IP Library Granted Patent US 12,597,419
Granted Patent B2
US 12,597,419 · App. 18/083,119 · Granted Apr 7, 2026

Natural language processing apparatus and natural language processing method

Inventor: Chang Woo Chun (Suwon-si, KR)
Assignees: HYUNDAI MOTOR COMPANY; KIA CORPORATION
G10L15/1815G06F40/166G06F40/30G10L15/22G10L2015/223
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,597,419
App. No.
18/083,119
Granted
Apr 7, 2026
Kind
B2
Abstract

The present disclosure provides a natural language processing apparatus and a natural language processing method that may determine whether a user's speech command is a compound sentence or a complex sentence based on an output of a natural language understanding module, and when the user's speech command is a compound sentence or a complex sentence, recursively call the natural language understanding module, thereby executing all of the plurality of functions expressed as a single sentence.

Claims (49)

1 . A natural language processing apparatus comprising:

at least one memory storing computer-executable instructions;

a processor coupled with the at least one memory, wherein the processor is configured to execute the computer-executable instructions;

a speech recognition module executed by the processor, the speech recognition module configured to convert a speech command of a user to a text;

a natural language understanding module executed by the processor, the natural language understanding module configured to classify an intent corresponding to the speech command and extract a slot included in the speech command based on the text; and

a control module executed by the processor, the control module configured to determine whether the speech command includes a compound sentence or a complex sentence based on the intent classified by the natural language understanding module, and when the speech command includes the compound sentence or the complex sentence, re-input the text corresponding to the speech command to the natural language understanding module again,

wherein a sentence including a plurality of intents is defined as the compound sentence or the complex sentence, and a sentence including a single intent is defined as a single sentence, and

among the plurality of intents, when two or more intents have different actions, the sentence is determined to be the complex sentence, and when actions of the plurality of intents are the same but slots are different, the sentence is determined to be the compound sentence.

2 . The natural language processing apparatus of claim 1 , wherein the natural language understanding module is configured to calculate a matching probability of the speech command for each of a plurality of pre-defined intents, to classify the intent corresponding to the speech command.

3 . The natural language processing apparatus of claim 2 , wherein the control module is configured to determine that the speech command includes the compound sentence or the complex sentence when a highest matching probability among the calculated matching probabilities is less than a first reference value and two or more matching probabilities are greater than a second reference value.

4 . The natural language processing apparatus of claim 3 , wherein the control module is configured to determine that the speech command includes the compound sentence when a top number of intents have an identical action and different slots based on the calculated matching probabilities, and

determine that the speech command includes the complex sentence when two or more intents among the top number of intents have different actions,

wherein the top number of intents is an integer greater than or equal to two.

5 . The natural language processing apparatus of claim 3 , wherein the natural language understanding module is configured to determine a final intent and a final slot for the re-input text, based on information about a top number of intents based on the calculated matching probabilities.

6 . The natural language processing apparatus of claim 5 , wherein the control module is configured to determine a consistency between the final intent and the final slot.

7 . The natural language processing apparatus of claim 6 , wherein the control module is configured to generate a result processing signal for performing a function corresponding to the final intent whose consistency with the final slot is acknowledged.

8 . The natural language processing apparatus of claim 1 , wherein, when the speech command includes the complex sentence, the control module is configured to divide the text converted from the speech command into a plurality of sentences and input the divided plurality of sentences to the natural language understanding module again.

9 . A natural language processing apparatus comprising:

at least one memory storing computer-executable instructions;

a processor coupled with the at least one memory, wherein the processor is configured to execute the computer-executable instructions;

a speech recognition module executed by the processor, the speech recognition module configured to convert a speech command of a user to a text;

a natural language understanding module executed by the processor, the natural language understanding module configured to classify an intent corresponding to the speech command and extract a slot included in the speech command based on the text; and

a control module executed by the processor, the control module configured to re-input a functional speech corresponding to the speech command, the functional speech indicating a function to be registered that matches a personal command to the natural language understanding module again, when the intent classified by the natural language understanding module is registration of the personal command,

wherein the natural language understanding module is re-invoked to perform intent classification and slot extraction for the functional speech,

wherein the control module is configured to determine whether the functional speech is a single sentence, a compound sentence, or a complex sentence,

wherein a sentence including a plurality of intents is defined as the compound sentence or the complex sentence, and a sentence including a single intent is defined as a single sentence, and

among the plurality of intents, when two or more intents have different actions, the sentence is determined to be the complex sentence, and when actions of the plurality of intents are the same but slots are different, the sentence is determined to be the compound sentence.

10 . The natural language processing apparatus of claim 9 , wherein the natural language understanding module is configured to classify an intent corresponding to the re-input functional speech and extract a slot included in the re-input functional speech.

11 . The natural language processing apparatus of claim 10 , wherein the control module is configured to store another function defined by the intent corresponding to the functional speech and the slot included in the functional speech, as the function corresponding to the personal command.

12 . The natural language processing apparatus of claim 11 , wherein the control module is configured to generate a control signal for performing the function corresponding to the personal command, when a speech command of the user input after the personal command is registered includes the registered personal command.

13 . A natural language processing method comprising:

converting, by a speech recognition module executed by a processor, a speech command of a user to a text;

classifying, by a natural language understanding module executed by the processor, an intent corresponding to the speech command and extracting a slot included in the speech command based on the text;

determining, by a control module executed by the processor, whether the speech command includes a compound sentence or a complex sentence based on the classified intent; and

when the speech command includes the compound sentence or the complex sentence, re-inputting, by the control module, the text corresponding to the speech command to the natural language understanding module again,

wherein a sentence including a plurality of intents is defined as the compound sentence or the complex sentence, and a sentence including a single intent is defined as a single sentence, and

among the plurality of intents, when two or more intents have different actions, the sentence is determined to be the complex sentence, and when actions of the plurality of intents are the same but slots are different, the sentence is determined to be the compound sentence.

14 . The natural language processing method of claim 13 , wherein the determining of whether the speech command includes the compound sentence or the complex sentence comprises calculating a matching probability of the speech command for each of a plurality of pre-defined intents, to classify the intent corresponding to the speech command.

15 . The natural language processing method of claim 14 , wherein the determining of whether the speech command includes the compound sentence or the complex sentence comprises determining that the speech command includes the compound sentence or the complex sentence, when a highest matching probability among the calculated matching probabilities is less than a first reference value and two or more matching probabilities are greater than a second reference value.

16 . The natural language processing method of claim 15 , wherein the determining of whether the speech command includes the compound sentence or the complex sentence comprises determining that the speech command includes the compound sentence, when a top number of intents have an identical action and different slots based on the calculated matching probabilities, and

determining that the speech command includes the complex sentence when two or more intents among the top number of intents have different actions,

wherein the top number of intents is an integer greater than or equal to two.

17 . The natural language processing method of claim 15 , further comprising:

determining a final intent and a final slot for the re-input text, based on information about a top number of intents based on the calculated matching probabilities.

18 . The natural language processing method of claim 17 , further comprising:

determining a consistency between the final intent and the final slot.

19 . The natural language processing method of claim 18 , further comprising:

generating a result processing signal for performing a function corresponding to the final intent whose consistency with the final slot is acknowledged.

20 . The natural language processing method of claim 13 , wherein the re-inputting of the text corresponding to the speech command to the natural language understanding module again comprises, when the speech command includes the complex sentence, dividing the text converted from the speech command into a plurality of sentences, and inputting the divided plurality of sentences to the natural language understanding module again.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Dec 16, 2022
From: CHUN, CHANG WOO
To: HYUNDAI MOTOR COMPANY; KIA CORPORATION
Reel/Frame 062131/0096 →
Priority Claims (1)
KR 10-2022-0023903 · Feb 23, 2022 · national
Continuity (1)
Related Publication 20230267923A1 · Aug 24, 2023
References Cited (12)
US 11735207B1 · Venkataraman · 2023 [cited by examiner]
US 20080256071A1 · Prasad · 2008 [cited by examiner]
US 20150370787A1 · Akbacak · 2015 [cited by examiner]
US 20170069315A1 · Chung · 2017 [cited by examiner]
US 20180374482A1 · Woo · 2018 [cited by examiner]
US 20190051288A1 · Lee et al. · 2019 [cited by applicant]
US 20200342874A1 · Teserra · 2020 [cited by examiner]
US 20210173862A1 · Chen · 2021 [cited by examiner]
US 20210217408A1 · Hakkani-Tur · 2021 [cited by examiner]
US 20210360109A1 · Rico Ródenas · 2021 [cited by examiner]
US 20220246141A1 · Mirrezaei · 2022 [cited by examiner]
KR 20190018282A · 2019 [cited by applicant]