BibTeX

@inProceedings{chiriatti-etal-2026-gold-367165,
	title        = {Gold-Preserving Synthetic Augmentation for Verbatim Swedish Automatic Speech Recognition},
	abstract     = {Verbatim transcripts, which keep the hesitations and interactional particles that ordinary automatic speech recognition (ASR) removes, are a prerequisite for speech-based cognitive assessment, and they are hardest to obtain for languages with little annotated speech. We study this problem for Swedish with a 1.02-hour corpus of manually transcribed spontaneous speech, too small for full fine-tuning of KB-Whisper-large. We adapt the model with a low-rank adapter and compare three ways of extending the training data: pseudo-labeling of untranscribed in-domain audio, unguided synthesis with a language model and a text-to-speech system, and a gold-preserving pipeline that reproduces the disfluency distribution of the gold corpus by construction and can be audited before training. Fine-tuning reduces disfluency-segment word error rate (WER) by 32% relative, with blockwise-bootstrap confidence intervals disjoint from the zero-shot baseline, while the three augmentation strategies are not separable by WER across five seeds. The gold-preserving pipeline needs no untranscribed audio, and its disfluency distribution can be verified in advance.},
	booktitle    = {The Proceedings of Speech and Audio Language Models Workshop (2nd Edition), October 28, 2026},
	author       = {Chiriatti, Giuseppe and Kokkinakis, Dimitrios and Bertini, Flavio and Muñoz Sánchez, Ricardo and Lange, Herbert},
	year         = {2026},
	publisher    = {Association for Computational Linguistics (ACL)},
	address      = {Kerrville, TX },
	ISBN         = {979-8-89176-496-5},
	pages        = {38--44},
}

@inProceedings{kokkinakis-etal-2026-longitudinally-367171,
	title        = {Longitudinally Informed Synthetic Speech and Language Data for Neuropsychological Research – a Pilot Study},
	abstract     = {Large longitudinal health cohorts provide a valuable basis for studying how language, cognition, and aging interact over time, and for identifying early markers of cognitive decline. In this proof-of-concept study, we examine how such cohorts can also support the generation of synthetic neuropsychological speech and language data that preserve clinically meaningful patterns while reducing privacy risks. Using clinically grounded score profiles across individuals and time, large language models can generate synthetic cases that reflect important structural relationships without corresponding to real participants. Our pilot findings suggest that predictive modelling of cognitive trajectories remains the primary value of longitudinal cohorts, while synthetic datasets offer an important complementary resource for privacy-preserving research, benchmarking, and translational evaluation. For the text component, we explored controlled synthetic generation for task types commonly used in dementia assessment, including the Boston Naming Test, semantic verbal fluency, phonemic verbal fluency, and picture description*. Prompting strategies were designed to capture lexical retrieval difficulties, circumlocution, hesitation, omission, reduced informativeness, repetition, and topic drift. For the speech component, we investigated pipelines for speech representation, annotation, transcription, and synthetic speech generation, focusing on features such as pausing, reduced fluency, speech rate, hesitations, prosody, and temporal structure. Important challenges remain, as both natural and pathological speech are highly complex, and clinically relevant variation is expressed not only through lexical content but also through rhythm, timing, pause structure, voice quality, and prosody. Although the present findings should be regarded as preliminary, they demonstrate the feasibility of this approach and motivate future work on realism, utility, and domain adaptation for speech and language technologies used in cognitive screening and dementia-related assessment. Future work will further examine emerging text-tospeech models, with particular attention to improved prosody control, pause modelling, voice quality, temporal dynamics, and systematic validation in clinically realistic synthetic speech for dementia-related assessment contexts. },
	booktitle    = {The 36th Alzheimer Europe conference},
	author       = {Kokkinakis, Dimitrios and Chiriatti, Giuseppe and Lange, Herbert and Muñoz Sánchez, Ricardo},
	year         = {2026},
}

@inProceedings{francis-etal-2026-chatgpt-364799,
	title        = {ChatGPT, why can’t anyone afford a house? On the Effects of LLM pre-annotation on Annotator Subjectivity},
	abstract     = {Large language models (LLMs) have often been proposed as substitutes for human annotators in a variety of tasks. At the same time, there has been increased focus on the role that human subjectivity and perspective plays in data annotation. To avoid eliminating the human role in annotation entirely, the use of LLMs for pre-annotation has been suggested as an alternative approach. In this paper, we explore to which degree this approach affects subjectivity of social media annotation in English. We focus on comments regarding the current status of the housing market and label them for concern level, factors affecting housing affordability, and aspects that authors claim either exacerbate or improve the situation. To investigate this, we design an experiment involving two rounds of annotation: the first, a dataset annotated by humans only; and the second, a dataset with LLM pre-annotations curated by the same human annotators. We observe that the second setting leads to much higher agreement, as well as significant changes in label distribution and co-occurrence. Similar shifts do not appear in the LLM labels. Our findings show that use of LLMs in the annotation process leads to convergence in annotations and, thus, to an erosion of human subjectivity.},
	booktitle    = {Proceedings of the the fifth edition of NLPerspectives 11-16 May 2026, Palma, Mallorca, Spain},
	author       = {Francis, Emilie and Leuzinger, Celine and Muñoz Sánchez, Ricardo and Gauthier, Lee D.},
	year         = {2026},
	publisher    = {ELRA Language Resources Association (ELRA)},
	pages        = {98--111},
}

@inProceedings{kokkinakis-etal-2026-disfluencies-364800,
	title        = {Disfluencies and ASR Performance on Swedish Spontaneous Speech from the ‘Trip to Stockholm’ Discourse Narrative Task},
	abstract     = {Automatic Speech Recognition (ASR) offers a scalable and cost-efficient alternative to manual transcription and is becoming increasingly relevant in clinical contexts, particularly for the detection of cognitive decline and mental health assessment. However, current ASR-systems still struggle with spontaneous speech, particularly when processing disfluencies, pauses, and speaker variability that often carry diagnostic value. This study evaluates state-of-the-art open ASR models targeting Swedish using recordings from the “Trip to Stockholm” discourse narrative task which elicits ecologically valid, cognitively demanding speech. Recognition quality is assessed using various metrics, alongside an analysis of linguistic and technical sources of error focused on disfluencies. Our findings show that disfluency-related phenomena degrade recognition performance. Possible post-processing strategies can improve specific error patterns emerging for filled pauses, word repetitions, and self-corrections. The results illustrate both the advances and ongoing limitations of ASR for spontaneous Swedish speech, emphasizing the need for models explicitly trained, or fine-tuned, on disfluent data to ensure robustness in clinical and research applications.},
	booktitle    = {    Proceedings of the Sixth Resources and ProcessIng of linguistic, para-linguistic and extra-linguistic Data from people with various forms of cognitive/psychiatric/developmental impairments in cooperation with the MENTAL.ai consortium, 11-16 May 2026, Palma, Mallorca, Spain},
	author       = {Kokkinakis, Dimitrios and Lange, Herbert and Muñoz Sánchez, Ricardo},
	year         = {2026},
	publisher    = {European Language Resources Association (ELRA)},
	pages        = {24–33},
}

@misc{siegert-etal-2026-proceedings-362841,
	title        = {Proceedings of the Joint Workshop on Legal and Ethical Issues in Human Language Technologies and Computational Approaches to Language Data Pseudonymization, Anonymization, De-identification, and Data Privacy (LEGAL2026 and CALD-pseudo 2026) @ LREC 2026. 11-16 May 2026, Palma, Mallorca, Spain},
	author       = {Siegert, Ingo and Szawerna, Maria Irena and Choukri, Khalid and Dobnik, Simon and Kamocki, Paweł and Lindström Tiedemann, Therese and Lison, Pierre and Muñoz Sánchez, Ricardo and Pilán, Ildikó and Södergård, Lisa and Talmoudi, Kossay and Volodina, Elena and Vu, Xuan-Son},
	year         = {2026},
	publisher    = {ELRA},
	address      = {Paris},
	ISBN         = { 978-2-493814-86-9},
}

@inProceedings{greco-etal-2026-stereobusters-361361,
	title        = {StereoBusters at GSI:detect: LLM-Based Detection and Human Qualitative Analysis of Gender Stereotypes in Italian Short Texts},
	booktitle    = {Proceedings of the 9th Evaluation Campaign of Natural Language Processing and Speech Tools for Italian, Bari, Italy, February 26th-27th, 2026. Final Workshop (EVALITA 2026)},
	author       = {Greco, Salvatore and La Quatra, Moreno and Marchiori Manerba, Marta and Muñoz Sánchez, Ricardo and Cignarella, Alessandra Teresa},
	year         = {2026},
	publisher    = {CEUR Workshop Proceedings},
	address      = {Bari, Italy},
}