Graph-based supervised Machine Learning search. Each document is converted into graph format that separates the technology’s features. The search query is also in graph format that can be build from scratch, using a publication number or copy-based free text or claim text.
The Neural Network is trained using millions of Patent Examiner citations to mimic the professional search flow.
On top of the deep learning search, a set of traditional boolean filters are available.
Iterative search: hits ranked by the user can be used to improve the results.