US 12,332,941 B2
Automated metadata asset creation using machine learning models
Kyle Miller, Durham, NC (US)
Assigned to Adeia Guides Inc., San Jose, CA (US)
Filed by Adeia Guides Inc., San Jose, CA (US)
Filed on Jan. 25, 2023, as Appl. No. 18/101,327.
Application 18/101,327 is a continuation of application No. 16/883,053, filed on May 26, 2020, granted, now 11,593,435.
Prior Publication US 2023/0244721 A1, Aug. 3, 2023
Int. Cl. G06F 16/245 (2019.01); G06F 16/334 (2025.01); G06F 16/903 (2019.01); G06F 18/214 (2023.01)
CPC G06F 16/90344 (2019.01) [G06F 16/3347 (2019.01); G06F 18/214 (2023.01)] 18 Claims
OG exemplary drawing
 
1. A method, comprising:
receiving a received database record from a remote database, wherein the received database record comprises metadata of a remote content item;
identifying, in a local database, a plurality of potential matching records that match the received database record;
determining, using a matching machine learning model, a respective probability of a match between each respective potential matching record of the plurality of potential matching records and the received database record based on inputting into the matching machine learning model: feature scores of the received database record and feature scores of each respective potential matching record;
inputting into an in-database matching machine learning model at least two of the respective probabilities of the match between each respective potential matching record and the received database record to cause the in-database matching machine learning model to output a probability score of the received database record corresponding to at least one record of the local database;
determining, using an out-of-policy machine learning model, a probability that the received database record fails to comply with inclusion policy rules based on inputting into the out-of-policy machine learning model the received database record and a set of inclusion policy rules; and
in response to determining that a combined probability score based on the probability score output by the in-database matching machine learning model and that the probability that the received database record fails to comply with inclusion policy rules is below a predetermined threshold, generating a new record in the local database comprising the received database record.