Identifying relational segments

Patent metadata

Field	Value
Publication number	US-11822888-B2
Application number	US-201916546941-A
Country	US
Kind code	B2
Filing date	Aug 21, 2019
Priority date	Oct 5, 2018
Publication date	Nov 21, 2023
Grant date	Nov 21, 2023

Abstract

Features, libraries, and techniques are provided herein for determining the kinds of relational language that are present. Applying audio, emojis, and sentiment shifts as features may be used to determine whether the customer is providing backstory, whether there is ranting, etc. Textual features may be considered, as well as audio features may be considered.

First claim

What is claimed:

1. A system comprising:
a computing device at a service provider configured to:
cause display of a virtual assistant on or adjacent to a website of the service provider,
receive input, comprising a query, from a user device of a user through the virtual assistant while the user is engaged in a session on the website of the service provider,

wherein the input comprises a plurality of segments,
detect relational language in a first subset of the plurality of segments and not in a second subset of the plurality of segments by:
segmenting the input based on punctuation symbols and conjunctions,
generating a set of hypothesis pairs based on the segmented inputs,
generating a respective confidence score for each hypothesis pair using an intent classifier, and
determining that (i) at least a first confidence score associated with the first subset of the plurality of segments does not meet or exceed a confidence threshold corresponding with an intent and (ii) at least a second confidence score associated with the second subset of the plurality of segments meets or exceeds the confidence threshold;

determine, based at least on the detected relational language, that a plurality of intents are present in the input,
determine a primary intent of the input using segments only from the second subset and a secondary intent of the input using segments from the first subset responsive to the determining that the plurality of intents are present in the input;
classify the secondary intent into at least one category using the relational language;
generate a response to the query based on the primary intent, the secondary intent and the at least one category using the relational language, and provide the response to the virtual assistant, wherein the response includes content and at least one action; and
the virtual assistant configured to interact with the user and provide to the user the content via the user device and to perform the at least one action, wherein the at least one action comprises the virtual assistant navigating the user device to a webpage that is responsive to the query.

2. The system of claim 1 , wherein the primary intent is a task and the secondary intent is a sentiment.

3. The system of claim 1 , wherein determining the secondary intent of the input uses at least one of audio, an emoji, or sentiment shift.

4. The system of claim 1 , wherein the input comprises audio features.

5. The system of claim 4 , wherein the computing device at the service provider is configured to extract pitch, energy, and frequency features from the input to detect emotions in the audio features.

6. The system of claim 1 , wherein the input comprises textual features.

7. The system of claim 1 , wherein the query comprises a fixed word or phrase.

8. A method comprising:
causing display of a virtual assistant on or adjacent to a website of a service provider;
receiving input, comprising a query, at a computing device of the service provider from a user device of a user through the virtual assistant while the user is engaged in a session on the website of the service provider, wherein the input comprises a plurality of segments;
detecting relational language in a first subset of the plurality of segments and not in a second subset of the plurality of segments by:
segmenting the input based on punctuation symbols and conjunctions,
generating a set of hypothesis pairs based on the segmented inputs,
generating a respective confidence score for each hypothesis pair using an intent classifier, and
determining that (i) at least a first confidence score associated with the first subset of the plurality of segments does not meet or exceed a confidence threshold corresponding with an intent, and (ii) at least a second confidence score associated with the second subset of the plurality of segments meets or exceeds the confidence threshold;

determining, based at least on the detected relational language, that a plurality of intents are present in the input;
determining a primary intent of the input using segments only from the second subset and a secondary intent of the input using segments from the first subset responsive to the determining that the plurality of intents are present in the input;
classifying the secondary intent into at least one category using the relational language;
generating a response to the query based on the primary intent, the secondary intent, and the at least one category using the relational language, and providing the response to the virtual assistant, wherein the response includes content and at least one action; and
providing to the user, from the virtual assistant, the response to the query via the user device, and performing the at least one action, wherein the at least one action comprises the virtual assistant navigating the user device to a webpage that is responsive to the query.

9. The method of claim 8 , further comprising taking action based on the secondary intent.

10. The method of claim 8 , wherein determining the secondary intent of the input uses at least one of audio, an emoji, or sentiment shift.

11. The method of claim 8 , wherein the input comprises audio features.

12. The method of claim 11 , further comprising extracting pitch, energy, and frequency features from the input to detect emotions in the audio features.

13. The method of claim 8 , wherein the input comprises textual features.

14. The method of claim 8 , wherein the query comprises a fixed word or phrase.

15. A non-transitory computer-readable medium storing instructions that when executed by at least one computing device of a system, cause the system to:
cause display of a virtual assistant on or adjacent to a website of a service provider,
receive input, comprising a query, from a user device of a user through the virtual assistant while the user is engaged in a session on the website of the service provider, wherein the input comprises a plurality of segments,
detect relational language in a first subset of the plurality of segments and not in a second subset of the plurality of segments by:
segmenting the input based on punctuation symbols and conjunctions,
generating a set of hypothesis pairs based on the segmented inputs,
generating a respective confidence score for each hypothesis pair using an intent classifier, and
determining that (i) at least a first confidence score associated with the first subset of the plurality of segments does not meet or exceed a confidence threshold corresponding with an intent, and (ii) at least a second confidence score associated with the second subset of the plurality of segments meets or exceeds the confidence threshold;

determine, based at least on the detected relational language, that a plurality of intents are present in the input,
determine a primary intent of the input using segments only from the second subset and a secondary intent of the input using segments from the first subset responsive to the determining that the plurality of intents are present in the input,
classify the secondary intent into at least one category using the relational language, and
generate a response to the query based on the primary intent, the secondary intent and the at least one category using the relational language, and provide the response to the virtual assistant, wherein the response includes content and at least one action; and
the virtual assistant configured to interact with the user and provide to the user the content via the user device and to perform the at least one action, wherein the at least one action comprises the virtual assistant

Assignees

• Verint Americas Inc

Inventors

• Beaver Ian
• Freeman Cynthia
• Pham Andrew T

Classifications

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06F3/0482

  Interaction with lists of selectable items, e.g. menus · CPC title

• G06F3/04817

  using icons (graphical or visual programming using iconic symbols G06F8/34) · CPC title

• G06F40/274

  Converting codes to words; Guess-ahead of partial word inputs · CPC title

• G10L15/22 Primary

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