US 12,032,627 B2
Hybrid transformer-based dialog processor
Jinchao Li, Redmond, WA (US); Lars H. Liden, Seattle, WA (US); Baolin Peng, Bellevue, WA (US); Thomas Park, Seattle, WA (US); Swadheen Kumar Shukla, Kirkland, WA (US); and Jianfeng Gao, Woodinville, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Nov. 15, 2021, as Appl. No. 17/526,806.
Prior Publication US 2023/0153348 A1, May 18, 2023
Int. Cl. G06F 16/60 (2019.01); G06F 16/63 (2019.01); G06N 20/00 (2019.01)
CPC G06F 16/63 (2019.01) [G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, the method comprising:
receiving a query, wherein the query is a part of a dialog representing at least a part of a task;
predicting, based on the query, a first candidate response to the query using a dialog embedding of a first model, wherein the first model includes a data-driven, pre-trained generative model;
predicting, based on the query and the task, a second candidate response to the query using a second model, wherein the second model includes a classifier, and the classifier determines the second candidate response by using a set of predetermined rules of the task;
determining, based on a combination of the first candidate response and the second candidate response, a third candidate response to the query; and
transmitting the third candidate response as a next action in the dialog.