Skill discovery for computerized personal assistant

Patent metadata

Field	Value
Publication number	US-11526369-B2
Application number	US-202016826492-A
Country	US
Kind code	B2
Filing date	Mar 23, 2020
Priority date	May 23, 2018
Publication date	Dec 13, 2022
Grant date	Dec 13, 2022

Abstract

A computerized personal assistant communicatively couples to a computer database including a plurality of available skills for the computerized personal assistant. The computerized personal assistant recognizes a current context of the user. The computerized personal assistant operates a previously-trained machine learning classifier to assess a match confidence for a candidate skill, the match confidence indicating a quality of match between the current context and a reference context previously associated with the candidate skill. The computerized personal assistant executes instructions defining an assistive action associated with the candidate skill responsive to the match confidence exceeding a predefined match confidence threshold. The computerized personal assistant executes the instructions defining a complementary help action associated with the candidate skill responsive to the match confidence not exceeding the predefined match confidence threshold.

First claim

The invention claimed is:

1. A method of facilitating the discovery of a skill of a computerized personal assistant, comprising:
receiving, from a user, a query via a natural language interface, the received query being received corresponding to a current context of the user;
upon receiving the query, for a first candidate skill of a plurality of candidate skills, determining a first match confidence between the current context corresponding to the received query and a reference context of a reference query associated with the first candidate skill, wherein the first match confidence for the first candidate skill is determined to be higher than a second match confidence of a second candidate skill of the plurality of candidate skills based, at least in part, on the first candidate skill having a lower utilization frequency than the second candidate skill of the plurality of candidate skills;
comparing the determined first match confidence of the first candidate skill to a predefined match confidence threshold; and
executing an assistive action associated with the first candidate skill based on determining that the first match confidence of the first candidate skill exceeds the predefined match confidence threshold.

2. The method of claim 1 , wherein the first candidate skill is associated with a complementary help action defined by instructions executable by the computerized personal assistant to output a predefined response to educate a user about the first candidate skill.

3. The method of claim 2 , further comprising:
executing the instructions defining the complementary help action responsive to the match confidence of the first candidate skill not exceeding the predefined match confidence threshold; and
outputting the predefined response.

4. The method of claim 1 , further comprising operating a previously-trained machine learning classifier to determine the first match confidence for the first candidate skill of the plurality of candidate skills, the first match confidence indicating a quality of match between the current context of the current query and the reference context of the reference query of the first candidate skill.

5. A method of facilitating discovery of a skill of a computerized personal assistant, comprising:
determining a plurality of candidate skills for the computerized personal assistant, wherein
at least one candidate skill of the plurality of candidate skills is associated with:
an assistive action that performs an operation to assist a user;
a reference context previously associated with the assistive action; and
a complementary help action associated with the assistive action, wherein the complementary help action outputs a predefined response to educate the user about the assistive action;

recognizing a current context associated with the user;
determining, for each available skill of the plurality of candidate skills, a match confidence indicating a quality of a match between the current context and the reference context of the candidate skill;
ranking the plurality of available skills by candidate match confidence, wherein the at least one candidate skill is an available skill having a highest candidate match confidence; and
causing execution of the complementary help action responsive to the match confidence not exceeding a predefined match confidence threshold.

6. The method of claim 5 , wherein:
the reference context includes a seed query defined by an exemplary natural language sentence;
the operation to assist the user pertains to the seed query; and
the current context of the user includes a user query received via a natural language user interface.

7. The method of claim 5 , wherein:
the reference context includes a historical activity context including state data of a computer service, the state data associated with a recognized activity of a historical user of the computer service;
the operation to assist the user pertains to the recognized activity; and
the current context of the user includes a current activity context including state data of the computer service, the state data indicating the user is using the computer service to engage in the recognized activity.

8. The method of claim 5 , wherein the instructions defining the complementary help action of the candidate skill are executed proactively, responsive to assessing the match confidence for the candidate skill.

9. The method of claim 5 , wherein:
the current context of the user includes a user preference; and
assessing the match confidence includes assessing a higher match confidence for the candidate skill based on the assistive action satisfying the user preference.

10. The method of claim 5 , wherein:
the current context of the user includes, for each available skill of the plurality of available skills, a utilization frequency indicating a degree of past utilization of the available skill by the user; and
assessing the match confidence includes assessing a higher match confidence for the candidate skill based on the candidate skill having a relatively lower utilization frequency.

11. The method of claim 5 , further comprising:
operating the previously-trained machine learning classifier to assess, for each available skill of the plurality of available skills, a candidate match confidence for the available skill;
wherein the candidate skill is an available skill having a highest candidate match confidence.

12. The method of claim 5 , wherein the computer database further includes an unavailable skill, the method further comprising:
operating the previously-trained machine learning classifier to assess an unavailable skill match confidence indicating a quality of match between the current context and a reference context previously associated with the unavailable skill; and
responsive to the unavailable skill match confidence exceeding the predefined match confidence threshold, outputting a response describing the unavailable skill and educating the user about configuring the computerized personal assistant to make the unavailable skill available.

13. The method of claim 5 , wherein outputting the predefined response to educate the user about the assistive action includes outputting, via a natural language user interface, a natural language description of the assistive action.

14. The method of claim 5 , wherein the predefined response to educate the user about the assistive action indicates an exemplary query responsive to which the computerized personal assistant is programmed to perform the assistive action.

15. The method of claim 5 , wherein the complementary help action further defines a disambiguating question, the method further comprising:
outputting, via a natural language user interface (NLUI), the disambiguating question;
receiving, via the NLUI, an answer to the disambiguating question; and
responsive to receiving the answer, operating the computerized personal assistant to execute the instructions defining the assistive action of the candidate skill.

16. The method of claim 5 , further comprising:
receiving a training example context;
for an available skill of the plurality of available skills:
operating the previously-trained machine learning classifier to assess a training confidence indicating a quality of match between the training example context and a first training reference context previously associated with the available skill, wherein the training confidence does not exceed the predefined match confidence threshold;
receiving, from a trainer, a second training reference context; and
updating the computer database to a

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Vangala Vipindeep
• Valecha Swati
• Shastri Suryanarayana
• Pande Nitin
• Menon Tulasi
• Jhanwar Madan Gopal
• Kumar Nishchay

Classifications

• G10L15/22 Primary

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title

• G06F16/24578

  using ranking · CPC title

• G06F3/167

  Audio in a user interface, e.g. using voice commands for navigating, audio feedback · CPC title

• G06F9/453 Primary

  Help systems · CPC title

• G06F16/243

  Natural language query formulation · CPC title