	US 12,205,579 B2
	Systems and methods for dynamically updating machine learning models that provide conversational responses
Tate Travaglini, New York, NY (US); Andrew Oestreicher, Washington, DC (US); Victor Alvarez Miranda, McLean, VA (US); Parag Jain, Falls Church, VA (US); and Rui Zhang, New York, NY (US)
Assigned to Capital One Services, LLC, McLean, VA (US)
Filed by Capital One Services, LLC, McLean, VA (US)
Filed on Oct. 18, 2023, as Appl. No. 18/489,482.
Application 18/489,482 is a continuation of application No. 18/169,162, filed on Feb. 14, 2023, granted, now 11,798,540.
Application 18/169,162 is a continuation of application No. 17/193,774, filed on Mar. 5, 2021, granted, now 11,605,375, issued on Mar. 14, 2023.
Prior Publication US 2024/0046922 A1, Feb. 8, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G10L 15/22 (2006.01); G06F 40/30 (2020.01); G10L 15/065 (2013.01); G10L 15/18 (2013.01); G06F 40/35 (2020.01); G10L 15/16 (2006.01); G10L 15/183 (2013.01)

CPC G10L 15/1822 (2013.01) [G06F 40/30 (2020.01); G10L 15/065 (2013.01); G10L 15/1815 (2013.01); G10L 15/22 (2013.01); G06F 40/35 (2020.01); G10L 15/16 (2013.01); G10L 15/18 (2013.01); G10L 15/183 (2013.01); G10L 2015/221 (2013.01); G10L 2015/223 (2013.01); G10L 2015/227 (2013.01)]

20 Claims

2. A method for dynamically updating machine learning models that provide conversational responses, the method comprising:

receiving an output from a machine learning model, wherein the machine learning model is trained to determine respective confidence levels for a plurality of specific intents, wherein each of the respective confidence levels indicates a respective likelihood that a user has a corresponding specific intent, and wherein the output comprises the respective confidence levels for the plurality of specific intents;

determining that a confidence level of a specific intent of the plurality of specific intents is below a threshold confidence level;

retrieving a substitute specific intent from a configuration file based on determining that the confidence level of the specific intent is below the threshold confidence level, wherein the configuration file comprises a mutation algorithm for the machine learning model that introduces a modification to rules of the machine learning model used to select specific intents;

selecting the substitute specific intent from the configuration file, wherein the substitute specific intent corresponds to a dynamic conversational response; and

generating for display, at a user interface, the dynamic conversational response during a conversational interaction.