US 11,941,508 B2
Dialog system with adaptive recurrent hopping and dual context encoding
Quan Tran, San Jose, CA (US); Franck Dernoncourt, San Jose, CA (US); and Walter Chang, San Jose, CA (US)
Assigned to ADOBE INC., San Jose, CA (US)
Filed by ADOBE INC., San Jose, CA (US)
Filed on Feb. 26, 2021, as Appl. No. 17/186,566.
Prior Publication US 2022/0277186 A1, Sep. 1, 2022
Int. Cl. G06F 16/00 (2019.01); G06F 16/242 (2019.01); G06N 3/042 (2023.01); G06N 3/044 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01)
CPC G06N 3/042 (2023.01) [G06F 16/243 (2019.01); G06N 3/044 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving a natural language query that relates to information in a knowledge base;
encoding the natural language query to obtain a first query vector;
performing a first memory hop based on the first query vector to obtain a first readout from the knowledge base;
generating a hop label based on the first readout using a meta network of an artificial neural network (ANN);
obtaining a second query vector from a hidden state of the artificial neural network (ANN), wherein the hidden state of the artificial neural network (ANN) is based on the first query vector and the first readout;
performing a second memory hop to obtain a second readout based on the hop label and the second query vector; and
generating a response to the natural language query based on the second readout.