On November 13th, the Hack-a-Little hackathon for linked data was organised by the Network Digital Heritage. This acted as a replacement for the yearly Hack-a-LOD, which did not occur in 2020 due to the Covid-19 pandemic. This story illustrates the results that were achieved during the event. In this data story, we are going to show how you can combine tools from four different international networks into a single pipeline to visualize data from a CSV onto a map image, whilst offering it as Linked Open Data.
Starting in the 16th century...
A long time ago in a galax... Well, actually, in this galaxy and in a place that we call 'The Netherlands' (at least nowadays). What we know from that era may not be a lot except for an important group of people, those who would rise in the ranks of religious orders and of which we now know quite a lot. For example, we know where they were born and knowing this is a bit more exciting that you would expect. Knowing this raises questions: was faith random? Did the divine call spread evenly across the country? Or were those of the cloth born in the vicinity of churches where the influence of religious orders were stronger?
Step 1: Transpose CSV to Linked Data
The first thing we did was to create Linked Data from a CSV file using the LDWizard: a tool brought to you by the Dutch Digital Heritage Network. Please see this demo to see how you can do it yourself.
One really cool feature of LDWizard is that, through its design, everyone is able to create their own 'flavour' of LDWizard. For this Hack-a-LOD, Jorrit from Kadaster created a geo-flavoured LDWizard which allows you to directly transpose geo-coordinates into properly defined Linked Data. Make sure you watch his demo. This allows us to easily visualise data and to perform geo-related SPARQL queries such as, for example, whether something is close to something else.
This feature proved to be crucial to our project: our 16th century place names, did not at all resemble contemporary place names: e.g. "Oculo" back then is "Schiermonnikoog" now. Not your average regex exercise...
Step 2: Match via Geographic Proximity
So, in order to relate anything contemporary to our 16th century data, we decided to match the historical places based on their geographic location to their respective contemporary places. For the contemporary places, we used Kadaster's BRT dataset which is already available via this endpoint. We uploaded the geo-LDWizard RDF representation of the birthplaces CSV file to an instance of TriplyDB. As a result we could write a federated query, retrieving both contemporary and historical information on the birth places of priests.
Step 3: Eye Candy or Not?
When presenting historic map visualizations, a often heard complaint at a conference is that contemporary maps are ugly (or even 'evil') as layer for the actual visualization. And to be fair, especially in the case of the Netherlands, it is awkward to read place names, see bridges and highways, where 500 years ago, there was nothing but sea. It cast a shadow on the accurateness of the academic work.
Now there are ton of tools that have dealt with this problem and actually a very good one is QGIS. But even QGIS requires, well.. QGIS yet another tool in the pipeline. So instead, Triply brought a new feature to their triple store, one that allows you to use maps as underlays as long as these maps are provided as a WMS service from a secure website (https://). This even works when you don't have access to the triplestore via a federated query. So how do we get a map?
Well, based on knowledge orginating from the New York Public Library, MapWarper is a very decent piece of tooling that allows you to host maps, georeference them and provide them as .kml and WMS (amongst others). For our use case, we decided to pick a map that qualifies for the golden raspberry amongst maps: https://mapwarper.net/maps/40981. That said, many kudos to mapwarper.net for providing this excellent service and don't forget to donate to this good cause. So brace your eyes and visit the Druid data story for this query.
Warping time by 502 years: 1518-2020 what's left of the Monasteries?
So, thus far we have recovered the places that priests are coming from and what we call those places now. But what about their religious institutions? What has become of those? To answer that question we use the 16th century coordinates of the locations of the monasteries and retrieve from the BRT what currently lies at that point: fast forwarding 500+ years in one federated query.
Wrapping the Mapping
Thus far in over four years of hack-a-LODs, the focus was always on a telling an important data story. The Hack-a-LOD has been of crucial importance to the (Dutch) LOD community and, as a result, this year we felt things have matured enough to tell a story about the networks themselves. There are now so many components in place that we are able to exchange tools and information from a variety of networks and suppliers without the need for adding much ourselves. To summarise: source data are provided by the IISG, VU; the Linked Data by Kadaster, the newly LDWizard derived LOD is hosted by CLARIAH and PLDN while the maplayers are provided by MapWarper and PDOK andTriply provided the WMS to SPARQL demonstrator.