Knowledge graphs are in the very heart of IPRally. They make it possible for the artificial intelligence to learn how the patent system works and to find the most relevant documents for you.

If you are familiar with "claim charts" (or "feature charts"), then you will immediately get the idea of the graphs: they are basically the same thing, but now in a structured and automated format.

Graphs contain two types of items: Features and Relations.

Features are basic concepts of the invention, like "vehicle", "water hose", "means for manipulating", "at least two elements".

Features can be nested under each other. A parent feature either contains or conceptually covers its child features. A root feature has no parents. A root is usually the main concept of the invention, typically the sales unit of the company, like "washing machine". Its children can comprise e.g. "rotatable water jet member" (the machine contains the member) or "dishwashing machine" (the machine is a dishwashing machine).

Relations help to define complex technical relationships, that cannot be expressed simply by nesting features. Relations are nested in the graphs under features but cannot be nested with each other. A relation must refer to its parent feature. A single-feature relation is for example "water hose for watering garden" (defining the purpose of the feature water hose). A multi-feature relation is for example "water hose is connected at first end to a water output and at second end to sprinkler means".

Building search graphs automatically using publication numbers or copy-pasted claims and inspecting them for a while will quickly demonstrate the graph approach. 

Did this answer your question?