Federated intelligent assistance

Patent metadata

Field	Value
Publication number	US-11631017-B2
Application number	US-201815866257-A
Country	US
Kind code	B2
Filing date	Jan 9, 2018
Priority date	Jan 9, 2018
Publication date	Apr 18, 2023
Grant date	Apr 18, 2023

Abstract

Because digital assistants tend to have different areas of expertise and/or different abilities to fulfill a given request, it is sometimes difficult for a user to know which digital assistant is best able to fulfill a request. Representative embodiments disclose mechanisms to increase federate digital assistants so that a user's request can be funneled to the digital assistant best able to fulfill the user's request. A meta-assistant gathers information on skills provided by a set of digital assistants. The meta-assistant also gathers completion data for requests for different digital assistants and user satisfaction information. A user submits a request to the meta-assistant. The meta-assistant extracts user intent from the request and redirects the user's request to the digital assistant best able to fulfill the request. Embodiments can utilize trained machine learning models or scored algorithmic approaches to select the digital assistant.

First claim

What is claimed is:

1. A method implemented on a computer executing instructions to provide a meta digital assistant to users via a computer network, the method comprising:
receiving, at the computer, input representing a request from a user intended to be filled by a digital assistant having machine executable instructions configured to accomplish a designed task when executed;
upon receiving the data representing the request, extracting, with the meta digital assistant, a user intent from the input representing the request, the user intent indicating a target task desired by the user;
identifying, with the meta digital assistant, a digital assistant capable of fulfilling the target task desired by the user from among a plurality of digital assistants, including:
submitting the extracted user intent to a trained machine learning model to identify two or more candidate digital assistants whose corresponding designed task matches the target task desired by the user from the extracted user intent, the machine learning model being trained on a set of designed task data gathered from the plurality of digital assistants and a plurality of user intents extracted from a plurality of requests from users, wherein each of the two or more identified candidate digital assistants is capable of intelligently interacting in conversation with the user to accomplish the target task; and
utilizing a set of rules to select the digital assistant from the two or more identified candidate digital assistants based on the extracted user intent; and
responsive to selecting the digital assistant, invoking, by the meta digital assistant, the selected digital assistant to execute corresponding machine executable instructions to interact in conversation with the user to accomplish the corresponding designed task to effectuate the extracted user intent, wherein each of the two or more identified candidate digital assistants has a skill that causes the digital assistant to be capable of intelligently interactively interacting in conversation with the user to accomplish at least one particular task, such that at least one of the two or more identified candidate digital assistants is capable of accomplishing a different particular task than another of the two or more identified candidate digital assistants.

2. The method of claim 1 , further comprising responsive to identifying the digital assistant, presenting the two or more identified candidate digital assistants to the user along with explanatory information.

3. The method of claim 2 further comprising:
receiving a selection from the user indicating that one of the two or more identified candidate digital assistants should be engaged; and
in response to the received selection, invoking, by the meta digital assistant, the identified digital assistant to execute corresponding machine executable instructions to accomplish the corresponding designed task to effectuate the extracted user intent.

4. The method of claim 1 further comprising collecting completion data comprising one or more of:
an indication of whether or not the request was completed;
a time measure of how long the request took until completion or abandonment; or
a dialog step measure of how many steps in a dialog until completion or abandonment.

5. The method of claim 1 further comprising collecting one or more of:
which one or more of the identified candidate digital assistants the user prefers;
a relative indication of whether the user prefers one digital assistant attribute over another; or
an indication of a digital assistant attribute the user prefers.

6. The method of claim 1 , further comprising enumerating skills implemented by the plurality of digital assistants and mapping the enumerated skills into a taxonomy used to describe all enumerated skills implemented by the plurality of digital assistants.

7. The method of claim 1 wherein the set of designed task data used to train the machine learning model comprises:
a request received from a user;
an intent derived from the request;
a domain associated with the intent;
whether or not the request was completed;
skills implemented by a digital assistant;
percentage representing a number of times the request was successfully completed by the digital assistant;
user feedback indicating a preference for the digital assistant;
user feedback indicating a preference for a characteristic of the digital assistant;
a time measure indicating how long the request took to completion or abandonment; and
a step measure indicating how many steps in a dialog were performed before completion or abandonment.

8. The method of claim 7 wherein at least some of the set of data used to train the machine learning model is aggregated on one or more of:
a per user basis;
a per cohort basis;
a user population basis;
a per intent basis;
a per intent group basis;
a per domain basis;
a per domain group basis; and
a per query basis.

9. A computer system, comprising:
a processor and device-storage media having executable instructions which, when executed by the processor, cause the computer system to provide a meta digital assistant and to perform operations comprising:
receiving, at the computer system, input representing a request from a user intended to be filled by a digital assistant having machine executable instructions configured to accomplish a designed task when executed;
upon receiving the data representing the request, extracting, with the meta digital assistant, a user intent from the input representing the request, the user intent indicating a target task desired by the user;
based on the extracted user intent, deriving an intended skill used to achieve the extracted user intent;
identifying, with the meta digital assistant, a digital assistant capable of fulfilling the target task desired by the user from among a plurality of digital assistants, including:
submitting the extracted user intent to a trained machine learning model to identify two or more candidate digital assistants whose corresponding designed task matches the target task from the extracted user intent, the machine learning model being trained on a set of designed task data gathered from the plurality of digital assistants and a plurality of user intents extracted from a plurality of requests from users, wherein each of the two or more identified candidate digital assistants is capable of intelligently interacting in conversation with the user to accomplish the target task;
assembling a score for performing the skill for each of the two or more candidate digital assistants;
ranking the two or more candidate digital assistants by the assembled score to identify the digital assistant as a top ranked digital assistant; and

responsive to the ranking, invoking, by the meta digital assistant, the identified top ranked digital assistant to execute corresponding machine executable instructions to interact in conversation with the user to accomplish the corresponding designed task to effectuate the user intent, wherein each of the two or more identified candidate digital assistants has a candidate skill that causes the digital assistant to be capable of intelligently interactively interacting in conversation with the user to accomplish at least one particular task, such that at least one of the two or more identified candidate digital assistants is capable of accomplishing a different particular task than another of the two or more identified candidate digital assistants.

10. The computer system of claim 9 wherein the operations further comprise presenting the top ranked digital assistant to the user along with explanatory information.

11. The computer system

Assignees

• Microsoft Technology Licensing Llc

Inventors

• White Ryen
• Nathan Girish Sthanu

Classifications

• G06F16/3329

  Natural language query formulation · CPC title

• G06Q10/10 Primary

  Office automation; Time management · CPC title

• G06N20/00

  Machine learning · CPC title

• G10L15/1822

  Parsing for meaning understanding · CPC title

• G10L15/26

  Speech to text systems (G10L15/08 takes precedence) · CPC title