KR20200122034A - Dialogue system and method for vehicle - Google Patents

Abstract

The present invention relates to a dialogue system and method for a vehicle. The dialogue system for a vehicle comprises: an input unit into which voice information and non-voice information are inputted; a dialogue management unit generating a selection question by identifying an interior and exterior situation of the vehicle based on at least one of the voice information and the non-voice information and determining a question type to be asked through system selection according to the interior and exterior situation of the vehicle; and an output unit outputting the selection question according to an instruction of the dialogue management unit.

Description

Vehicle dialogue system and method {DIALOGUE SYSTEM AND METHOD FOR VEHICLE}

The present invention relates to a vehicle conversation system and method.

The voice conversation system is in the spotlight as a next-generation interface in which users and systems communicate in natural language. The voice conversation system can be classified into user-priority speech and system-priority speech according to the subject initiating the conversation. Here, the user-first utterance refers to a method that the user first speaks, and the system-first utterance refers to a method that the system first speaks. Existing voice chat systems are being developed based on user-first utterance that analyzes the user's utterance intention and calculates an appropriate response if the user speaks first, that is, if the user speaks first. It is defined to start only in context.

An object of the present invention is to provide a vehicle dialogue system and method for determining system-priority ignition by grasping internal and external conditions of a vehicle and generating a question for prior ignition.

In the vehicle conversation system according to an embodiment of the present invention, an input unit into which voice information and non-voice information are input, the voice information and the non-voice information identify a vehicle internal and external situation based on at least one information, and the vehicle internal and external situation And a conversation management unit configured to generate a selection question by determining a question type to be selected through system selection according to the system selection method, and an output unit outputting the selection question according to an instruction of the conversation management unit.

The conversation management unit may include a first determination unit for determining the system selection if there is a selection rule that satisfies the vehicle interior and exterior conditions among the previously stored selection rules, and a selection timing of the system based on the vehicle interior and exterior conditions. It characterized in that it comprises a second determination unit to determine.

The first determination unit may include a question generation unit including a question type classifier for classifying question types according to internal and external conditions of the vehicle, and a question generator for generating questions for each question type classified by the question type classifier.

The question type is characterized in that it is classified into a fixed event question, a user intention grasp question, and a user information acquisition question.

The question generator may include a first question generator that searches for and outputs a question corresponding to a situation inside and outside the vehicle from among previously stored question rules.

The question generator constructs a hypothesis by analyzing user behavior based on the information input through the input unit, obtains a score for the hypothesis using a learning model, and generates a question for grasping user intention based on the obtained score. It characterized in that it comprises a second question generator.

The question generator includes a third question generator for generating a list of necessary information based on information input through the input unit, calculating a score of each information, and generating a question for acquiring user information based on the calculated score. It is characterized.

The question generator is selected by a candidate question selector for selecting a candidate question from among one or more questions output from at least one of the first question generator, the second question generator, and the third question generator, and the candidate question selector. It characterized in that it further comprises a question synthesizer for generating the selection question by using the candidate question.

On the other hand, in the conversation method of the vehicle conversation system according to an embodiment of the present invention, the first step of receiving voice information and non-voice information, and the internal and external conditions of the vehicle based on at least one of the voice information and the non-voice information. A second step of grasping, and a third step of generating a selection question by determining a type of question to be asked through system selection according to the internal and external conditions of the vehicle.

Before the third step, it is characterized in that it is determined whether or not to select a system by checking whether there is a selection rule that satisfies the interior and exterior conditions of the vehicle among the previously stored selection rules.

The third step includes classifying a question type according to internal and external conditions of the vehicle, generating a question for each question type, selecting one or more candidate questions from among the questions for each question type, and using the one or more candidate questions. And generating the selection question.

The question type is characterized in that it is classified into a fixed event question, a user intention grasp question, and a user information acquisition question.

The generating of the question for each question type may include searching for and outputting a question corresponding to a situation inside and outside the vehicle from among previously stored question rules.

In the step of generating a question for each question type, a hypothesis is established by analyzing user behavior based on input information, and a score for the hypothesis is obtained using a learning model, and a question for identifying user intention is generated based on the obtained score. Characterized in that.

In the step of generating a question for each question type, a list of necessary information is generated based on the input information, a score of each information is calculated, and a question for acquiring user information is generated based on the calculated score.

After the third step, determining a selection time point based on the internal and external conditions of the vehicle, and outputting the selection questionnaire at the selection time point.

According to the present invention, since the system prioritizes utterance by grasping the internal/external situation of the vehicle, and generates a question to be uttered first, the system can obtain necessary user information through the prior utterance, thereby providing an accurate and convenient conversation service.

1 is a block diagram showing a vehicle conversation system according to an embodiment of the present invention.
Figure 2 is a configuration diagram of a conversation management unit shown in Figure 1;
3 is a view showing the configuration of a first determination unit shown in FIG.
4 is a diagram showing the configuration of a first determination unit and an output processing unit shown in FIG. 2;
5 is a flowchart illustrating a conversation method of a vehicle conversation system according to an embodiment of the present invention.
6 is a flowchart illustrating a question generation process shown in FIG. 5.
7 is a block diagram showing a computing system executing a conversation method according to an embodiment of the present invention.

Hereinafter, some embodiments of the present invention will be described in detail through exemplary drawings. In adding reference numerals to elements of each drawing, it should be noted that the same elements are assigned the same numerals as possible even if they are indicated on different drawings. In addition, in describing an embodiment of the present invention, when it is determined that a detailed description of a related known configuration or function interferes with an understanding of the embodiment of the present invention, the detailed description thereof will be omitted.

In describing the constituent elements of the exemplary embodiment of the present invention, terms such as first, second, A, B, (a), (b) may be used. These terms are only used to distinguish the component from other components, and the nature, order, or order of the component is not limited by the term. In addition, unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted as an ideal or excessively formal meaning unless explicitly defined in this application. Does not.

The present invention relates to a technology for determining system-first utterance (hereinafter, ``selection'') based on a situation inside and outside a vehicle, and generating a question to be asked through selection.

1 is a configuration diagram illustrating a vehicle conversation system according to an embodiment of the present invention, FIG. 2 is a configuration diagram of a conversation management unit shown in FIG. 1, and FIG. 3 is a configuration diagram of a first determination unit shown in FIG. FIG. 4 is a diagram illustrating a configuration of a first determination unit and an output processing unit shown in FIG. 2.

Referring to FIG. 1, a vehicle conversation system (hereinafter, a conversation system) is a system that is installed in a vehicle and supports a user to communicate with a vehicle using natural language. Such a conversation system may communicate with electric control units (ECUs) mounted on a vehicle through an in-vehicle network (IVN). The in-vehicle communication network may be implemented as a Controller Area Network (CAN), a Media Oriented Systems Transport (MOST) network, a Local Interconnect Network (LIN), and/or Flexray (X-by-Wire).

The conversation system includes an input unit 100, an input processing unit 200, a storage unit 300, a conversation management unit 400, an output processing unit 500, and an output unit 600.

The input unit 100 may include a first input unit 110 for receiving voice information uttered by a user and a second input unit 120 for receiving non-voice information inside and outside the vehicle.

The first input unit 110 acquires voice information uttered by a user (eg, a driver and/or a passenger) in the vehicle. The first input unit 110 may include a microphone (microphone) for receiving a user's voice and converting it into an electrical signal, and a signal processor for removing noise input along with the voice. The microphone may be implemented as a dynamic microphone, a condenser microphone, a crystal microphone, a carbon microphone, or a directional microphone. The first input unit 110 may acquire voice information spoken by a person outside the vehicle (eg, a pedestrian, etc.). In other words, the first input unit 110 may collect voice information generated inside the vehicle and/or outside the vehicle.

The second input unit 120 acquires non-voice information inside the vehicle and/or outside the vehicle, and includes a camera, a vehicle sensor, a navigation terminal, and/or an electronic control unit (ECU). The second input unit 120 may acquire image information of a user (eg, a driver) and/or the surroundings of a vehicle through a camera. In addition, the second input unit 120 may obtain a vehicle speed, a vehicle position, a steering angle, an acceleration, an impact strength, and/or whether or not a specific function of the vehicle is activated using a vehicle sensor and/or an electronic control device.

The second input unit 120 communicates with electronic devices (eg, smartphones and/or wearable devices, etc.), a service server outside the vehicle, and/or a content server, and provides information related to the user (user information), for example, Schedule and payment information can be collected. In this case, the second input unit 120 may use at least one of communication technologies such as wireless Internet technology, mobile communication technology, and vehicle communication technology (V2X). As a wireless Internet technology, WLAN (Wireless LAN) (WiFi), Wibro (Wireless broadband), and Wimax (World Interoperability for Microwave Access) may be used. As a mobile communication technology, Code Division Multiple Access (CDMA), Global System for Mobile communication (GSM), Long Term Evolution (LTE), LTE-Advanced, and International Mobile Telecommunication (IMT)-2020 may be used.

The input processing unit 200 processes (data processing) voice information and non-voice information input through the input unit 100 to output valid information, that is, situation information for grasping the internal and external conditions (context) of the vehicle. Here, the context information may include user information, vehicle information, and vehicle surrounding information. The input processing unit 200 includes a first input processing unit 210 and a second input processing unit 220.

The first input processing unit 210 includes an automatic speech recognition (ASR) module that recognizes a speech signal (speech) input through the first input unit 110 as a sentence (text) and analyzes the meaning of the recognized sentence. Includes Spoken Language Understanding (SLU) module. In other words, the first input processing unit 210 analyzes voice information input through the first input unit 110 to determine the meaning of the user's speech, that is, a speech act.

The second input processing unit 220 processes the non-voice information input from the second input unit 120 to output context information. The second input processing unit 220 may output information such as a driver status (eg, fatigue level), a vehicle status, and a driving status (eg, congestion and slowness) as situation information based on the non-voice information.

The storage unit 300 may store voice information and non-speech information input through the input unit 100, and store voice information and non-speech information processed by the input processing unit 200, that is, context information in a database (DB). Can be saved in the form. The storage unit 300 may store system selection rules and question rules (questions corresponding to fixed events) in a DB format. The system selection rule (hereinafter, the selection rule) consists of a key and a value. The storage unit 300 may also store a question knowledge base. The question knowledge base is knowledge information for classifying question types (types), that is, rules for determining the type of question.

The storage unit 300 may store information in a short-term storage space or a long-term storage space according to a predetermined criterion when storing information. For example, the storage unit 300 stores the information for a short period of time when permanence is not guaranteed, such as the user's today's schedule, current location, previous conversation content, surroundings, driving situation, and driver status, etc. Save it in space. On the other hand, the storage unit 300 stores the information in a long-term storage space when permanence is guaranteed such as information related to a user's disposition, preference, schedule, education, job, and family, and so on for a long term.

The storage unit 300 includes a flash memory, a hard disk, a secure digital card (SD card), a random access memory (RAM), a static random access memory (SRAM), and a read only memory. , ROM), PROM (Programmable Read Only Memory), EEPROM (Electrically Erasable and Programmable ROM), EPROM (Erasable and Programmable ROM), register, at least one storage medium among storage media such as removable disk and web storage It can be implemented as a (recording medium).

The conversation management unit 400 establishes a conversation progress strategy in consideration of a conversation purpose and a conversation situation, based on the voice information and non-voice information processed by the input processing unit 200. The conversation management unit 400 includes a first determination unit 410 and a second determination unit 420 as shown in FIG. 2. Each of the first determining unit 410 and the second determining unit 420 includes a memory (not shown) and a processor (not shown) that processes instructions stored in the memory (not shown). The processor is a subject that controls the operation of each determination unit 410.

The first determination unit 410 determines whether the system selection condition is satisfied by grasping the interior and exterior conditions of the vehicle based on the situation information output from the input processing unit 200. More specifically, the processor (not shown) of the first determination unit 410 grasps the internal and external conditions of the vehicle based on the voice information and non-voice information input through the input processing unit 200. The processor (not shown) of the first determination unit 410 checks whether there is a selection rule that satisfies among the selection rules stored in the storage unit 300 for the identified vehicle interior and exterior conditions. In other words, the first determination unit 410 searches the selection rule DB stored in the storage unit 300 for a selection rule matching the internal and external conditions of the vehicle. The processor (not shown) of the first determination unit 410 determines that the vehicle interior and exterior conditions satisfy the system selection condition when a selection rule matching the vehicle interior and exterior conditions is searched.

The first determination unit 410 outputs the searched selection rule and rule attribute when the inside and outside of the vehicle satisfies the system selection condition. Here, the rule attribute may include a selection type, response length, and priority. The selection type refers to the type of sentence to be selected, and can be classified into a plain preface or an interrogative sentence. In addition, the processor (not shown) of the first determination unit 410 is for activating the operation of the question generator 4100 included in the first determination unit 410 when the selection type of the rule attribute is'question question'. It generates an enable signal. That is, the first determination unit 410 operates the question generation unit 4100 in the first determination unit 410 when the selection type of the selection rule corresponding to the internal and external conditions of the vehicle is “question question”.

As shown in FIG. 3, the question generator 4100 may include a question type classifier 4110, a question generator 4120, and a final question generator 4130.

The question type classifier 4110 uses information related to the user stored in the context information and long and short (long/short) storage spaces of the storage unit 300 and a question knowledge base stored in the storage unit 300, that is, a question classification rule. Thus, it is determined whether or not to ask a question to the user through selection. The question type classifier 4110 determines whether a situation inside and outside the vehicle is a situation in which a question needs to be asked through system selection. At this time, the question type classifier 4110 determines that it is a situation in which a question must be asked through system selection when the following situation determination criteria are satisfied.

[Criteria for judging the situation]

1) When the user information required to provide a specific service is unclear, that is, when additional user information is required to provide a specific service

2) When generating a control (action) command, the information level required for determining the parameter value (control factor) has not been reached

3) If you want to obtain advance information necessary for providing a service that is not currently being provided

For example, when the clarity of the service provision situation is less than the reference value due to the lack of user information necessary to determine the provision of a specific service, the question type classifier 4110 determines a situation in which a question should be asked through system selection. Meanwhile, the question type classifier 4110 determines a situation in which a question should be asked through system selection when the control factor clarity is less than the reference value due to lack of information necessary to determine the control factor included in a specific control command.

When the question type classifier 4110 is determined as a situation in which a question is to be asked through system selection, the selection is performed using the situation information, user information stored in the long/short storage space, and question classification rules stored in the storage unit 300. Determine the type (type) of the question to be asked, that is, the selection question. The form of the question may be classified into a fixed event question, a user intention grasp question, and a user information acquisition question. Here, the fixed event question refers to a specific fixed event, that is, a question in a predetermined form corresponding to a situation inside and outside the vehicle. The user intention grasp question is a question to clearly grasp the user's intention, and is to acquire information for providing a service to the user based on the current point in time. The user information acquisition question is a question for acquiring user status information, and is for acquiring information for providing a new service based on a specific point in the future.

In other words, the question type classifier 4110 determines the question type as a fixed event question if the current internal and external situation of the vehicle is a predefined fixed event. On the other hand, the question type classifier 4110 may determine a question type as a user information acquisition question or a user intention grasp question according to a current vehicle interior or exterior situation when the current interior and exterior situations are repeated or a new event occurs.

For example, the question type classifier 4110 determines the question type as a user information acquisition question when it is necessary to acquire user information such as taste, preference, satisfaction, and body information (eg, fatigue level) of the user.

On the other hand, when the question type classifier 4110 needs to inform the user of information, when a repetitive input or event occurs, the user performs a simple command difficult and there is a shortcut for the command, the user If you continue to fail to enter the command, if you cannot determine what the user is talking about, if the user fails to execute the function (if the function execution has failed several times), or if you request a function that is not supported, you need to understand the intention of the user. It is determined that it is, and it is decided as a question to understand user intention.

The question generator 4120 generates a question for each question type determined by the question type classifier 4110. The question generator 4120 includes a first question generator 4121 that generates a fixed event question, a second question generator 4122 that generates a user intent grasping question, and a third question generator 4123 that generates a user information acquisition question. Can include.

The first question generator 4121 searches for a question corresponding to the inside and outside of the vehicle (situation information) from the question rule, and transmits the searched result to the candidate question selector 4131. For example, if there is a lot of congestion in front of the vehicle due to an accident, the first question generator (14121) searches for "an accident has occurred and is severely congested. Would you like to tell me another way?" in the question rule to select a candidate question to be described later. It is transmitted to the machine 4131.

The second question generator 4122 analyzes user behavior results based on situation information such as vehicle information, non-vehicle information, and conversation history, and user information stored in the short/long-term storage space, and hypothesizes (predicts) the user's intention for user behavior. Stand with The second question generator 4122 calculates and outputs a score (hypothesis score) for a hypothesis using the learning model. The second question generator 4122 selects a question candidate, for example, a hypothesis having the highest score, for determining user intention based on the hypothesis score, and transmits it to the candidate question selector 4131. In other words, the second question generator 4122 outputs a key keyword of a question for grasping user intention.

For example, the second question generator 4122 may establish a hypothesis that "the user wants ventilation" from the user's opening and closing a window, and obtains a score of 0.8 for the hypothesis using the learning model.

The third question generator 4123 acquires information related to the user's status in the short/long-term storage space, checks whether the information has been previously acquired, and generates a list of information that needs to be acquired (a list of necessary information). The third question generator 4123 calculates a score (score) of each item by calculating the correlation and importance between the individual item of the required information list and the service (action information and factor information) provided by the dialog system. For example, the third question generator 4123 determines how many times an individual item appears in the entire service, how important an individual item has in an individual service, and what importance of a service that can be provided using an individual item. , And/or the score of each item is calculated in consideration of when the update was made.

The third question generator 4123 selects a question candidate for checking the current status of a user and acquiring user information, that is, information (information name) that needs to be acquired, based on the score of each item in the required information list, and selecting a candidate question selector 4131 ). For example, when a user pays for a card at a mart and exits or goes to the mart and gets into the car again, the third question generator 4123 checks the purchase list through the user's mobile phone or card details, Did you live?" And "You often buy item A, do you like item A?" Can be selected as a question candidate.

The question sentence generator 4130 generates a question sentence, that is, a selection question, using at least one question among questions output from the question generator 4120 based on the question policy and question rule. When the question sentence generator 4130 cannot grasp the user's intention based on the context information, the system knows the user's intention based on the context information, but the system does not know by itself in order to provide a service suitable for the user's intention. Questions can be created when there is something to do, when a user's decision is required, or when the surrounding situation needs to be considered. The question sentence generator 4130 includes a candidate question selector 4131 and a question synthesizer 4132.

The candidate question selector 4131 selects one or more questions output from at least one or more of the first question generator 4121, the second question generator 4122, and the third question generator 4123 as a candidate question. Then, a list of selected questions is generated and printed. The candidate question selector 4131 changes various types of questions or words inputted from the multiple question generators 4121 to 4123 into a common format used in the system, and normalizes individual scores (scores) using a question policy. Converts to a number that can be compared with each other. The candidate question selector 4131 selects one or more questions to be provided to the user by using the converted score.

For example,'Restaurant' and 45 points and'color' and 34 points from the third question generator 4123,'confirm new directions' and 60 points from the first question generator 4121, and the second question generator 4122 ), the candidate question selector (4131) changes the question format to'Restaurant Information Request','New Route Guide Conform', and'Perform Ventilation Function', and individual scores are 45 points. , Normalized to 60 and 70 points. The candidate question selector 4131 determines'ventilation function performance' and'new route guidance conform' as question candidates based on the score.

The question synthesizer 4132 combines one or more question(s) selected based on the selected question list, and transforms it into a complete question for providing a conversation service. In this case, the question synthesizer 4132 may generate a question sentence using a natural language generation (NLG) program. For example, if'Ventilation function performance' and'New route guidance conform' are selected by the candidate question selector (4131), the question synthesizer (4132) is'Indoor air is not good according to the question rule. Do you want to ventilate? Or, can I guide you to a pleasant path?'

In the above, a case in which the candidate question selector 4131 and the question synthesizer 4132 are disposed in the question generator 4100 has been described, but is not limited thereto. For example, as shown in FIG. 4, the candidate question selector 510 and the question synthesizer 520 may be disposed in the output processor 500.

The second determination unit 420 stores the results (selection rules and rule attributes) output from the first determination unit 410 in the storage unit 300. In this case, the output result of the first determination unit 410 may be stored based on the priority of the rule attribute.

The second determination unit 420 determines the selection timing of the system based on the internal and external conditions of the vehicle. In other words, the second determination unit 420 may determine the system selection time based on the voice information and non-voice information input through the input processing unit 200 to determine the internal and external conditions of the vehicle and consider the identified internal and external conditions of the vehicle. have. For example, the second determination unit 420 may determine a random time point, a time point at which a specific event occurs, or a time point at which a user's request is not interpreted as a system selection time point.

The output processing unit 500 generates the generated question as visual information and auditory information. The output processing unit 500 converts a text corresponding to a question to be selected by the system into speech using a text-to-speech program and outputs the converted text.

The output unit 600 may include a speaker for amplifying and outputting the converted voice output from the output processing unit 500 to an appropriate level. In addition, the output unit 600 may include a display for displaying a question in text. The display is a liquid crystal display (LCD), a thin film transistor-liquid crystal display (TFT LCD), an organic light-emitting diode (OLED) display, a flexible display, It may include one or more of a 3D display, a transparent display, a head-up display (HUD), a touch screen, and a cluster. The display is implemented as a touch screen combined with a touch sensor and can be used as an input device as well as an output device. As the touch sensor, a touch film or a touch pad may be used.

5 is a flowchart illustrating a conversation method of a vehicle conversation system according to an embodiment of the present invention.

Referring to FIG. 5, the conversation management unit 400 of the conversation system receives voice information and non-voice information (S110). The input processing unit 200 processes voice information and non-speech information input through the input unit 100 to output context information obtained by data processing to the conversation management unit 400.

The conversation management unit 400 determines the inside and outside of the vehicle based on the voice information and the non-voice information (S120). In other words, the conversation management unit 400 identifies the inside and outside of the vehicle based on the situation information.

The conversation management unit 400 checks whether the inside and outside of the vehicle satisfies the system selection condition (S130). The conversation management unit 400 searches the storage unit 300 for a system selection rule corresponding to a situation inside and outside the vehicle. The conversation management unit 400 determines that the system selection condition has been satisfied when searching for a selection rule corresponding to an internal and external situation of the vehicle among system selection rules stored in the storage unit 300.

The conversation management unit 400 generates a selection question when the inside and outside of the vehicle satisfies the system selection condition (S140). When a question through system selection is determined, the first determination unit 410 of the conversation management unit 400 generates a question to be performed through selection. In this case, the first determination unit 410 determines a question type to be generated according to the internal and external conditions of the vehicle, and generates one or more question(s) for each determined question type. The first judgment unit 410 selects one or more candidate questions based on a predetermined criterion (rule) among the generated one or more question(s), and generates one question sentence, that is, a selection question, using the selected question. do.

The conversation management unit 400 determines a system selection time point based on internal and external conditions of the vehicle (S150). The second determination unit 420 of the conversation management unit 400 determines the selection timing in consideration of the context information and the rule attribute output from the first determination unit 410. That is, the second determination unit 420 determines whether to immediately ask a question through selection or to postpone the question point to a specific point in time in consideration of situation information and rule attributes.

The conversation management unit 400 outputs a selection question at the time of selection (S160). The conversation management unit 400 transmits the selection question generated at the time of selection to the output processing unit 500. The output processing unit 500 converts the selection question into voice and GUI (Graphic User Interface) and outputs the converted question to the output unit 600.

6 is a flowchart illustrating a process of generating a question shown in FIG. 5.

When the vehicle interior and exterior conditions satisfy the system selection condition, the first determination unit 410 checks whether the vehicle interior or exterior situation is a situation in which a question should be asked through system selection (S141). The first determination unit 410 determines whether to ask a question through selection to the user using context information, user information stored in the long/short storage space, and question classification rules stored in the storage unit 300.

The first determination unit 410 determines a question type based on the question classification rule stored in the storage unit 300 when the vehicle interior and exterior situations are situations in which a question must be asked through system selection (S142).

The first determining unit 410 generates a question for each determined question type (S143). The question generation unit 4100 of the first determination unit 410 searches the storage unit 300 for a question corresponding to the vehicle interior/exterior situation when the vehicle interior/exterior situation is a predefined fixed event. In addition, the question generator 4100 may generate a user intention identification question or a user information acquisition question according to a situation inside and outside the vehicle.

The first determination unit 410 generates a selection question by using one or more of the generated questions (S144). The question generation unit 4100 of the first determination unit 410 selects one or more question candidates from among questions generated for each question type, and generates a final question sentence, that is, a selection question, using the selected question candidate.

7 is a block diagram showing a computing system for executing a conversation method according to an embodiment of the present invention.

Referring to FIG. 7, the computing system 1000 includes at least one processor 1100, a memory 1300, a user interface input device 1400, a user interface output device 1500, and storage connected through a bus 1200. (1600), and a network interface (1700).

The processor 1100 may be a central processing unit (CPU) or a semiconductor device that processes instructions stored in the memory 1300 and/or the storage 1600. The memory 1300 and the storage 1600 may include various types of volatile or nonvolatile storage media. For example, the memory 1300 may include read only memory (ROM) and random access memory (RAM).

Accordingly, the steps of the method or algorithm described in connection with the embodiments disclosed herein may be directly implemented by hardware, software modules, or a combination of the two executed by the processor 1100. Software modules reside in storage media (i.e., memory 1300 and/or storage 1600) such as RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM. You may. An exemplary storage medium is coupled to the processor 1100, which is capable of reading information from and writing information to the storage medium. Alternatively, the storage medium may be integral with the processor 1100. The processor and storage media may reside within an application specific integrated circuit (ASIC). The ASIC may reside within the user terminal. Alternatively, the processor and storage medium may reside as separate components within the user terminal.

The above description is merely illustrative of the technical idea of the present invention, and those of ordinary skill in the art to which the present invention pertains will be able to make various modifications and variations without departing from the essential characteristics of the present invention. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain the technical idea, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of the present invention.

100: input unit 110: first input unit
120: second input unit 200: input processing unit
210: first input processing unit 220: second input processing unit
300: storage unit 400: conversation management unit
410: first judgment portion 420: second judgment portion
4100: question generator 4110: question type classifier
4120: question generator 4121: first question generator
4122: second question generator 4123: third question generator
4130: question sentence generator 4131, 510: candidate question selector
4132, 520: question synthesizer 500: output processing unit
600: output

Claims (16)

An input unit into which voice information and non-voice information are input,
A conversation management unit for generating a selection question by identifying a vehicle internal and external situation based on at least one of the voice information and the non-voice information, and determining a question type to be asked through system selection according to the internal and external conditions of the vehicle, and
A vehicle conversation system comprising an output unit for outputting the selection question according to the instruction of the conversation management unit.
The method of claim 1,
The conversation management unit,
A first determination unit that determines the system selection if there is a selection rule that satisfies the interior and exterior conditions of the vehicle among the previously stored selection rules, and
And a second determination unit for determining a selection timing of the system based on the internal and external conditions of the vehicle.
The method of claim 2,
The first judgment unit,
And a question generation unit including a question type classifier for classifying question types according to internal and external conditions of the vehicle, and a question generator for generating a question for each question type classified by the question type classifier.
The method of claim 3,
The question type is classified into a fixed event question, a user intention grasp question, and a user information acquisition question.
The method of claim 4,
And the question generator comprises a first question generator that searches for and outputs a question corresponding to a situation inside and outside the vehicle from among pre-stored question rules.
The method of claim 4,
The question generator constructs a hypothesis by analyzing user behavior based on the information input through the input unit, obtains a score for the hypothesis using a learning model, and generates a question for grasping user intention based on the obtained score. A conversation system for a vehicle comprising a second question generator.
The method of claim 4,
The question generator includes a third question generator for generating a list of necessary information based on information input through the input unit, calculating a score of each information, and generating a question for acquiring user information based on the calculated score. Vehicle conversation system characterized by.
The method according to any one of claims 5 to 7,
The question generator,
A candidate question selector for selecting a candidate question among one or more questions output from at least one or more of the first question generator, the second question generator, and the third question generator, and
And a question synthesizer for generating the selection question by using the candidate question selected by the candidate question selector.
The first step of receiving voice information and non-voice information,
A second step of identifying a situation inside and outside the vehicle based on at least one of the voice information and the non-voice information, and
And a third step of generating a selection question by determining a type of question to be selected through system selection according to the internal and external conditions of the vehicle.
The method of claim 9,
Before the third step,
And determining whether to select a system by checking whether there is a selection rule that satisfies the interior and exterior conditions of the vehicle among the previously stored selection rules.
The method of claim 9,
The third step,
Classifying a question type according to internal and external conditions of the vehicle,
Generating a question for each question type,
Selecting one or more candidate questions from among the questions for each question type, and
And generating the selection question by using the one or more candidate questions.
The method of claim 11,
The question type is classified into a fixed event question, a user intention grasp question, and a user information acquisition question.
The method of claim 11,
The step of generating a question for each question type,
A conversation method of a vehicle conversation system, characterized in that, among pre-stored question rules, a question corresponding to the inside and outside of the vehicle is searched for and output.
The method of claim 11,
The step of generating a question for each question type,
Conversation in a vehicle dialogue system, characterized in that a hypothesis is established by analyzing user behavior based on input information, a score for the hypothesis is obtained using a learning model, and a question to grasp user intention is generated based on the obtained score. Way.
The method of claim 11,
The step of generating a question for each question type,
A conversation method of a vehicle conversation system, comprising generating a list of necessary information based on input information, calculating a score of each information, and generating a question for acquiring user information based on the calculated score.
The method of claim 9,
After the third step,
Determining a selection time based on the inside and outside of the vehicle, and
And outputting the selection question at the selection time point.

Images