In Cologne it feels like autumn has arrived and somehow fittingly our Summer School “Deep Learning for Language Analysis #DL4LASS” came to an end. In fact, we were able to celebrate the fifth edition of this event this year and we managed – after two online years – also to return to the University of Cologne campus. Since one always forgets the impressions so quickly, I would like to report briefly here on the rather entertaining – and hopefully also instructive – week.
We made the decision to organise an on-site school again sometime in spring after two virtual summer schools that actually worked quite well. Of course, we had to weigh up many advantages and disadvantages. Traditionally there is quite an international audience for whom an online format makes it easier to participate in our school. For a third consecutive online year, we could have built relatively comfortably on the last two editions, and the organisation would have been a much smaller effort. And yes, of course, the Covid-19 pandemic status was and is somehow not really clear yet. On the other hand, we were keen to test out once again whether we could revive the very special flair of our Summer School in the University’s Philosophikum, with all-day opportunities for exchange over coffee or tea in the Atrium and a very Cologne-specific social event in the Brauhaus Pütz.
The early decision in favour of an on-site event was then briefly put to the test again when it became clear that Christian Fischer (staff member at Competence Area III – Quantitative Modelling of Complex Systems), who had supported us in the last two editions, was leaving the university and that it would probably not be possible to find a successor in the short time available. Christian was already the third CAIII coordinator in four years of Summer School, and we are beginning to feel like we were wearing out teachers for “Defence against the Dark Arts” at Hogwarts. After a bit of internal restructuring, however, we found someone who was able to take over the coordinator’s job (me) and act in a dual role at the Summer School (alongside my role as one of the long-time scientific organisers).
In the end, however, the effort paid off, at least as far as I can tell. For the entire duration of the school, there was a work-promoting, harmonious and innovative atmosphere, even if the start was not the best, because unfortunately the keynote had to be cancelled due to illness. But it was quickly replaced by an additional session by Nils Reiter and Paul Wallbott, who – as last year – provided the main act of the Summer School with their parallel workshops “Text Analysis with Deep Learning” and “Deep Learning with Audio & Speech Data”, which lasted from Monday afternoon until Thursday morning. This year’s tutors were Moritz Eßer and Nina Eckertz, both students of the MA Information Processing offered by our Department of Digital Humanities. In order to be able to provide the participants with adequate computing machines, we were once again able to use the two Nvidia DGX-1s from the Center of Data and Simulation Science.
On Tuesday evening, in good tradition, we (teachers, participants, organisers, tutors, staff) visited the Brauhaus Pütz together. To be on the safe side, I had already briefed the participants in the morning about how it is with the small Kölsch glasses that never get empty. In breweries, you don’t have to be proactive to get another Kölsch. It comes on its own. To curb this, you have to put your beer mat on the glass, otherwise it goes on and on until the beer taps. Since I had to discuss the payment modalities with the landlord and the Köbes (that is the name of a waiter in a Cologne brewery) beforehand, I was there first and accordingly finished my Kölsch first. The exchange (empty glass against filled glass) was immediately demonstrated by the Köbes, much to the delight of the group of students from the UCL, which was already present. The cheerfulness was preserved until the end of the evening, and some people from our international audience showed impressive courage in their choice of food (Himmel un Ääd, Heringspöttche rut un wieß).
As in previous years, we were able to win NVIDIA as a partner for our school and were allowed to offer one of the courses developed there as part of our Summer School. This year we decided on “Building Conversational AI Applications“, and I am looking forward to the feedback we get from the participants. Initial reports lead me to believe that the course was very instructive, but possibly too specialised for participants from the scientific community, since NVIDIA not only supplies the hardware here, but also markets its own software product.
We will collect the various impressions and take them into account when planning our next Deep Learning Summer School. In fact, we have already started thinking about how we can further develop ourselves and the school. That there will be an edition in 2023 seems pretty certain to me. What exactly it will look like – let’s see. For now, we would like to thank all teacher, tutors, helpers and participants for the – at least from our point of view – very successful Summer School and look forward to further public or private feedback!
For those who are interested – here are the blog posts on #DL4LASS from the past four years:
- Report from 2021
- Diary from 2020
- Announcement from 2019 / Review of 2018 (german)
- Announcement from 2018 (german)