Abstracts
Petra Brandoburová. From paper-pen to automated analysis from speech : the story of neuropsychological assessment of neurodegenerative diseases in Slovakia.
Simona Krakovská. Assessment is not enough: cognitive training for people with Alzheimer’s disease.
Sara Majsniarová. Why memory and proper methods of assessment matter for Alzheimer’s disease.
Viktória Kevická. The Role of a Speech Therapist in the Diagnostics and/or Rehabilitation of Neurodegenerative Diseases and the Possibilities for Task Automation.
The role of speech therapists in the intervention of neurodegenerative diseases has become increasingly acknowledged as essential. We are focusing to highlight their contributions within the realms of diagnostics and rehabilitation, with a particular focus on the specific language processing difficulties associated with these conditions. Read more
Benko Tomáš, Kilvádyová Dominika, Strapcová Tatiana, Pánisz Pavol, Garrard Peter, Cséfalvay Zsolt. Proposal for a digitalized Slovak version of the Mini Language State Examination-work in progress.
Early diagnosis and analysis of language deficits in patients with primary progressive aphasia is a challenge even for experienced speech-language therapists. MLSE (Mini Language State ExaminationI) is a newly developed diagnostic tool, which has the potential to evaluate all important areas of language processes in a short time. Our goal was to create a proposal for a digitalized version of the Slovak adaptation of the MLSE, which would facilitate the administration of MLSE, as well as the automatic calculation of all test scores. Read more
Branislav Gerazov. Review of apps for treating Alzheimer and other neurodegenerative diseases.
This review aims to synthesize existing literature on mobile apps specifically designed for AD and other neurodegenerative diseases, focusing on their features, efficacy, usability, and acceptability. e variety of apps analyzed, focus primarily on cognitive training, mood monitoring, and disease management education. Read more
Lucia Mareková, Štefan Beňuš. Robots in the eyes of Alzheimer patients
This presentation synthesizes perceptions of social robots from AD patients, their relatives, and formal or informal caregivers, offering insights into how robots should behave, communicate, and interact to meet the needs of this population. The synthesis is based on 10 relevant studies published between 2019 and 2024, identified using the PRISMA method. Read more
Róbert Sabo, Marian Trnka, Milan Rusko. The use of a social robot as an assistant for recording speech of patients with Alzheimer Dissese.
This paper presents the utilization of a social robot, Furhat, as a tool for speech data collection from patients with Alzheimer’s disease. Our primary objective was to employ a social robot to autonomously collect speech data from patients, ensuring a user-friendly experience while gathering data suitable for both individual patient diagnosis. In our experiment, we recorded 30 elderly individuals conversing with the Furhat robot. Read more
Rami Kammoun, Mohamed Salah Al-Radhi, Géza Németh. Enhancing Expressive TTS Synthesis for Multilingual Low-Resource Languages: Challenges and Applications
Despite the significant progress in TTS technology for rich-resource languages, such as English, which benefits from a large quantity of high-quality audio and transcription datasets, there remain considerable challenges in developing expressive TTS for low-resource languages. These languages often lack sufficient, high-quality data, making it difficult to achieve the same level of naturalness and expressivity as rich-resource counterparts. This research focuses on addressing
these challenges, particularly for languages such as the Arabic-Tunisian dialect, emotional French, Kazakhstani, and German. This study aims to push the boundaries of expressive TTS synthesis for low-resource languages by leveraging state-of-the-art models and exploring their applicability in multilingual contexts. Read more
Attila Zoltán Jenei, Gábor Kiss, Dávid Sztahó Integrating Multilingual Acoustic and NLP Models for Scalable and Explainable Dementia Diagnosis: Challenges and Future Proposal.
In our starting CELSA project, called “Identification of language-independent speech and language biomarkers characteristic of dementia”, we propose to address the current limitations of automatic AD detection, namely: language dependent models, small sample sizes, integrating methods with human intervention and establishing explainable features sets. Read more
Arthur Janicki et al. …to be specified………. Read more
Milan Rusko, Marian Trnka.Taxonomy of the acoustic space of the streamed computer games.
Within the framework of international cooperation, the SAS Institute of Informatics is involved in the research of phenomena related to the behavior and communication of computer game players and their consumers – viewers on streaming platforms. The basic principles of communication, voice expressions, speech and text expressions of participants in parasocial communication on the Twitch streaming platform will be examined. This article is a first step towards the analysis of the audio side of streamed computer games, and its aim is to define a taxonomy of acoustic manifestations occurring in streamed gaming and to suggest possible avenues of their analysis. The taxonomy is realized from the point of view of sound source criteria, sound intention and acoustic characteristics. Read more
Ceren Cerasi A Look At The Parasocial Relationship Between The Streamer And The Viewers: A Sentiment Analysis Study.
In order to examine the current situation, this study will examine the parasocial relationship between the streamer and the viewer on a widely used social media platform called Twitch. The study will be carried out using sentiment analysis, a natural language processing method. Read more
Branislav Gerazov, Zoran Ivanovski, and Dimitar Taskovski. Macedonian assistive speech technologies for a more inclusive world.
The Speech Group at FEEIT, UKIM in Skopje, has been actively involved in developing assistive technologies for Macedonian, a language spoken by approximately 3 million people worldwide. We worked on the successful localization of UNICEF’s Augmentative and Alternative Communication (AAC) tool, Cboard, to Macedonian and Albanian. We developed Suze, a free high-quality Text-to-Speech (TTS) voice for Macedonian, which has been welcomed by AT users from the blind community. Read more
Laxmi Kantham Durgam, Ravi Kumar Jatoth, Daniel Hladek, Stanislav Ondas, Matúš Pleva and Jozef Juhar. Age and Gender Estimation from Speech using various Deep
Learning and Dimensionality Reduction Techniques.
Identifying a person’s age and gender from speech signal characteristics poses a significant challenge in personal identity recognition systems, particularly when security considerations are involved. In signal processing applications such as speaker recognition, biometric identification, human-machine interface (HMI), and telecommunication, age and gender estimation from voice is a crucial and demanding problem. In several signal processing domains, deep learning models have demonstrated remarkable effectiveness. In this paper, we propose a new deep learning system for the identification of speakers age and gender from speech using various speech features. Read more
Eva Kiktova. Speech tests for hearing assessments in Romany language.
This study is devoted to speech audiometry in the Romani language, spoken by a substantial minority population in Slovakia. A carefully selected set of 50 Romani words, encompassing 10 nouns, 10 verbs, 10 numerals, 10 adjectives, and 10 objects, serves as the foundation for the test material. Read more
Pleva, Hládek. Question Answering Dataset for Information Retrieval in Slovak.
We introduce a novel benchmark for information retrieval in the Slovak language, with a unique twist that could be of particular interest to researchers in speech recognition. By leveraging a question-answering dataset, we fine-tune and evaluate sentence transformers with direct implications for speech-based applications. The dataset, named Retrieval SkQuAD, has been integrated into two prominent evaluation frameworks, MTEB and BEIR. It consists of 19,000 manually annotated answers to 1,134 questions, each rated with a relevance score, alongside metadata on document utility for answer generation. Read more
Eugen Ružický, Ján Lacko, Milan Rusko.Taxonomy of the acoustic space of the streamed computer games.
This paper shows two examples of the use of an original methodology for creating databases using virtual reality and speech recognition for use in healthcare and industry. The new proposed methodology is based on the combination and interaction of advanced machine learning techniques, the use of virtual reality and speech recognition. Read more
