![]() |
ISCApad Archive » 2022 » ISCApad #294 » ISCA News » Message from ISCA president Prof. Sebastian Möller |
ISCApad #294 |
Saturday, December 10, 2022 by Chris Wellekens |
Dear friends of ISCA,
This time around the turn of the year is usually a busy period where many of us are dealing with housekeeping activities. This is similar for ISCA: We recently had a workshop for the installation of a new association management software, which will hopefully provide new and long-wished services for all of our community. Along the same lines, we are targeting for INTERSPEECH a new conference management system which requires us to update our reviewer database. In the future, we hope that this system will be able to assign more accurately submitted papers to reviewers with the support of algorithms learning from prior publications of the reviewers, and to better support session building and program set-up.
A number of INTERSPEECH 2022 participants pointed out their dissatisfaction concerning the overemphasis of (special) sessions focussing on end-to-end deep-learning approaches. Whereas the impact and importance of artificial intelligence and machine learning (ML) in our field is acknowledged, it is felt that a too much focus on such approaches would impede new and fresh ideas. It was advocated that there should be more sessions in which the scientific question comes first. In fact, in some areas it was observed that reviewers propose to reject papers if they do not provide quantitative results which go beyond the current state-of-the-art. Whereas challenges have their merit in comparing different approaches to the same task, they seem to dominate the program at times. A stronger balance which emphasizes the importance of speech and language sciences would help balance future programs and will likely lead to greater leaps forward in speech technology and our understanding of the human speech production and perception systems.
The ISCA board takes these concerns serious, and we have started a discussion about the impact of ML-based challenges on our program. The opinions are widespread, ranging from a complete ban of such sessions from the main INTERSPEECH program (thereby encouraging the activities to be organized in terms of more focussed workshops, perhaps run as satellites to INTERSPEECH), through modifying the review process, criteria or paper templates, to collecting and distributing data for such challenges as part of our ISCA services, but outside the conference program. If you have pronounced opinions in this respect, the board would be happy to hear and consider them during the planning of future INTERSPEECH events.
I hope that you will have a good finish of the year 2022, and some quiet days of rest and recovery before starting into a healthy and successful year 2023. I wish you a good read of this informative ISCApad, as usual kindly compiled by Chris Wellekens.
Sebastian Möller ISCA President
|
![]() | Back | ![]() | Top |