JP6643468B2 - Response control device, control program, information processing method, and communication system - Google Patents

Description

  The present invention relates to a response control device, a control program, an information processing method, and a communication system including a server and a response control device. This application claims the priority based on Japanese Patent Application No. 2006-099488 filed on May 18, 2016. The entire contents described in the Japanese patent application are incorporated herein by reference.

  2. Description of the Related Art Conventionally, a voice interaction device for performing a voice interaction with a user has been known. In a spoken dialogue apparatus, in order to respond to a wide range of questions from a user, it is necessary to store a large amount of question contents and response contents in advance, but even if such preparation is made, unexpected questions are asked. Sometimes it was impossible to get a perfect fit from the start.

  For this reason, for example, Japanese Patent Application Laid-Open No. 2004-109323 (Patent Document 1), when recognizing a question that is not stored in advance, asks the user for a response to the question and uses the answer from the user as the response. A speech dialogue device that stores and learns to use for subsequent dialogues is disclosed.

JP 2004-109323 A

  However, when recognizing a question that is not stored in advance, the conventional voice interaction device immediately performs a process of inquiring the user of the response content. For this reason, even if the actual question from the user is a stored question, if the question is incorrectly recognized as an unstored question due to, for example, voice recognition failure, a process of returning the question is performed. I will.

  As a result, the conventional voice interaction device cannot perform the process of inquiring as intended unless the accuracy of voice recognition is 100%, which may cause the user to be disappointed. Also, if the process of returning the question is performed due to the failure of the speech recognition, the question is based on the misrecognition and the question is not intended by the user. There was a risk that the answer would be inappropriate.

  The present disclosure has been made in view of the above problems, and an object in one aspect thereof is to provide a response control device capable of preventing inappropriate inquiry and learning from being performed, a control program, and an information processing method. It is to provide a method and a communication system.

  According to a certain aspect, the response control device includes a voice receiving unit that receives a voice input, and a phrase specified from the voice received by the voice receiving unit is one of a plurality of predetermined types of phrases. In the case where the phrase specified from the voice received by the voice processing unit and the response process execution unit that executes the response process corresponding to the phrase is not any of the plural types of phrases, When the frequency has reached a predetermined frequency, there is provided a learning process executing means for executing a learning process for specifying a response process corresponding to the phrase thereafter.

  According to another aspect, the control program causes a computer to function as a response control device, and causes the computer to function as each of the above units.

  According to yet another aspect, the information processing method includes the steps of: receiving a voice input; and, when the phrase specified from the received voice is one of a plurality of predetermined types of phrases, Performing a response process corresponding to the phrase, and when the phrase identified from the received voice is not any of the plurality of types of phrases, and when the frequency of the phrase has reached a predetermined frequency, And thereafter performing a learning process for enabling a response process corresponding to the phrase to be specified.

  According to yet another aspect, a communication system includes a server and a response control device capable of communicating with the server. The response control device includes: a voice receiving unit that receives a voice input; a communication unit that transmits voice information corresponding to the voice received by the voice receiving unit and receives response information from the server; Response processing executing means for executing response processing. The server includes a storage unit that stores response information corresponding to each of the plurality of predetermined types of phrases, and a storage unit that stores the phrase specified from the voice information from the response control device when the phrase is one of the plurality of types of phrases. A response information transmitting unit that transmits response information corresponding to the phrase; and a case where the phrase identified from the voice information from the response control device is not one of the plural types of phrases, and the frequency of the phrase is A learning process executing means for executing a learning process for enabling a response process corresponding to the phrase to be specified when the predetermined frequency is reached.

  According to one aspect, inappropriate inquiry and learning can be prevented from being performed.

FIG. 1 is a diagram for explaining a schematic configuration of a communication system. It is a figure showing an example of exchange of a conversation between a user and a terminal. FIG. 3 is a diagram illustrating an example of a hardware configuration of a terminal. FIG. 3 is a diagram illustrating an example of a hardware configuration of a server device. FIG. 3 is a functional block diagram for describing a functional configuration of a terminal. FIG. 4 is a diagram for explaining a schematic configuration of a response phrase DB, a speech recognition result DB, and a learning result DB stored in a storage unit. It is a flowchart for demonstrating the flow of the response process at the time of a voice input. It is a functional block diagram for explaining functional composition of a terminal and a server device. FIG. 4 is a diagram for explaining a schematic configuration of a response phrase DB in which response phrases are stored in correspondence with regular phrases and phrases that are close to each other.

  Hereinafter, embodiments will be described with reference to the drawings. In the following description, the same components are denoted by the same reference numerals. Their names and functions are the same. Therefore, detailed description thereof will not be repeated.

[Embodiment 1]
<A. System Configuration>
FIG. 1 is a diagram for explaining a schematic configuration of a communication system according to the present embodiment. Referring to FIG. 1, communication system 1 includes a mobile terminal 100 (hereinafter, also referred to as terminal 100) and a server device 200. The terminal 100 is an example of a response control device, and performs a process of outputting a response phrase to a user's voice (voice input response process). Hereinafter, as the terminal 100, a terminal (a so-called robot-type terminal) capable of automatically moving a movable part forming a housing by executing a program will be described as an example.

  Specifically, the terminal 100 includes a hand, a foot, a head, a torso, and the like. The terminal 100 is typically configured as a walkable autonomous mobile body. The head is configured to be rotatable within a predetermined angle with respect to the trunk. A camera is built in the head. The terminal 100 is not limited to a humanoid robot as described above.

  The terminal 100 is used in various places by being carried by the user 700. The terminal 100 communicates with the server device 200 via the base station 500 and the network 600.

<B. Overview of processing>
Hereinafter, an outline of the processing in the communication system 1 will be described. The terminal 100 identifies the phrase by performing voice recognition based on the voice emitted from the user 700. The phrase refers to, for example, a phrase, a word, a group of words, and the like. Speech recognition refers to converting input speech data into a corresponding phrase. The terminal 100 stores in advance response phrases corresponding to a plurality of types of assumed phrases (also referred to as regular phrases). The terminal 100 outputs the response phrase when the response phrase corresponding to the phrase specified from the voice is stored.

  In addition, the terminal 100 may erroneously recognize a part of the voice due to the influence of the user's pronunciation, surrounding noise, and the like. In preparation for such an erroneous recognition, the terminal 100 also stores in advance a response phrase corresponding to a phrase that is similar to an assumed regular phrase. Phrases that are similar to similar phrases differ from regular phrases only in, for example, the presence or absence of a cloud point ("゛"), the presence or absence of a prompt ("tsu"), the presence or absence of a long note ("-"), and the like. Say. For this reason, even when the terminal 100 does not store the response phrase corresponding to the specified phrase, the terminal 100 outputs the response phrase when storing the response phrase corresponding to the phrase similar to the phrase. I do.

  Further, when the learning condition is satisfied, the terminal 100 performs a learning process for learning a response phrase. The learning condition is satisfied when a phrase that is not stored is specified at a predetermined frequency (for example, specified twice in succession), or when a predetermined learning start phrase is specified. The outline is described below.

  FIG. 2 is a diagram illustrating an example of a conversation exchange between the user 700 and the terminal 100. The speech balloons (steps M01 to M12) shown in FIG. 2 indicate a voice emitted from the user 700 or a voice output from the terminal 100. The boxes (steps S01 to S13) shown in FIG. 2 show the outline of the processing executed by the terminal 100.

  First, a case where a response phrase corresponding to a phrase specified from a voice is stored in terminal 100 will be described.

  As shown in step M01, it is assumed that the user 700 has uttered a voice saying “height?” To the terminal 100. On the other hand, the terminal 100 performs voice recognition of the message from the user 700, extracts and outputs a response phrase corresponding to the result of the voice recognition. The terminal 100 stores a phrase specified as a result of the voice recognition as a history.

  In FIG. 2, as shown in step S01, the terminal 100 recognizes that the voice from the user 700 is “Shinchoha?”, And recognizes that “Shincho” has been asked based on the phrase.

  In step S02, a response phrase that matches "Shincho" which is the object to be queried is extracted. In this example, it is assumed that “height is about 19 cm” is stored as a response phrase that matches “Shincho”. Therefore, as shown in Step M02, the terminal 100 outputs a response phrase such as "Height is about 19 cm."

  Next, the response phrase corresponding to the phrase specified from the voice is not stored in terminal 100, but the response phrase corresponding to the phrase similar to the phrase specified from the voice (also referred to as phrase (approximate)) is stored. Will be described.

  As shown in Step M03, it is assumed that the user 700 has uttered a voice saying “Weight?” To the terminal 100. On the other hand, as shown in step S03, it is assumed that the terminal 100 has erroneously recognized that "Daijoha?"

  However, a phrase that matches a word that does not make sense of “daijou” is not assumed or stored in the design stage. In such a case, the terminal 100 determines whether or not a phrase similar to the phrase identified from the voice is stored, and if the similar phrase is stored, the response phrase corresponding to the approximate phrase is stored. Is extracted and output. The terminal 100 determines whether or not a phrase different from the phrase specified based on the result of the voice recognition in terms of the presence or absence of a voiced voice, a prompt, a long note, and the like is stored.

  In the example of FIG. 2, it is assumed that “Taijou” excluding the first character “Da” of the “Daijou” is stored. In this case, the terminal 100 determines that “daiju” is similar to “taiju”, and extracts a response phrase that matches the approximate “taiju” as shown in Step M04. In this example, it is assumed that “possibly weight? Weight is about 300 g” is stored as a response phrase that matches the approximate “taiju”. For this reason, as shown in Step M04, the terminal 100 outputs a response phrase such as "perhaps your weight? Weight is about 300 g."

  Next, a case where neither the response phrase corresponding to the phrase specified from the voice nor the response phrase corresponding to the approximate phrase is stored will be described.

  As shown in Step M05, it is assumed that the user 700 has uttered a voice saying "What is your foot size?" On the other hand, as shown in step S05, it is assumed that the terminal 100 correctly recognizes that the voice from the user 700 is “Ashinoxah?”.

  However, when it is not assumed at the design stage that “Asino Oxaha?” Is asked, the phrase “Ashino Oxaha?” Is not stored, and a phrase similar to the phrase is not stored.

  In this case, the terminal 100 determines that there is no response phrase corresponding to the phrase specified in step S06, and in step S07, stores the result of the voice recognition “Ashinoooxaha?” As history. Then, the terminal 100 outputs a response phrase when the unknown phrase is specified. As a response phrase in the case where an unknown phrase is specified, a phrase urging re-utterance is defined. For example, as shown in Step M06, a phrase “I did not hear well” is defined. The terminal 100 outputs the response phrase and drives the head to take a pose of tilting the neck.

  As shown in step M07, in a situation where the unknown phrase is specified, if the user 700 again utters the voice "What is your foot size?", The terminal 100 returns to the previous time as shown in steps S08 and S09. It is determined that the response phrase is not stored in the same manner as in. Subsequently, it is determined whether or not the result of the current speech recognition matches the latest (previous) speech recognition result stored in step S07.

  As shown in step S <b> 10, when the terminal 100 determines that the result of the current speech recognition matches the result of the previous speech recognition, the terminal 700 does not misrecognize the speech but intentionally “ It is highly probable that the user is saying "how big is it?", So the following learning process is performed.

  First, as shown in Step M08, based on the identified unknown phrase, the terminal 100 should answer "If you ask" Ashinohoxaha? " Is output. In this way, since a response is made like a parrot return, it is possible to prevent the user from asking what the meaning is not understood. In response to this inquiry, it is assumed that the user 700 has uttered a voice saying “5 cm” as shown in Step M09.

  As shown in step S11, the terminal 100 recognizes that the voice from the user 700 is "gocentimeter dayo", and based on the result, as shown in step M10, "" Should I answer? Is output. In response to this inquiry, as shown in step M11, it is assumed that the user 700 has uttered a sound “OK”. On the other hand, as shown in step S12, when the terminal 100 recognizes the voice as “OK”, as shown in step S13, the terminal 100 outputs the phrase “Gocmidayo” as a response phrase of “Ashinokioxaha?”. After the storage, a response phrase such as "I understand" is output as shown in step M12.

  By performing such a learning process, the terminal 100 can thereafter output “Gocentimeter Dayo” stored as a response phrase when “Asino Oxaha?” Is specified as a result of speech recognition. . Also, when the result of the speech recognition is similar to “ashinooxa” (for example, “azinooooxa”), “possibly asinooxa” or “ashinooxa” is output as a response phrase. You may make it.

  As described above, when the response phrase corresponding to the phrase specified based on the voice emitted from the user 700 is stored, the terminal 100 outputs the response phrase (Steps S01, S02, and M02). Further, even when the response phrase corresponding to the specified phrase is not stored, terminal 100 outputs the response phrase when the response phrase corresponding to the phrase similar to the phrase is stored (step S03, S04, M04). Furthermore, when the response phrase corresponding to the approximate phrase is not stored, and when the specified phrase is recognized at a predetermined frequency (for example, two consecutive recognitions), the terminal 100 (Steps S05 to S13, M06 to M12).

<C. Hardware Configuration>
FIG. 3 is a diagram illustrating an example of a hardware configuration of the terminal 100. Referring to FIG. 3, terminal 100 includes, as main components, CPU (Central Processing Unit) 151 that executes a program, ROM (Read-Only Memory) 152 that stores data in a nonvolatile manner, and a program executed by CPU 151. RAM (Random Access Memory) 153 for volatilely storing data generated by execution or data input via the input device, flash memory 154 for storing data in a nonvolatile manner, and LED (Light Emitting Diode) 155, operation keys 156, switch 157, GPS (Global Positioning System) receiver 158, communication IF (Interface) 159, power supply circuit 160, touch screen 161, microphone 162, speaker 163, camera 164, a driving device 165, and antennas 1581 and 1591. Each component is mutually connected by a data bus.

  The touch screen 161 includes a display 1611 and a touch panel 1612. The antenna 1581 is an antenna for the GPS receiver 158. The antenna 1591 is an antenna for the communication IF 159.

  The LEDs 155 are various display lamps indicating the operation state of the terminal 100. For example, the LED 155 indicates the on / off state of the main power supply of the terminal 100, the read / write state of the flash memory 154, and the like.

  The operation key 156 is a key (operation button) for the user of the terminal 100 to turn on or off the main power. The switch 157 is a main power switch for switching whether or not to supply power to the power supply circuit 160, and other various push button switches.

  The GPS receiver 158 acquires position information of the current position of the terminal 100 based on radio waves from four or more GPS satellites. The position information acquired by the GPS receiver 158 is transmitted to the server device 200 via the communication ID 159. The start timing of the acquisition of the position information by the terminal 100 will be described later.

  Communication IF 159 performs a process of transmitting data to server device 200 and a process of receiving data transmitted from server device 200.

  The power supply circuit 160 is a circuit that steps down the voltage of the commercial power supply received via the outlet and supplies power to each unit of the terminal 100.

  The touch screen 161 is a device for displaying various data and accepting inputs. The display 1611 is configured to include a screen for displaying an image.

  Microphone 162 collects sounds around terminal 100. For example, microphone 162 collects voice based on the utterance of user 700.

  Speaker 163 outputs a sound corresponding to the response phrase. The speaker 163 utters in a certain situation for communication with a user or the like.

  The camera 164 is an imaging device for imaging an object around the terminal 100. Image data obtained by imaging with the camera 164 is transmitted to the server device 200 via the communication ID 159.

  The driving device 165 is a driving mechanism for driving the hands, feet, and head of the terminal 100. Note that the terminal 100 walks when the feet are driven by the driving device 165. In addition, when the head is rotated with respect to the trunk by the driving device 165, the direction of the camera 164 is changed. In addition, the terminal 100 can perform a pose of tilting the neck by changing the angle of the head by the driving device 165.

  The processing in terminal 100 (for example, voice input response processing) is realized by each hardware and software (control program) executed by CPU 151. Such software may be stored in the flash memory 154 in advance. The software may be stored in another storage medium and distributed as a program product. Alternatively, the software may be provided as a downloadable program product by an information provider connected to the so-called Internet. Such software is temporarily stored in the flash memory 154 after being read from the storage medium by the reading device or downloaded via the communication IF 159 or the like. The software is read from the flash memory 154 by the CPU 151 and stored in the RAM 153 in the form of an executable program. CPU 151 executes the program.

  Each component constituting the terminal 100 shown in FIG. 1 is a general one. Therefore, it can be said that an essential part of the present disclosure is the software stored in the RAM 153, the flash memory 154, the storage medium, or the software downloadable via the network. Since the operation of each piece of hardware of terminal 100 is well known, detailed description will not be repeated.

  The recording medium is not limited to a DVD (Digital Versatile Disc) -RAM, but may be a DVD-ROM, a CD (Compact Disc) -ROM, a FD (Flexible Disc), a hard disk, a magnetic tape, a cassette tape, an optical disk, an EEPROM (Electrically). A medium that fixedly holds the program, such as a semiconductor memory such as an Erasable Programmable ROM or a flash ROM, may be used. The recording medium is a non-transitory medium in which the program or the like can be read by a computer. In addition, the program referred to here includes not only a program directly executable by the CPU but also a program in a source program format, a compressed program, an encrypted program, and the like.

  FIG. 4 is a diagram illustrating an example of a hardware configuration of the server device 200. Referring to FIG. 4, server device 200 includes, as main components, CPU 251 for executing a program, ROM 252 for storing data in a nonvolatile manner, and data generated by execution of the program by CPU 251 or an input device. RAM 253 for volatilely storing input data, HDD (Hard Disc Drive) 254 for nonvolatilely storing data, LED 255, switch 256, communication IF (Interface) 257, power supply circuit 258, A display 259 and operation keys 260 are included. Each component is mutually connected by a data bus.

  The power supply circuit 258 is a circuit that steps down the voltage of the commercial power supply received via the outlet and supplies power to each unit of the server device 200. The switch 256 is a switch for a main power supply for switching whether or not to supply power to the power supply circuit 258, and other various push button switches. The display 259 is a device for displaying various data.

  Communication IF 257 performs a process of transmitting data to terminal 100 and a process of receiving data transmitted from terminal 100.

  The LEDs 255 are various display lamps indicating the operation state of the server device 200. For example, the LED 255 indicates the on / off state of the main power supply of the server device 200, the read / write state of the HDD 254, and the like. The operation keys 260 are keys (keyboard) used by a user of the server device 200 to input data to the server device 200.

  The processing in the server device 200 is realized by each hardware and software executed by the CPU 251. Such software may be stored in the HDD 254 in advance. The software may be stored in another storage medium and distributed as a program product. Alternatively, the software may be provided as a downloadable program product by an information provider connected to the so-called Internet. Such software is temporarily stored in the HDD 254 after being read from the storage medium by the reading device or downloaded via the communication IF 257 or the like. The software is read from the HDD 254 by the CPU 251 and stored in the RAM 253 in the form of an executable program. CPU 251 executes the program.

  Each component constituting the server device 200 shown in the figure is a general one. Therefore, it can be said that an essential part of the present disclosure is software stored in the RAM 253, the HDD 254, the storage medium, or software downloadable via the network. Since the operation of each piece of hardware of server device 200 is well known, detailed description will not be repeated.

  Note that the recording medium is not limited to a DVD-RAM, and a fixed program such as a DVD-ROM, a CD-ROM, an FD, a hard disk, a magnetic tape, a cassette tape, an optical disk, a semiconductor memory such as an EEPROM, a flash ROM, or the like. Media may be used. The recording medium is a non-transitory medium in which the program or the like can be read by a computer. In addition, the program referred to here includes not only a program directly executable by the CPU but also a program in a source program format, a compressed program, an encrypted program, and the like.

<D. Functional configuration>
FIG. 5 is a functional block diagram for describing a functional configuration of terminal 100. Referring to FIG. 5, terminal 100 includes control unit 111, storage unit 112, driving unit 115, audio input unit 116, audio output unit 117, and communication processing unit 118. Note that the terminal 100 also includes the position information acquisition unit 113, the imaging unit 114, and the like, and may have another functional configuration.

  The sound input unit 116 collects sounds around the terminal 100 and sends the collected sounds to the control unit 111 as sound data. The voice input unit 116 is constituted by, for example, a microphone 162. The voice output unit 117 outputs a voice corresponding to the response phrase. The audio output unit 117 is configured by, for example, a speaker 163.

  The storage unit 112 stores various control programs, and has a response phrase DB (Data Base) 1121, a speech recognition result DB 1122, and a learning result DB 1123. The storage unit 112 includes, for example, a RAM 153 and the like.

  The response phrase DB 1121 stores response phrases corresponding to a plurality of types of phrases (regular phrases) assumed from the design stage, response phrases when each of the plurality of types of phrases (regular phrases) is approximated, and the like. The voice recognition result DB 1122 stores a result of voice recognition based on a voice uttered by the user. The learning result DB 1123 stores a response phrase corresponding to the phrase newly added by the learning process. A specific example will be described below.

  FIG. 6 is a diagram for explaining a schematic configuration of the response phrase DB 1121, the speech recognition result DB 1122, and the learning result DB 1123 stored in the storage unit 112. Referring to FIG. 6A, response phrase DB 1121 includes a phrase and a response phrase corresponding to the phrase.

  Examples of the phrase include plural types of phrases (phrase (match)) such as “Shincho” and “Taiju”, phrases (phrases (approximate)) approximating each of the plural types of phrases, and “Henzio Boete” as a learning start phrase. In addition, a response phrase is stored for each phrase, for example, for “Shincho (match)”, as shown by M02 in FIG. The message "Taijou (approximate)" is stored, as shown by M04 in FIG. 2, "Is it probably weight? Weight is about 300g." For the learning start phrase, "Okay, first let me know the words to learn." Message is stored to say. Others, for unknown phrases or phrases in the training process, for example, steps of FIG. 2 M06, M08, M10, M12 response phrases as shown in the like have been stored.

  Referring to FIG. 6B, speech recognition result DB 1122 includes a phrase specified by the result of speech recognition. In the example of FIG. 6B, the phrase specified by the result of the voice recognition in steps S01, S03, S05, S08, S11, and S12 of FIG. 2 is stored.

  Referring to FIG. 6C, learning result DB 1123 includes a phrase added by learning and a response phrase corresponding to the phrase. In the example of FIG. 6C, the response phrase “Gocentimeter Dayo.” Is stored with respect to the additional phrase “Ashinohoxaha?” In step S13 of FIG. 2.

  The storage information of the response phrase DB 1121 is updated based on update data periodically transmitted from the server device 200. Thereby, the phrase and the response phrase can be updated. The update data is data for specifying a phrase and a response phrase input by an administrator who manages the server device 200 or the like.

  The information stored in the learning result DB 1123 can be transmitted to the server device 200. The server device 200 transmits update data including information stored in the learning result DB 1123 from the terminal 100 to another terminal. Thus, the content learned on the terminal 100 can be reflected on other terminals.

  Returning to FIG. 5, the control unit 111 controls the overall operation of the terminal 100. The control unit 111 includes a speech recognition unit 1110, an utterance content determination unit 1111, an approximation determination unit 1112, a learning function unit 1113, a drive control unit 1114, and a display control unit 1115. The control unit 111 includes, for example, a CPU 151 and the like.

  The voice recognition unit 1110 has a function of performing voice recognition for specifying a phrase based on voice data input by the voice input unit 116.

  The utterance content determining unit 1111 has a function of determining a response phrase output from the audio output unit 116. Specifically, the utterance content determination unit 1111 determines whether or not a response phrase corresponding to the phrase specified by the voice recognition unit 1110 is stored in the response phrase DB 1121 or the learning result DB 1123. Determine the response phrase.

  The approximation determination unit 1112 determines whether or not a regular phrase approximate to the phrase specified by the speech recognition unit 1110 is stored in the response phrase DB 1121 or the learning result DB 1123. Specifically, the approximation determination unit 1112 determines whether or not a phrase that is different from the phrase specified by the voice recognition unit 1110 only in the presence or absence of a voiced voice, a prompt sound, a long note, or the like is stored.

  Even when the response phrase corresponding to the specified phrase is not stored, the utterance content determination unit 1111 determines that the approximate determination unit 1112 determines that a phrase similar to the phrase is stored. A response phrase corresponding to the approximate phrase is determined.

  The learning function unit 1113 has a function of executing a learning process when a learning condition is satisfied. When the learning function unit 1113 determines that the learning condition is satisfied, for example, by recognizing a phrase that is not stored in the response phrase DB 1121 and the learning result DB 1123 twice consecutively, the learning function unit 1113 responds to the corresponding phrase. The phrase is stored in the learning result DB 1123.

  The drive control unit 1113 has a function of driving the drive unit 115 of the terminal 100. Thereby, the terminal 100 can move the movable part. The display control unit 1115 has a function of displaying various types of information on the display unit 119 of the terminal 100.

  The communication processing unit 118 is used for communication with the server device 200 via the network 600. The communication processing unit 118 includes a transmitting unit 1181 for transmitting data to the server device 200, and a receiving unit 1182 for receiving data from the server device 200.

<E. Details of processing>
FIG. 7 is a flowchart illustrating the flow of the voice input response process executed by CPU 151 of terminal 100. The CPU 151 executes voice input response processing when voice is emitted from the user 700 and voice data is input from the voice input unit 116. The voice recognition unit 1110 of the CPU 151 performs voice recognition based on the input voice data and specifies a phrase.

  Referring to FIG. 7, in step S100, a process of searching for the specified phrase is performed. Specifically, it is determined whether or not the specified phrase or an approximate phrase is stored in the response phrase DB 1121 and the learning result DB 1123.

  In step S101, the CPU 151 determines whether a phrase that matches the specified phrase itself is stored. When it is determined in step S101 that the matching phrase is stored, the CPU 151 outputs the response phrase stored corresponding to the phrase in step S102. Thus, the response phrase can be output from the audio output unit 117. Accordingly, terminal 100 can respond to a phrase specified from the utterance from user 700.

  On the other hand, when it is determined in step S101 that no matching phrase is stored, in step S103, the CPU 151 determines whether or not a phrase similar to the specified phrase is stored. When it is determined in step S103 that an approximate phrase is stored, in step S104, the CPU 151 outputs a response phrase stored corresponding to the approximate phrase. As a result, even when a phrase specified from an utterance from the user 700 is not stored, the terminal 100 can respond to a phrase similar to the phrase. As a result, even when a response phrase corresponding to the phrase itself specified by the voice is not prepared, the terminal 100 can respond to the user 700 and compensate for the lack of the response phrase.

  When it is determined in step S103 that an approximate phrase is not stored, in step S105, the CPU 151 determines whether the specified phrase is a learning start phrase for starting a learning process. The learning start phrase is, for example, "remember reply (henji boete)", "remember words (kotoba oboete)" or the like.

  When it is not determined in step S105 that the phrase is the learning start phrase, in step S106, the CPU 151 stores the phrase as the result of the current voice recognition in the voice recognition result DB 1122. Thereby, the terminal 100 can accumulate the history of the result of the voice recognition. Even when a matching phrase or a similar phrase is stored, the result of voice recognition is accumulated as a history.

  In step S107, the CPU 151 determines whether or not the result of the current speech recognition matches the result of the previous speech recognition. That is, it is determined whether the same phrase has been specified twice consecutively. If it is determined in step S107 that the result does not match the result of the previous speech recognition, in step S109, the CPU 151 outputs “I did not hear well” as a response phrase and sets the head to pose to tilt. Drive unit. Thereby, the user can be prompted to speak again.

  On the other hand, when it is determined in step S107 that the result of the current speech recognition matches the result of the previous speech recognition and the learning condition is satisfied, the CPU 151 shifts the control to step S108 and executes a learning process. In the learning process executed by specifying the same phrase twice consecutively, the CPU 151 performs the utterance / response exemplified in steps M08 to M12 and S11 to S13 in FIG. The response phrase corresponding to the phrase based on is stored in the learning result DB 1123. As described above, since the learning process is executed when the unknown phrase is specified at a predetermined frequency, it is possible to prevent inappropriate inquiry and learning from being performed. Thereafter, when the phrase specified from the voice becomes the phrase added in the learning process, the terminal 100 can output the corresponding response phrase.

Returning to step S105, when it is determined that the phrase is a learning start phrase, the CPU 151 shifts the control to step S108 and executes a learning process. In the learning process executed by becoming the learning start phrase, for example, the following utterance / response is performed.
Response content of terminal 100: OK, first tell me the words to learn.
The content of the utterance of the user 700: “○ △ × □”.
Response content of terminal 100: "○ △ × □", right? What should I do if I say "○ △ × □"? Please tell me the words to reply.
The utterance content of the user 700: “△△ × ※” is sufficient.
The response content of the terminal 100: “△△ × ※”. Okay, I remember.

  By performing such a learning process, the terminal 100 stores the response phrase (△△ × ※) corresponding to the phrase (△△ □) based on the result of the voice recognition in the learning result DB 1123. Accordingly, when a phrase (△ Δ △) added in the learning process thereafter, a corresponding response phrase (△△ × ※) can be output.

[Embodiment 2]
In the first embodiment, the example in which the terminal 100 can execute the voice input response process alone has been described. However, the present invention is not limited to this, and the terminal 100 can communicate with the server device 200 to execute the voice input response process. You may do so.

  For example, the server device 200 may include the storage unit 112, the utterance content determination unit 1111, and the learning function unit 1113 illustrated in FIG. The functional configuration of the terminal 100 and the server device 200 in this case will be exemplified. FIG. 8 is a functional block diagram for explaining the functional configuration of terminal 100 and server device 200.

  As shown in FIG. 8, the server device 200 includes a control unit 211, a storage unit 212, and a communication processing unit 213. The storage unit 212 has a response phrase DB 2121, a speech recognition result DB 2122, and a learning result DB 2123. The response phrase DB 2121, the speech recognition result DB 2122, and the learning result DB 2123 correspond to the response phrase DB 1121, the speech recognition result DB 1122, and the learning result DB 1123, respectively, in the first embodiment.

  The control unit 211 includes, for example, an utterance content determination unit 2111, an approximation determination unit 2112, and a learning function unit 2113. The utterance content determination unit 2111, the approximation determination unit 2112, and the learning function unit 2113 correspond to the utterance content determination unit 1111, the approximation determination unit 1112, and the learning function unit 1113, respectively, in the first embodiment.

  The communication processing unit 213 is used for communication with the terminal 100 via the network 600. The communication processing unit 213 includes a transmitting unit 2131 for transmitting data to the terminal 100 and a receiving unit 2132 for receiving data from the terminal 100.

  Next, an outline of the voice input response process will be described. The terminal 100 specifies a phrase based on a voice emitted from the user 700 and transmits phrase data that can specify the phrase to the server device 200 via the transmission unit 1181.

  When the server device 200 receives the phrase data, the speech content determination unit 2111 determines a response phrase based on the phrase specified from the phrase data. The utterance content determination unit 2111 determines whether or not a response phrase corresponding to the specified phrase is stored in the storage unit 212, and specifies the response phrase when the response phrase corresponding to the phrase is stored. The possible response data is transmitted to the terminal 100 via the transmission unit 2131. Thereby, the terminal 100 can output a response phrase corresponding to the specified phrase.

  Further, even when the response phrase corresponding to the specified phrase is not stored, the utterance content determination unit 2111 stores the response phrase corresponding to the phrase determined to be approximate to the phrase by the approximation determination unit 2112. If so, response data that can specify the response phrase is transmitted to the terminal 100 via the transmission unit 2131. Thereby, the terminal 100 can output a response phrase corresponding to a phrase that is similar to the specified phrase.

  Further, even when the utterance content determination unit 2111 determines that the response phrase corresponding to the specified phrase and the approximate phrase is not stored, the specified phrase is recognized at a predetermined frequency (for example, twice). When the recognition is continuously performed, the learning function unit 2113 performs a learning process for learning a response phrase corresponding to the phrase. Specifically, the learning function unit 2113 executes a process for performing an utterance / response exemplified in steps M08 to M12 and S11 to S13 in FIG.

  Also, the utterance content determination unit 2111 executes the process for performing the utterance / response at the time of the learning start phrase determination described in the first embodiment, even when the specified phrase is the learning start phrase.

  In this case, the storage information of the learning result DB 2123 may be stored so as to be identifiable for each terminal (for example, for each identification number capable of identifying the terminal), and may be shared among all terminals. May be stored.

  In addition, as an example in which the response process at the time of voice input can be executed by communicating with the server device 200, not only the storage unit 112, the utterance content determination unit 1111 and the learning function unit 1113 shown in FIG. May be included in the server device 200. In this case, terminal 100 transmits audio data that can specify the audio emitted from user 700 to server device 200 via transmitting section 1181. When receiving the voice data, the server device 200 may recognize the voice by the voice recognition unit based on the voice data, specify the phrase, and execute a process based on the phrase.

[Embodiment 3]
In the first and second embodiments, as the approximate phrase, a phrase that is different only in the presence or absence of a cloud point or the like is exemplified. However, instead of or in addition to this, some phrases included in the regular phrase are included. The phrase may be a phrase that approximates the phrase. For example, phrases similar to “sincho” may include “sincho” or “ncho” instead of or in addition to “gincho” or the like. Further, the phrase approximated to “Taiju” may include “Taiju”, “Iju”, or the like instead of or in addition to “Daiju”.

  Further, in the first and second embodiments, the example in which the approximation determining unit is provided and the approximation determining unit determines whether or not the phrase is an approximation has been described. , The response phrase DB may be configured such that the response phrase is stored for the approximate phrase itself.

  FIG. 9 is a diagram for explaining a schematic configuration of a response phrase DB in which response phrases are stored corresponding to regular phrases and approximate phrases as phrases. For example, a response phrase corresponding to a regular phrase such as "Shincho" or "Taijou" is stored, and a response phrase corresponding to "Gincho", "Sincho", "Ncho" or the like as a phrase similar to "Shincho" is stored. In addition, a response phrase corresponding to “daiju”, “taiju”, “juju” or the like is stored as a phrase similar to “taiju”.

  When the response phrase DB is configured as described above, the utterance content determination unit includes an approximation determination unit by determining whether the phrase specified by the result of the voice recognition is stored in the response phrase DB. At least, it is possible to extract not only a response phrase corresponding to a regular phrase but also a response phrase corresponding to an approximate phrase.

  Further, the response phrase corresponding to the phrase approximate to the regular phrase may store a common (shared) response phrase regardless of the approximate phrase. More specifically, as a common response phrase in the case of approximation, "maybe ...?" Is stored as a response phrase, and as a response phrase, a regular phrase itself is inserted in the part of "...". The response phrase corresponding to the phrase may be added thereafter. For example, "maybe ..." is defined for phrases that approximate "Shincho" or "Taijou", and when a phrase that approximates "Shincho" is identified, " Maybe you're a Sincho? He's about 19cm tall. " Thereby, the storage capacity for storing the response data corresponding to the approximate phrase can be reduced.

[Embodiment 4]
Although the learning process in the first to third embodiments is described as being executed when the unknown phrase is specified twice consecutively, the learning process is executed when the unknown phrase is specified at a predetermined frequency. If there is, it is not limited to this. The learning process may be performed, for example, by identifying the same unknown phrase three times in the last ten histories in the speech recognition result DB. Further, the determination may be made not only in the number of times but also in a period in which the interval between the first and second times is within one minute.

[Others]
In the first to fourth embodiments, an example has been described in which the response phrase is determined by either the terminal 100 or the server device 200. However, the present invention is not limited to this. The response phrase may be determined. In this case, the terminal 100 also obtains a response phrase responding to the voice from the user from the server device 200, and determines the importance (response level) of the information from the response phrase and the response phrase determined by itself. A higher response phrase may be selected and output as a response phrase to be output.

  In the first to fourth embodiments, an example has been described in which, when an approximate phrase is specified, a response phrase corresponding to the approximate phrase is output. However, the present invention is not limited to this, and an approximate phrase is specified at a predetermined frequency. When the specified phrase is specified (for example, when specified twice in succession), the specified phrase may be assumed to be a legitimate phrase, and a response phrase to the specified phrase may be learned.

  In the first to fourth embodiments, the processing for outputting the response phrase and the learning processing are illustrated as the response processing corresponding to the phrase specified from the voice. However, the processing is not limited to this as long as the processing is associated in advance. For example, a process of driving the terminal 100 in a predetermined mode, a process of capturing an image with the camera 164, and the like may be used.

[Summary]
Hereinafter, main processes among the processes described above and advantages obtained by the processes will be described.

  (1) The terminal 100 outputs the response phrase corresponding to the phrase specified from the voice from the user 700 when the phrase is one of the response phrase DB and the phrase stored in the learning result DB. Learning to execute a process to perform a response phrase and to output a response phrase corresponding to the unknown phrase when the unknown phrase is not stored and the unknown phrase is specified at a predetermined frequency. Execute the process. Thus, the learning process is not executed immediately when the unknown phrase is specified, and the learning process can be executed on the assumption that the user intentionally emits the unknown phrase when the frequency reaches a predetermined frequency. As a result, inappropriate questions and learning can be prevented from being performed while allowing the user to respond widely to utterances in the future.

  (2) The terminal 100 has a response phrase DB that stores response phrases corresponding to a plurality of types of supposed phrases and approximate phrases, and a learning result DB that updates and stores response phrases corresponding to the phrases by learning processing. . As a result, the time from when the user utters the voice to when the user responds can be minimized.

  (3) The predetermined frequency that triggers the learning process is a frequency that is reached when an unknown phrase that is not stored is specified twice consecutively. Thus, for example, there is no need to perform a special voice or operation for starting the learning process, so that the learning hurdle can be reduced. As a result, the learning frequency can be improved.

  (4) When an unknown phrase that is not stored is specified, a response such as “I did not hear well” is output as shown in M06 in FIG. Thus, the user can be prompted to speak again.

  (5) The learning process is a process of prompting the utterance of a response phrase corresponding to the unknown phrase that triggered the learning (M08 in FIG. 2), and the phrase itself identified from the user's utterance corresponds to the unknown phrase. (S13 in FIG. 2) for storing as a response phrase. Thus, any phrase can be stored as a response phrase.

  (6) When the phrase is a learning start phrase, a response phrase corresponding to the phrase can be learned thereafter. Thereby, learning can be positively performed based on the intention of the user.

  (7) The terminal 100 outputs a response phrase corresponding to the regular phrase when the phrase specified from the voice from the user 700 matches the regular phrase stored in the response phrase DB or the learning result DB. When the approximate phrase is approximated to the regular phrase stored in the response phrase DB or the learning result DB, a process of outputting a response phrase corresponding to the approximate case is performed. This allows a response even when a response phrase corresponding to the phrase itself specified from the voice of the user 700 is not prepared. As a result, the shortage of the response phrase can be compensated.

  (8) In the embodiment having the approximation determination unit, it is determined whether or not the phrase specified from the voice of the user 700 approximates one of the regular phrases. When the phrase specified from the voice of the user 700 approximates one of the regular phrases, the utterance content determination unit outputs the stored response phrase corresponding to the approximate case. Thus, by preparing response phrases that match and approximate a regular phrase, it is possible to respond widely to the utterance of the user.

  (9) In the embodiment having no approximation determination unit, some phrases included in the regular phrase are simulated as phrases approximating the regular phrase, and then as shown in the response phrase DB in FIG. By preparing a response phrase corresponding to a regular phrase and a response phrase corresponding to some of the phrases included in the regular phrase, it is possible to respond widely to the utterance of the user while reducing the processing load. Can be.

  (10) Of the regular phrases, for example, the response phrase when approximating “Shincho” and the response phrase when approximating “Taijou” include a common phrase such as “by chance”.

  Also, when approximating “Shincho”, for example, using a common “Hyotto” and a response phrase corresponding to “Shincho”, for example, “Hitto is a Shincho? Height is about 19 cm. Is output. This makes it possible to reduce the storage capacity for storing the response phrases as compared with the case where a different response phrase is prepared for each similar phrase.

  (11) The communication system includes the server device 200 and the terminal 100 capable of communicating with the server device 200. Then, terminal 100 in Embodiments 2 and 3 transmits voice information (for example, phrase data, voice data, and the like specified from the result of voice recognition) corresponding to the voice from the user, and thereafter, server device A process of outputting a response phrase based on the response information (response data) transmitted from 200 is executed.

  On the other hand, when the phrase specified from the voice information from the terminal 100 is one of the response phrase DB and the phrase stored in the learning result DB, the server device 200 transmits the response information corresponding to the phrase. Executing an output process to output response information corresponding to the unknown phrase when the unknown phrase is not stored and the unknown phrase is specified at a predetermined frequency. Execute the learning process. Thereby, the learning process is not immediately executed when the unknown phrase is specified, and the learning process can be executed on the assumption that the user intentionally utters the unknown phrase when the predetermined frequency is reached. As a result, it is possible to respond to the utterance from the user in a wide range in the future, while preventing inappropriate inquiry and learning from being performed.

  Further, when the phrase specified from the voice information from terminal 100 matches the phrase stored in response phrase DB or learning result DB, server device 200 performs a process of outputting response information corresponding to the phrase. When executed, when the phrase approximates to the phrase stored in the response phrase DB or the learning result DB, a process of outputting response information corresponding to the approximate phrase is executed. Accordingly, a response can be made even when response information corresponding to the phrase itself specified from the voice of the user is not prepared. As a result, shortage of response information can be compensated.

  The embodiments disclosed this time are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  1 communication system, 100 communication terminals, 111, 211 control units, 112, 212 storage units, 115 drive units, 116 audio input units, 117 audio output units, 118, 213 communication processing units, 119 display units, 158 GPS receivers, 162 microphone, 163 speaker, 164 camera, 165 driving device, 200 server device, 500 base station, 600 network, 700 user, 1110 voice recognition unit, 1111, 111 utterance content determination unit, 1112, 2112 approximation determination unit, 1113, 2113 Learning function unit, 1114 drive control unit, 1115 display control unit, 1121, 121 response phrase DB, 1122, 2122 speech recognition result DB, 1123, 2123 learning result DB, 1181, 2131 transmission unit, 1182, 2132 reception unit.

Claims (9)

A voice receiving unit for receiving voice input;
When the phrase specified from the voice received by the voice receiving unit is any of a plurality of predetermined types of phrases, a response process execution unit that performs a response process corresponding to the phrase,
When the phrase specified from the voice received by the voice receiving unit is not any of the plurality of types of phrases, and when the frequency of the phrase has reached a predetermined frequency, the phrase A response control device comprising: a learning process execution unit that executes a learning process for enabling a corresponding response process to be specified. A storage unit that stores information that can specify a response process corresponding to each of the plurality of types of phrases,
The response process execution unit executes a response process corresponding to a phrase specified from the voice received by the voice reception unit, based on the storage information of the storage unit,
The response control according to claim 1, wherein the learning process execution unit updates the storage information of the storage unit so that a corresponding response process can be specified from the phrase that has reached the predetermined frequency by the learning process. apparatus.   3. The response control device according to claim 1, wherein the predetermined frequency is a frequency that is reached when a voice in which a phrase that is not any of the plurality of types of phrases is specified is continuously received a predetermined number of times. 4.   When the phrase specified from the voice received by the voice receiving unit is not any of the plurality of types of phrases, the voice control unit includes a voice promoting unit that prompts the user of the response control device to voice the voice. The response control device according to claim 1. The response process that can be specified by the learning process is a process of outputting a phrase specified from the voice received by the voice receiving unit as a response phrase,
The learning process includes:
A process for prompting the user of the response control device to utter a voice corresponding to the response phrase,
5. The processing according to claim 1, further comprising: specifying a phrase specified from the voice received by the voice receiving unit as a response phrase corresponding to a phrase that triggered the learning process. The response control device according to any one of the above.   The learning process execution unit, when the phrase specified from the voice received by the voice receiving unit is a predetermined learning phrase, then to the phrase specified from the voice received by the voice receiving unit The response control device according to claim 1, wherein the response control device executes a process for enabling a corresponding response process to be specified.   A control program for causing a computer to function as the response control device according to claim 1, wherein the control program causes the computer to function as each of the units. Accepting voice input,
When the phrase specified from the received voice is one of a plurality of predetermined types of phrases, executing a response process corresponding to the phrase;
When the phrase specified from the received voice is not any of the plural types of phrases, and when the frequency of the phrase has reached a predetermined frequency, a response process corresponding to the phrase is thereafter performed. Performing a learning process for enabling identification. A communication system comprising a server and a response control device capable of communicating with the server,
The response control device,
A voice receiving unit for receiving voice input;
A communication unit that transmits voice information corresponding to the voice received by the voice reception unit and receives response information from the server;
A response processing execution unit that performs response processing based on the received response information,
The server is
A storage unit for storing response information corresponding to each of a plurality of predetermined types of phrases,
When the phrase specified from the voice information from the response control device is any of the plurality of types of phrases, a response information transmission unit that transmits response information corresponding to the phrase,
When the phrase specified from the voice information from the response control device is not any of the plural types of phrases, and when the frequency of the phrase has reached a predetermined frequency, the phrase corresponding to the phrase thereafter A learning process executing unit that executes a learning process for enabling a response process to be performed to be specified.

Images