Speech recognition device and method therefor

Abstract

In embodiments, a speech recognition device and method thereof includes a microphone that extracts a user's utterance, a memory that stores a scenario-adapted intent extracted from the utterance, and a processor that searches for the scenario based on the utterance and performs a speech recognition function. The processor may extract a first intent from a first utterance and extract a second intent from a second utterance. The processor may separate the first intent and the second intent into partial intent units using separators and generate a final intent by combining partial intents of the first intent and the second intent such that duplicate partial intents are deleted depending on definitions of the separators.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to a speech recognition apparatus and a method therefor. BACKGROUND

[0002] A speech recognition device refers to a system capable of identifying a user intent contained in a user's utterance and providing a service corresponding to the identified user intent. The speech recognition device controls the corresponding device in conjunction with a specific device depending on the user intent and provides specific information depending on the user intent.

[0003] Voice recognition devices are widely used due to the convenience of user input. For example, with the development of the automotive industry, a system that automatically performs an action that satisfies user intent using voice recognition in a vehicle is continuously being developed to provide convenience to a driver. Services that use voice recognition are provided in various devices such as a smartphone and an artificial intelligence (AI) speaker, as well as a vehicle. For example, various service providers such as Apple's Siri, Amazon's Alexa, KT's Genie, SK's NUGU, and the like provide voice recognition services.

[0004] The speech recognition device can improve speech recognition by using AI. However, when a user's utterance is implied, it may be difficult for the speech recognition device to clearly determine the intent corresponding to the utterance.

[0005] To recognize a wide range of utterances, it is necessary to define scenarios that match as many utterances as possible. In this case, the number of cases is too large, thus increasing storage capacity. SUMMARY

[0006] The present disclosure relates to a speech recognition apparatus and method thereof, and more particularly to a technology for unambiguously determining the user's intention.

[0007] The present disclosure has been made to solve the above-mentioned problems encountered in the prior art while keeping intact the advantages achieved by the prior art.

[0008] An embodiment of the present disclosure provides a speech recognition device capable of uniquely determining a user utterance and a method embodiment thereof.

[0009] An embodiment of the present disclosure provides a speech recognition apparatus capable of performing speech recognition based on various user utterances while reducing the memory capacity, and a method embodiment thereof.

[0010] Technical problems to be solved by an embodiment of the present disclosure are not necessarily limited to the problems mentioned above, and any other technical problems not mentioned here can be clearly understood by those skilled in the art to which the present disclosure relates from the following description.

[0011] According to one embodiment of the present disclosure, a speech recognition device includes a microphone that extracts a user's utterance, a memory that stores a scenario-adapted intent extracted from the utterance, and a processor that searches for the scenario based on the utterance and performs a speech recognition function. The processor can extract a first intent from a first utterance and a second intent from a second utterance, separate the first intent and the second intent into partial intent units using predetermined separators, and generate a final intent by combining partial intents of each of the first intent and the second intent such that duplicate partial intents are deleted depending on definitions of the separators.

[0012] According to one embodiment, the processor may generate the final intent based on the first utterance and the second utterance if an action, a goal, an entity, or any combination thereof is missing in the second utterance determined after the first utterance.

[0013] According to one embodiment, the processor may assign a separator among the separators of an action, a target, and entities included in each of the first intent and the second intent.

[0014] According to one embodiment, the processor may assign a first separator to a first sub-intent pointing to the action or the goal among first sub-intent extracted from the first intent, and may assign a second separator to a second sub-intent pointing to the action or the goal opposite to the sub-intent to which the first separator is assigned, among second sub-intent extracted from the second intent.

[0015] According to one embodiment, the processor may generate the final intent by deleting the first sub-intent and the second sub-intent that are equal to each other if the first sub-intent to which the first separator is assigned is the same as the second sub-intent to which the second separator is assigned.

[0016] According to one embodiment, the processor may assign a third separator to an entity among the first sub-intent and may assign a fourth separator to a sub-intent that negates the entity to which the third separator is assigned among the second sub-intent.

[0017] According to one embodiment, the processor may generate the final intent by deleting the entities that are equal to each other if the entity to which the third separator is assigned is the same as the entity to which the fourth separator is assigned.

[0018] According to one embodiment, the processor may generate the final intent by arranging the first intent and the second intent in a reverse order of utterances.

[0019] According to one embodiment, the processor may generate the final intent by deleting actions other than the most previous action among the sub-intent(s).

[0020] According to one embodiment, the processor may generate the final intent by deleting sub-intent(s) that match the deleted sub-intent(s) among the sub-intent(s).

[0021] According to an embodiment of the present disclosure, a speech recognition method includes extracting a first intent from a first utterance and extracting a second intent from a second utterance, separating the first intent and the second intent into partial intent units using predetermined separators, generating a final intent by combining partial intents of each of the first intent and the second intent such that duplicate partial intents are deleted depending on definitions of the separators, and performing a speech recognition function based on a scenario consistent with the final intent.

[0022] According to one embodiment, extracting the second intent from the second utterance may be performed after extracting the first intent.

[0023] According to one embodiment, separating the first intent and the second intent into the sub-intent units may include assigning a separator among the separators to an action, a target, and entities included in each of the first intent and the second intent.

[0024] According to one embodiment, separating the first intent and the second intent into the sub-intent units may include assigning a first separator to a first sub-intent pointing to the action or the goal among first sub-intent extracts from the first intent, and assigning a second separator to a second sub-intent pointing to the action or the goal opposite to the sub-intent to which the first separator is assigned among second sub-intent extracts from the second intent.

[0025] According to one embodiment, generating the final intent may include deleting the first sub-intent and the second sub-intent that are equal to each other if the first sub-intent to which the first separator is assigned is the same as the second sub-intent to which the second separator is assigned.

[0026] According to one embodiment, separating the first intent and the second intent into the sub-intent units may comprise assigning a third separator to an entity among the first sub-intent units and assigning a fourth separator to the sub-intent that negates the entity to which the third separator is assigned among the second sub-intent units.

[0027] According to one embodiment, generating the final intent may include deleting the entities that are equal to each other if the entity to which the third separator is assigned is the same as the entity to which the fourth separator is assigned.

[0028] According to one embodiment, generating the final intent may further comprise arranging the first intent and the second intent in a reverse order of utterances.

[0029] According to one embodiment, generating the final intent may further comprise deleting actions other than the most previous action among the sub-intent(s).

[0030] According to one embodiment, generating the final intent may further comprise deleting sub-intent(s) that match the deleted sub-intent(s) among the sub-intent(s). BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The above and other features and advantages of the present disclosure will become more apparent from the following detailed description taken in conjunction with the accompanying drawings in which: Fig. 1 is a block diagram illustrating a configuration of a speech recognition apparatus according to an embodiment of the present disclosure; Fig. 2 is a diagram illustrating an internal configuration of a vehicle equipped with a voice recognition device according to an embodiment of the present disclosure; Fig. 3 is a flowchart illustrating a speech recognition method according to an embodiment of the present disclosure; Fig. 4 is a diagram for describing an intent extraction process according to an embodiment of the present disclosure; Fig. 5 is a diagram for describing a rearrangement process of a user's intent according to an embodiment of the present disclosure; Fig. 6 is a diagram for describing a method for generating a final intent according to an embodiment of the present disclosure; Fig. 7 is a diagram for describing a method for generating a final intent according to an embodiment of the present disclosure; Fig. 8 is a diagram for describing a method for generating a final intent according to an embodiment of the present disclosure; and Fig. 9 is a block diagram illustrating a computing system according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF ILLUSTRATIVE EMBODIMENTS

[0032] Hereinafter, some embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. When attaching reference numerals to components of each drawing, it should be noted that the same components have the same reference numerals even though they are indicated in a different drawing. Furthermore, in describing the embodiments of the present disclosure, detailed descriptions associated with well-known functions or configurations will be omitted when they may unnecessarily obscure matters of the present disclosure.

[0033] When describing elements of an embodiment of the present disclosure, the terms "first," "second," "A," "B," "(a)," "(b)," and the like may be used herein. These terms may be used merely to distinguish one element from another, but do not necessarily limit the corresponding elements regardless of the type, order, or priority of the corresponding elements. Furthermore, unless otherwise defined, all terms, including technical and scientific terms, used herein may be interpreted as is customary in the field to which the present disclosure relates.It is to be understood that terms used herein may be interpreted to have a meaning consistent with their meaning in the context of the present disclosure and the relevant field, and are not to be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0034] Below, various embodiments of the present disclosure are described with reference to the Fig. 1 to Fig. 9 described in detail.

[0035] Fig. 1 is a block diagram illustrating a configuration of a speech recognition apparatus according to an embodiment of the present disclosure. Fig. 2 is a diagram illustrating an internal configuration of a vehicle equipped with a voice recognition device according to an embodiment of the present disclosure. An embodiment in which a voice recognition device is mounted on a vehicle will be mainly described, but the voice recognition device can be applied to any system for a voice recognition function.

[0036] With reference to the Fig. 1 and Fig. 2, a speech recognition device according to an embodiment of the present disclosure may include a microphone 110, a processor 120, and a memory 130.

[0037] The microphone 110 can receive a voice signal and can be positioned in a location where it is easy to receive a user's voice in a vehicle. For example, the microphone 110 can be positioned around a ceiling of the vehicle where a room mirror 60 is located, and a location of the microphone 110 may not be limited thereto.

[0038] The processor 120 may perform a speech recognition function based on a user voice obtained through the microphone 110.

[0039] Processor 120 may extract a first intent from a first utterance and may extract a second intent from a second utterance. The first utterance and the second utterance may be utterances obtained from the user through microphone 110. The second utterance may be performed after the first intent is extracted for the first utterance. For example, the second utterance may refer to an utterance obtained after a speech recognition function is performed on the first utterance.

[0040] The processor 120 may separate the first intent and the second intent into sub-intent units using preset / selected / predetermined separators. The sub-intent unit may be a piece of speech or a word. Further, the processor 120 may separate the first intent and the second intent by positioning a separator to precede sub-intent units. Two or more separators may be included, and each of the separators may separate goals and / or actions that are opposite to each other. For example, a first separator may be assigned to a sub-intent indicating an operation for a radio, and a second separator may be assigned to a sub-intent to stop the operation for the radio or radio device. A specific embodiment of this will be described later.

[0041] Processor 120 may generate a final intent by deleting duplicate sub-intent(s) depending on the definition(s) of separators. The duplicate sub-intent(s) may be sub-intent(s) extracted from different intents. For example, one of a pair of duplicate sub-intent(s) may be extracted from the first intent, and the other may be extracted from the second intent. If the duplicate sub-intent(s) match separators that are opposite to each other, processor 120 may delete all duplicate sub-intent(s). For example, if there is a navigation assigned the first separator and a navigation assigned the second separator, the navigation may be deleted from the final intent (because the first and second separators are opposite).

[0042] The processor 120 may perform a speech recognition function based on a scenario consistent with the final intent.

[0043] Additionally, if an action, goal, entity, or any combination thereof is missing from the second utterance, the processor 120 may generate the final intent based on the first utterance and the second utterance.

[0044] The processor 120 can perform AI training on utterance data obtained by the microphone 110 to perform the speech recognition function. For this purpose, the processor 120 can include an artificial intelligence (hereinafter referred to as "AI") processor. The AI ​​processor can train a neural network using a pre-stored program. The neural network for detecting a target vehicle and a dangerous vehicle can be designed to simulate a human brain structure on a computer and can include multiple network nodes, each of which has a weight and simulates the neurons of a human neural network. The plurality of network nodes can exchange data depending on each connection relationship, so that neurons simulate synaptic activity of neurons that exchange signals through synapses.The neural network may include a deep learning model developed from a neural network model. In the deep learning model, multiple network nodes can exchange data depending on a convolutional connection relationship while residing at different layers. Examples of neural network models may include various deep learning techniques, such as deep neural networks (DNN), convolutional neural networks (CNN), recurrent neural networks (RNN), restricted Boltzmann machines (RBM), deep belief networks (DBN), and deep Q-networks.

[0045] In addition, the processor 120 can control automotive devices 200 mounted in a vehicle through a communication interface 140.

[0046] The memory 130 may store a scenario-adapted intent extracted based on an utterance. Furthermore, the memory 130 may store an algorithm for operating the processor 120 and the AI ​​processor. The memory 130 may use a hard disk drive, flash memory, electrically erasable programmable read-only memory (EEPROM), static random access memory (SRAM), ferroelectric random access memory (FRAM), phase change random access memory (PRAM), magnetic random access memory (MRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double date rate SDRAM (DDR-SDRAM), and the like, or any combination thereof.

[0047] The automotive device 200 may be controlled based on voice recognition and may include an audiovisual navigation (AVN) device mounted on an instrument panel 51. The AVN device may include an audio or navigation device and may include an AVN display 21 that displays an operating status while receiving user input in addition to voice recognition.

[0048] A cluster display 22 that displays vehicle status or driving-related information may be positioned in an area of ​​the instrument panel 51 near a driver's seat.

[0049] Further, a speaker 70 for supporting a voice recognition process by outputting a voice in the form of a conversation with a user while performing the voice recognition function may be positioned within a vehicle.

[0050] Fig. 3 is a flowchart illustrating a speech recognition method according to an embodiment of the present disclosure. Fig. The method shown in Figure 3 can be implemented by the Fig. 1 shown processor.

[0051] With reference to Fig. 3, a speech recognition method according to an embodiment of the present disclosure is described as follows.

[0052] In operation S310, the processor 120 may extract a first intent from a first utterance and may extract a second intent from a second utterance.

[0053] The second utterance can be obtained after the first intention for the first utterance has been extracted.

[0054] In operation S320, the processor 120 may separate the first intention and the second intention into partial intention units using preset / selected / predetermined separators.

[0055] The sub-intention unit can be an action, a goal, or an entity contained in the intention.

[0056] The processor 120 may separate the first intent and the second intent into sub-intent units by assigning a separator to actions, goals, and entities included in the first intent and the second intent.

[0057] The actions and goals in the first intention and the second intention can be identified by a first separator and a second separator. The first separator can be assigned to actions and goals of the first intention and the second intention. The second separator can be assigned to actions and goals among the actions and goals of the second intention that overlap those extracted from the first intention and that contain negative meanings.

[0058] Entities in the first intent and the second intent may be identified by a third separator and a fourth separator. Processor 120 may assign the third separator to an entity under first sub-intent. Further, processor 120 may assign the fourth separator to a sub-intent that negates an entity assigned the third separator under second sub-intent.

[0059] In operation S330, the processor 120 may generate a final intent by deleting duplicate sub-intent(s) depending on the definition of separators.

[0060] If separators that are opposite to each other are assigned to the duplicate sub-intent, the processor 120 may remove the duplicate sub-intent. According to one embodiment, the first separator and the second separator may be opposite to each other, and the third separator and the fourth separator may be opposite to each other.

[0061] Additionally, the processor 120 may obtain the final intent by arranging the first intent and the second intent in the reverse order of the time an utterance is received.

[0062] In addition, in a process of generating the final intent, the processor 120 may delete actions other than the most previous action among sub-intent(s).

[0063] Furthermore, in the process of generating the final intent, the processor 120 may delete sub-intent(s) that match the deleted sub-intent(s) among the sub-intent(s).

[0064] In operation S340, the processor 120 may perform a speech recognition function based on a scenario that matches or is adapted to the final intent.

[0065] A specific embodiment of each method is described below.

[0066] Fig. 4 is a diagram for describing an intent extraction process according to an embodiment of the present disclosure. Referring to the Fig. 1 and Fig. 4, the intent extraction process is as follows.

[0067] Microphone 110 may receive a user's utterance, and processor 120 may extract the user's intent using a deep learning algorithm. For example, processor 120 may convert a user utterance into utterance text and determine the user intent corresponding to the utterance text.

[0068] To convert the user utterance into the utterance text, the processor 120 may use a speech-to-text (STT) engine. The processor 120 may extract a feature vector from the user utterance and may obtain the utterance text by using the extracted feature vector and the trained reference pattern.

[0069] To determine the user intent contained in the utterance text, processor 120 may use natural language understanding (NLU) technology. Processor 120 may include an NLU engine that determines user intent by applying NLU technology to an input sentence.

[0070] The processor 120 can perform NLU as follows.

[0071] Processor 120 may recognize a named entity from the utterance text. The named entity may be a proper noun, such as a person's name, a place name, an organization's name, time, date, currency, or the like. Named entity recognition refers to a task of identifying the named entity in a sentence and determining the type of the identified named entity. The meaning of the sentence can be captured by extracting important keywords from a sentence through named entity recognition.

[0072] Processor 120 may determine a domain from the utterance text. A domain may identify the subject of a user utterance. For example, domains representing various subjects, such as controlling electronic devices, managing schedules, providing information about weather or traffic conditions, sending messages, making phone calls, navigation, and vehicle control, may be determined based on the utterance text.

[0073] Processor 120 may analyze the speech act of the utterance text. Speech act analysis refers to a task of analyzing the intent of the utterance and is used to determine the intent of the utterance, such as whether a user is asking a question, making a request, responding, or simply expressing emotion.

[0074] The processor 120 may determine the intent and an entity required to perform the intent based on information such as a domain, a named entity, and a speech act extracted from the utterance text.

[0075] For example, if the utterance text is “Turn on the air conditioner,” the domain may be [vehicle control], the intent may be [Turn on the air conditioner], and the entity required to control the air conditioner may be [temperature, wind volume].

[0076] Alternatively, if the utterance text is "send a message", the domain may be [send / receive messages], the intent may be [send message], and the entity required to send and receive messages may be [content].

[0077] The intention can be determined by the action and the goal. The action can specify an operation of the domain, and the goal can indicate a subject or object of the action. For example, the action becomes "turn on" and the goal becomes "air conditioning." The action can be referred to as the "operator" and the goal can be referred to as the "object."

[0078] Through this process, the processor 120 can extract various intentions from the user utterance and can extract user intentions such as [call phone book], [search POI], [play radio broadcast] and [play music].

[0079] Fig. 5 is a diagram for describing a user intent rearrangement process according to an embodiment of the present disclosure.

[0080] With reference to Fig. 5, the processor 120 may extract sub-intent based on a plurality of utterances and may reorder the sub-intent. The process of reordering sub-intent can be understood as an intermediate process of generating final intentions.

[0081] The processor 120 can distinguish utterances received at different times. Each of the utterances can be classified based on a state in which the intent is extracted from the utterance. For example, at a first time t1, the processor 120 can generate a first sub-intent based on a first utterance. Thereafter, at a second time t2, the processor 120 can generate a second sub-intent based on a second utterance. Similarly, at a third time t3, the processor 120 can generate a third sub-intent based on a third utterance.

[0082] The first partial intent can be obtained by separating a first intent extracted based on the first utterance using a separator. Similarly, the second partial intent can be obtained by separating a second intent extracted based on the second utterance using a separator. The third partial intent can be obtained by separating a third intent extracted based on the third utterance using a separator.

[0083] Processor 120 may separate the intent by assigning a separator to actions, goals, and entities. A process of assigning a separator may include a process of marking a separator before each action, goal, or entity. The separator may be omitted from the first action, goal, or entity in respective intents.

[0084] The separators may contain first to fourth separators.

[0085] The first separator can be expressed as [+] and can be assigned to an action or goal. For example, intentions such as [call phone book], [search POI], [play radio broadcast], and [play music] can be separated using the first separator into [call + phone book], [search POI + POI], [play radio broadcast +], [play music +].

[0086] The second separator can be expressed as [-] and can be assigned to an action or goal that negates the action or goal assigned to the first separator. Specifically, the second separator can be used in a process of negating the action or goal assigned to the first separator in the preceding intent. For example, in the first utterance, a sub-intent of [+Navi] can be generated depending on the intent of [destination search]. If the intent of [not route guidance] is extracted from the second utterance, the sub-intent of [-Navi] can be generated.

[0087] A third separator may be expressed as [*] and may be assigned to an entity. For example, if a singer or track entity is tagged as an intent to play music, processor 120 may generate a partial intent, such as *{singer} and *{track}.

[0088] A fourth separator can be expressed as [÷] and can be assigned to an entity that negates the entity to which the first separator is assigned. The fourth separator can be used in a process of negating an entity to which the second separator is assigned in the preceding intent.

[0089] Below is an example of generating a final intent based on a user's utterances as follows.

[0090] Fig. 6 is a diagram for describing a method for generating a final intent according to an embodiment of the present disclosure.

[0091] With reference to Fig. 6, in operation S61, the processor 120 may extract an intention based on a first utterance at the first time t1. Similarly, the processor 120 may extract a second intention based on a second utterance at the second time t2, and may extract a third intention based on a third utterance at the third time t3. For example, the processor 120 may extract the first intention of [Search POI] based on the first utterance of "search for a destination" and may extract the second intention of [Play music {Name: Yanghwa Bridge}] based on the second utterance of "play the song by Yanghwa Bridge." Furthermore, the processor 120 may extract the third intention of [Change] based on the third utterance of "something else."

[0092] In operation S62, the processor 120 may separate the first intention, the second intention, and the third intention using separators. [Search POI] may be separated into an action of "search" and a destination of "POI." The processor 120 may separate the first intention into [Search + POI] using a first separator. [Play Music {Name: Yanghwa Bridge}] may be separated into an action of "play," a destination of "music," and an entity of {Name: Yanghwa Bridge}. The processor 120 may separate the second intention into [Play + Play Music * {Name: Yanghwa Bridge}] using the first separator and a third separator. Since [Change] also includes an action, the processor 120 may determine [Change] as a sub-intent. In [Search + POI], each of "search" and "POI" may be referred to as a "first sub-intent."In [Play + Music * {Name: Yanghwa Bridge}], each of "Play Music" and {Name: Yanghwa Bridge} can be referred to as the "second partial intention." In [Change], "Change" can be referred to as the "third partial intention."

[0093] In operation S63, the processor 120 may reorder sub-intent. The processor 120 may order sub-intent in the reverse order of utterances. For example, the processor 120 may place first sub-intent received first in the last order and may place third sub-intent received last in the first order.

[0094] In operation S64, the processor 120 may remove actions other than the most recent action. For example, "Play" and "Search" may be removed from [Change Play Music {Name: Yanghwa Bridge} Search + POI], while retaining only "Change," which is the most recent action.

[0095] In operation S65, processor 120 may remove a mismatched or adjusted sub-intent. For example, +POI may be interpreted as a target that matches a "search" action, and the "search" action may be removed after operation S64. If there is a target from which the action was removed, processor 120 may also remove the corresponding target.

[0096] As a result, the processor 120 can generate a final intention of [Change + Play Music * {Name: Yanghwa Bridge}].

[0097] Like in a Fig. In the embodiment shown in FIG. 6, because an action or target is unclear in the third utterance of "something else," a situation in which a speech recognition function cannot be accurately performed using only the third utterance may conventionally occur. According to an embodiment of the present disclosure, even when ambiguous utterances are obtained, the final intent capable of accurately reflecting a user's intention can be generated based on the previous utterances, and the speech recognition function can be accurately performed based on the final intent.

[0098] Fig. 7 is a diagram for describing a method for generating a final intention according to another embodiment of the present disclosure.

[0099] With reference to Fig. 7, in operation S71, the processor 120 may extract an intention based on a first utterance at the first time t1. Similarly, the processor 120 may extract a second intention based on a second utterance at the second time t2, and may extract a third intention based on a third utterance at the third time t3. For example, the processor 120 may extract the first intention of [Play Music] based on the first utterance of "Play Music" and may extract the second intention of [Route Navi {Name: Yanghwa Bridge}] based on the second utterance of "Yanghwa Bridge." Furthermore, the processor 120 may extract the third intention of [Play Music Not Navi] based on the third utterance of "Play Music, Not Route Guidance."

[0100] In operation S72, the processor 120 may separate the first intent, the second intent, and the third intent using separators.

[0101] Since [Play Music] includes an action of “Play” and a target of “Music,” the processor 120 can separate the first intent into [Play + Music].

[0102] Since [Route Navi {Name: Yanghwa Bridge}] includes an action of "Route", a destination of "Navi", and an entity of {Name: Yanghwa Bridge}, the processor 120 can separate the second intent into [Route + Navi * {Name: Yanghwa Bridge}].

[0103] [Play music not navigation] may include an action of "Play," a target of "Music," and a target of "not navigation." In the third intent, "not navigation" may negate "navigation," which is the preceding second sub-intent. As a result, processor 120 may split the third intent into [Play + Music - Navigation].

[0104] In operation S73, the processor 120 may reorder sub-intent. The processor 120 may order sub-intent in the reverse order of utterances. For example, the processor 120 may place first sub-intent received first in the last order and may place third sub-intent received last in the first order.

[0105] In operation S74, the processor 120 may remove actions other than the most recent action. For example, the processor 120 may remove "Route" of the second sub-intent and "Play" of the first sub-intent from [Play + Music - Navi Route + Navi {Name: Yanghwa Bridge} Play Music], while retaining only "Play," which is the most recent action.

[0106] In operation S75, the processor 120 may combine sub-intent(s) from which an action is removed, and may remove duplicate sub-intent(s) during the combining process. The duplicate sub-intent(s) may be identical sub-intent(s) marked with separators that are opposite to each other. The first separator and the second separator may be a pair of separators that are opposite to each other. The third separator and the fourth separator may be a pair of separators that are opposite to each other. For example, if sub-intent(s) pointing to the same target, such as "-Navi" and "+Navi," are separated by separators that are opposite to each other, the processor 120 may determine "-Navi" and "+Navi" as duplicate sub-intent(s) and delete "-Navi" and "+Navi."

[0107] As a result, the processor 120 can generate a final intention of [Play + Music * {Name: Yanghwa Bridge}].

[0108] As in Fig. As illustrated in Figure 7, it may be unclear to identify the intention because the action is omitted in the second utterance. Furthermore, it may be unclear to identify the intention because the entity is omitted in the third utterance.

[0109] According to an embodiment of the present disclosure, even when ambiguous utterances are obtained, the final intention capable of accurately reflecting a user's intention can be generated based on the previous utterances, and the speech recognition function can be accurately performed based on the final intention.

[0110] Fig. 8 is a diagram for describing a method for generating a final intention according to another embodiment of the present disclosure.

[0111] With reference to Fig. 8, in operation S81, the processor 120 may extract an intent based on a first utterance at the first time t1, and may extract a second intent based on a second utterance at the second time t2. For example, the processor 120 may extract the first intent of [Route Navi {Name: Yanghwa Bridge}] based on the first utterance of "Show me a route to Yanghwa Bridge" and may extract the second intent of [{Name: Olympic Park} not {Name: Yanghwa Bridge}] based on the second utterance of "Olympic Park not Yanghwa Bridge."

[0112] In operation S82, the processor 120 may separate the first intent, the second intent, and the third intent using separators.

[0113] Since [Route Navi {Name: Yanghwa Bridge}] includes an action of "Route", a destination of "Navi", and an entity of {Name: Yanghwa Bridge}, the processor 120 can separate the first intent into [Route + Navi * {Name: Yanghwa Bridge}].

[0114] Since [{Name: Olympic Park} not {Name: Yanghwa Bridge}] contains an entity of {Name: Olympic Park} and an entity of not {Name: Yanghwa Bridge}, the processor 120 may separate the second intent into [* {Name: Olympic Park} ÷ {Name: Yanghwa Bridge}].

[0115] In operation S83, the processor 120 may combine sub-intent(s) from which an action is removed, and may remove duplicate sub-intent(s) during the combination process. For example, * {Name: Yanghwa Bridge} is a sub-intent assigned a third separator, and ÷ {Name: Yanghwa Bridge} is a sub-intent assigned a fourth separator. Since the two entities are equal to each other and have opposite separators "*" and "÷", the two entities may be deleted.

[0116] As a result, the processor 120 may generate a final intent of [Route + Navi * {Name: Olympic Park}].

[0117] As in the Fig. 8, even if the user's intention is not clear because the action of the second utterance is unclear, according to an embodiment of the present disclosure, the final intention capable of accurately reflecting the user's intention can be generated.

[0118] Fig. 9 illustrates a computing system according to an embodiment of the present disclosure.

[0119] With reference to Fig. 9, a computing system 1000 may include at least a processor 1100, a memory 1300, a user interface input device 1400, a user interface output device 1500, a memory 1600, and a network interface 1700, which may be interconnected via a bus 1200.

[0120] Processor 1100 may be a central processing unit (CPU) or a semiconductor device that processes instructions stored in memory 1300 and / or storage 1600. Both memory 1300 and storage 1600 may include various types of volatile or non-volatile storage media. For example, memory 1300 may include read-only memory (ROM) 1310 and random access memory (RAM) 1320.

[0121] Accordingly, operations of a method or algorithm described in connection with an embodiment disclosed in the specification may be implemented directly with a hardware module, a software module, or a combination of the hardware module and the software module executed by the processor 1100. The software module may be located on a storage medium (i.e., the memory 1300 and / or the memory 1600), such as random access memory (RAM), flash memory, read-only memory (ROM), erasable programmable read-only memory (EPROM), electrical EPROM (EEPROM), a register, a hard disk drive, a removable disk, a compact disc-ROM (CD-ROM), or any combination thereof.

[0122] The storage medium may be coupled to the processor 1100. The processor 1100 may read information from the storage medium and may write information to the storage medium. Alternatively, the storage medium may be integrated with the processor 1100. The processor and the storage medium may be implemented with an application-specific integrated circuit (ASIC). The ASIC may be provided in a user terminal. Alternatively, the processor and the storage medium may be implemented with separate components in the user terminal.

[0123] The above description is merely an example of a technical idea of ​​an embodiment of the present disclosure, and various modifications and modifications can be made by a person skilled in the art without departing from the essential feature of the present disclosure.

[0124] Accordingly, embodiments of the present disclosure are intended to illustrate rather than limit technical ideas of the present disclosure, and the scope and spirit of the present disclosure are not necessarily limited by the above embodiments. The scope of the present disclosure may be interpreted by the appended claims, and all equivalents thereof may be construed as being included within the scope of the present disclosure.

[0125] According to an embodiment of the present disclosure, even if a user utterance is unclear, a clearer intent can be generated based on the intent of previous utterances.

[0126] Furthermore, according to an embodiment of the present disclosure, a speech recognition function may be performed in response to various user utterances while reducing the number of scenarios that match a user utterance.

[0127] A variety of effects understood directly or indirectly by the present disclosure can be provided.

[0128] Although the present disclosure has been described above with reference to exemplary embodiments and the accompanying drawings, the present disclosure is not necessarily limited thereto, but may be variously modified and changed by those skilled in the art to which the present disclosure relates without departing from the spirit and scope of the present disclosure as claimed in the following claims. LIST OF REFERENCE SYMBOLS FOR THE FIGURES

[0129] Fig. 1 100 speech recognition device 110 Microphone 120 processor 130 storage 140 Communication interface 200 motor vehicle equipment Fig. 9 1100 processor 1300 storage 1400 User Interface Input Device 1500 User Interface Output Device 1600 memory 1700 network interface

Claims

[1] Speech recognition device comprising: a microphone configured to extract a user's utterance; a memory configured to store a scenario-adapted intent extracted from the utterance; and a processor configured to search for the scenario based on the utterance and perform a speech recognition function, the processor configured to: to extract a first intention from a first utterance and to extract a second intention from a second utterance, to separate the first intention and the second intention into partial intentions using separators, and to create a final intent by combining sub-intent(s) from each of the first intent and the second intent such that duplicate sub-intent(s) are deleted depending on definitions of the separators. [2] The apparatus of claim 1, wherein the processor is further configured to generate the final intent based on the first utterance and the second utterance in response to one or any combination of an action, a goal, or an entity being missing from the second utterance determined after the first utterance. [3] The apparatus of claim 1, wherein the processor is further configured to assign a separator among the separators to each of one or any combination of an action, a target, and an entity included in each of the first intent and the second intent. [4] The apparatus of claim 3, wherein the processor is further configured to: assign a first separator to first sub-intentions, the first separator pointing to the action or goal included among the first sub-intentions extracted from the first intention; and assign a second separator to second sub-intent(s) included among the second sub-intent(s) extracted from the second intent, wherein the second separator points to the action or goal opposite to the first sub-intent(s) to which the first separator has been assigned. [5] The apparatus of claim 4, wherein the processor is further configured to generate the final intent by deleting the first sub-intent and the second sub-intent that are equal to each other in response to the first sub-intent assigned the first separator being the same as the second sub-intent assigned the second separator. [6] The apparatus of claim 3, wherein the processor is further configured to: to assign a third separator to an entity contained among the first partial intentions; and to assign a fourth separator to the entity contained under the second sub-intent, where the fourth separator negates the entity to which the third separator was assigned. [7] The apparatus of claim 6, wherein the processor is further configured to generate the final intent by deleting the entities that are equal to each other and to which the third separator is assigned is the same as the entity to which the fourth separator is assigned. [8] The apparatus of claim 1, wherein the processor is configured to generate the final intent by arranging the first intent and the second intent in a reverse order of utterances. [9] The apparatus of claim 8, wherein the processor is configured to generate the final intention by deleting actions other than a most previous action among the sub-intentions. [10] The apparatus of claim 9, wherein the processor is configured to generate the final intent by deleting sub-intent(s) matching the deleted sub-intent(s) among the sub-intent(s). [11] Speech recognition method, the method comprising: Extracting a first intention from a first utterance; Extracting a second intention from a second utterance; Separating the first intention and the second intention into partial intention units using separators; Creating a final intent by combining the sub-intent units of each of the first intent and the second intent such that duplicate sub-intent units are deleted based on definitions of the separators; and Performing a speech recognition function based on a scenario that matches the final intent. [12] The method of claim 11, wherein extracting the second intent from the second utterance is performed after extracting the first intent. [13] The method of claim 11, wherein separating the first intent and the second intent into the sub-intent units comprises assigning a separator among the separators to one or any combination of an action, a goal, and an entity included in each of the first intent and the second intent. [14] The method of claim 13, wherein separating the first intention and the second intention into the partial intention units comprises: Assigning a first separator to a first sub-intent contained in the first intent, the first separator pointing to the action or goal contained in the first sub-intent; and Assigning a second separator to a second sub-intent contained in the second intent, the second separator pointing to the action or goal contained in the second sub-intent that is opposite to the first sub-intent to which the first separator was assigned. [15] The method of claim 14, wherein generating the final intent comprises deleting the first sub-intent and the second sub-intent that are equal to each other and for which the first sub-intent assigned the first separator is the same as the second sub-intent assigned the second separator. [16] The method of claim 13, wherein separating the first intention and the second intention into the partial intention units comprises: Assigning a third separator to an entity among the first sub-intent(s); and Assigning a fourth separator to the entity contained in the second intent, where the fourth separator points to the entity contained in the second intent that negates the entity contained in the first sub-intent to which the third separator was assigned. [17] The method of claim 16, wherein generating the final intent comprises deleting the entities that are equal to each other and for which the entity to which the third separator is assigned is the same as the entity to which the fourth separator is assigned. [18] The method of claim 11, wherein generating the final intent further comprises arranging the first intent and the second intent in a reverse order of utterances. [19] The method of claim 18, wherein generating the final intent further comprises deleting actions other than a most previous action among the sub-intent(s). [20] The method of claim 19, wherein generating the final intent further comprises deleting sub-intent(s) that match the deleted sub-intent(s) among the sub-intent(s).

Images