US 11,853,362 B2
Using a multi-task-trained neural network to guide interaction with a query-processing system via useful suggestions
Corby Louis Rosset, Seattle, WA (US); Chenyan Xiong, Bellevue, WA (US); Paul Nathan Bennett, Redmond, WA (US); Saurabh Kumar Tiwary, Bellevue, WA (US); Daniel Fernando Campos, Seattle, WA (US); Xia Song, Redmond, WA (US); and Nicholas Eric Craswell, Seattle, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Apr. 16, 2020, as Appl. No. 16/850,886.
Prior Publication US 2021/0326742 A1, Oct. 21, 2021
Int. Cl. G06F 16/00 (2019.01); G06F 16/9032 (2019.01); G06N 20/00 (2019.01); G06F 17/16 (2006.01); G06N 5/04 (2023.01); G06F 18/24 (2023.01)
CPC G06F 16/90324 (2019.01) [G06F 17/16 (2013.01); G06F 18/24 (2023.01); G06N 5/04 (2013.01); G06N 20/00 (2019.01)] 22 Claims
OG exemplary drawing
 
1. A computer-implemented method for assisting a user in performing a task, comprising:
receiving a query provided by a user computing device, via a computer network;
generating at least one suggestion based on the query using a first machine-trained model, the first machine-trained model being trained using a multi-task training framework that performs machine training based on plural groups of training examples,
the plural groups of training examples being produced by plural example-generating methods that differ from each other,
for a first group of training examples, a particular training example providing a sequence of plural queries submitted in a prior search session and stored in a historical log that have been determined by the training framework to exhibit a coherent task-related intent by: using a second machine-trained model to identify intents associated with the respective queries; and then determining relationships among the intents;
generating output information to be sent to the user computing device that includes said at least one suggestion, and sending the output information to the user computing device; and
receiving a reply from the user to said at least one suggestion in response to interaction by the user with a user interface provided by the user computing device,
the computer-implemented method overall guiding the user towards a completion of an inferred task through a succession of suggestions.