Knowledge Graph Creation and Analytics

At today's Cognitive Systems Institute Group meeting a presentation by Costas Bekas of IBM  Research:  Knowledge Graph Creation and Analytics for Cognitive Systems. Slides here.  You can get the recording here.

An improved methodology for knowledge graph analytics that is approximately Order (N).  (Technical).  Previous methods for analytics on complex knowledge graphs were of Order (N Cubed) making them computationally impractical for typical knowledge graphs.

Typical use would be to get a measure of similarity approximation between two graphs.  An example of metallurgy in materials science was shown.  This method will eventually be included in the Watson Discovery Advisor, and likely among the online blue mix services by the end of this year. Quite impressive development, look forward to challenging it in other domains.

(Update) Further ongoing discussion on this and related topics in Linkedin Group.