Linking data in Amsterdam

The 2016 SWAT4LS conference was held in the city of Amsterdam which has some symbolic connection with the subject of the conference. Like the canals, binding cities and villages in Netherlands together in a net of water ways. Data islands are linked together using semantic technologies allowing data islands to be combined, fertilizing each other with enriched information and knowledge.

The first day of the conference offered five different tutorials. One tutorial was about “HLT7 on FHIR” and how it can be represented using RDF. The most interesting aspect in this for the Uppsala Monitoring Centre is how information about drug usage can be transferred from the sender to the receiver without loss of specificity. Introducing URIs to represent drugs with the possibility to lookup the details of the drug by following the URI is the obvious solution. This fits very well with the strategy at the UMC to make a simple RESTful drug lookup tool available leveraging drug lists containing URIs as identifiers.

The second workshop showed how DICOM data could be “semanticized” and enriched with related information from other systems to provide a rich source of easily accessible information for clinical care and research.

Much of the focus during the conference itself was on research projects aiming at mapping and combining data using semantic technologies to discover unknown patterns and making the total bigger than the individual parts. There were also a number of presentations where different tools and data stores where mentioned. Both existing and newly developed. Various topics like gene data sets, topological pathways, wikidata linkage, structuring of scientific literature, ontological annotations and much more were covered.

One very interesting presentation, by Ontoforce, covered the problem of benchmarking different RDF engines. During the tests some odd results where noted, indicating that poorly equipped servers would perform better than powerful, parallel systems. After some research it was discovered that the “too good to be true” results were caused by the queries crashing without causing an error. The queries simply returned an empty result set and an “ok” which caused the benchmarking tool to record a very high performance score. The wisdom and key message in this presentation was to not only measure performance but also measure correctness. It is also clear that many RDF stores are error prone and do have performance problems especially when the work load is high.

SENCKENBERG presented FLOPO which is an ontology integrating lots of information about plants including images. They do have links to, or at least information from, KEW which makes it interesting when it comes to enriching the WHODrug Herbal with more information about herbal substances.

Marco Roos had a presentation where he tried to define the need of a new role, between data owner and database engineer (tech) which he called a FAIR Data Steward. The FAIR Data Steward should be responsible for the data and how it is maintained as it becomes part of the Linked Open Data cloud.

To get away from the conference environment the first day of the conference was ended with a very nice boat trip along the canals of Amsterdam. We had the pleasure of enjoying the “Amsterdam Light Festival” during the trip as we passed a number of exciting light installations. After the boat trip we were all invited to the official conference dinner at an Italian “inspired” restaurant along one of the canals.

Day number two had a more practical focus than the first day. I presented “WHODrug as Linked Data” as mentioned before during a five-minute flash presentation. The interest in what WHODrug as Linked Data, introducing URIs as drug identifiers, can bring to the world of Linked Open Data, is great. The comments and discussions following the presentation is very encouraging as we continue to explore the field at the Uppsala Monitoring Centre.

The key note session in the afternoon was held by Dietrich Rebholz-Shuhmann. He presented lots of different projects his team at the Insight Centre for Data Analytics was involved in and he specifically pointed that “Bringing the data to the semantic web is the key”. You need to put semantics on top of the “Big Data” to be able to analyse it. Dietrich also showed some very nice graphics for multidimensional analysis.

In the afternoon there was a talk by Tomas Janssen from the Cancer Institute where the conference was held, showing how semantic technologies actually bring benefits in Clinical Practice. We also had an interesting presentation by Laura Slaughter from the Oslo University Hospital showing how they link different data sources to provide an overview of information about drugs, interactions side effects and much more for pharmacists in Norway. In this project there should be a role for both WHODrug as well as VigiAccess to play. It was not entirely clear but probably the interest from Legemiddelverket in supporting the WHO centre in Oslo to leverage the ATC classification as Linked Open data is related to this project.

As a summary it can be concluded that yet another SWAT4LS conference has been successfully leveraged. There were many interesting topics of various kinds and also many specific research presentations, hard for someone with a background in Engineering Physics - like the author, to grasp. As usual though, the interactions with the participants attending the conference during breaks and the social events have been just as fruitful as the conference itself.