Systems and methods for intent-based natural language processing

Patent metadata

Field	Value
Publication number	US-12499327-B2
Application number	US-202318296304-A
Country	US
Kind code	B2
Filing date	Apr 5, 2023
Priority date	Apr 27, 2021
Publication date	Dec 16, 2025
Grant date	Dec 16, 2025

Abstract

A method for intent-based natural language processing may include a conversation engine receiving text of an utterance in a conversation from a conversation program, storing a session identifier for the conversation and a conversation state for the conversation, selecting one of a plurality a priority of intents based on the text of the utterance, communicating the text of the utterance to a natural language understanding computer program, receiving a plurality of potential intents and a confidence score for each potential intent from the natural language understanding computer program, selecting a selected intent based on the selected priority of intents, the potential intents, and the confidence scores, executing intent logic associated with the selected intent, updating the conversation state in the database based on the executed intent logic, and returning a response to the conversation program based on the executed intent logic.

First claim

What is claimed is:

1 . A method for intent-based natural language processing, comprising:
receiving, at a conversation engine, text of an utterance from a user in a conversation from a conversation program executed on an electronic device; storing, by the conversation engine, a conversation state for the conversation in a database with a unique identifier for the conversation and a channel identifier for the electronic device, wherein the conversation state comprises a stack of processes that are added and subtracted as information is requested or obtained from the user; determining, by the conversation engine, that the text of the utterance does not comprise a question; selecting, by the conversation engine, one of a plurality of priority of intents selected from the group comprising a flow intent, a command intent, and an unexpected answer/last resorts intent in response to the determination; communicating, by the conversation engine, the text of the utterance to a natural language understanding computer program; receiving, by the conversation engine, a plurality of potential intents and a confidence score for each potential intent from the natural language understanding computer program; selecting, by the conversation engine, a selected intent based on the selected priority of intents, the potential intents, and the confidence scores; executing, by the conversation engine, intent logic associated with the selected intent; updating, by the conversation engine, the conversation state in the database based on the executed intent logic; and returning, by the conversation engine, a response to the conversation program based on the executed intent logic.

2 . The method of claim 1 , wherein the electronic device comprises an Internet of Things appliance.

3 . The method of claim 1 , wherein the stack of processes comprises a base process and a sub-process.

4 . The method of claim 1 , wherein the intents in a first of the plurality of priority of intents are ordered as a flow intent, a command intent, and an unexpected answer intent, and the intents in a second of the plurality of priority intents are ordered as an expected answer intent, a command intent, a flow intent, and an unexpected answer intent.

5 . The method of claim 1 , wherein the selected priority of intents and the potential intents are programmatically weighted.

6 . A method for intent-based natural language processing, comprising:
receiving, at a conversation engine, text of an utterance from a user in a conversation from a conversation program executed on an electronic device; storing, by the conversation engine a conversation state for the conversation in a database with a unique identifier for the conversation and a channel identifier for the electronic device, wherein the conversation state comprises a stack of processes that are added and subtracted as information is requested or obtained from the user; determining, by the conversation engine, that the text of the utterance comprises a question; selecting, by the conversation engine, one of a plurality of priority of intents selected from the group comprising an expected answer intent, a command intent, a flow intent, and an unexpected answer/last resort intent; communicating, by the conversation engine, the text of the utterance to a natural language understanding computer program; receiving, by the conversation engine, a plurality of potential intents and a confidence score for each potential intent from the natural language understanding computer program; selecting, by the conversation engine, a selected intent based on the selected priority of intents, the potential intents, and the confidence scores; executing, by the conversation engine, intent logic associated with the selected intent; updating, by the conversation engine, the conversation state in the database based on the executed intent logic; and returning, by the conversation engine, a response to the conversation program based on the executed intent logic.

7 . The method of claim 6 , wherein the electronic device comprises an Internet of Things appliance.

8 . The method of claim 6 , wherein the stack of processes comprises a base process and a sub-process.

9 . The method of claim 6 , wherein the intents in a first of the plurality of priority of intents are ordered as a flow intent, a command intent, and an unexpected answer intent, and the intents in a second of the plurality of priority intents are ordered as an expected answer intent, a command intent, a flow intent, and an unexpected answer intent.

10 . The method of claim 6 , wherein the selected priority of intents and the potential intents are programmatically weighted.

11 . A server electronic device, comprising:
a computer processor; and a memory comprising a conversation engine computer program; wherein the conversation engine computer program is configured to:
receive text of an utterance from a user in a conversation from a conversation program executed on an electronic device;
store a conversation state for the conversation in a database with a unique identifier for the conversation and a channel identifier for the electronic device, wherein the conversation state comprises a stack of processes comprising a base process and a sub-process that are added and subtracted as information is requested or obtained from the user;
determine that the text of the utterance does not comprises a question;
select one of a plurality of priority of intents selected from the group comprising an expected answer intent, a command intent, a flow intent, and an unexpected answer/last resort intent;
communicate the text of the utterance to a natural language understanding computer program;
receive a plurality of potential intents and a confidence score for each potential intent from the natural language understanding computer program;
select a selected intent based on the selected priority of intents, the potential intents, and the confidence scores;
execute intent logic associated with the selected intent;
update the conversation state in the database based on the executed intent logic; and
return a response to the conversation program based on the executed intent logic.

12 . The server electronic device of claim 11 , wherein the electronic device comprises an Internet of Things appliance.

13 . The server electronic device of claim 11 , wherein the intents in a first of the plurality of priority of intents are ordered as a flow intent, a command intent, and an unexpected answer intent, and the intents in a second of the plurality of priority intents are ordered as an expected answer intent, a command intent, a flow intent, and an unexpected answer intent.

14 . The server electronic device of claim 11 , wherein the selected priority of intents and the potential intents are programmatically weighted.

15 . The method of claim 1 , further comprising:
continuing the conversation on a second electronic device using the unique identifier.

16 . The method of claim 6 , further comprising:
continuing the conversation on a second electronic device using the unique identifier.

17 . The server electronic device of claim 11 , wherein the wherein the conversation engine is further configured to continue the conversation on a second electronic device using the unique identifier.

Assignees

• Jpmorgan Chase Bank Na

Inventors

• White Ben
• Stikkel Michael C
• Cook James

Classifications

• G06F40/279

  Recognition of textual entities · CPC title

• G06F40/35 Primary

  Discourse or dialogue representation · CPC title

• G06F16/3329

  Natural language query formulation · CPC title

• G06F40/295

  Named entity recognition · CPC title

• G06F40/44 Primary

  Statistical methods, e.g. probability models · CPC title