Institute of Formal and Applied Linguistics
Charles University, Czech Republic
Faculty of Mathematics and Physics
Czech Speech & NLP Day 2025
Czech Speech & NLP Day 2025 is a lovely bi-annual meetup of the community of Speech and/or NLP researchers with relation to Czech Republic. It is open to all researchers, developers, programmers, data specialists, teachers, managers, and other professionals from both academia and industry who are engaged in research of fields of speech and natural language processing.
The date 28.11.2025 is confirmed. The exact schedule is preliminary.
Program
Morning (appx 10:00-13:00):
-
Keynote lecture by Ahmed Ali: Beyond the Hype: The Unfinished Journey of Scaling Arabic Language Technologies
Abstract:
The last ten years have transformed Arabic Language Technologies. We have progressed from single-task models to the dawn of powerful, large-scale AI systems. We moved beyond vanilla speech recognition and robotic text-to-speech to models that perform impressively in formal Arabic. Today's Large Language Models promise a great leap forward, aiming to integrate speech, text, and documents into a single, intelligent model. This talk will look closely at this transformation, moving beyond the hype to reveal the realities on the ground.Despite this significant progress, the journey remains unfinished because our systems still struggle with the language of daily life. This includes the persistent challenge of dialect identification recognizing which of the many regional forms of Arabic is being used, and handling seamless code-switching, where speakers fluidly mix languages or dialects. Crucially, the next generation of AI must also be culturally aware, understanding the local customs, values, and context that vary across the Arab world.
We will highlight the critical challenge ahead: evolving from AI that merely processes Arabic to AI that truly understands the language in all its complex, nuanced, and culturally rich reality. Our success in the next decade will be judged not by the size of our models, but by the depth of their comprehension in real-world applications.
Bio:
Ahmed Ali leads the Speech group at HUMAIN, Saudi Arabia’s national AI company, and is the founder of ArabicSpeech.org. With over two decades of experience in speech and NLP, his research is dedicated to advancing technology for low-resource and multidialectal languages. He is the co-founder of KANARI AI and has driven innovation at the intersection of academia and industry, holding key roles at IBM, Nuance, QCRI, and SDAIA. His work is documented in over 100 peer-reviewed publications and patents, all aimed at bridging the gap for speech and language technologies.
- Then some plenary talks. To be specified.
Lunch
Afternoon (appx 14:30-17:00):
- Open poster and demo session. Bring your recent interesting poster or demo that you already published somewhere or are planning to, and show it to colleagues. But be aware that this is a meetup of colleagues and not a conference with proceedings.
Venue
Karolinum, Vlastenecký sál
Address: Ovocný trh 3, Praha 1
Registration
Participation is free but limited because of venue. The participants from industry are encouraged to sponsor this event.
Organizers
Program commitee: Dominik Macháček (ÚFAL), Jan Černocký (FIT VUT), Ondřej Bojar (ÚFAL)
Administration: Jiří Valach (ÚFAL)
Contact persons: Dominik Macháček for program, Jiří Valach for venue and administration, both if you're in doubt.
Emails: machacek@ufal.mff.cuni.cz , valach@ufal.mff.cuni.cz .
Phone: Jiří Valach, +420 739 771 932.
Last editions