovenian over non-Slovenian names. We use the dataset to evaluate social bias in three computational language models (large language models and an encoder-only transformer) to perform sentiment analysis—specifically, valence. We analyse the results in terms of differences in sentiment between minimally-distant groups of sentences and inferential tests. We found limited evidence for social bias with regard to ethnicity, and no evidence for gender bias, in any of the employed models.

This Ph.D. proposal introduces a plan to develop a computational framework to identify Self-aspects in text. The Self is a multifaceted construct and it is reflected in language. While it is described across disciplines like cognitive science and phenomenology, it remains underexplored in natural language processing (NLP). Many of the aspects of the Self align with psychological and other well-researched phenomena (e.g., those related to mental health), highlighting the need for systematic NLP-based analysis. In line with this, we plan to introduce an ontology of Self-aspects and a gold-standard annotated dataset. Using this foundation, we will develop and evaluate conventional discriminative models, generative large language models, and embedding-based retrieval approaches against four main criteria: interpretability, ground-truth adherence, accuracy, and computational efficiency. Top-performing models will be applied in case studies in mental health and empirical phenomenology.

Dehumanisation involves the perception and/or treatment of a social group’s members as less than human. This phenomenon is rarely addressed with computational linguistic techniques. We adapt a recently proposed approach for English, making it easier to transfer to other languages and to evaluate, introducing a new sentiment resource, the use of zero-shot cross-lingual valence and arousal detection, and a new method for statistical significance testing. We then apply it to study attitudes to migration expressed in Slovene newspapers, to examine changes in the Slovene discourse on migration between the 2015-16 migration crisis following the war in Syria and the 2022-23 period following the war in Ukraine. We find that while this discourse became more negative and more intense over time, it is less dehumanising when specifically addressing Ukrainian migrants compared to others.

This paper presents our ensembling solutions for detecting signs of depression in social media text, as part of the Shared Task at LT-EDI@RANLP 2023. By leveraging social media posts in English, the task involves the development of a system to accurately classify them as presenting signs of depression of one of three levels: “severe”, “moderate”, and “not depressed”. We verify the hypothesis that combining contextual information from a language model with local domain-specific features can improve the classifier’s performance. We do so by evaluating: (1) two global classifiers (support vector machine and logistic regression); (2) contextual information from language models; and (3) the ensembling results.