Research leader name
Dr. Márk Jelasity
On the one hand, the HUN-REN-SZTE Artificial Intelligence Research Group focuses on theoretical issues related to the security and explainability of AI systems, and in terms of practical applications, the focus is on language and speech technology.
Downloads
Our main tasks are improving the exact and heuristic verification of machine learning models, examining their vulnerabilities, and understanding the knowledge representation of robustly trained deep neural networks, increasing the efficiency of robust (adversarial) learning. In addition, in the area of interpretability, we translate the operation of graph neural networks and traditional deep neural networks into predetermined explanatory languages and structures.
Applied research is also carried out, primarily in the fields of language and speech technology. In the field of language technology, we work on the applications of dialogue systems (e.g. customer service), for which the goal is the Hungarian informal spoken language and its semantic and pragmatic analysis. In the field of speech technology, the research group primarily focuses on non-verbal communication and medical applications of automatic speech processing (e.g. recognition of mild cognitive impairment, Alzheimer’s disease, multiple sclerosis, and analysis of patients’ language use).
The applied language and speech technology research of the research group is straightforwardly connected to the examination of the interpretability of machine learning models. For example, in medical applications, where the goal is to diagnose diseases, the medical explanation of the diagnoses is crucial to the acceptance of the technology.
