Hi, if you wish to collaborate to a project to accelerate science and knowledge discovery, PVT me.
I’ve been using multi-partite graph database modelling, with benchmarks on factual knowledge, IP, scientific articles. Advantage of this model allow to achieve good results even without recommender systems (bi-partite networks) of users and items. Next step, is to study how they evolve and predict knowledge: deep learning applied to graph topology.

Currently self-funded project.

I developed also interface to navigate knowledge networks, below a public demo.

http://nifty.works/