Harm Delva

PhD student
IDLab, Ghent University – imec
AA Tower, Technologiepark-Zwijnaarde 122, 9052 Gent

About me

I am a PhD student at IDLab, affiliated with both Ghent University and imec. My original research subject lay at the intersection of the Semantic Web technologies and Linked Open Data publishing, but with a focus on researching how applications can find and use this data - beyond the traditional SPARQL usecase. I have worked on a flexible and multi-modal route planner, as well as an autocompletion client that can query several heterogeneous data sources at the same time. Our route planner is able to find subjectively pleasant routes by asking the user what they want to see while walking; see it in action here. And, because all computations are done on the client, there are no data-hoarding behemoths that are learning your opinions on shopping streets.

Other than that, I have developed an appreciation for the intricacies of Geographic Information Systems (GIS), especially for Volunteered Geographic Information (VGI) projects such as OpenStreetmap and, closer to home, Telraam. Of course, it is not enough to simply publish Open Data: people must be able to find this data. Whenever time permits it, I like to work on making data discoverable, and try to follow the developments within the FAIR community. Alternatively, I would like to help pave the way towards more mature WebAssembly tooling for working with RDF data — my applications have a need for speed.

Last, but not least, I would love to explore Semantic Web applications that focus on pen-and-paper games, such as as Dungeons & Dragons. Why spend years modelling and gathering data from the real world, when there are perfectly suitably imaginary worlds out there? By avoiding pedantic discussions (what is an address really?), we could skip straight to the good stuff such as semantic interoperability.
Curious to hear what I have in mind? Feel free to reach out, I will always find time to discuss this.

Professional Experience

Additional Academic Experience
  • Assisted with the practical sessions for the Programming, Web Development and Big Data Science courses at Ghent University.
  • Mentored Master Thesises:
  • Jeroen Flipts. Fragmenting Public Transport Timetables On the Web In Tiles for Serverless Route Planning. 2020.
  • Thomas Neuser. Een Routeplanner Op Basis Van Open Data: Hoe Hypermedia Controls De Semantische Beschrijving Van Datasets In Planner.js Kunnen Verbeteren. 2020.
  • Pieter-Jan Vandenberghe. Designing a Web Interface for Scalable Public Transport Routing Through Spatiotemporal Fragmentation of Public Transport Networks. 2019.
  • Tim Vanhee. Studying the Technique of Contraction Hierarchies for Routing On Web-scale. 2019.
  • Program committee member at WSDM 2020, SIGIR 2020, SIGIR 2021
  • Publications

    2020

    2019

    Projects

    Planner.js github
    A client-side route planner that uses Linked Data Fragments to compute routes just the way you like them. My own contributions here have been to optimize the route planner to be competive with server-side alternatives. Computing long routes isn't entirely there yet, but it's definitely useable for city-level routes. Have a look at our Regions of Interest demo for example, where you can tweak a walk through the city of Ghent to your heart's content. Or if you want to see how longer routes can still be feasible, we have a Highway Hierarchies-like demo available.
    TREE spec
    The glue between our Linked Data Fragments; the TREE specification can be used to create paginated, ordered, collections. But unlike similar vocabularies, TREE also describes what can be found by following a specific link. Supported descriptors include time ranges, geospatial regions, and textual patterns. For an example of the latter, have a look at the substring fragmenter that we created together with Netwerk Digitaal Erfgoed.
    TREE+Comunica github
    Comunica is already a great framework for querying the decentralized Web, and we added TREE support to make it even better. Development is still ongoing, but we have a pretty cool autocompletion demo up and running. All the resources this demo uses are hosted on Github Pages; it's like the classic jQuery autocompletion but better.
    NGSI-LDF github
    Do you have NGSI-LD data you want to make public, but the thought of having to manage a rate limiting and authentication service is holding you back? Consider publishing your data through a read-only view of cacheable data fragments. Clients can use these to answer any NGSI-LD query you may have; or you could even use SPARQL instead if you like semantics.

    Education

    Righteous Dudes

    Good people working on good things.

  • Jonathan Peck
  • Robbert Gurdeep Singh
  • Pieter Colpaert
  • Julián Rojas
  • Brecht Van de Vyvere
  • Ruben Dedecker
  • Ruben Verborgh