US 12,032,915 B2
Creating and interacting with data records having semantic vectors and natural language expressions produced by a machine-trained model
Li Huang, Sammamish, WA (US); and Congyong Su, Sammamish, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Jul. 24, 2022, as Appl. No. 17/871,968.
Application 17/871,968 is a continuation of application No. 16/869,336, filed on May 7, 2020, granted, now 11,429,792.
Prior Publication US 2022/0358294 A1, Nov. 10, 2022
Int. Cl. G06F 17/27 (2006.01); G06F 16/25 (2019.01); G06F 40/30 (2020.01); G06N 5/04 (2023.01); G06N 20/00 (2019.01)
CPC G06F 40/30 (2020.01) [G06F 16/258 (2019.01); G06N 5/04 (2013.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for processing input information, comprising:
receiving the input information, the input information including one or more original words;
converting the input information into a natural language expression by using a machine-trained model to predict one or more new words that are not included in the one or more original words, and to insert the one or more new words among the one or more original words;
converting the natural language expression into a generated semantic vector in a vector space using the machine-trained model; and
providing output information that is produced based on the generated semantic vector and/or the natural language expression,
the machine-trained model being produced based on training examples generated in a mining process performed on network-accessible pages.