Datasets and useful relationships

[decentral1se]

ACTION: @nickdickinson to pick datasets and useful relationships

• We leaned towards would be the most minimal examples of some insight that can be used to talk people through or show in brief demos during talks/visits/calls etc. in order to get a feedback loop going with potential end-users.

[decentral1se]

Following from https://github.com/nickdickinson/WASHWeb/issues/1#issuecomment-461866250, I've found a nice document which I'll drop here. Looks like we need to use the thinking contained in this document (or if you have others, please share) in order to inform what we want to learn from this issue. Here's the page (from the very nice Neo4J documentation):

https://neo4j.com/developer/guide-data-modeling/

So, if we can draw it on a whiteboard, we're pretty much there ;)

[nickdickinson]

I can explain the following and we can draw them out. The entities are:

• International commitments
• Country commitments (multi-stakeholder)
• Individual commitments (one stakeholder / constituency)
• Actions / programmes contributing to achieving those commitments
• Source documents that embody / encode those commitments or provide context (background research, conference statements, policy documents)
• Targets / milestones that define progress towards that specific goal and have properties about their achievement
• External data sources that we can query to shed light on targets / commitments (these may be outside of the graph)

This would be the graph.

However, a lot of knowledge out there is not in the shape of a graph. The external data sources like water point data (GPS coordinates / tables of data) add a lot of rich information.

As soon as we start look at full text analysis, since lots of information isn't formally organized, then we would potentially be looking at creating a taxonomy of WASH terms and linking them to the network above and using that to link concepts/entities in texts to the graphdb.

This is getting pretty complex but I'm curious if there are already tools out there to link for example graphdbs to elastic search or natural language processing and if so, then maybe it only sounds hard but it is actually doable.

[decentral1se]

OK, closing this one out until we gather more steam again!

(cleaning out my issue tracking list ...)