Speakers & ABSTRACTS

Olya Kudina (Delft University of Technology) 

In her keynote, Olya will argue that voice-based interfaces urgently need to integrate diversity and inclusion dimensions because of the value they give to society in interacting with technologies by means other than typing or swiping.

By navigating several case studies, such as the social integration of voice assistants, Olya will show how voice-based interfaces are not ethically neutral technologies as they help to shape who gets to be heard, how we are expected to be heard, and fundamentally, what it is that we hear. By employing an ethics-as-accompaniment lens, Olya will provide several ideas on how to think with a diversity and inclusion lens in mind throughout the design process. 

Badal Marhé and Jorik van der Hoek (Achmea / Interpolis) 

In our presentation, we discuss Achmea's advanced use of voice technology. We'll highlight how this innovation is transforming customer experiences, with a focus on practical applications like our smart IVR and voicebot for claims.  The presentation also touches on the future potential for Achmea.

Arjan van Hessen (NOTAS / University of Twente)  

Launched in autumn 2022, OpenAI Whisper: a very good open source speech recognition that can recognize approx. 100 different languages (very) well. It can be run on your own computer, in a cluster at the department or company and is available via the cloud from e.g. SURF. Because it is an open-source model, many different groups are working on improving, speeding up and enriching what can be done with Whisper.

In this presentation we will show how you can install it yourself, run it and make use of the efforts of other groups trying to improve Whisper.

Berend Jutte (Attendi)  

Our presentation will discuss language tech to make healthcare better, focusing on how it can help healthcare professionals and the challenges of using it. The basis of good care, for Attendi, is a strong relationship between healthcare professionals and clients. This involves having time and attention for each other and being able to engage in meaningful conversations.

Hubert van Beusekom (TNO)  

TNO, SURF and NFI will jointly develop an open Dutch language model: GPT-NL. These parties, in collaboration with parties who will use GPT-NL, want to take an important step towards reliable, transparent and fair use of (generative) AI in the Netherlands and in Europe. In this presentation we will tell you more about the ambitions and possibilities with GPT-NL and we would like to discuss possible applications of GPT-NL in your own organization.

Dragoș Bălan (University of Twente) 

Numerous studies have been conducted on Dutch speech recognition, encompassing a wide range of speaker and speech categories such as read speech, conversations, adults, children, the elderly, and individuals with speech disorders. Despite this, a comprehensive repository containing all the results and their corresponding experimental setups is currently lacking. Past initiatives, like the N-Best 2008 Dutch Evaluation Challenge, are more restricted in the domains evaluated and do not reflect the current state-of-the-art in Dutch ASR. SOS (Stichting Open Spraaktechnologie) presents a joint effort to enhance the existing Dutch ASR Benchmark landscape, with a particular emphasis on replicability, transparency, and accessibility. This solution is open-source and with the goal of being driven by the community, inviting every researcher, company, or developer to contribute, whether it be results or suggestions for improvement.

Matt Coler (University of Groningen) 

In this presentation I will  describe multiple graduate student projects related to inclusive speech technology, emphasizing its role in enhancing accessibility for diverse populations including the elderly, those with speech impairments, and speakers of minority languages like Frisian. I highlight the development of tailored recognition systems to address varied speech patterns and environmental challenges and foment reflection on the importance of linguistic diversity and social inclusion in technological innovation. I close by underscoring how speech technology is pivotal in bridging communication gaps and fostering a more equitable digital world.

Grzegorz Chrupała  (Tilburg University) 

In this talk I highlight recent developments in modeling that bridge the gap between speech processing and NLP and call for the unification of these fields, emphasizing the potential for more efficient and human-like systems, promoting true natural language processing beyond textual communication.

Martijn Bentum (Radboud University, Nijmegen)

A model is a great way to test your understanding of a phenomenon. However, recent advances in the performance of AI models do not necessarily correspond with a deeper understanding of the underlying problem (e.g. speech recognition). Explainable AI tries to elucidate the opaque workings of deep neural networks. In this talk I will discuss the challenges with explainable AI and give some examples from my research in the processing of lexical stress by Wav2Vec 2.0.