Granted Patent B2

US 12,494,193 · App. 18/076,039 · Granted Dec 9, 2025

Inventors: Sung Soo Park (Seoul, KR); Kyowoong Choo (Seoul, KR); Chang Woo Chun (Suwon-si, KR)

Assignees: HYUNDAI MOTOR COMPANY; KIA CORPORATION

G10L15/08G10L15/26G10L2015/088

Abstract

Claims (42)

1 . A method for determining a vehicle domain, the method comprising:

converting a user's speech into text; and

classifying the user's speech into a vehicle domain or an external domain based on the text,

wherein the classifying of the user's speech into the vehicle domain or the external domain comprises classifying a domain of the user's speech based on previously stored keyword-related information and then classifying the domain of the user's speech based on a trained domain classification model,

wherein the trained domain classification model comprises an intent classification model configured to classify an intent of the user's speech, and

wherein the classifying of the domain of the user's speech based on the trained domain classification model comprises:

when the intent classified by the intent classification model is not included in an external intent, extracting a slot value from the user's speech, and

when the extracted slot value is not included in a vehicle-related slot, classifying the user's speech into the external domain.

2 . The method of claim 1 , wherein when the user's speech is not classified into the external domain based on the previously stored keyword-related information, the classifying of the user's speech into the vehicle domain or the external domain comprises classifying the domain of the user's speech based on the trained domain classification model.

3 . The method of claim 2 , wherein the classifying of the user's speech into the vehicle domain or the external domain further comprises adjusting a confidence value of the trained domain classification model.

4 . The method of claim 3 , wherein the adjusting of the confidence value of the trained domain classification model uses a re-trained domain classification model to adjust the confidence value.

5 . The method of claim 3 , wherein when the user's speech is not classified into the external domain based on the trained domain classification model, the classifying of the user's speech into the vehicle domain or the external domain re-adjusts the confidence value of the trained domain classification model.

6 . The method of claim 1 , wherein the previously stored keyword-related information includes a vehicle-related keyword, and

wherein the classifying of the domain of the user's speech based on the previously stored keyword-related information comprises classifying the user's speech into the vehicle domain, when the vehicle-related keyword or a combination of the vehicle-related keywords is included in the user's speech.

7 . The method of claim 1 , wherein the previously stored keyword-related information includes a pattern defined by a vehicle-related keyword and a predicate, and

wherein the classifying of the domain of the user's speech based on the previously stored keyword-related information comprises classifying the user's speech into the vehicle domain, when the defined pattern is included in the user's speech.

8 . The method of claim 1 , wherein the previously stored keyword-related information includes a vehicle-related entity name, and

wherein the classifying of the domain of the user's speech based on the previously stored keyword-related information comprises classifying the user's speech into the vehicle domain, when the vehicle-related entity name is included in the user's speech.

9 . The method of claim 1 , wherein the trained domain classification model comprises an area classification model that classifies a result of sentence embedding of the text, converted from the user's speech, into a vehicle area or another area within an embedding space, and

wherein the classifying of the domain of the user's speech based on the trained domain classification model comprises classifying the user's speech into the external domain, when the result of sentence embedding of the text is not included in the vehicle area.

10 . The method of claim 1 ,

wherein the classifying of the domain of the user's speech based on the trained domain classification model comprises classifying the user's speech into the external domain, when the intent classified by the intent classification model is included in the external intent.

11 . A speech recognition system for a vehicle, comprising:

a speech recognition module configured to convert a user's speech into text; and

a domain classification module configured to classify the user's speech into a vehicle domain or an external domain based on the text,

wherein the domain classification module is configured to classify a domain of the user's speech based on previously stored keyword-related information, and then classify the domain of the user's speech based on a trained domain classification model,

wherein the trained domain classification model comprises an intent classification model configured to classify an intent of the user's speech, and

wherein to classify the domain of the user's speech based on the trained domain classification model, the domain classification module is configured to:

when the intent classified by the intent classification model is not included in an external intent, extract a slot value from the user's speech, and

when the extracted slot value is not included in a vehicle-related slot, classify the user's speech into the external domain.

12 . The speech recognition system for the vehicle of claim 11 , wherein when the user's speech is not classified into the external domain based on the previously stored keyword-related information, the domain classification module is configured to classify the domain of the user's speech based on the trained domain classification model.

13 . The speech recognition system for the vehicle of claim 12 , wherein the domain classification module is configured to adjust a confidence value of the trained domain classification model.

14 . The speech recognition system for the vehicle of claim 13 , wherein the domain classification module uses a re-trained domain classification model to adjust the confidence value.

15 . The speech recognition system for the vehicle of claim 13 , wherein when the user's speech is not classified into the external domain based on the trained domain classification model, the domain classification module is configured to re-adjust the confidence value of the trained domain classification model.

16 . The speech recognition system for the vehicle of claim 11 , wherein the previously stored keyword-related information includes a vehicle-related keyword, and

wherein the domain classification module is configured to classify the user's speech into the vehicle domain, when the vehicle-related keyword or a combination of the vehicle-related keywords is included in the user's speech.

17 . The speech recognition system for the vehicle of claim 11 , wherein the previously stored keyword-related information includes a pattern defined by a vehicle-related keyword and a predicate, and

wherein the domain classification module is configured to classify the user's speech into the vehicle domain, when the defined pattern is included in the user's speech.

18 . The speech recognition system for the vehicle of claim 11 , wherein the previously stored keyword-related information includes a vehicle-related entity name, and

wherein the domain classification module is configured to classify the user's speech into the vehicle domain, when the vehicle-related entity name is included in the user's speech.

19 . The speech recognition system for the vehicle of claim 11 , wherein the trained domain classification model comprises an area classification model that classifies a result of sentence embedding of the text, converted from the user's speech, into a vehicle area or another area within an embedding space, and

wherein the domain classification module is configured to classify the user's speech into the external domain, when the result of sentence embedding of the text is not included in the vehicle area.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (26)