“THALES” is a Deep Neural Network (DNN) that was developed entirely by the Hellenic Industrial Property Organisation (OBI), specifically within the Search Directorate, by a team of scientific experts with knowledge or studies in the field of Artificial Intelligence (AI).
In very general terms, a Neural Network attempts to mimic the functioning of the human brain, which is composed of billions of neurons interconnected in a dense network along connections called Synapses.
After long-term testing tailored to the technical/scientific texts of Patent Applications, the team arrived at an architecture that, without going into technical details, includes a total of 483 neurons (or units) distributed across 5 hidden layers, with a total of 31 million trainable parameters.
The goal of developing “THALES” was to automate the initial review and classification process of incoming patent applications based on the technical/scientific field they relate to, and to subsequently route/assign them to the appropriate examining departments of OBI. This process used to take the working team several days to complete.
Today, “THALES” has fully replaced the previous working team, and the entire process of reviewing and assigning new patent applications now takes only 2-3 minutes.
