Master theses

Building a data strategy for the Deparment of Education in Flanders

Keywords: AI, Artificial Intelligence, Data Retrieval, Linked Data, Semantic Web, Indexing, open data

Supervision: Pieter Colpaert

Students: max 1

The Flemish government maintains base registries, such as the list of addresses, or all schools in Flanders. As names for schools or a full text representation of an address is difficult to compare, they maintain official identifiers. However, when looking at how other people organizations then reuse lists of schools or addresses, we see that basic problems occur: type-ahead fields show multiple results with the same label, and the data is not exchanged with their official identifiers.

When third parties need to reuse official identifiers, and API is needed to go from a description of a thing to the official identifier. APIs like this exist, such as GraphQL, SPARQL, or simply a type-ahead API. However, these APIs are expensive to host, and currently the only fall-back are data dumps, which require everyone to do their own data integration, indexing, and querying.

The goal of this master thesis is to republish the data of all addresses in Flanders (for Informatie Vlaanderen, supervised by Raf Buyle) and all schools in Belgium (for Departement Onderwijs, supervised by Bart Colpaert), according to the state of the art in hypermedia Web APIs, and following the ideas of Linked Data Fragments. We started working a hypermedia specification called TREE, which enables describing hypermedia interfaces based on multidimensional indexes. When the same client code can be used to autocomplete schools and addresses, or find them based on a description, while they are published from different servers, the master thesis succeeded.