US 12,282,486 B2
Address matching from single string to address matching score
Iraklis Psaroudakis, Zurich (CH); Giulia Carocari, Redwood Shores, CA (US); Andrea Ziani, Redwood Shores, CA (US); and Miroslav Cepek, Redwood Shores, CA (US)
Assigned to Oracle International Corporation, Redwood Shores, CA (US)
Filed by ORACLE INTERNATIONAL CORPORATION, Redwood Shores, CA (US)
Filed on Apr. 29, 2022, as Appl. No. 17/733,011.
Prior Publication US 2023/0350903 A1, Nov. 2, 2023
Int. Cl. G06F 16/2457 (2019.01); G06F 16/2458 (2019.01); G06F 16/29 (2019.01); G06N 3/08 (2023.01); G06N 3/10 (2006.01)
CPC G06F 16/24578 (2019.01) [G06F 16/2468 (2019.01); G06F 16/29 (2019.01); G06N 3/08 (2013.01); G06N 3/105 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving an address string;
generating parsed address data based on the address string;
wherein the parsed address data comprises a plurality of subfield values, said plurality of subfield values including at least one street-level subfield value and at least one non-street level subfield value;
using a global toponym database, identifying a plurality of field candidates for a subset of the plurality of subfield values of the parsed address data;
generating a plurality of address tuples based on the plurality of field candidates;
wherein the plurality of address tuples comprises combinations of field candidates of the plurality of field candidates;
generating a plurality of queries for the global toponym database based on the plurality of address tuples;
providing, to the global toponym database, at least one first query of the plurality of queries corresponding to at least one first address tuple of the plurality of address tuples;
in response to the global toponym database failing to identify a canonical street locale for the at least one first address tuple, providing, to the global toponym database, a second query of the plurality of queries corresponding to a second address tuple of the plurality of address tuples;
receiving, from the global toponym database, one or more standardized street locales, each of said one or more standardized street locales being an actual non-street level address associated with the second address tuple;
generating a query for a global address database to: exactly match one or more non-street-level subfield values included in a particular standardized street locale of said one or more standardized street locales and fuzzy match a street-level value of said at least one street-level subfield value, the fuzzy match excluding said one or more non-street-level subfield values;
receiving, from the global address database, a set of candidate addresses that are identical to one of said one or more standardized street locales on a non-street level;
generating a score for each candidate address of the set of candidate addresses, each score of said each candidate address comprising a probability that said each candidate address represent a same place or location as a combination of said at least one street level subfield and the respective standardized street locale of said each candidate address.