Utterance processing method, electronic device for performing said method, and recording medium

Abstract

Disclosed are an utterance processing method and an electronic device for performing the method. The electronic device, according to various embodiments, may comprise: a processor; and a memory that is electrically connected to the processor and stores instructions executable by the processor. When the instructions are executed by the processor, the processor may receive a plurality of utterances from a user through a voice input device. The processor may generate a plurality of utterance vectors corresponding to the plurality of utterances. The processor may calculate a first similarity between each of a plurality of reference vectors stored in the memory and each of the plurality of utterance vectors. The processor may determine, from among the plurality of utterance vectors on the basis of the first similarity, training utterance vectors for training a first model for identifying a user-specified wake-up word. The processor may train the model by using the training utterance vectors.

Description

An ignition processing method and an electronic device performing the method, and a recording medium

[0001] The disclosure below relates to a method for processing ignition and an electronic device for performing the method, and a recording medium.

[0002] Voice activation is a common feature of electronic devices, whereby a user's utterance of a predefined word initiates an interaction with a voice-based agent to perform a specific task.

[0003] Custom wake up or custom wake up allows users to set their own words to activate voice.

[0004] Predefined word recognition models are typically trained on large amounts of data and can be tuned for recognition accuracy. When custom wake-up is implemented on electronic devices, small-capacity recognition models must be trained using small amounts of user-provided data (or small amounts of audio input or speech).

[0005] In learning custom wake-ups, the quality of user input can affect the performance of the recognition model, as individual user inputs affect the performance of the recognition model.

[0006] The above information may be provided as background art to aid in understanding the present disclosure. No claim or determination is made as to whether any of the above is applicable as prior art related to the present disclosure.

[0007] An electronic device according to various embodiments may include a processor and a memory electrically connected to the processor and storing instructions executable by the processor. The processor may receive a plurality of utterances from a user through a voice input device when the instructions are executed by the processor. The processor may generate a plurality of utterance vectors corresponding to the plurality of utterances. The processor may calculate a first similarity between each of the plurality of utterance vectors and each of the plurality of reference vectors stored in the memory. The processor may determine, based on the first similarity, training utterance vectors for training a first model for identifying a user-specified wake word among the plurality of utterance vectors. The processor may train the model using the training utterance vectors.

[0008] An electronic device according to various embodiments may include a processor and a memory electrically connected to the processor and storing instructions executable by the processor. The processor may receive a plurality of utterances from a user through a voice input device when the instructions are executed by the processor. The processor may generate a plurality of utterance vectors corresponding to the plurality of utterances. The processor may calculate a first similarity between each of the plurality of utterance vectors and each of the plurality of reference vectors stored in the memory. The processor may determine, based on the first similarity, training utterance vectors for training a first model for identifying a user-specified wake word among the plurality of utterance vectors. The processor may train the model using the training utterance vectors. The processor may provide a first interface for requesting an additional utterance to the user when the first similarity is greater than or equal to a first threshold value. The first interface may include a type of a reference vector for which the first similarity is greater than or equal to the first threshold value.

[0009] A method for processing speech according to various embodiments may include an operation of receiving a plurality of speeches from a user through a voice input device, an operation of generating a plurality of speech vectors corresponding to the plurality of speeches, an operation of calculating a first similarity between each of the plurality of speech vectors and a plurality of reference vectors stored in a memory, an operation of determining learning speech vectors for training a first model for identifying a user-specified call word among the plurality of speech vectors based on the first similarity, and an operation of training the first model using the learning speech vectors.

[0010] A method for processing speech according to various embodiments may include receiving a plurality of utterances from a user through a voice input device, generating a plurality of utterance vectors corresponding to the plurality of utterances, calculating a first similarity between each of the plurality of utterance vectors and a plurality of reference vectors stored in the memory, determining learning utterance vectors for training a first model for identifying a user-specified call word among the plurality of utterance vectors based on the first similarity, training the model using the learning utterance vectors, and providing a first interface for requesting an additional utterance to the user when the first similarity is equal to or greater than a first threshold value. The first interface may include a type of a reference vector for which the first similarity is equal to or greater than the first threshold value.

[0011] FIG. 1 is a block diagram of an electronic device within a network environment according to various embodiments.

[0012] FIG. 2 is a block diagram illustrating an integrated intelligence system according to one embodiment.

[0013] FIG. 3 is a diagram showing a form in which relationship information between concepts and actions is stored in a database according to one embodiment.

[0014] FIG. 4 is a diagram illustrating a user terminal displaying a screen for processing voice input received through an intelligent app, according to one embodiment.

[0015] FIG. 5 is a flowchart of an operation of a method for processing ignition performed by an electronic device according to various embodiments.

[0016] FIG. 6 is a flowchart illustrating an operation of an electronic device according to various embodiments to provide an interface according to a first similarity.

[0017] FIG. 7 is a diagram illustrating a plurality of ignition vectors and a plurality of reference vectors according to various embodiments.

[0018] FIG. 8 is a flowchart illustrating an operation of an electronic device according to various embodiments to provide an interface according to a second similarity.

[0019] FIG. 9 is a diagram illustrating multiple ignition vectors according to various embodiments.

[0020] FIG. 10 is a diagram illustrating a plurality of utterance vectors for training a first model and a second model determined based on user input according to various embodiments.

[0021] FIG. 11 is a diagram illustrating a plurality of utterance vectors for training a first model determined based on user input according to various embodiments.

[0022] FIGS. 12, 13, and 14 are diagrams illustrating interfaces provided to a user by an electronic device according to various embodiments.

[0023] Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the attached drawings, identical components are assigned the same reference numerals regardless of the drawing numbers, and redundant descriptions thereof will be omitted.

[0024] In this document, each of the phrases "A or B", "at least one of A and B", "at least one of A or B", "A, B, or C", "at least one of A, B, and C", and "at least one of A, B, or C" may include any one of the items listed together in that phrase, or all possible combinations thereof.

[0025] FIG. 1 is a block diagram of an electronic device (101) within a network environment (100) according to various embodiments. Referring to FIG. 1, in the network environment (100), the electronic device (101) may communicate with the electronic device (102) via a first network (198) (e.g., a short-range wireless communication network), or may communicate with at least one of the electronic device (104) or the server (108) via a second network (199) (e.g., a long-range wireless communication network). According to one embodiment, the electronic device (101) may communicate with the electronic device (104) via the server (108). According to one embodiment, the electronic device (101) may include a processor (120), a memory (130), an input module (150), an audio output module (155), a display module (160), an audio module (170), a sensor module (176), an interface (177), a connection terminal (178), a haptic module (179), a camera module (180), a power management module (188), a battery (189), a communication module (190), a subscriber identification module (196), or an antenna module (197). In some embodiments, the electronic device (101) may omit at least one of these components (e.g., the connection terminal (178)), or may have one or more other components added. In some embodiments, some of these components (e.g., the sensor module (176), the camera module (180), or the antenna module (197)) may be integrated into one component (e.g., the display module (160)).

[0026] The processor (120) may, for example, execute software (e.g., a program (140)) to control at least one other component (e.g., a hardware or software component) of the electronic device (101) connected to the processor (120) and perform various data processing or operations. According to one embodiment, as at least a part of the data processing or operations, the processor (120) may store commands or data received from other components (e.g., a sensor module (176) or a communication module (190)) in a volatile memory (132), process the commands or data stored in the volatile memory (132), and store result data in a non-volatile memory (134). According to one embodiment, the processor (120) may include a main processor (121) (e.g., a central processing unit or an application processor) or an auxiliary processor (123) (e.g., a graphics processing unit, a neural processing unit (NPU), an image signal processor, a sensor hub processor, or a communication processor) that can operate independently or together with the main processor (121). For example, when the electronic device (101) includes the main processor (121) and the auxiliary processor (123), the auxiliary processor (123) may be configured to use less power than the main processor (121) or to be specialized for a given function. The auxiliary processor (123) may be implemented separately from the main processor (121) or as a part thereof.

[0027] The auxiliary processor (123) may control at least a portion of functions or states associated with at least one component (e.g., a display module (160), a sensor module (176), or a communication module (190)) of the electronic device (101), for example, on behalf of the main processor (121) while the main processor (121) is in an inactive (e.g., sleep) state, or together with the main processor (121) while the main processor (121) is in an active (e.g., application execution) state. In one embodiment, the auxiliary processor (123) (e.g., an image signal processor or a communication processor) may be implemented as a part of another functionally related component (e.g., a camera module (180) or a communication module (190)). In one embodiment, the auxiliary processor (123) (e.g., a neural network processing unit) may include a hardware structure specialized for processing artificial intelligence models. The artificial intelligence models may be generated through machine learning. This learning can be performed, for example, on the electronic device (101) itself where the artificial intelligence model is executed, or can be performed through a separate server (e.g., server (108)). The learning algorithm can include, for example, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but is not limited to the examples described above. The artificial intelligence model can include multiple artificial neural network layers.The artificial neural network may be one of a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a restricted Boltzmann machine (RBM), a deep belief network (DBN), a bidirectional recurrent deep neural network (BRDNN), a deep Q-network, or a combination of two or more of the above, but is not limited to the examples described above. In addition to, or alternatively to, a hardware structure, an artificial intelligence model may include a software structure.

[0028] The memory (130) can store various data used by at least one component (e.g., processor (120) or sensor module (176)) of the electronic device (101). The data can include, for example, software (e.g., program (140)) and input data or output data for commands related thereto. The memory (130) can include volatile memory (132) or non-volatile memory (134).

[0029] The program (140) may be stored as software in the memory (130) and may include, for example, an operating system (142), middleware (144), or an application (146).

[0030] The input module (150) can receive commands or data to be used in a component of the electronic device (101) (e.g., a processor (120)) from an external source (e.g., a user) of the electronic device (101). The input module (150) can include, for example, a microphone, a mouse, a keyboard, a key (e.g., a button), or a digital pen (e.g., a stylus pen).

[0031] The audio output module (155) can output audio signals to the outside of the electronic device (101). The audio output module (155) can include, for example, a speaker or a receiver. The speaker can be used for general purposes, such as multimedia playback or recording playback. The receiver can be used to receive incoming calls. In one embodiment, the receiver can be implemented separately from the speaker or as part of the speaker.

[0032] The display module (160) can visually provide information to an external party (e.g., a user) of the electronic device (101). The display module (160) may include, for example, a display, a holographic device, or a projector and a control circuit for controlling the device. According to one embodiment, the display module (160) may include a touch sensor configured to detect a touch, or a pressure sensor configured to measure the intensity of a force generated by the touch.

[0033] The audio module (170) can convert sound into an electrical signal, or vice versa, convert an electrical signal into sound. According to one embodiment, the audio module (170) can acquire sound through the input module (150), output sound through the sound output module (155), or an external electronic device (e.g., electronic device (102)) (e.g., speaker or headphone) directly or wirelessly connected to the electronic device (101).

[0034] The sensor module (176) can detect the operating status (e.g., power or temperature) of the electronic device (101) or the external environmental status (e.g., user status) and generate an electrical signal or data value corresponding to the detected status. According to one embodiment, the sensor module (176) can include, for example, a gesture sensor, a gyro sensor, a barometric pressure sensor, a magnetic sensor, an acceleration sensor, a grip sensor, a proximity sensor, a color sensor, an IR (infrared) sensor, a biometric sensor, a temperature sensor, a humidity sensor, or an illuminance sensor.

[0035] The interface (177) may support one or more designated protocols that may be used to directly or wirelessly connect the electronic device (101) with an external electronic device (e.g., the electronic device (102)). In one embodiment, the interface (177) may include, for example, a high definition multimedia interface (HDMI), a universal serial bus (USB) interface, an SD card interface, or an audio interface.

[0036] The connection terminal (178) may include a connector through which the electronic device (101) may be physically connected to an external electronic device (e.g., electronic device (102)). According to one embodiment, the connection terminal (178) may include, for example, an HDMI connector, a USB connector, an SD card connector, or an audio connector (e.g., a headphone connector).

[0037] The haptic module (179) can convert electrical signals into mechanical stimuli (e.g., vibration or movement) or electrical stimuli that a user can perceive through tactile or kinesthetic sensations. According to one embodiment, the haptic module (179) can include, for example, a motor, a piezoelectric element, or an electrical stimulation device.

[0038] The camera module (180) can capture still images and videos. According to one embodiment, the camera module (180) may include one or more lenses, image sensors, image signal processors, or flashes.

[0039] The power management module (188) can manage power supplied to the electronic device (101). According to one embodiment, the power management module (188) can be implemented as, for example, at least a part of a power management integrated circuit (PMIC).

[0040] A battery (189) may power at least one component of the electronic device (101). In one embodiment, the battery (189) may include, for example, a non-rechargeable primary battery, a rechargeable secondary battery, or a fuel cell.

[0041] The communication module (190) may support the establishment of a direct (e.g., wired) communication channel or a wireless communication channel between the electronic device (101) and an external electronic device (e.g., electronic device (102), electronic device (104), or server (108)), and the performance of communication through the established communication channel. The communication module (190) may operate independently from the processor (120) (e.g., application processor) and may include one or more communication processors that support direct (e.g., wired) communication or wireless communication. According to one embodiment, the communication module (190) may include a wireless communication module (192) (e.g., a cellular communication module, a short-range wireless communication module, or a global navigation satellite system (GNSS) communication module) or a wired communication module (194) (e.g., a local area network (LAN) communication module, or a power line communication module). Among these communication modules, the corresponding communication module can communicate with an external electronic device (104) via a first network (198) (e.g., a short-range communication network such as Bluetooth, wireless fidelity (WiFi) direct, or infrared data association (IrDA)) or a second network (199) (e.g., a long-range communication network such as a legacy cellular network, a 5G network, a next-generation communication network, the Internet, or a computer network (e.g., a LAN or WAN)). These various types of communication modules can be integrated into a single component (e.g., a single chip) or implemented as multiple separate components (e.g., multiple chips). The wireless communication module (192) can verify or authenticate the electronic device (101) within a communication network such as the first network (198) or the second network (199) by using subscriber information (e.g., an international mobile subscriber identity (IMSI)) stored in the subscriber identification module (196).

[0042] The wireless communication module (192) can support 5G networks and next-generation communication technologies following the 4G network, such as NR access technology (new radio access technology). The NR access technology can support high-speed transmission of high-capacity data (eMBB (enhanced mobile broadband)), minimization of terminal power and connection of multiple terminals (mMTC (massive machine type communications)), or high reliability and low latency (URLLC (ultra-reliable and low-latency communications)). The wireless communication module (192) can support, for example, a high-frequency band (e.g., mmWave band) to achieve a high data transmission rate. The wireless communication module (192) can support various technologies for securing performance in a high-frequency band, such as beamforming, massive multiple-input and multiple-output (MIMO), full dimensional MIMO (FD-MIMO), array antenna, analog beam-forming, or large scale antenna. The wireless communication module (192) can support various requirements specified in the electronic device (101), an external electronic device (e.g., the electronic device (104)), or a network system (e.g., the second network (199)). According to one embodiment, the wireless communication module (192) can support a peak data rate (e.g., 20 Gbps or more) for eMBB realization, a loss coverage (e.g., 164 dB or less) for mMTC realization, or a U-plane latency (e.g., 0.5 ms or less for downlink (DL) and uplink (UL), or 1 ms or less for round trip) for URLLC realization.

[0043] The antenna module (197) can transmit or receive signals or power to or from an external device (e.g., an external electronic device). In one embodiment, the antenna module (197) may include an antenna including a radiator formed of a conductor or a conductive pattern formed on a substrate (e.g., a PCB). In one embodiment, the antenna module (197) may include a plurality of antennas (e.g., an array antenna). In this case, at least one antenna suitable for a communication method used in a communication network, such as the first network (198) or the second network (199), may be selected from the plurality of antennas, for example, by the communication module (190). A signal or power may be transmitted or received between the communication module (190) and an external electronic device via the at least one selected antenna. In some embodiments, in addition to the radiator, another component (e.g., a radio frequency integrated circuit (RFIC)) may be additionally formed as a part of the antenna module (197).

[0044] According to various embodiments, the antenna module (197) may form a mmWave antenna module. In one embodiment, the mmWave antenna module may include a printed circuit board, an RFIC disposed on or adjacent a first side (e.g., a bottom side) of the printed circuit board and capable of supporting a designated high-frequency band (e.g., a mmWave band), and a plurality of antennas (e.g., an array antenna) disposed on or adjacent a second side (e.g., a top side or a side side) of the printed circuit board and capable of transmitting or receiving signals in the designated high-frequency band.

[0045] At least some of the above components can be interconnected and exchange signals (e.g., commands or data) with each other via a communication method between peripheral devices (e.g., a bus, GPIO (general purpose input and output), SPI (serial peripheral interface), or MIPI (mobile industry processor interface)).

[0046] According to one embodiment, commands or data may be transmitted or received between the electronic device (101) and an external electronic device (104) via a server (108) connected to a second network (199). Each of the external electronic devices (102 or 104) may be the same or a different type of device as the electronic device (101). According to one embodiment, all or part of the operations executed in the electronic device (101) may be executed in one or more of the external electronic devices (102, 104, or 108). For example, when the electronic device (101) is to perform a certain function or service automatically or in response to a request from a user or another device, the electronic device (101) may, instead of or in addition to executing the function or service itself, request one or more external electronic devices to perform the function or at least a part of the service. One or more external electronic devices that receive the request may execute at least a portion of the requested function or service, or an additional function or service related to the request, and transmit the result of the execution to the electronic device (101). The electronic device (101) may process the result as is or additionally and provide it as at least a portion of a response to the request. For this purpose, cloud computing, distributed computing, mobile edge computing (MEC), or client-server computing technology may be used, for example. The electronic device (101) may provide an ultra-low latency service by using distributed computing or mobile edge computing, for example. In another embodiment, the external electronic device (104) may include an Internet of Things (IoT) device. The server (108) may be an intelligent server utilizing machine learning and / or a neural network. According to one embodiment, the external electronic device (104) or the server (108) may be included in the second network (199).The electronic device (101) can be applied to intelligent services (e.g., smart home, smart city, smart car, or healthcare) based on 5G communication technology and IoT-related technology.

[0047]

[0048] FIG. 2 is a block diagram illustrating an integrated intelligence system according to one embodiment.

[0049] Referring to FIG. 2, an integrated intelligent system of one embodiment may include an electronic device (101), an intelligent server (200) (e.g., server (108) of FIG. 1), and a service server (300) (e.g., server (108) of FIG. 1).

[0050] The electronic device (101) of one embodiment may be a terminal device (or electronic device) that can connect to the Internet, and may be, for example, a mobile phone, a smart phone, a personal digital assistant (PDA), a laptop computer, a TV, white goods, a wearable device, an HMD, or a smart speaker.

[0051] According to the illustrated embodiment, the electronic device (101) may include an interface (177), an input module (150), an audio output module (155), a display module (160), a memory (130), or a processor (120). The above-listed components may be operatively or electrically connected to each other.

[0052] An interface (177) of one embodiment may be configured to connect to an external device and transmit and receive data. An input module (150) of one embodiment may receive sound (e.g., user speech) and convert it into an electrical signal. An audio output module (155) of one embodiment may output the electrical signal as sound (e.g., voice).

[0053] The display module (160) of one embodiment may be configured to display an image or video. The display module (160) of one embodiment may also display a graphical user interface (GUI) of a running app (or application program). The display module (160) of one embodiment may receive touch input via a touch sensor. For example, the display module (160) may receive text input via a touch sensor in an on-screen keyboard area displayed within the display module (160).

[0054] In one embodiment, the memory (130) may store a client module (151), a software development kit (SDK) (153), and a plurality of apps (146). The client module (151) and the SDK (153) may constitute a framework (or solution program) for performing general functions. In addition, the client module (151) or the SDK (153) may constitute a framework for processing user input (e.g., voice input, text input, touch input).

[0055] The plurality of apps (146) stored in the memory (130) of one embodiment may be programs for performing a specified function. According to one embodiment, the plurality of apps (146) may include a first app (146-1) and a second app (146-2). According to one embodiment, each of the plurality of apps (146) may include a plurality of operations for performing a specified function. For example, the apps may include an alarm app, a message app, and / or a schedule app. According to one embodiment, the plurality of apps (146) may be executed by the processor (120) to sequentially execute at least some of the plurality of operations.

[0056] The processor (120) of one embodiment can control the overall operation of the electronic device (101). For example, the processor (120) can be electrically connected to an interface (177), an input module (150), an audio output module (155), and a display module (160) to perform a specified operation.

[0057] The processor (120) of one embodiment may also execute a program stored in the memory (130) to perform a designated function. For example, the processor (120) may execute at least one of the client module (151) or the SDK (153) to perform the following operations for processing user input. The processor (120) may control the operations of multiple apps (146), for example, through the SDK (153). The following operations described as operations of the client module (151) or the SDK (153) may be operations executed by the processor (120).

[0058] The client module (151) of one embodiment can receive user input. For example, the client module (151) can receive a voice signal corresponding to a user utterance detected through the input module (150). Alternatively, the client module (151) can receive a touch input detected through the display module (160). Alternatively, the client module (151) can receive a text input detected through a keyboard or a visual keyboard. In addition, the client module (151) can receive various forms of user input detected through an input module included in the electronic device (101) or an input module connected to the electronic device (101). The client module (151) can transmit the received user input to the intelligent server (200). The client module (151) can transmit status information of the electronic device (101) to the intelligent server (200) together with the received user input. The status information can be, for example, execution status information of an app.

[0059] The client module (151) of one embodiment can receive a result corresponding to the received user input. For example, the client module (151) can receive a result corresponding to the received voice input when the intelligent server (200) can produce a result corresponding to the received user input. The client module (151) can display the received result on the display module (160). In addition, the client module (151) can output the received result as audio through the audio output module (155).

[0060] In one embodiment, the client module (151) can receive a plan corresponding to the received user input. The client module (151) can display the results of executing multiple operations of the app according to the plan on the display module (160). For example, the client module (151) can sequentially display the results of executing multiple operations on the display and output audio through the audio output module (155). The electronic device (101) can, for example, display only some results of executing multiple operations (e.g., the result of the last operation) on the display module (160) and output audio through the audio output module (155).

[0061] According to one embodiment, the client module (151) may receive a request from the intelligent server (200) to obtain information necessary to produce a result corresponding to a user input. According to one embodiment, the client module (151) may transmit the necessary information to the intelligent server (200) in response to the request.

[0062] The client module (151) of one embodiment can transmit result information of executing multiple operations according to a plan to the intelligent server (200). The intelligent server (200) can use the result information to confirm that the received user input has been processed correctly.

[0063] The client module (151) of one embodiment may include a voice recognition module. According to one embodiment, the client module (151) may recognize voice inputs that perform limited functions through the voice recognition module. For example, the client module (151) may execute an intelligent app to process voice inputs to perform organic actions through designated inputs (e.g., "Wake up!").

[0064] An intelligent server (200) of one embodiment can receive information related to a user voice input from an electronic device (101) via a communication network. According to one embodiment, the intelligent server (200) can change data related to the received voice input into text data. According to one embodiment, the intelligent server (200) can generate a plan for performing a task corresponding to the user voice input based on the text data.

[0065] In one embodiment, the plan may be generated by an artificial intelligence (AI) system. The AI ​​system may be a rule-based system, a neural network-based system (e.g., a feedforward neural network (FNN) or a recurrent neural network (RNN)), or a combination of the above or another AI system. In one embodiment, the plan may be selected from a set of predefined plans or may be generated in real time in response to a user request. For example, the AI ​​system may select at least one plan from a plurality of predefined plans.

[0066] An intelligent server (200) of one embodiment may transmit a result according to a generated plan to an electronic device (101), or transmit the generated plan to the electronic device (101). According to one embodiment, the electronic device (101) may display a result according to the plan on a display module (160). According to one embodiment, the electronic device (101) may display a result of executing an operation according to the plan on a display module (160).

[0067] An intelligent server (200) of one embodiment may include a front end (210), a natural language platform (220), a capsule database (230), an execution engine (240), an end user interface (250), a management platform (260), a big data platform (270), or an analytic platform (280).

[0068] The front end (210) of one embodiment can receive user input from the electronic device (101). The front end (210) can transmit a response corresponding to the user input.

[0069] According to one embodiment, the natural language platform (220) may include an automatic speech recognition module (ASR module) (221), a natural language understanding module (NLU module) (223), a planner module (225), a natural language generator module (NLG module) (227), or a text to speech module (TTS module) (229).

[0070] The automatic speech recognition module (221) of one embodiment can convert voice input received from the electronic device (101) into text data. The natural language understanding module (223) of one embodiment can identify the user's intention using the text data of the voice input. For example, the natural language understanding module (223) can identify the user's intention by performing syntactic analysis or semantic analysis on user input in the form of text data. The natural language understanding module (223) of one embodiment can identify the meaning of words extracted from the user input using linguistic features (e.g., grammatical elements) of morphemes or phrases, and can determine the user's intention by matching the meaning of the identified words to the intent. The natural language understanding module (223) can obtain intent information corresponding to the user's utterance. The intent information may be information indicating the user's intent determined by interpreting text data. The intent information may include information indicating an action or function that the user intends to execute using the device.

[0071] The planner module (225) of one embodiment can generate a plan using the intent and parameters determined by the natural language understanding module (223). According to one embodiment, the planner module (225) can determine a plurality of domains necessary to perform a task based on the determined intent. The planner module (225) can determine a plurality of operations included in each of the plurality of domains determined based on the intent. According to one embodiment, the planner module (225) can determine parameters necessary to execute the determined plurality of operations or result values ​​output by the execution of the plurality of operations. The parameters and the result values ​​can be defined as concepts of a specified format (or class). Accordingly, the plan can include a plurality of operations and a plurality of concepts determined by the user's intent. The planner module (225) can determine the relationship between the plurality of operations and the plurality of concepts in a stepwise (or hierarchical) manner. For example, the planner module (225) can determine the execution order of a plurality of actions based on the user's intention based on a plurality of concepts. In other words, the planner module (225) can determine the execution order of a plurality of actions based on parameters required for the execution of the plurality of actions and results output by the execution of the plurality of actions. Accordingly, the planner module (225) can generate a plan including association information (e.g., ontology) between the plurality of actions and the plurality of concepts. The planner module (225) can generate the plan using information stored in a capsule database (230) in which a set of relationships between concepts and actions is stored.

[0072] The natural language generation module (227) of one embodiment can convert specified information into text format. The information converted into text format may be in the form of natural language speech. The text-to-speech conversion module (229) of one embodiment can convert information in text format into information in speech format.

[0073] According to one embodiment, some or all of the functions of the natural language platform (220) may also be implemented in the electronic device (101).

[0074] The capsule database (230) may store information on the relationships between a plurality of concepts and actions corresponding to a plurality of domains. According to one embodiment, a capsule may include a plurality of action objects (or action information) and concept objects (or concept information) included in a plan. According to one embodiment, the capsule database (230) may store a plurality of capsules in the form of a concept action network (CAN). According to one embodiment, the plurality of capsules may be stored in a function registry included in the capsule database (230).

[0075] The capsule database (230) may include a strategy registry that stores strategy information required when determining a plan corresponding to a voice input. The strategy information may include reference information for determining a single plan when there are multiple plans corresponding to a user input. According to one embodiment, the capsule database (230) may include a follow-up registry that stores information on follow-up actions for suggesting follow-up actions to a user in a given situation. The follow-up actions may include, for example, follow-up utterances. According to one embodiment, the capsule database (230) may include a layout registry that stores layout information of information output through the electronic device (101). According to one embodiment, the capsule database (230) may include a vocabulary registry that stores vocabulary information included in capsule information. According to one embodiment, the capsule database (230) may include a dialog registry that stores information on dialogue (or interaction) with a user. The capsule database (230) may update stored objects through a developer tool. The developer tool may include, for example, a function editor for updating action objects or concept objects. The developer tool may include a vocabulary editor for updating vocabulary. The developer tool may include a strategy editor for creating and registering strategies that determine plans. The developer tool may include a dialog editor for creating dialogues with the user.The developer tool may include a follow-up editor that activates follow-up goals and allows editing of follow-up utterances that provide hints. The follow-up goals may be determined based on currently set goals, user preferences, or environmental conditions. In one embodiment, the capsule database (230) may also be implemented within the electronic device (101).

[0076] The execution engine (240) of one embodiment can produce a result using the generated plan. The end user interface (250) can transmit the produced result to the electronic device (101). Accordingly, the electronic device (101) can receive the result and provide the received result to the user. The management platform (260) of one embodiment can manage information used in the intelligent server (200). The big data platform (270) of one embodiment can collect user data. The analysis platform (280) of one embodiment can manage the quality of service (QoS) of the intelligent server (200). For example, the analysis platform (280) can manage the components and processing speed (or efficiency) of the intelligent server (200).

[0077] A service server (300) of one embodiment may include CP service A (301), CP service B (302), and CP service C. The service server (300) of one embodiment may provide a service (e.g., food ordering or hotel reservation) specified for an electronic device (101). According to one embodiment, the service server (300) may be a server operated by a third party. The service server (300) of one embodiment may provide information for generating a plan corresponding to a received user input to the intelligent server (200). The provided information may be stored in the capsule database (230). In addition, the service server (300) may provide result information according to the plan to the intelligent server (200).

[0078] In the integrated intelligence system (10) described above, the electronic device (101) can provide various intelligent services to the user in response to user input. The user input may include, for example, input via a physical button, touch input, or voice input.

[0079] In one embodiment, the electronic device (101) may provide a voice recognition service through an intelligent app (or voice recognition app) stored internally. In this case, for example, the electronic device (101) may recognize a user utterance or voice input received through the microphone and provide the user with a service corresponding to the recognized voice input.

[0080] In one embodiment, the electronic device (101) may perform a designated operation, either alone or in conjunction with the intelligent server and / or service server, based on the received voice input. For example, the electronic device (101) may execute an app corresponding to the received voice input and perform a designated operation through the executed app.

[0081] In one embodiment, when an electronic device (101) provides a service together with an intelligent server (200) and / or a service server, the user terminal may detect a user utterance using the input module (150) and generate a signal (or voice data) corresponding to the detected user utterance. The user terminal may transmit the voice data to the intelligent server (200) using the interface (177).

[0082] According to one embodiment, an intelligent server (200) may generate a plan for performing a task corresponding to a voice input received from an electronic device (101), or a result of performing an operation according to the plan. The plan may include, for example, a plurality of operations for performing a task corresponding to a user's voice input, and a plurality of concepts related to the plurality of operations. The concept may define parameters input to the execution of the plurality of operations, or result values ​​output by the execution of the plurality of operations. The plan may include association information between the plurality of operations and the plurality of concepts.

[0083] An electronic device (101) of one embodiment can receive the response using an interface (177). The electronic device (101) can output a voice signal generated within the electronic device (101) to the outside using the audio output module (155), or can output an image generated within the electronic device (101) to the outside using the display module (160).

[0084]

[0085] FIG. 3 is a diagram showing a form in which relationship information between concepts and actions is stored in a database according to various embodiments.

[0086] The capsule database (e.g., the capsule database (230) of FIG. 2) of the intelligent server (e.g., the intelligent server (200) of FIG. 2) may store capsules in the form of a CAN (concept action network). The capsule database may store operations for processing tasks corresponding to a user's voice input and parameters necessary for the operations in the form of a CAN (concept action network).

[0087] The capsule database may store a plurality of capsules (capsule (A) (401), capsule (B) (404)) corresponding to each of a plurality of domains (e.g., applications). According to one embodiment, one capsule (e.g., capsule (A) (401)) may correspond to one domain (e.g., location (geo), application). In addition, one capsule may correspond to at least one service provider (e.g., CP 1 (402), CP 2 (403), CP 3 (406), or CP 4 (405)) for performing a function for a domain related to the capsule. According to one embodiment, one capsule may include at least one operation (410) and at least one concept (420) for performing a specified function.

[0088] The above natural language platform (e.g., the natural language platform (220) of FIG. 2) can generate a plan for performing a task corresponding to a received voice input using capsules stored in a capsule database. For example, the planner module of the natural language platform (e.g., the planner module (225) of FIG. 2) can generate a plan using capsules stored in a capsule database. For example, a plan (407) can be generated using actions (4011, 4013) and concepts (4012, 4014) of capsule A (410) and actions (4041) and concepts (4042) of capsule B (404).

[0089]

[0090] FIG. 4 is a diagram showing a screen for processing voice input received through an intelligent app by a user terminal according to various embodiments.

[0091] The electronic device (101) can execute an intelligent app to process user input through an intelligent server (e.g., the intelligent server (200) of FIG. 2).

[0092] According to one embodiment, on screen 310, when the electronic device (101) recognizes a designated voice input (e.g., wake up!) or receives an input via a hardware key (e.g., a dedicated hardware key), the electronic device (101) may execute an intelligent app for processing the voice input. For example, the electronic device (101) may execute an intelligent app while executing a schedule app. According to one embodiment, the electronic device (101) may display an object (e.g., an icon) (311) corresponding to the intelligent app on the display module (160). According to one embodiment, the electronic device (101) may receive a voice input by a user's speech. For example, the electronic device (101) may receive a voice input such as "Tell me my schedule for this week!" According to one embodiment, the electronic device (101) can display a user interface (UI) (313) (e.g., an input window) of an intelligent app displaying text data of received voice input on a display module (e.g., the display module (160) of FIG. 2).

[0093] According to one embodiment, on the 320 screen, the electronic device (101) can display a result corresponding to the received voice input on the display module (160). For example, the electronic device (101) can receive a plan corresponding to the received user input and display 'this week's schedule' on the display module (160) according to the plan.

[0094]

[0095] FIG. 5 is a flowchart of an operation of a ignition processing method performed by an electronic device (e.g., the electronic device (101) of FIG. 1 and FIG. 2) according to various embodiments.

[0096] In the following examples, the operations may be performed sequentially, but are not necessarily sequential. For example, the order of the operations may be changed, and at least two operations may be performed in parallel.

[0097] According to one embodiment, operations (510) to (550) of FIG. 5 may be understood to be performed in a processor (120) (e.g., processor (120) of FIG. 1, FIG. 2) of an electronic device (101) (e.g., electronic device (101) of FIG. 1, FIG. 2).

[0098] When a user-defined wake-up word (or wake-up word) is registered with a user-defined word, various error checks can be performed on each user input (e.g., each received utterance). For example, the electronic device (101) can perform checks on the number of syllables in a word, ambient noise conditions, the size of a voice signal, and the speaking speed. For example, if the check result corresponds to a set condition (e.g., when the number of syllables in a word is less than a set number, when the ambient noise conditions exceed a set value, when the size of a voice signal is less than a set value, when the speaking speed is outside a set speed range), the electronic device (101) can determine that there is an error in the utterance.

[0099] If an error is determined in the utterance, the electronic device (101) can provide appropriate feedback to the user. The user can input additional utterances based on the feedback, and the electronic device (101) can use the additional utterances to train a model for recognizing the user-specified wake word.

[0100] For example, the electronic device (101) may receive multiple utterances from the user via the voice input device in operation (510). For example, the electronic device (101) may provide the user with an interface (e.g., displaying a screen on a display module (e.g., display module (160) of FIG. 1), outputting sound using an audio output module (e.g., audio output module (155) of FIG. 1)) that requests the user to utter a custom wake-up word (or custom wake-up word) a set number of times.

[0101] For example, the electronic device (101) can generate a plurality of utterance vectors corresponding to a plurality of utterances in operation (520).

[0102] For example, the electronic device (101) may generate a plurality of utterance vectors corresponding to a plurality of utterances using a base model. For example, the base model may represent a model trained as a word classifier. The base model may be trained to classify words using audio (or utterance, or voice signal) containing words and labels associated with the words. The number of output nodes in the base model may be determined based on the number of words to be classified.

[0103] For example, the base model could be a statistical model based on artificial neural networks. The base model could output an utterance vector containing numerical values ​​that compress the linguistic and acoustic features of the input speech signal.

[0104] For example, a basic model can transform audio input (or utterance) into an embedding vector (or fixed-length vector). For example, the audio input can be represented in the time-frequency domain as mel-frequency cepstral coefficients (MFCCs).

[0105] For example, an utterance vector can be understood as an embedded, compressed representation (or vector) of a fixed length. For example, an utterance vector can represent a fixed-length vector output from the layer immediately preceding the output layer (or final layer) of the base model, inputting the utterance into the base model.

[0106] For example, the electronic device (101) can input each of a plurality of utterances into a basic model and generate a plurality of utterance vectors corresponding to each of the plurality of utterances.

[0107] [Mathematical Formula 1]

[0108]

[0109] [Equation 2]

[0110]

[0111] The above mathematical formulas 1 and 2 are merely examples to aid understanding and are not limited thereto, and can be modified, applied, or expanded in various ways.

[0112] The above mathematical expression 2 is the utterance of the T frame of mathematical expression 1. as the basic model The utterance vector of the i-th utterance transformed (or processed) according to represents the ignition vector according to mathematical formula 2 can be understood as representing user utterance in an N-dimensional space. For example, N can represent the number of elements of the utterance vector. Here, It could be. is a matrix containing F-dimensional frequency information (e.g., MFCC) of T frames, can be expressed as an N-dimensional vector.

[0113] In the above mathematical expressions 1 and 2, i is the index of the utterance, N is a value set according to the desired base model size, and can be set to various values, for example, 64 for a very small base model and 1024 for a very large base model. F can represent the number of coefficients of MFCC, and can be, for example, 13, 26, or 39. T can represent the temporal context or number of frames of audio provided as input to the base model.

[0114] For example, the electronic device (101) can convert a plurality of utterances to generate a plurality of utterance vectors according to the above mathematical equations 1 and 2.

[0115] For example, the electronic device (101) can calculate a first similarity between each of a plurality of ignition vectors and each of a plurality of reference vectors stored in the memory (130) in operation (530).

[0116] For example, multiple reference vectors may represent vectors stored for comparison with multiple utterance vectors. Multiple reference vectors may represent vectors embedded by inputting everyday noises or commonly used words (e.g., greetings, people's names, pronouns) into a reference model. For example, multiple reference vectors may represent vectors embedded using a reference model of sounds that can occur in everyday life, such as human voices, laughter, coughs, animal sounds, the sound of a door closing, and environmental noise.

[0117] The sound corresponding to the reference vector is not limited to the examples above, and may include various sounds such as the sound of running water, the sound of a vacuum cleaner, the sound of a dishwasher, the sound of birds, living noise, and traffic noise.

[0118] For example, the reference vector can be embedded from a recording (or audio, sound) of a sound category, or can be computed from multiple recordings by a specific summary function such as mean, weighted mean, or median.

[0119] [Equation 3]

[0120]

[0121] For example, similarity Is th utterance vector and The th reference vector can be defined as the inner product of .

[0122] The above mathematical formula 3 is merely an example to aid understanding and is not limited thereto, and can be modified, applied, or expanded in various ways.

[0123] For example, the electronic device (101) can calculate the first similarity between each of the plurality of ignition vectors and each of the plurality of reference vectors, as in the above mathematical expression 3.

[0124] In the above mathematical expression 3, among the N utterance vectors is the i-th utterance vector, among the M reference vectors. can represent the jth reference vector. can represent the first similarity between the ith utterance vector and the jth reference vector.

[0125] For example, the electronic device (101) can calculate the first similarity of each of the plurality of ignition vectors and each of the plurality of reference vectors in real time as in the above mathematical expression 3.

[0126] For example, the electronic device (101) can calculate the first similarity using the inner product, cosine distance, Manhattan distance, and Euclidean distance of each of the plurality of ignition vectors and each of the plurality of reference vectors.

[0127] For example, if similarity is defined as the distance between an utterance vector and a reference vector (e.g., Manhattan distance, Euclidean distance, etc.), the first similarity may have a smaller value as the distance between the two vectors is closer, and a larger value as the distance between the two vectors is farther. When comparing similarity with a reference value (or threshold), the direction of the inequality between the reference value and the similarity may change depending on the method of calculating the similarity and / or the reference.

[0128] For example, the electronic device (101) can determine the first similarity as the inner product value of the ignition vector and the reference vector is larger. For example, the electronic device (101) can determine the first similarity as the distance between the ignition vector and the reference vector is smaller.

[0129] For example, the electronic device (101) may determine learning utterance vectors for training a first model for identifying a user-specified call word among a plurality of utterance vectors based on the first similarity in operation (540).

[0130] For example, the electronic device (101) may determine, among a plurality of utterance vectors, utterance vectors other than utterance vectors whose first similarity with at least one of the plurality of reference vectors is greater than or equal to a set threshold value as utterance vectors for learning.

[0131] For example, the electronic device (101) can train the first model using learning utterance vectors in the operation (550).

[0132] [Equation 4]

[0133]

[0134] The above mathematical formula 4 is merely an example to aid understanding and is not limited thereto, and can be modified, applied, or expanded in various ways.

[0135] For example, the electronic device (101) can learn (or register, or model) the first model as an average of learning utterance vectors, as in the above mathematical expression 4.

[0136] If the input audio quality is high, the electronic device (101) can check the quality of the input audio, assuming that the performance of the model for recognizing the custom call word is high.

[0137] Even when training a model by collecting high-quality utterances under ideal conditions, false acceptance or FA (false accept) issues can occur. For example, if the utterances collected from a user (or a registered custom wake word) are words commonly used in everyday life, the electronic device (101) may have trouble recognizing the custom wake word even if the user does not utter the registered custom wake word.

[0138] In relation to the above FA problem, the electronic device (101) can compare a plurality of utterance vectors with a plurality of reference vectors to determine a plurality of first similarities, and determine learning utterance vectors according to the plurality of first similarities. By not including an utterance vector determined to be similar to at least one of the plurality of reference vectors (or having a first similarity with at least one of the plurality of reference vectors greater than a threshold value) in the learning utterance vector, the electronic device (101) can reduce and / or alleviate the probability of occurrence of the FA problem and improve the user experience.

[0139] For example, the electronic device (101) can identify a user-specified wake word included in an utterance received from a user (after the first model is trained) using the trained first model. For example, the electronic device (101) can process the received utterance using the basic model. The electronic device (101) can embed the received utterance to generate an utterance vector. For example, the electronic device (101) can compare the utterance vector with the first model to determine a score. For example, the score can indicate the degree to which the utterance vector is close to the first model (or a modeled vector of the first model) (e.g., angular proximity in the case of cosine distance, linear proximity in the case of Euclidean distance or Manhattan distance). The closer the utterance vector is to the first model, the higher the score can be. For example, the score can be normalized to a value greater than or equal to 0 and less than or equal to 1.

[0140] For example, the electronic device (101) can determine a score by comparing the vector of the first model with the utterance vector. The vector of the first model can represent, for example, an average vector of the utterance vectors for learning.

[0141] For example, when a user's speech is received in real time, the electronic device (101) may generate a plurality of audio frames according to a sliding window. The electronic device (101) may embed the plurality of audio frames using a basic model, as in Equations 1 and 2, and generate a plurality of speech vectors. The electronic device (101) may compare the plurality of speech vectors generated according to each sliding window with a first model (or a vector of the first model) to calculate a plurality of scores. The electronic device (101) may determine whether the speech received from the user includes a user-specified wake word using a weighted average of the plurality of scores calculated according to each sliding window.

[0142] With regard to the operation of determining whether the utterance received from the user includes a user-specified call word, the electronic device (101) is not limited to the above-described examples and can use various methods to determine whether the utterance received from the user includes a user-specified call word.

[0143]

[0144] FIG. 6 is a flowchart illustrating an operation of an electronic device (e.g., the electronic device (101) of FIG. 1 and FIG. 2) according to various embodiments providing an interface according to a first similarity.

[0145] In the following examples, the operations may be performed sequentially, but are not necessarily sequential. For example, the order of the operations may be changed, and at least two operations may be performed in parallel.

[0146] According to one embodiment, operations (610) to (630) of FIG. 6 may be understood to be performed by a processor (120) of an electronic device (101) (e.g., the processor (120) of FIG. 1, FIG. 2).

[0147] For example, after operation (530) of FIG. 5, the electronic device (101) may compare the first similarity with the first threshold value in operation (610). The electronic device (101) may determine whether the first similarity is greater than or equal to the first threshold value in operation (610). For example, operation (610) is an example of an operation of the electronic device (101) when the first similarity is calculated as the inner product of the utterance vector and the reference vector.

[0148] Operation (610) is an example of comparing the first similarity with the first threshold value, and is not limited to operation (610) of FIG. 6. For example, when determining the first similarity based on the distance between each of a plurality of utterance vectors and each of a plurality of reference vectors, the electronic device (101) can determine whether the first similarity is less than or equal to the first threshold value.

[0149] For example, if the first similarity is determined to be greater than or equal to the first threshold value in operation (610), the electronic device (101) may provide an interface requesting additional utterance from the user in operation (620).

[0150] For example, an interface requesting additional utterances may include a type of reference vector whose first similarity is greater than or equal to a first threshold.

[0151] For example, the type of a reference vector can indicate the sound label for generating the reference vectors. The types of reference vectors can be categorized into, for example, commonly used words, everyday noises, and living noises.

[0152] For example, if the first similarity between the utterance vector and the reference vector is greater than or equal to a first threshold value and the reference vector is an embedded vector corresponding to a sound labeled as a commonly used word, the electronic device (101) may provide the user with an interface including the type (or label) of the reference vector.

[0153] Users can check what kind of sounds their custom wake word (or entered utterance) resembles.

[0154] For example, the electronic device (101) may generate a plurality of utterance vectors corresponding to additional utterances received in response to the interface in operation (630).

[0155] For example, the electronic device (101) may generate a plurality of utterance vectors corresponding to additional utterances substantially in the same manner as generating a plurality of utterance vectors corresponding to the plurality of utterances in operation (520) of FIG. 5.

[0156] For example, the electronic device (101) can perform operation (530) of FIG. 5 after operation (630). The electronic device (101) can perform operation (530), operation (610) to operation (630) of FIG. 5 using a plurality of ignition vectors corresponding to additional ignition.

[0157] For example, when performing operation (530) after operation (630), the electronic device (101) can calculate a first similarity between each of a plurality of utterance vectors corresponding to additional utterances and each of a plurality of reference vectors.

[0158] For example, when performing operation (530) of FIG. 5 after operation (630), the electronic device (101) can calculate a first similarity between each of a plurality of utterance vectors corresponding to additional utterances and a plurality of reference vectors.

[0159] For example, when performing operation (540) of FIG. 5 after operation (630), the electronic device (101) can determine a learning utterance vector from among a plurality of utterance vectors corresponding to a plurality of previously received utterances and an additional utterance based on the first similarity.

[0160] For example, the electronic device (101) may compare the number of utterance vectors having a first similarity greater than or equal to a first threshold value with a set number of utterances (e.g., 5). For example, if the number of utterance vectors having a first similarity greater than or equal to the first threshold value is greater than or equal to the set number of utterances, the electronic device (101) may perform operation (540) of FIG. 5 after operation (630).

[0161] For example, if it is determined that the first similarity is less than the first threshold value in operation (610), the electronic device (101) may perform operation (540) of FIG. 5.

[0162]

[0163] Fig. 7 is a diagram illustrating a plurality of ignition vectors (710, 720) and a plurality of reference vectors (730, 740, 750) according to various embodiments. Fig. 7 is an example of representing an N-dimensional embedding space as a two-dimensional embedding space.

[0164] Referring to FIG. 7, the electronic device (101) can determine a first similarity based on the distance between each of the plurality of ignition vectors (710, 720) and each of the plurality of reference vectors (730, 740, 750).

[0165] For example, in FIG. 7, the distance between the first ignition vector (710) and the first reference vectors (730) may be within a set threshold value. For example, the distance between the second ignition vector (720) and each of the first reference vectors (730), the second reference vectors (740), and the third reference vectors (750) may exceed the set threshold value.

[0166] In FIG. 7, since the distance between the first utterance vector (710) and the first reference vectors (730) is within a set threshold value (or, the first similarity is greater than or equal to the first threshold value), the electronic device (101) can exclude the first utterance vector (710) from the learning utterance vectors.

[0167] For example, since the distance between the first utterance vector (710) and the first reference vectors (730) is within a set threshold value, the electronic device (101) may provide an interface for requesting additional utterance from the user.

[0168] For example, the electronic device (101) may provide an interface including the type of the first reference vectors (730). For example, if the first reference vectors (730) are vectors generated by embedding sounds classified as vacuum cleaner noise (or sounds labeled as vacuum cleaner noise), the electronic device (101) may provide the user with information through the interface that the utterance corresponding to the first utterance vector (710) is similar to the vacuum cleaner noise.

[0169] In FIG. 7, since the distance of the second utterance vector (720) from each of the first reference vectors (730), the second reference vectors (740), and the third reference vectors (750) is greater than or equal to a set threshold value (or the first similarity is less than the first threshold value), the electronic device (101) can determine the second utterance vector (720) as a learning utterance vector.

[0170] When calculating the first similarity based on the distance between the utterance vector and the reference vector, a smaller distance can be judged to indicate a higher first similarity. A distance between the utterance vector and the reference vector that is within a set threshold can be understood as being substantially equivalent to a first similarity that is greater than or equal to the set first threshold.

[0171] FIG. 7 is an example of a two-dimensional embedding space in which each of the reference vectors (730, 740, 750) is grouped according to a label, and the types of the reference vectors (730, 740, 750) grouped according to the label are determined. The embodiment illustrated in FIG. 7 is not limited, and the types of the plurality of reference vectors (730, 740, 750) may be determined according to various methods.

[0172] For example, the electronic device (101) can cluster a plurality of reference vectors in an N-dimensional embedding space. The electronic device (101) can determine the type of each cluster based on the labels of the reference vectors included in each cluster. For example, the electronic device (101) can determine the label with the highest frequency among the labels of the reference vectors included in each cluster as the type of the cluster. For example, the electronic device (101) can determine the type of the cluster by applying a weight according to the distance between the labels of the reference vectors included in each cluster and the corresponding reference vector from the center of the cluster.

[0173]

[0174] FIG. 8 is a flowchart illustrating an operation of an electronic device (e.g., the electronic device (101) of FIG. 1 and FIG. 2) according to various embodiments providing an interface according to a second similarity.

[0175] In the following examples, the operations may be performed sequentially, but are not necessarily sequential. For example, the order of the operations may be changed, and at least two operations may be performed in parallel.

[0176] According to one embodiment, operations (810) to (850) of FIG. 8 may be understood to be performed by a processor (120) of an electronic device (101) (e.g., the processor (120) of FIG. 1 and FIG. 2).

[0177] For example, the electronic device (101) may calculate a second similarity between the plurality of utterance vectors in operation (810) after operation (530) of FIG. 5. For example, the electronic device (101) may calculate the second similarity between the plurality of utterance vectors in substantially the same manner as calculating the first similarity between each of the plurality of utterance vectors and each of the plurality of reference vectors in operation (530) of FIG. 5.

[0178] For example, the electronic device (101) may compare the second similarity with the second threshold in operation (820). For example, the electronic device (101) may determine whether the second similarity exceeds (or is greater than) the second threshold.

[0179] For example, if it is determined that the second similarity exceeds the second threshold value in operation (820), the electronic device (101) may perform operation (540) of FIG. 5.

[0180] For example, if it is determined in operation (820) that the second similarity is less than or equal to the second threshold, the electronic device (101) may provide an interface for the user to check in operation (830) whether to include an utterance vector among a plurality of utterance vectors having a second similarity less than or equal to the second threshold in the utterance vectors for learning.

[0181] For example, the electronic device (101) may train a second model using an utterance vector having a second similarity less than or equal to a second threshold based on a user input received through the interface in operation (840). For example, if the user decides to include an utterance vector having a second similarity less than or equal to the second threshold in the utterance vector for training, the electronic device (101) may train a second model separate from the first model (or model the second model, register the second model) using the utterance vector.

[0182] For example, the electronic device (101) can train the first model using an utterance vector whose second similarity exceeds a second threshold value in operation (850).

[0183] FIG. 8 is a diagram illustrating an example of performing an operation of determining a learning utterance vector among a plurality of utterance vectors according to a second similarity after operation (530) among various embodiments. The operations of FIG. 8 are not limited to being performed after operation (530) as in FIG. 8. For example, the electronic device (101) may perform operations (810) to (850) after operation (520) of FIG. 5.

[0184] The electronic device (101) can use the learned second model to determine whether an utterance received from the user includes a user-specified call word. The description of the operation of determining whether an utterance received from the user includes a user-specified call word using the first model can be substantially identically applied to the operation of determining whether an utterance received from the user includes a user-specified call word using the second model.

[0185]

[0186] FIG. 9 is a diagram illustrating a plurality of utterance vectors (910, 920, 930) according to various embodiments. FIG. 10 is a diagram illustrating a plurality of utterance vectors (910, 920, 930) for training a first model and a second model determined based on user input according to various embodiments. FIG. 11 is a diagram illustrating a plurality of utterance vectors (910, 930) for training a first model determined based on user input according to various embodiments.

[0187] For example, the electronic device (101) can calculate a second similarity between a plurality of utterance vectors (910, 920, 930). For example, the electronic device (101) can calculate the second similarity using an inner product or distance between the plurality of utterance vectors (910, 920, 930). With respect to the operation of calculating the second similarity between the plurality of utterance vectors (910, 920, 930), the description of the operation of calculating the first similarity between each of the plurality of utterance vectors (910, 920, 930) and the plurality of reference vectors (e.g., the first reference vectors (730), the second reference vectors (740), and the third reference vectors (750) of FIG. 7) can be substantially identically applied.

[0188] Figure 9 is an example showing a two-dimensional embedding space (900).

[0189] In the embedding space (900), the distance between the first utterance vector (910) and the utterance vectors (930) may be within a set threshold value. For example, the second similarity between the first utterance vector (910) and the utterance vectors (930) may be greater than or equal to the set threshold value.

[0190] In the embedding space (900), the distance between the second utterance vector (920) and the utterance vectors (930) may be greater than or equal to a set threshold. For example, the second similarity between the second utterance vector (920) and the utterance vectors (930) may be less than a set threshold.

[0191] If the second similarity between the first utterance vector (910) and at least one of the respective utterance vectors (930) is greater than or equal to a set threshold value, the electronic device (101) can determine the first utterance vector (910) as a learning utterance vector.

[0192] For example, the electronic device (101) may determine a learning utterance vector from among a plurality of utterance vectors (910, 920, 930) based on the first similarity and the second similarity. For example, the electronic device (101) may determine an utterance vector in which a first similarity between the utterance vector (910, 920, 930) and each of the plurality of reference vectors (730, 740, 750) is less than a set first threshold value, and a second similarity between the utterance vector and each of the remaining plurality of utterance vectors is greater than or equal to the second threshold value, as a learning utterance vector.

[0193] If the second similarity of at least one of the second utterance vectors (920) and each of the utterance vectors (930) is less than a set threshold, the electronic device (101) may provide an interface for the user to check whether to include the utterance vector (920) among the plurality of utterance vectors (910, 920, 930) whose second similarity is less than the set threshold in the utterance vectors for learning.

[0194] For example, the electronic device (101) may provide a user with an interface to determine whether to include the second utterance vector (920) in the learning utterance vectors.

[0195] FIG. 10 is an example showing an embedding space (1000) when an electronic device (101) receives an input from a user to include the second utterance vector (920) of FIG. 9 in a learning utterance vector.

[0196] For example, the electronic device (101) can train the first model using the first utterance vector (910) and utterance vectors (930). For example, the vector (1010) can represent the learned result of the first model (e.g., the average of the learning utterance vectors for training the first model). For example, the electronic device (101) can determine the vector (1010) using the average of the first utterance vector (910) and utterance vectors (930).

[0197] For example, the electronic device (101) can train a second model using a second utterance vector (920).

[0198] During the process of registering a custom wake word, the electronic device (101) may receive utterances (e.g., three to five) from the user. If there is a significant variation between multiple user utterances (910, 920, 930), the model for recognizing the custom wake word may not be properly trained.

[0199] For example, if a word that is pronounced with two or more different pronunciations, intonations, and speeds is registered as a custom trigger word, even if the user utters the same word, at least one of the similarities between multiple utterance vectors may be low.

[0200] For example, in FIG. 10, when training the first model using the first utterance vector (910), the second utterance vector (920), and the utterance vectors (930), the performance of the first model may be reduced because the similarity between the second utterance vector (920) and the first utterance vector (910) and the utterance vectors (930) is low. For example, when modeling the first model as the average of the first utterance vector (910), the second utterance vector (920), and the utterance vectors (930), the modeled vector of the first model may be significantly different from the first utterance vector (910), the second utterance vector (920), and the utterance vectors (930).

[0201] The electronic device (101) can register a user-specified wake word suitable for the user's intention by training a first model using a first utterance vector (910) and utterance vectors (930) with high similarity and training a second model using a second utterance vector (920) with low similarity. The electronic device (101) can recognize a user-specified wake word using the trained first model and second model.

[0202] FIG. 11 is an example of an embedding space (1100) when an electronic device (101) receives an input from a user not to include the second utterance vector (920) of FIG. 9 in the learning utterance vector.

[0203] The electronic device (101) can train the first model using the remaining utterance vectors (e.g., the first utterance vector (910), utterance vectors (930)) excluding the second utterance vector (920) of FIG. 9. For example, the learning result of the first model may be a vector (1010). For example, the electronic device (101) can model the first model as a vector (1010).

[0204] Referring to FIGS. 9 to 11, the electronic device (101) can cluster a plurality of utterance vectors. For example, the electronic device (101) can use various clustering algorithms (e.g., LBG (linde-buzo-gray), K-means) to classify the plurality of utterance vectors into a plurality of clusters. The electronic device (101) can train a model corresponding to each cluster using the average of the utterance vectors included in each cluster.

[0205]

[0206] FIGS. 12, 13, and 14 are diagrams illustrating interfaces provided to a user by an electronic device (e.g., the electronic device (101) of FIGS. 1 and 2) according to various embodiments.

[0207] FIG. 12 is a drawing showing a screen (1200) provided when an electronic device (101) according to various embodiments detects an error in a speech received from a user.

[0208] For example, the electronic device (101) can detect errors (or errors) in received speech. For example, the electronic device (101) can detect errors in speech by comparing information about the speech, such as the number of syllables in a word included in the speech, ambient noise conditions, voice volume, and speech rate, with set conditions.

[0209] When an error is detected in which the ambient noise is higher than the set condition and the voice volume is smaller than the set condition, the electronic device (101) can provide an interface (1210) including information about the detected error.

[0210] The interface (1210) shown in Fig. 12, the detected error is exemplary and is not limited to the above-described example, and errors can be detected using various speech information (e.g., information on a voice signal corresponding to speech) and the detected error can be provided through the interface.

[0211] FIG. 13 is a drawing showing a screen (1300) that provides an interface (1310) for requesting an additional utterance to a user when the first similarity of each of a plurality of utterance vectors and each of a plurality of reference vectors is higher than a set threshold value, according to various embodiments of the present invention.

[0212] Referring to FIG. 13, the electronic device (101) may provide an interface including a type of a reference vector having a first similarity higher than a set threshold value. For example, if the label of a reference vector having a first similarity higher than the threshold value (or a category or classification of sounds corresponding to the reference vector) is 'frequently used words', the electronic device (101) may provide an interface (1310) including a type of the reference vector.

[0213] FIG. 14 is a screen (1400) showing an interface (1410) provided to a user by an electronic device (101) according to various embodiments when a second similarity between a plurality of ignition vectors is below a set threshold value.

[0214] For example, the electronic device (101) may sequentially receive a first utterance, a second utterance, and a third utterance from the user. A second similarity between an utterance vector corresponding to the second utterance and the utterance vector corresponding to the first utterance may be higher than a threshold. If the second similarity between an utterance vector corresponding to the third utterance and the utterance vectors corresponding to the first utterance and / or the second utterance is lower than the threshold, the electronic device (101) may provide the user with an interface (1410) for determining whether to include the utterance vector corresponding to the third utterance as a learning utterance vector, such as the interface (1410).

[0215]

[0216] An electronic device (e.g., an electronic device (101) of FIGS. 1 and 2) according to various embodiments may include a processor (e.g., a processor (120) of FIGS. 1 and 2) and a memory (e.g., a memory (130) of FIGS. 1 and 2) electrically connected to the processor (120) and storing instructions executable by the processor (120). The processor (120) may receive a plurality of utterances from a user through a voice input device when the instructions are executed by the processor (120). The processor (120) may generate a plurality of utterance vectors (e.g., a first utterance vector (710) and a second utterance vector (720) of FIG. 7, a first utterance vector (910) and a second utterance vector (920) of FIG. 9) corresponding to the plurality of utterances. The processor (120) may calculate a first similarity between each of the plurality of utterance vectors and each of the plurality of reference vectors (e.g., the plurality of reference vectors (730, 740, 750) of FIG. 7) stored in the memory (130). Based on the first similarity, the processor (120) may determine learning utterance vectors for training a first model for identifying a user-specified call word among the plurality of utterance vectors (710, 720, 910, 920, 930). The processor (120) may train the model using the learning utterance vectors.

[0217] The processor (120) may provide an interface (e.g., interface (1310) of FIG. 13) for requesting additional utterances from the user when the first similarity is equal to or greater than a set first threshold value. The processor (120) may generate a plurality of utterance vectors (710, 720, 910, 920, 930) corresponding to the additional utterances received in response to the interface (1310).

[0218] The above processor (120) can provide the interface (1310) including the type of the reference vector whose first similarity is greater than or equal to the first threshold value.

[0219] The processor (120) can calculate a second similarity between the plurality of utterance vectors (710, 720, 910, 920, 930). The processor (120) can determine the learning utterance vectors based on the first similarity and the second similarity.

[0220] The processor (120) may provide the user with an interface (e.g., interface (1410) of FIG. 14) for checking whether to include an utterance vector among the plurality of utterance vectors (710, 720, 910, 920, 930) whose second similarity is less than or equal to the second threshold value in the learning utterance vectors, if the second similarity is less than or equal to the set second threshold value. The processor (120) may train a second model using an utterance vector whose second similarity is less than or equal to the second threshold value, based on a user input received through the interface (1410).

[0221] The processor (120) can determine the first similarity based on the distance between each of the plurality of ignition vectors (710, 720, 910, 920, 930) and each of the plurality of reference vectors (730, 740, 750).

[0222] An electronic device (101) according to various embodiments may include a processor (120) and a memory (130) electrically connected to the processor (120) and storing instructions executable by the processor (120). The processor (120) may receive a plurality of utterances from a user through a voice input device when the instructions are executed by the processor (120). The processor (120) may generate a plurality of utterance vectors (710, 720, 910, 920, 930) corresponding to the plurality of utterances. The processor (120) may calculate a first similarity between each of the plurality of utterance vectors (710, 720, 910, 920, 930) and each of the plurality of reference vectors (730, 740, 750) stored in the memory (130). The processor (120) may determine learning utterance vectors for training a first model for identifying a user-specified call word among the plurality of utterance vectors (710, 720, 910, 920, 930) based on the first similarity. The processor (120) may train the model using the learning utterance vectors. The processor (120) may provide a first interface (1310) for requesting an additional utterance to the user in a first case where the first similarity is equal to or greater than a set first threshold value. The first interface (1310) may include a type of a reference vector whose first similarity is equal to or greater than the first threshold value.

[0223] The processor (120) can calculate a second similarity between the plurality of utterance vectors (710, 720, 910, 920, 930). The processor (120) can determine the learning utterance vectors based on the first similarity and the second similarity.

[0224] The processor (120) may provide the user with a second interface (1410) to check whether an utterance vector having a second similarity less than or equal to the second threshold value among the plurality of utterance vectors (710, 720, 910, 920, 930) is to be included in the learning utterance vectors, if the second similarity is less than or equal to the set second threshold value. The processor (120) may train a second model using an utterance vector having a second similarity less than or equal to the second threshold value, based on a user input received through the second interface (1410).

[0225] The processor (120) can determine the first similarity based on the distance between each of the plurality of ignition vectors (710, 720, 910, 920, 930) and each of the plurality of reference vectors (730, 740, 750).

[0226] A method for processing speech according to various embodiments may include an operation of receiving a plurality of speeches from a user through a voice input device, an operation of generating a plurality of speech vectors (710, 720, 910, 920, 930) corresponding to the plurality of speeches, an operation of calculating a first similarity between each of the plurality of speech vectors (710, 720, 910, 920, 930) and a plurality of reference vectors (730, 740, 750) stored in a memory (130), an operation of determining learning speech vectors for training a first model for identifying a user-specified call word among the plurality of speech vectors (710, 720, 910, 920, 930) based on the first similarity, and an operation of training the first model using the learning speech vectors.

[0227] The above utterance processing method may further include an operation of providing an interface (1310) for requesting an additional utterance from the user when the first similarity is equal to or greater than a set first threshold value, and an operation of generating a plurality of utterance vectors (710, 720, 910, 920, 930) corresponding to the additional utterance received in response to the interface (1310).

[0228] The operation of providing the above interface may provide the interface (1310) including the type of the reference vector whose first similarity is greater than or equal to the first threshold value.

[0229] The above-described speech processing method may further include an operation of calculating a second similarity between the plurality of speech vectors (710, 720, 910, 920, 930). The operation of determining the speech vectors for learning may include an operation of determining the speech vectors for learning based on the first similarity and the second similarity.

[0230] The above-described speech processing method may further include an operation of providing an interface (1410) for confirming to the user whether an utterance vector having the second similarity less than or equal to the second threshold value among the plurality of utterance vectors (710, 720, 910, 920, 930) is to be included in the learning utterance vectors, if the second similarity is less than or equal to a set second threshold value. The operation of training the model may include an operation of training a second model using an utterance vector having the second similarity less than or equal to the second threshold value, based on a user input received through the interface (1410).

[0231] The operation of calculating the first similarity may include an operation of determining the first similarity based on the distance between each of the plurality of utterance vectors (710, 720, 910, 920, 930) and each of the plurality of reference vectors (730, 740, 750).

[0232] Electronic devices according to the various embodiments disclosed in this document may take various forms. Electronic devices may include, for example, portable communication devices (e.g., smartphones), computer devices, portable multimedia devices, portable medical devices, cameras, wearable devices, or home appliances. Electronic devices according to the embodiments of this document are not limited to the aforementioned devices.

[0233] The various embodiments of this document and the terminology used therein are not intended to limit the technical features described in this document to specific embodiments, but should be understood to include various modifications, equivalents, or substitutes of the embodiments. In connection with the description of the drawings, similar reference numerals may be used for similar or related components. The singular form of a noun corresponding to an item may include one or more of the items, unless the context clearly indicates otherwise. In this document, each of the phrases "A or B", "at least one of A and B", "at least one of A or B", "A, B, or C", "at least one of A, B, and C", and "at least one of A, B, or C" can include any one of the items listed together in the corresponding phrase among those phrases, or all possible combinations thereof. Terms such as "first," "second," or "first" or "second" may be used merely to distinguish one component from another, and do not limit the components in any other respect (e.g., importance or order). When a component (e.g., a first component) is referred to as "coupled" or "connected" to another (e.g., a second component), with or without the terms "functionally" or "communicatively," it means that the component can be connected to the other component directly (e.g., wired), wirelessly, or through a third component.

[0234] The term "module" used in various embodiments of this document may include a unit implemented in hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit. A module may be an integral component, or a minimum unit or part of such a component that performs one or more functions. For example, according to one embodiment, a module may be implemented in the form of an application-specific integrated circuit (ASIC).

[0235] Various embodiments of the present document may be implemented as software (e.g., a program (140)) including one or more instructions stored in a storage medium (e.g., an internal memory (136) or an external memory (138)) readable by a machine (e.g., an electronic device (101)). For example, a processor (e.g., a processor (120)) of the machine (e.g., an electronic device (101)) may call at least one instruction among the one or more instructions stored from the storage medium and execute it. This enables the machine to operate to perform at least one function according to the at least one called instruction. The one or more instructions may include code generated by a compiler or code executable by an interpreter. The machine-readable storage medium may be provided in the form of a non-transitory storage medium. Here, 'non-transitory' simply means that the storage medium is a tangible device and does not contain signals (e.g., electromagnetic waves), and the term does not distinguish between cases where data is stored semi-permanently or temporarily on the storage medium.

[0236] According to one embodiment, the method according to various embodiments disclosed in this document may be provided as included in a computer program product. The computer program product may be traded as a product between a seller and a buyer. The computer program product may be distributed in the form of a machine-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or may be distributed online (e.g., downloaded or uploaded) through an application store (e.g., Play Store™) or directly between two user devices (e.g., smart phones). In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily generated in a machine-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or an intermediary server.

[0237] According to various embodiments, each component (e.g., a module or a program) of the above-described components may include one or more entities, and some of the entities may be separated and arranged in other components. According to various embodiments, one or more components or operations of the aforementioned components may be omitted, or one or more other components or operations may be added. Alternatively or additionally, a plurality of components (e.g., a module or a program) may be integrated into a single component. In such a case, the integrated component may perform one or more functions of each of the plurality of components identically or similarly to those performed by the corresponding component among the plurality of components prior to the integration. According to various embodiments, the operations performed by a module, program, or other component may be executed sequentially, in parallel, iteratively, or heuristically, or one or more of the operations may be executed in a different order, omitted, or one or more other operations may be added.

Claims

1. In an electronic device (101), Processor (120); and A memory (130) electrically connected to the processor (120) and storing instructions executable by the processor (120) Including, The above processor (120) When the above command is executed by the processor (120), a plurality of utterances are received from a user through a voice input device; Generating a plurality of utterance vectors (710; 720; 910; 920; 930) corresponding to the above plurality of utterances; Calculating the first similarity between each of the plurality of utterance vectors (710; 720; 910; 920; 930) and each of the plurality of reference vectors (730; 740; 750) stored in the memory (130); Based on the first similarity, learning utterance vectors for training a first model for identifying a user-specified call word are determined among the plurality of utterance vectors (710; 720; 910; 920; 930); Using the above learning utterance vectors, the model is trained. Electronic device (101).

2. In paragraph 1, The above processor (120) If the first similarity is greater than or equal to the set first threshold value, an interface (1310) is provided for requesting additional utterance from the user; Generating a plurality of utterance vectors (710; 720; 910; 920; 930) corresponding to the additional utterance received in response to the above interface (1310). Electronic device (101).

3. In any one of paragraphs 1 and 2, The above processor (120) Providing the interface (1310) including the type of the reference vector whose first similarity is greater than or equal to the first threshold value; Electronic device (101).

4. In any one of paragraphs 1 to 3, The above processor (120) Compute the second similarity between the above plurality of utterance vectors (710; 720; 910; 920; 930); Based on the first similarity and the second similarity, the learning utterance vectors are determined. Electronic device (101).

5. In any one of paragraphs 1 to 4, The above processor (120) If the second similarity is less than or equal to a set second threshold value, an interface (1410) is provided for the user to check whether to include an utterance vector among the plurality of utterance vectors (710; 720; 910; 920; 930) whose second similarity is less than or equal to the second threshold value in the utterance vectors for learning; Based on the user input received through the above interface (1410), a second model is trained using the utterance vector whose second similarity is less than or equal to the second threshold value. Electronic devices (101).

6. In any one of paragraphs 1 to 5, The above processor (120) Determining the first similarity based on the distance between each of the plurality of utterance vectors (710; 720; 910; 920; 930) and each of the plurality of reference vectors (730; 740; 750). Electronic devices (101).

7. An action of receiving multiple utterances from a user via a voice input device; An operation of generating a plurality of utterance vectors (710; 720; 910; 920; 930) corresponding to the plurality of utterances; An operation of calculating a first similarity between each of the plurality of utterance vectors (710; 720; 910; 920; 930) and each of the plurality of reference vectors (730; 740; 750) stored in the memory (130); An operation of determining learning utterance vectors for training a first model for identifying a user-specified call word among the plurality of utterance vectors (710; 720; 910; 920; 930) based on the first similarity; An operation of training the first model using the above learning utterance vectors. Including, How to handle ignition.

8. In paragraph 7, An operation of providing an interface (1310) for requesting additional utterance to the user when the first similarity is greater than or equal to the set first threshold; and An operation of generating a plurality of utterance vectors (710; 720; 910; 920; 930) corresponding to the additional utterance received in response to the interface (1310). Including more, How to handle ignition.

9. In any one of paragraphs 7 to 8, The operation of providing the above interface (1310) is: Providing the interface (1310) including the type of the reference vector whose first similarity is greater than or equal to the first threshold value; How to handle ignition.

10. In any one of paragraphs 7 to 9, An operation of calculating a second similarity between the above plurality of utterance vectors (710; 720; 910; 920; 930). Including more, The operation of determining the above learning utterance vectors is as follows: An operation of determining the learning utterance vectors based on the first and second similarities. Including, How to handle ignition.

11. In any one of paragraphs 7 to 10, If the second similarity is lower than or equal to the set second threshold value, an operation of providing an interface (1410) for the user to check whether to include an utterance vector among the plurality of utterance vectors (710; 720; 910; 920; 930) whose second similarity is lower than or equal to the second threshold value in the utterance vectors for learning Including more, The action of training the above model is, An operation of training a second model using an utterance vector having a second similarity less than or equal to the second threshold value based on a user input received through the interface (1410). Including, How to handle ignition.

12. In any one of paragraphs 7 to 11, The operation of calculating the first similarity is as follows: An operation of determining the first similarity based on the distance between each of the plurality of utterance vectors (710; 720; 910; 920; 930) and each of the plurality of reference vectors (730; 740; 750). Including, How to handle ignition.

13. A computer-readable recording medium having recorded thereon a program for executing the ignition processing method of any one of clauses 7 to 12.

Images