US 11,869,490 B1
Model configuration
Rahul Gupta, Waltham, MA (US); Jwala Dhamala, Boston, MA (US); Melanie C B Gens, Seattle, WA (US); Sachin Midha, Bothell, WA (US); Jennifer Yuen, Seattle, WA (US); Dewan Muhammed Ibtesham, Redmond, WA (US); Wael Hamza, Yorktown Heights, NY (US); Xinhong Zhang, Mercer Island, WA (US); and Md Humayun Arafat, Bellevue, WA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Aug. 14, 2020, as Appl. No. 16/993,482.
Int. Cl. G10L 15/183 (2013.01); G10L 15/06 (2013.01); G06N 3/08 (2023.01); G06N 20/00 (2019.01)
CPC G10L 15/183 (2013.01) [G06N 3/08 (2013.01); G06N 20/00 (2019.01); G10L 15/063 (2013.01)] 21 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
generating, using a first machine learning (ML) model configured to perform natural language processing first output data responsive to a first user input, wherein the first ML model is associated with a first hyper-parameter having a first value;
determining the first output data corresponds to an undesired response to the first user input;
based at least in part on determining the first output data corresponds to the undesired response to the first user input, generating a second ML model corresponding to the first ML model, wherein the second ML model is associated with the first hyper-parameter having a second value instead of the first value;
processing a first set of user inputs using the second ML model;
determining a first metric value corresponding to processing of the first set of user inputs using the second ML model;
determining a number of values to test for the first hyper-parameter;
determining the number of values is satisfied by processing of the first set of user inputs using the second ML model; and
based on the first metric value and the number of values being satisfied, storing the second ML model for processing of future user inputs.