The project I carried out at Oracle Labs Zurich during an internship was patented.
Abstract: This project presents techniques for matching an address string to a standardized address representation and determining the likelihood of a match. The approach involves parsing the address string into distinct components, normalizing the components to a standard format, and identifying candidate addresses that share similarities with the standardized representation. Each candidate address is assigned a score that reflects the probability of it corresponding to the original address. This matching methodology enables accurate and efficient address resolution, even in the presence of variations in address formatting or colloquial terminology.
Checkout the full paper: