Czechia is global powerhouse in AI research. This is demonstrated by international workshop JSALT 2025 at FIT BUT

(PR, Brno, June 17, 2025) The Faculty of Information Technology at Brno University of Technology (FIT BUT) will host the 32nd edition of the prestigious international JSALT 2025 workshop from June 23 to August 1. This renowned event in the field of artificial intelligence research, long organized by Johns Hopkins University in the United States, will bring global leaders in natural language and speech processing to Brno. The workshop can be seen as a recognition of the long-term work of the Brno research group BUT Speech@FIT and its leader, Professor Jan Černocký.

In the summer weeks, Brno and the Czech Republic will confirm their position as a global center for research in one of AI’s key areas. From June 23 to August 1, 2025, FIT BUT will host more than 100 world experts in speech and language technologies. The important workshop Jelinek Summer Workshop on Speech and Language Technology (JSALT) is organized by Johns Hopkins University in collaboration with partners from the European ESPERANTO project. The fact that Brno’s Faculty of Information Technology is the host is no coincidence – it is recognition of the successful work of the local research group BUT Speech@FIT and its head Jan Černocký. The group’s scientific director is Lukáš Burget. Černocký, Burget, and former member Pavel Matějka are ranked among the world’s top 100 researchers in speech recognition.

Mining data from speech, or more precisely, research into machine learning applied to human speech and language – this is how Černocký and his colleagues' long-term work at Brno University of Technology can be described. Their success has placed Czech research among the top five in the world. What can actually be extracted from speech? Keywords, the speaker's age, gender, emotions, socio-economic data… Human-machine communication via speech or text is no longer an abstract scientific theme – every web browser and social network uses it. Most of us use voice assistants, speech-to-text messaging, navigation, and many have encountered speaker verification (e.g. in banking services). Research results can also be applied to automatic subtitling, emergency call handling, psychotherapy session optimization, communication between air traffic controllers and pilots, and even the security sector. In recent years, new topics such as the fight against voice deepfakes have emerged. Behind recent technological breakthroughs, there is often a Czech footprint.

Czech and Brno footprint in speech processing

The contribution of Czech scientists to the development of speech and language technologies is not as well known to the public as it deserves to be. Few areas have seen such global success. The Brno-based group BUT Speech@FIT is a flagship of the field. Its origins go back to 1997, when Jan Černocký was working at the Faculty of Electrical Engineering and Computer Science at BUT. The beginnings were organizationally challenging, as Černocký recalls: “The 1990s were very difficult, especially in terms of university funding.” The turning point came with international collaboration. Černocký completed his PhD in France, established contacts there, and in 1999, his team joined its first major European project focused on speech data databases. A real breakthrough came with his meeting with Professor Hynek Heřmanský, who has been working in the U.S. since the 1980s – Heřmanský remains a global authority and a true legend in speech processing. Thanks to him, the young Brno researchers gained their first internships in the U.S. and became involved in major projects. In 2002, the group moved to the newly established Faculty of Information Technology at BUT, and success quickly followed. In 2005, Brno’s “speech guys” surprised the scientific world in international evaluations of language recognition systems, and the following year did the same with speaker identification. Suddenly, everyone wanted to know where Brno was. These successes opened the doors to prestigious projects funded by U.S. government agencies, among others. They also impacted the Czech economy: companies such as Phonexia, which originated around the Brno researchers and was the first in the world to offer a highly accurate commercial biometric speaker identification technology, now operate in 60 countries.

Currently, Černocký’s team includes 30 members of ten nationalities, ranging from experienced researchers to PhD students. Internationalization is essential for scientific progress, according to Černocký: “We’ve always pushed hard in this area, something we learned early on from Hynek Heřmanský. It’s probably our only chance in the Czech Republic. Internationalization creates feedback and pays off – professionally and personally.” Although Brno isn’t Oxford, local science excels in some fields. The BUT Speech@FIT group certainly doesn’t lack interest from international researchers.

An event where future is born

Černocký often says that for such a small country, the Czech Republic plays a strong role in his field. In the international community, it’s well known that speech technologies are being developed in the U.S., China… and the Czech Republic. Thanks to the BUT Speech@FIT group, Brno has already hosted world-class events. Černocký and his team previously brought to the Czech Republic the world’s largest conference on speech processing, Interspeech (held in 2021). And this year, FIT BUT is hosting the top-tier international workshop JSALT 2025, organized by Johns Hopkins University. Over its more than 25-year history, the workshop has mostly been held in the U.S. and Western Europe – with Prague hosting in 2014. “Yes, hosting JSALT is one of the metaphorical medals our group wears around its neck,” Černocký acknowledges. The six-week research workshop will be attended by a hundred global experts. Notable participants include Sanjeev P. Khudanpur (Johns Hopkins University), Ricardo Marxer (University of Toulon), and Ramani Duraiswami (University of Maryland). The participants will be divided into four teams focusing on selected research themes. Topics include improving the ability of large language models (e.g. the well-known ChatGPT) to consistently follow assigned roles during longer interactions (such as doctor–patient simulations), or simplifying and streamlining speech recognition in complex scenarios such as meetings with many participants in noisy environments.

The aim of the workshop, held with support from the OP JAK project and under the auspices of the City of Brno and the South Moravian Region, is to push the boundaries of research, foster international scientific collaboration, and involve doctoral and master’s students who will benefit from invaluable expert mentoring. Many technologies we use today have their roots in research and collaborations that started at JSALT. One example is the automatic translation system used by services like Google Translate. “The workshop connects talented people from all over the world. And sometimes, within a year, a result emerges that turns existing technology on its head,” concludes Černocký.

By the way, JSALT carries a Czech legacy in its very name, which honors Czech-American scientist Frederick (Bedřich) Jelinek, a native of Kladno and a pioneer in speech recognition research using statistical methods. After the communist coup, Jelinek’s family emigrated from Czechoslovakia to the United States. Jelinek led speech research at IBM and later at Johns Hopkins University for many years. It is symbolic that an event bearing the name of this Czech native is being held at FIT BUT, a faculty that has become one of the world’s leading research centers for AI applications in speech and natural language.