Electronic device for processing user's inquiry, and operation method of the electronic device

Patent metadata

Field	Value
Publication number	US-12277390-B2
Application number	US-202217677396-A
Country	US
Kind code	B2
Filing date	Feb 22, 2022
Priority date	Feb 2, 2021
Publication date	Apr 15, 2025
Grant date	Apr 15, 2025

Abstract

An electronic device for providing a corrected response message according to an utterance intention of a user by using a keyword included in an input inquiry input by the user, and an operation method of the electronic device are provided. The electronic device includes receiving the input inquiry input by the user, identifying a representative inquiry according to an utterance intention of the user by analyzing the input inquiry by using a natural language understanding (NLU) model, extracting a keyword from the input inquiry by comparing a vector value of a first embedding vector of the input inquiry changed through the NLU model with a vector value of a second embedding vector of the representative inquiry, and correcting a response message mapped to correspond to the representative inquiry, by using the extracted keyword.

First claim

The invention claimed is:

1. A method, performed by an electronic device, of processing an inquiry of a user, the method comprising:
receiving an input inquiry input by the user;
obtaining a label value representing an intent of the input inquiry by analyzing the input inquiry by using a natural language understanding (NLU) model;
identifying a representative inquiry mapped to correspond to the label value from among a plurality of representative inquiries that are stored in a response message database;
comparing a vector value of a first embedding vector of the input inquiry changed during the identifying of the representative inquiry with a vector value of a second embedding vector of the representative inquiry;
identifying a keyword from the input inquiry, based on a result of the comparing; and
correcting a response message mapped to correspond to the representative inquiry, by using the identified keyword,
wherein the response message database stores the plurality of representative inquiries for a plurality of intents previously obtained before the input inquiry is received from the user, and
wherein each of the plurality of representative inquiries is mapped to correspond to respective label values for each of the plurality of intents.

2. The method of claim 1 , wherein the identifying of the keyword from the input inquiry comprises identifying the keyword from the input inquiry, based on a position relationship in a virtual vector space according to the vector value of the first embedding vector for one or more words included in the input inquiry changed by the NLU model and the vector value of the second embedding vector.

3. The method of claim 2 , wherein the identifying of the keyword from the input inquiry comprises:
calculating a cosine similarity between the first embedding vector and the second embedding vector; and
identifying the keyword, based on a position proximity of an embedding vector in the virtual vector space and the calculated cosine similarity.

4. The method of claim 2 , wherein the identifying of the keyword from the input inquiry comprises identifying the keyword from the input inquiry, based on a distance change rate in the virtual vector space between the first embedding vector changed by the NLU model and the second embedding vector.

5. The method of claim 1 , wherein the correcting of the response message comprises correcting the response message by replacing or changing a word included in the response message by using the identified keyword or adding the identified keyword to the response message.

6. The method of claim 1 , wherein the correcting of the response message comprises:
generating an answer start message comprising at least one of additional information and an additional description related to the identified keyword; and
adding the answer start message to the response message.

7. The method of claim 1 , further comprising:
identifying the response message mapped to correspond to the label value representing the intent of the input inquiry.

8. An electronic device for processing an inquiry of a user, the electronic device comprising:
a communication interface configured to perform data transmission/reception with another device;
a memory storing a program comprising one or more instructions; and
a processor configured to execute the one or more instructions of the program stored in the memory,
wherein the processor is further configured to:
receive an input inquiry input by the user through the communication interface,
obtain a label value representing an intent of the input inquiry by analyzing the input inquiry by using a natural language understanding (NLU) model,
identify a representative inquiry mapped to correspond to the label value from among a plurality of representative inquiries that are stored in a response message database,
compare a vector value of a first embedding vector of the input inquiry changed during the identifying of the representative inquiry with a vector value of a second embedding vector of the representative inquiry,
identify a keyword from the input inquiry, based on a result of the comparing, and
correct a response message mapped to correspond to the representative inquiry, by using the identified keyword,
wherein the response message database stores the plurality of representative inquiries for a plurality of intents previously obtained before the input inquiry is received from the user, and
wherein each of the plurality of representative inquiries is mapped to correspond to respective label values for each of the plurality of intents.

9. The electronic device of claim 8 , wherein the processor is further configured to identify the keyword from the input inquiry, based on a position relationship in a virtual vector space according to the vector value of a first embedding vector for one or more words included in the input inquiry changed by the NLU model and a vector value of the second embedding vector.

10. The electronic device of claim 9 , wherein the processor is further configured to:
calculate a cosine similarity between the first and second embedding vectors, and
identify the keyword, based on a position proximity of the embedding vectors in the virtual vector space and the calculated cosine similarity.

11. The electronic device of claim 9 , wherein the processor is further configured to identify the keyword from the input inquiry, based on a distance change rate in the virtual vector space between the first embedding vector changed by the NLU model and the second embedding vector.

12. The electronic device of claim 8 , wherein the processor is further configured to correct the response message by replacing or changing a word included in the response message by using the identified keyword or adding the identified keyword to the response message.

13. The electronic device of claim 8 , wherein the processor is further configured to generate an answer start message comprising at least one of additional information and an additional description related to the identified keyword and add the answer start message to the response message.

14. The electronic device of claim 8 , wherein the processor is further configured to:
identify the response message mapped to correspond to the label value representing the intent of the input inquiry.

15. A non-transitory computer program product comprising a computer-readable storage medium, the computer-readable storage medium comprising instructions executed by a device to perform:
receiving an input inquiry input by a user;
obtaining a label value representing an intent of the input inquiry by analyzing the input inquiry by using a natural language understanding (NLU) model;
identifying a representative inquiry mapped to correspond to the label value from among a plurality of representative inquiries that are stored in a response message database;
comparing a vector value of a first embedding vector of the input inquiry changed during the identifying of the representative inquiry with a vector value of a second embedding vector of the representative inquiry;
identifying a keyword from the input inquiry, based on a result of the comparing; and
correcting a response message mapped to correspond to the representative inquiry, by using the identified keyword,
wherein the response message database stores the plurality of representative inquiries for a plurality of intents previously obtained before the input inquiry is received from the user, and
wherein each of the plurality of representative inquiries is mapped to correspond to respective label values for each

Assignees

• Samsung Electronics Co Ltd

Inventors

• Yang Dongil
• Kim Hoyoung
• An Jaehyung
• Lee Jiyeon
• Jung Cheolseung
• Choi Wonjong

Classifications

• G06V30/18143

  Extracting features based on salient regional features, e.g. scale invariant feature transform [SIFT] keypoints · CPC title

• G06V30/19013

  Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching (specially adapted for image segmentation G06T7/10; specially adapted for the analysis of motion G06T7/20; specially adapted for image alignment G06T7/30; specially adapted for the calculation of depth from stereo images G06T7/50; specially adapted for position determination G06T7/70) · CPC title

• G06V30/19093

  Proximity measures, i.e. similarity or distance measures · CPC title

• G06F18/24

  Classification techniques · CPC title

• G06F40/205

  Parsing · CPC title