Granted Patent B2

US 12,437,755 · App. 18/132,630 · Granted Oct 7, 2025

Inventor: Ji Hoon Kim (Seoul, KR)

Assignees: Hyundai Motor Company; Kia Corporation

G10L15/1815G10L15/22G10L15/30G10L2015/223

Abstract

Claims (33)

1. A dialogue processing method, which is performed by a dialogue system including one or more processors and a memory configured to store one or more programs executed by the one or more processors, the method comprising:

obtaining an intent classification result for a voice command of a user by use of a trained intent classification model;

obtaining utterance history information of the user;

adjusting the intent classification result based on the obtained utterance history information of the user;

generating a system response based on the adjusted intent classification result; and

transmitting the generated system response to a user terminal,

wherein the utterance history information of the user includes information on a frequency of an intent used by the user through utterance of the voice command,

wherein the intent classification result includes a score of each of predetermined intents, and the adjusting of the intent classification result includes adjusting the score of each of the predetermined intents based on the utterance history information of the user, and

wherein the user terminal outputs to the outside, a content signal or a system utterance corresponding to the system response through at least one of a speaker or a display.

2. The dialogue processing method of claim 1 , wherein the adjusting of the intent classification result further includes adjusting an intent with a higher frequency of use by the user through the voice command to have a higher score, among the predetermined intents.

3. The dialogue processing method of claim 1 , wherein the adjusting of the intent classification result further includes redetermining the score by assigning a high weight to a score of an intent with a high frequency of use by the user through utterance of the voice command, among the predetermined intents.

4. The dialogue processing method of claim 1 , wherein the adjusting of the intent classification result further includes decreasing a score which is less than a threshold and increasing a score of an intent with a high frequency of use by the user through utterance of the voice command, among the scores of the predetermined intents.

5. The dialogue processing method of claim 4 , wherein an amount of the decreasing in the score which is less than the threshold and an amount of the increasing in the score of the intent with the high frequency of use by the user through utterance of the voice command are proportional to each other.

6. The dialogue processing method of claim 1 , wherein the utterance history information of the user further includes at least one of information on an utterance time of the user's voice command or an utterance location of the user's voice command.

7. The dialogue processing method of claim 6 , wherein the adjusting of the intent classification result further includes redetermining the scores by assigning a high weight to a score of an intent with a high frequency of use at the utterance time or in the utterance location matching a current utterance time or a current utterance location, among the predetermined intents.

8. The dialogue processing method of claim 1 , wherein the adjusting of the intent classification result further includes determining an intent including a highest score among the adjusted scores to be a final intent for the voice command of the user, and performing slot tagging on the voice command of the user based on the final intent.

9. A computer-readable non-transitory recording medium storing a program for executing a dialogue processing method is recorded, the dialogue processing method comprising:

obtaining an intent classification result for a voice command of a user using a trained intent classification model;

obtaining utterance history information of the user;

adjusting the intent classification result based on the obtained utterance history information of the user;

generating a system response based on the adjusted intent classification result; and

transmitting the generated system response to a user terminal;

wherein the utterance history information of the user includes information on a frequency of an intent used by the user through utterance of the voice command,

wherein the intent classification result includes a score of each of predetermined intents, and

the adjusting of the intent classification result includes adjusting the score of each of the predetermined intents based on the utterance history information of the user,

wherein the user terminal outputs to the outside, a content signal or a system utterance corresponding to the system response through at least one of a speaker or a display.

10. The computer-readable non-transitory recording medium of claim 9 , wherein the adjusting of the intent classification result further includes adjusting an intent with a higher frequency of use by the user through the voice command to have a higher score, among the predetermined intents.

11. The computer-readable non-transitory recording medium of claim 9 , wherein the adjusting of the intent classification result further includes redetermining the score by assigning a high weight to a score of an intent with a high frequency of use by the user through utterance of the voice command, among the predetermined intents.

12. The computer-readable non-transitory recording medium of claim 9 , wherein the adjusting of the intent classification result further includes decreasing a score which is less than a threshold and increasing a score of an intent with a high frequency of use by the user through utterance of the voice command, among the scores of the predetermined intents.

13. The computer-readable non-transitory recording medium of claim 12 , wherein an amount of the decreasing in the score which is less than the threshold and an amount of the increasing in the score of the intent with the high frequency of use by the user through utterance of the voice command are proportional to each other.

14. The computer-readable non-transitory recording medium of claim 9 , wherein the utterance history information of the user further includes at least one of information on an utterance time of the user's voice command or an utterance location of the user's voice command.

15. The computer-readable non-transitory recording medium of claim 14 , wherein the adjusting of the intent classification result further includes redetermining the scores by assigning a high weight to a score of an intent with a high frequency of use at the utterance time or in the utterance location matching a current utterance time or a current utterance location, among the predetermined intents.

16. The computer-readable non-transitory recording medium of claim 9 , wherein the adjusting of the intent classification result further includes determining an intent including a highest score among the adjusted scores to be a final intent for the voice command of the user, and performing slot tagging on the voice command of the user based on the final intent.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (11)