SPECIAL SESSIONS
Automatic Assessment of Atypical Speech (AAAS)
Automatic Assessment of Atypical Speech (AAAS) explores assessment of pronunciation and speaking skills of children, language learners, and speakers with speech sound disorders and methods to provide automatic rating and feedback using ASR and LLMs. Automatic speaking assessment (ASA) is a key technology for developing AI tools to self-practise second and foreign language skills and provide more complex feedback about fluency, vocabulary and grammar of the recorded speech. ASA is also very relevant for the detection and quantification of speech disorders and for developing speech exercises that can be performed independently at home.
Organizers:
Mikko Kurimo (Aalto University)
Tamás Grósz (Aalto University)
Giampiero Salvi (NTNU)
Sofia Strömbergsson (Karolinska Institutet)
Sari Ylinen (Tampere University)
Minna Lehtonen (University of Turku)
Torbjørn Svendsen (NTNU)
Website: https://teflon.aalto.fi/mlsp-aaas-2025/
Large Vision Language Models (LVLMs) and their Application to Document Understanding
Document Understanding (DU) in Arabic presents significant challenges due to its linguistic complexity, diverse script structures, and the demands of long-context processing. While Large Vision-Language Models (LVLMs) perform well in short-context DU tasks in English, their effectiveness declines when handling Arabic documents, especially in long-context scenarios. This special session will explore innovative approaches to improving LVLM performance for Arabic, with broader implications for multilingual and long-context DU.
Organizers:
Sherif Mohamed (SDAIA)
Ahmed Masry (York University)
Enamul Hoque Prince (York University)
Parisa Kordjamshidi (Michigan State University)
Website: https://www.
LEAP: Low-Energy AI For Edge Learning and Processing
This special session focuses on sustainable and distributed approaches to address the escalating energy demands of centralized ML solutions. It seeks innovative ML solutions that enable efficient, scalable, and sustainable on-device training/inference of traditional and new ML approaches. We encourage submissions on topics such as energy-aware algorithms, in-memory computing hardware, communication, and software brain-inspired models, aiming to highlight energy-efficient solutions that bridge the gap between state-of-the-art ML techniques and their deployment in resource-constrained settings.
Organizers:
Roberto Pereira (CTTC/CERCA, Department of Sustainable AI, Barcelona, Spain)
Paolo Dini (CTTC/CERCA, Department of Sustainable AI, Barcelona, Spain)
Website: https://leap2025.github.io/mlsp/
Applications of AI in the Analysis of Cultural and Artistic Heritage
Artificial Intelligence (AI) is transforming the study of cultural and artistic heritage by enabling advanced analysis, classification, and restoration of historical artifacts. This special session will showcase the latest AI-driven methodologies for analyzing, understanding, and enhancing historical and artistic materials, with a particular focus on machine learning techniques applied to textual, visual, and multimodal data. By bringing together experts from AI, computer vision, and digital humanities, it aims to foster interdisciplinary collaboration and highlight innovative approaches to preserving and understanding humanity's artistic and cultural legacy.
Organizers:
Sinem Aslan (University of Milan, Department of Historical Studies, Milan, Italy)
Hazım Kemal Ekenel (Istanbul Technical University, Department of Computer Engineering, Istanbul, Turkey)
Gennaro Vessio (University of Bari Aldo Moro, Department of Computer Science, Bari, Italy)
Hassan Ugail (University of Bradford, School of Informatics, Bradford, UK)
Anthony Bourached (University College London, Department of Neurology, London, UK)
Gattiglia Gabriele (University of Pisa, Department of Civilization and Forms of Knowledge, Pisa, Italy)
Website: https://ai-cah2025.github.io/main/
Sign Language Translation in the era of Large Language Model - Beyond English
This special session addresses sign language translation challenges at the intersection of AI and accessibility, introducing a novel dataset co-developed with sign language users and interpreters. The session will provide a setup to discuss the generalization capabilities of sign language translation models, addressing a fundamental limitation in the field.
Organizers:
Yazeed Alharbi (SDAIA)
Ahmed Ali (SDAIA)
Marek Hruz (University of West Bohemia)
Lale Akarun (Bogazici University)
Ivan Gruber (University of West Bohemia)
Ali Al Hejab (SDAIA)
Website: https://signforall.github.io/challenge/
Decoding the Brain Time Series
Over the past decade, machine learning has revolutionized many areas, yet its application to time-series data like EEG remains slow and under exploration. This session will discuss how this dual challenge from EEG Decoding calls for clinical neuroscience expertise and machine-learning skills to handle complex signals and how we could explore emerging models that may offer promising solutions.
Organizers:
Bruno Aristimunha (Université Paris-Saclay)
Florian Yger (LITIS, INSA-Rouen Normandy)
Marie-Constance Corsi (ICM, Inria NERV, INSERM, Paris-Sorbonne Université)
Sylvain Chevallier (Université Paris-Saclay, Inria)