I got to do a little research while on my Masters before I joined the industry.
I'd love to come back to this someday.
In Linked Data navigational queries are expensive to compute as it requires navigating the graph across many possibly servers.
We found a way to use the query itself to derive a consistent heuristic that makes A* available for computing Property Paths more effectively than uninformed Search Algorithms, which helps reducing the amount of nodes expanded to get the answers to a given query.
I'm experimenting with a way to abstract search spaces into a hierarchic space that can provide sub-goals and better heuristics.