I’ve sketched a proposed workflow how to convert data from existing datasets (that are distributed and do not know of each other) into linked datasets with distributed attributes.
I’ve sketched the manual way, where each POI is manually merged via a proposed TransforMap Editor software by the user. Automatic merges are a looong more way to go and very complicated - postponed for the moment.
The goal is to transform 1-3★ - datasets into full 5★ Linked Open Data!
So how should the system work?
Before the editor can work, we have to collect existing open datasets (OSM, GreenMap, and other ones from our collection), and create inter-linked taxonomies between them (into our Taxonomy DB). This is needed because we want to compare the different attributes of different datasets and merge them.
We also have to specify APIs, how to fetch spatial data from different distributed databases for each of the datasets used as input.
What is the workflow - described in words:
- The user wants to add a POI, searches either via name or by clicking on a map.
- The editor queries all existing databases for POIs near the found/specified coordinates
- A list of found datasets from different databases is presented to the user, the user checks all that match the POI he want to add to TransforMap.
- The editor merges the different attributes from different databases (looks up corresponding fields in our Taxonomy DB), and presents the user to choose between attributes (if they differ), and/or add new ones.
- When the user is finished, the editor uploads the different attributes to the databases where they belong to. It also uploads links to the object on other databases to each database, creating a linked dataset.
The source files for this graph can be found on github, called “transformation_to_linked_data”.
“the editor” is the key part in merging attributes. Of course it should be possible to add completely new POIs too. Its proposed operating modes are described in another Discourse post.