Semgine's myMap for exploring semantic networks of information

At SemanticCamp the other day, one topic that I raised several times was the need for a decent graphical tool for navigating large graphs of data: focusing on the information that you're interested while remaining connected to the whole data web.

TouchGraph is fairly good at this: double-clicking on a node will dynamically load in all the information connected to that node, but it's indiscriminate about what information is loaded. It also recalculates its own layout as new information is added so nodes never stay fixed at the same point. The same is true for Moritz Stefaner's Relation Browser.

There's also Tim Berners-Lee's Tabulator, but this suffers - as you expand the graph - from showing the same information at multiple points in the tree. An ideal tool would collapse nodes that share the same identifiers.

None of the existing RDF or graph browsers that I've tried really fit the bill.

Until Thomas Meyer came into the office recently to demonstrate Semgine's Sciplorer (which is interesting enough as a domain-specific search engine, you can try it online), and most importantly a prototype of a new tool they're developing as part of Sciplorer, called myMap, for creating mind maps of information extracted from a semantic web of data.

In myMap, you start with a node - a term chosen from an extensive ontology, eg 'von Willebrand factor'. From there, you draw an edge. You choose what kind of edge it is, eg 'is carrier of', from a pop-up box showing all the known relations for that item. The pop-up box also contains all the known items that could be placed at the end of that edge, and their supporting statements extracted from the literature: in this case you might choose 'Coagulation factor VIII'.

Continue doing this and you end up with a map containing only the information you're interested in, laid out in a way that makes sense to you.

2008-04-10-Imind

The obvious next step to this would be to automatically join nodes together when they share identifiers, and then offer to automatically find the shortest paths across the graph to connect two selected nodes together: knowledge discovery, in other words.

I'm excited about this interface, and can see it being applicable in lots of situations where there's a large web of semantic data to explore.