US 11,775,565 B2
Systems and methods for database reconciliation
SungMin Park, Seoul (KR); Daeil Kim, Seoul (KR); and Sungyeol Bae, Seoul (KR)
Assigned to COUPANG CORP., Seoul (KR)
Filed by COUPANG CORP., Seoul (KR)
Filed on Oct. 14, 2020, as Appl. No. 17/70,600.
Prior Publication US 2022/0114197 A1, Apr. 14, 2022
Int. Cl. G06F 16/28 (2019.01); G06Q 30/0201 (2023.01); G06Q 30/0601 (2023.01); G06T 7/00 (2017.01); G06F 16/904 (2019.01); G06Q 10/0833 (2023.01); G06Q 10/083 (2023.01); G06Q 10/0835 (2023.01); G06K 7/10 (2006.01)
CPC G06F 16/285 (2019.01) [G06F 16/904 (2019.01); G06Q 30/0201 (2013.01); G06Q 30/0643 (2013.01); G06T 7/0002 (2013.01); G06K 7/10297 (2013.01); G06Q 10/0833 (2013.01); G06Q 10/0835 (2013.01); G06Q 10/0838 (2013.01); G06Q 30/0627 (2013.01); G06Q 30/0629 (2013.01); G06Q 30/0633 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/30168 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A method for database reconciliation, comprising:
receiving, from one or more sources, attribute data of a plurality of attribute data types representing aspects of a product, wherein each of the one or more sources has a source score, based on a source status for each of the one or more sources and a length of active time period of each of the one or more sources, and wherein the attribute data are associated with a corresponding source score;
generating a plurality of attribute categories based on the received attribute data types, each attribute category corresponding to one of the received attribute data types and containing all attribute data of the one of the received attribute data types;
training one or more machine learning models to process data to make predictions based on image characteristics;
determining data scores for each of the attribute data contained in each of the plurality of attribute categories, based on the source score and a product specification, the determining comprising:
analyzing each of at least one image, by the one or more trained machine learning models; and
determining, by one of the one or more trained machine learning models, image match between the at least one image and a representation of a corresponding product in at least one image of the product;
assigning a data score based on the image match determined by the one of the one or more trained machine learning models;
generating reconciled data for each of the attribute categories, the reconciled data being the attribute data in each of the attribute categories having the highest data score and image match data score based on information in the product specification indicating higher quality data;
storing in a database, the plurality of attribute categories each containing reconciled data corresponding to the product;
providing, from the database and via a database-to-database network connection, the reconciled data for display on a display interface with interactive user interface elements, wherein the interactive user interface elements include at least a tab enabling switching between interfaces; and
in response to a user interaction with the interactive user interface elements, transmitting a request to initiate a purchase to an external front end system.