SEMANTiCS2025 Trip report
As I wasn’t able to get my family logistics in the first week of school sorted out in such a way I would be able to attend the whole of SEMANTiCS2025, I attended the first day online. It’s great about hybrid conferences nowadays that it gives you these options. This way I also had both experiences of the hybrid conference.
The keynote of Hannah Bast
The keynote of Hannah Bast was as thought-provoking as impressive. At the start of my PhD on linked transport data her publications on route planning were instrumental for my own research. Today she’s shaking up the world of triple stores with her work on the QLever SPARQL engine that is able to beat the state of the art on many fronts for querying truly big knowledge graphs such as OpenStreetMap, Uniprot, Pubchem or Wikidata. During the keynote she was very down to earth about benchmarks: although QLever scored pretty well according to their own tests on these benchmarks, they are still working on yet another SPARQL benchmark to have a more honest comparison. The work on SPARQLoscope is to be presented at ISWC2025 later this year.
A line during the keynote that certain stuck was: “professors don’t code”. This was given as a reason why academic software is often of low quality: students have to learn the coding part by themselves. Hannah Bast herself is certainly an exception to this statement. I personally did not have the impression that “professors don’t code” within our field, certainly as I had her as an example during my PhD. Also my own supervisor at the time, Ruben Verborgh, I often found deep-down in coding projects.
Nevertheless, I do not think my own strenghts can be found in my coding qualities. I do write code, but that’s more incidental—like peeling potatoes when you want to make fries. While I have tried to position decent software in the past, my focus today is on building interoperable dataspace ecosystems through specifications (see previous blog post). Luckily in our Knowledge on Web-Scale team we have multiple professors, each with a different set of qualities. I would have liked to talk this over with Hannah Bast over a coffee break, but wasn’t able to talk to her before she had to leave—that will have to happen on a next occasion.
Our team in an organizing role
Our team took on quite some responsibilities for organizing workshops and tutorials:
- It’s certain that the world of Linked Data needs better developer tooling. Ruben Taelman, Jerven Bolleman and Jindřich Mynarz co-organized the Developer Workshop.
- A note to self: if you let your PhD students organize a tutorial at a conference, the tooling they are organizing this tutorial for is all of a sudden going to be pollished. Ieben Smessaert and Arthur Vercruysse organized the RDF-Connect tutorial for building data pipelines across environments. You can still do the tutorial yourself at home and ask them for feedback. The tutorial is going to extended as a full-day tutorial at the ISWC2025 conference.
- The Semantics for Transport workshop was organized by a new organization committee this year. I hope it’s the start of renewed interest in the topic!
- The NXDG workshop on next generation data governance, about technologies like the Data Privacy Vocabulary (DPV) and the Open Digital Rights Language (ODRL) organized by Harshvardhan Pandit and Beatriz Esteves.
A fishbowl session on dataspaces and whether semantics still play a role
I brought an opening statement in the fishbowl in which I referred to the Eclipse dataspace protocol. This specification uses JSON-LD, ODRL and DCAT. How can anyone claim that semantics would not be of interest to dataspaces? They use our work!
What remains however is that, on the data plane (the ODRL and DCAT is in the “control plane”, to set up the data connection), we still have to fight the same fight. If we want interoperable instance data, we will still need to bring semantics to the individual domains. That’s not a characteristic of dataspaces specifically, but of working with data in general.
In fact, the dataspaces protocol adds an important missing piece of the puzzle that we often forget within the Linked Data world. Next to vocabularies and application profiles, we should also work on interaction patterns, such as the contract negotiation steps in the dataspaces protocol.
The DiSHACLed lunch
We have a Flemish funded project going on called DiSHACLed. For that project we’re working on (i) extending DCAT with SHACL shapes for discovy algorithms in data portals, (ii) on SHACL based data pipelines with RDF-Connect, (iii) on extending SHACL with UI features (Ieben from our team is following the W3C data shapes working group for that reason), and (iv) looking at business models for dataspace actors. We’re doing this with Inuits’ Elody team, RedPencil and Sirus: three high-potential companies in the world of semantics and interoperable data services.
RedPencil is well-known within the SEMANTiCS community already. For Elody, it was a first time encounter. We had a lunch together to discuss the conference and discuss how to approach certain tasks within the project differently now.
Jelly and nanopublications
Probably the biggest aha-erlebnis I got during the conference was with Jelly of my namesake Piotr Sowiński. He had a presentation in which he showed the great speed ups (proceedings yet to be published...) of implementing this within nanopubs. During a coffee break I expressed my skepticism that it is really the binary format that did the trick. However I was shown to be wrong: with the help of an LLM I coded up a Python script (I’m not a python developer, and Jelly doesn’t have NodeJS support at this moment), that clearly showed an impressive increase in throughput. The test is a bit simplistic, but even if you would add the complexity, the increase will remain substantial. Let’s make sure Linked Data Event Streams can work with Jelly.
P.S.
Next year, we’ll be the local host of SEMANTiCS and hope to welcome you all in Ghent!
Notes to self for organizing next year:
- Make sure people’s names are on both sides of the name tag
- Make sure the proceedings are available before the conference: not being able to go through the paper while attending talks breaks how some people, including myself, try to attend talks efficiently
- Try to get a brass band during the conference dinner (Thanks Jean-Marc Acke for the suggestion)