JP6963778B2 - Service provision system and program - Google Patents

Description

The present invention relates to a service providing system that provides a service for a user and artificial intelligence to have various dialogues, and a program executed by a computer to receive a service for the user to have a dialogue with artificial intelligence .

For example, a voice control system recognizes a spoken command and associated words (eg, "call mom at home") and causes a selected application (eg, telephone dialer) to execute the command (eg, telephone dialer). There was one that caused a data processing system such as a smartphone to execute an operation based on (look up moms phone number at home and dial it to estimate call) (for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2013-73240

An object of the present invention is to enable an artificial intelligence for task processing having knowledge specialized in a specific field to respond to a question or request from the user when the user processes a task. be.

Specific examples of means for solving problems and their effects

Next, an example of the correspondence between various means for solving the problem and the embodiment is shown in parentheses below.

The present invention is a service providing system that provides a service for a user and artificial intelligence to have a dialogue.
A dialogue model (for example, the dialogue model of FIG. 19A), which is a model for dialogue generated by modeling using machine learning and is a general model used when a user interacts with artificial intelligence. , A task processing model for task processing having knowledge specialized in a specific field (for example, the task processing model of FIG. 21A), and a storage means for storing (for example, artificial intelligence DB17, etc.). ,
Service providing means (for example, S241 to S244, S247, S248, S261 to S264, S267 to S269, etc.) that provide a service for the user and artificial intelligence to have a dialogue using the model.
Through interaction with Ruyu over The by the said interaction model, the personalized means to personalize the interaction model dialogue model suitable for those該Yu chromatography THE (e.g., S245, S246, etc.) and includes a
The dialogue model includes a plurality of dialogue models (for example, FIG. 19A) including a specific talent dialogue model generated by using machine learning using dialogue data of an actual talent (for example, talent A) as learning data. ), Can be selected by the user,
The personalization means obtains the selected dialogue model through a dialogue between a selected dialogue model (for example, S242, S243) selected by the user among the plurality of dialogue models and a user who has selected the selected dialogue model. Personalize the dialogue model suitable for the user (for example, S245, S246, etc.).
The service providing means
Using the selected interactive models personalized by the personalization unit, a personal service providing means for providing a service for performing interactive the user who has selected the selected interaction model (e.g., updated Execute S244 etc. according to the personal weight),
When the user is interacting with the selected dialogue model, the user can take turns interacting with the task processing model according to the content of the dialogue, and can use the knowledge specific to a specific field of the task processing model. Specialized knowledge utilization means (for example, S247, S249, S267 to S269, etc.) and the like are included.

According to such a configuration, when a user processes a task, artificial intelligence for task processing having knowledge specialized in a specific field can respond to a question or request from the user. be able to.

Preferably, the dialogue model has emotional change control means that change emotions in response to dialogue with the user.

Another aspect of the invention is a program executed by a computer (eg, mobile communication device 3) to receive a service for the user to interact with artificial intelligence.
A dialogue model (for example, the dialogue model of FIG. 19A), which is a model for dialogue generated by modeling using machine learning and is a general model used when a user interacts with artificial intelligence. , Receive a service for the user to interact with artificial intelligence using a task processing model for task processing (for example, the task processing model of FIG. 21 (a)) having knowledge specialized in a specific field. Processing steps (for example, S235 to S240, S265, S266) for
Let the computer run
The dialogue model includes a plurality of dialogue models (for example, FIG. 19A) including a specific talent dialogue model generated by using machine learning using dialogue data of an actual talent (for example, talent A) as learning data. ), Can be selected by the user,
The processing step
Selected selected conversation model by the user among a plurality of said interaction model (e.g., S237, S238) through interaction with the user selecting the selected interaction model, the selected interaction model to those該Yu over THE personal service receiving step (for example, receive a response in accordance with the updated personal weights S239, etc. for receiving a pre-service Kiyu over tHE performs the conversation using the appropriate personalized the selected dialogue model interaction model Etc.) and
When the user is interacting with the selected dialogue model, the user can take turns interacting with the task processing model according to the content of the dialogue, and can use the knowledge specific to a specific field of the task processing model. Includes specialized knowledge utilization steps (eg, S265, S266, etc.) and so on.

It is a system diagram which shows the overall configuration of an IoT device intermediary system and a service provision system using machine learning. (A) is a block diagram showing a control circuit of an IoT device, and (b) is a block diagram showing a control circuit of a mobile communication device. It is a block diagram which shows the control circuit of a PC and a server. It is a flowchart which shows the main routine program of a mobile communication device and an artificial intelligence server. It is a flowchart showing a subroutine program of IoT processing and a flowchart of an IoT device (sensor) and an IoT server. It is a flowchart showing a subroutine program of IoT processing and a flowchart of an IoT device (actuator) and an IoT server. It is a flowchart showing a subroutine program of IoT processing and a flowchart of an IoT device (actuator) and an IoT server. It is a flowchart showing a subroutine program of processing for IoT and a flowchart of a wireless sensor network and an IoT server. It is explanatory drawing which shows the outline of the service provision system using machine learning. It is a functional block diagram which shows the service provision system using machine learning. It is a flowchart which shows the external display processing and the subroutine program of the display processing, and the flowchart of the digital menu board. (A) is a flowchart showing a subroutine program of expert data processing, and (b) is a flowchart showing a subroutine program of maintenance data processing and a flowchart of a maintenance specialist's PC. (A) is a flowchart showing a subroutine program of around data terminal processing and around data server processing, and (b) is a flowchart showing a subroutine program of personal service terminal processing and personal service server processing. It is a flowchart which shows the subroutine program of the learning terminal processing for a user and the learning server processing for a user. It is explanatory drawing explaining the general model creation method. It is a flowchart which shows the subroutine program of learning service terminal processing and learning service server processing. It is a flowchart which shows the subroutine program of the muscle training data processing, and the flowchart of the PC of the muscle training specialist. (A) is a flowchart showing a subroutine program of muscle training terminal processing and muscle training server processing, and (b) is a flowchart showing a subroutine program of muscle training service terminal processing and muscle training service server processing. (A) is a diagram showing a specific example of the dialogue model stored in the artificial intelligence DB, and (b) is a diagram showing a specific example of the personalized data for dialogue stored in the user DB. It is a flowchart which shows the subroutine program of the dialogue terminal processing and the dialogue server processing. (A) is a diagram showing a specific example of a task processing model stored in the artificial intelligence DB, and (b) is a diagram showing a specific example of personalized data for task processing stored in the user DB. It is a flowchart which shows the subroutine program of a task terminal processing and a task server processing. It is a flowchart which shows the subroutine program of a task terminal processing and a task server processing.

[Device mediation system for IoT]
First, the device mediation system for IoT will be described with reference to FIGS. 1 to 8. IoT is an abbreviation of Internet of Things, and things are networked by the Internet protocol, and things send signals on the Internet by themselves. The IoT device mediation system is a user's mobile communication device 3 mediates communication with a group 2 of IoT devices (various sensors and actuators for IoT, etc.) that do not have a function for connecting to the Internet and communicating. , This is a system for connecting the IoT device group 2 to the Internet 1 via the mobile communication device 3 to enable communication.

With reference to the entire system of FIG. 1, the Internet 1 includes a mobile communication device 3 represented by a user's wearable computer, a robot 6 at the user's home, and a personal computer (hereinafter referred to as “PC”) 7 at the user's home. , The IoT server 8, the artificial intelligence server 9, the PC 10s of various specialists, and the SNS server 11 are connected to each other so as to be able to communicate with each other. The mobile communication device 3 has a function for communicating with the device group 2 for IoT, the wireless sensor network 4, and the display group 5 such as a digital menu board. Communication methods include, for example, Wi-Fi (registered trademark) , Bluetooth (registered trademark) , Wi-Fi Direct (registered trademark) , Zigbee (registered trademark), Z-wave (registered trademark) , Ant + (registered trademark), etc. Is assumed. It also supports iOS, Android (registered trademark) , Linux (registered trademark), TIZEN (registered trademark) , and other real-time OS. Uses IEEE 802.11 as a wireless standard for transmission and reception. In addition, IPv6 (Internet Protocol Version 6) will be adopted as the Internet Protocol.

The wireless sensor network 4 is a wireless network in which a plurality of wireless terminals with sensors are scattered in a space, and they can cooperate with each other to collect an environment or a physical condition. For example, a sensor device is made of energy harvesting, M2M, or a battery, and for example, deterioration of metal fatigue is constantly monitored by a pressure sensor or gauge sensor, and the change is notified. It is mainly installed in buildings such as bridges and tunnels. Generally, it includes a plurality of sensor nodes and a gateway sensor node. These nodes usually consist of one or more sensors, a wireless chip, a microprocessor, and a power source (such as a battery). The hardware configuration of the node control circuit is the same as that shown in FIG. 2 (a). Wireless sensor networks typically have an ad hoc function and a routing algorithm for sending data from each node to a central node. In other words, there is a function to autonomously reconstruct another communication path when communication between nodes fails. Since the nodes cooperate as a group, there is also an element of distributed processing. In addition, there is also a function that operates for a long period of time without receiving power supply from the outside, and therefore has a power saving function or a self-power generation function. In this embodiment, the wireless sensor network 4 is a kind of IoT device group 2.

The IoT server 8 can write and read data to and from the IoT device DB 15 which is a database of IoT devices (hereinafter referred to as “DB”). The artificial intelligence server 9 can write and read data to the artificial intelligence DB 17, the user DB 12, and the learning DB 60. The SNS server can write and read data to and from the SNS DB 13.

Next, the hardware configuration of the control circuit of the IoT device 2 will be described with reference to FIG. 2A. The IoT device group 2 is sensors and actuators installed all over the earth, and is a CPU (Central Processing Unit) 18 for controlling the whole and a ROM (Read) for storing programs for executing various functions. Only Member) 20, RAM (Random Access Memory) 19 which is a work area of CPU 18, and wireless communication interface unit 22 using, for example, Wi-Fi, Bluetoot h, Wi-Fi Direct, Zigbee, Zwave, Ant +, etc. include. When the IoT device 2 is a sensor, the sensor unit 14 is included. On the other hand, when the IoT device 2 is an actuator, an actuator unit 21 is provided in place of or in addition to the sensorb 14. The CPU 18, the RAM 19, the ROM 20, the wireless communication interface unit 22, the sensor unit 14, and the actuator unit 21 are connected so that signals can be exchanged.

Next, the hardware configuration of the control circuit of the wearable computer as an example of the mobile communication device 3 will be described with reference to FIG. 2 (b). The wearable computer 3 is represented by smart glasses or the like, and a CPU 23, a RAM 24, a ROM 25, and an EEPROM (Electronically Erasable and Programmable Read Only Memory) 26 are connected by a bus 27. The functions of the CPU 23, the RAM 24, and the ROM 25 are basically the same as those described above in the IoT device 2. The EEPROM 26 stores an application program or the like downloaded through the Internet 1.

Various devices are connected to the bus 27 via the interface unit 28. For example, a digital video camera input unit 29 that captures images around the user (images in the direction of the user's line of sight, etc.), a display unit 30 that overlays information and the like on the lens unit of smart glasses, and wireless communication with a base station via the Internet. A wireless communication processing unit 31 that communicates data with a server or the like via 1, an input operation unit 32 for allowing the CPU 23 of the wearable computer 3 to execute a desired function, and a voice output unit 33 for the user to make a voice call. And voice input unit 34, position information acquisition unit 35 for acquiring the current position based on GPS information from satellites, radio waves from base stations, and wireless radio waves from wireless LAN access points, Wi-Fi and Bluetooth, Wireless communication interface unit 36 that communicates with IoT device group 2 using Wi-Fi Direct, Zigbee, Zwave, Ant +, etc., detects the user's line-of-sight position, and the user in the image taken by the digital video camera input unit 29. A line-of-sight position detection unit 37, various sensors 38, and the like for specifying the line-of-sight position are connected to the interface unit 28.

The input operation unit 32 accepts not only the manual operation by the user but also the instruction by the user's gesture, a specific wink, or the like. The user's voice instruction is received by the voice input unit 34. The various sensors 38 are sensors for temperature, humidity, illuminance, ultraviolet rays, etc. attached to the wearable computer 3. The hardware configuration of the robot control circuit is obtained by deleting various sensors 38 and adding a movement drive circuit to the hardware configuration shown in FIG. 2A.

Next, the hardware configuration of the control circuits of the PC and various servers 7 to 11 will be described with reference to FIG. Similar to the above, the CPU 40, the RAM 41, and the ROM 42 are connected by the bus 43. The interface unit 44 to which the bus 43 is connected is connected to a communication unit 45 with the Internet 1 and the like, a display unit 46 for displaying images and information to the operator, and an input operation unit 47 for receiving operations from the operator. ..

Next, a flowchart of a main program of control processing executed by the CPU 23 of the mobile communication device 3 and the CPU 40 of the artificial intelligence server 9 will be described with reference to FIG. The mobile communication device 3 performs IoT processing by S1. In this method, the mobile communication device 3 mediates and connects the IoT device group 2 and the Internet 1, transmits the detection data of the IoT device (sensor) 2 to the IoT server 8, and commands a command signal (command signal) from the IoT server 8. This is a process for transmitting the actuator control signal) to the IoT device (actuator) 2.

Next, the mobile communication device 3 performs an external display process by S2, and the artificial intelligence server 9 performs a display process by S3. This is a process for displaying a moving image, a still image, or the like uploaded by the user on the Internet on a display 5 such as a digital menu board. Next, the mobile communication device 3 performs the around data terminal processing in S4, and the artificial intelligence server 9 performs the around data server processing in S5. This is a process for collecting video and audio around the user by the mobile communication device 3 such as a wearable computer on the artificial intelligence server 9 and using it as information for machine learning.

Next, the mobile communication device 3 performs personal service terminal processing in S7, and the artificial intelligence server 9 performs personal service server processing in S8. This is a process for providing a personalized service to a user by utilizing the result of machine learning by artificial intelligence. Next, the mobile communication device 3 performs the dialogue terminal processing in S9, and the artificial intelligence server 9 performs the dialogue server processing in S10. This is a process for the user and artificial intelligence to engage in various dialogues. Next, the mobile communication device 3 performs task terminal processing in S11, and the artificial intelligence server 9 performs task server processing in S12. This is a process for artificial intelligence that has completed machine learning to respond to user's work questions and consultations.

Next, a flowchart of the subroutine program for IoT processing shown in S1 will be described with reference to FIGS. 5 to 8. First, with reference to FIG. 5, the processes performed by the IoT device (sensor) 2, the mobile communication device 3, and the IoT server 8 will be described.

The CPU 18 of the IoT device (sensor) 2 performs a process of storing the detection data by the sensor unit 14 in the RAM 19 by S20, determines whether or not a transmission command signal has been received by S21, and if not received. Returns to S20 and repeatedly circulates in the loop of S20 → 21 → S20.

On the other hand, the CPU 23 of the mobile communication device 3 transmits the position information based on the GPS information, the radio wave from the base station, the radio wave from the wireless LAN access point, and the like to the IoT server 8 in S22. The IoT server 8 that has received it in S24 searches the IoT device DB 15 in S25, and determines in S26 whether or not the IoT device 2 to be received exists in the vicinity of the mobile communication device 3 that has transmitted the location information. judge. Ideally, the IoT server 8 collects the detection data of the IoT device (sensor) 2 on a regular basis, searches the IoT device DB 15 and divides the IoT device (sensor) 2 whose regular reception time has come. put out. When the IoT device 2 to be received exists, the determination of YES is made by S26, the control proceeds to S27, and the command signal for reception is transmitted to the mobile communication device 3.

On the other hand, when the IoT device 2 to be received does not exist, NO is determined by S26, the control proceeds to S38, and the IoT device (actuator) 2 to transmit the command signal (actuator control signal) exists in the vicinity. Determine whether or not to do so. If present, the control proceeds to S39, and the transmission command signal and the transmission information (information for controlling the actuator unit 21) are transmitted to the mobile communication device 3. The mobile communication device 3 that has received the command signal from S27 or S39 determines YES in S23, advances control to S28, and determines whether or not the received command signal is for reception. In the case of the transmission command signal transmitted by S38, the control shifts to S47, but in the case of the reception command signal transmitted by S27, the control proceeds to S29 and the transmission command signal is in the vicinity. Send to IoT device (sensor) 2.

The mobile communication device 3 that has received it in S21 transmits the detection data stored in the RAM and the device ID to the mobile communication device 3 by S30, and returns to S20. The mobile communication device 3 that has received it in S31 transmits the received data to the IoT server 8 in S32. The mobile communication device 3 that has received it in S33 updates the stored data of the IoT device corresponding to the received device ID to the received data among the stored data of the IoT device DB 15 by S34.

Next, the IoT server 8 transmits the points of an example of the privilege to the mobile communication device 3 by S35, and then the control shifts to S61. The mobile communication device 3 that has received it in S36 adds and updates the points stored in the EEPROM 26 (S37), and then the control shifts to S60. The points stored in the EEPROM 26 can be used to receive various services from the operator of the IoT server 8. For example, you can use points to pay for highways and buy JR and private railway tickets and commuter passes. In addition, points may be redeemed. This can give the user an incentive to mediate the device for IoT.

On the other hand, if NO is determined by S38, the control proceeds to S40, and it is determined whether or not there are various sensors 38 of the mobile communication device 3 to be received. That is, it is determined whether or not any of the various sensors 38 provided in the mobile communication device 3 located in the position information received in S24 is desired to transmit the detection data at the geographical position. If not, control shifts to S24, but if there is, control shifts to S62.

Next, processing for the IoT device (actuator) 2 will be described with reference to FIG. The IoT device (actuator) 2 performs an operation control process based on the transmission information transmitted from the IoT server 8 and stored in the RAM 19 (S45). Next, S46 determines whether or not the reception command signal from the mobile communication device 3 has been received. If not received, the control returns in S45 and repeatedly circulates in the loop of S45 → S46 → S45. In this state, the mobile communication device 3 that has received the transmission command signal and the transmission information transmits the reception command message and the transmission information to the IoT device (actuator) 2 by S47. Upon receiving this, the IoT device (actuator) 2 determines YES in S46, the control proceeds to S48, and the received transmission information (information for controlling the actuator unit 21) is stored in RAM 19 and S49. The reception completion signal and the device ID are transmitted to the mobile communication device 3.

The mobile communication device 3 that has received it in S50 transmits a completion signal to the IoT server 8 in S51. The IoT server 8 that receives it in S52 searches the IoT device DB based on the device ID by S53, and updates the data corresponding to the device ID. For example, update such as "○ hour ○ minute △ × transmission information has been sent". Then, after the points are transmitted to the mobile communication device 3 by S54, the control shifts to S61. The mobile communication device 3 that receives it in S55 adds and updates the points stored in the EEPROM 26 (S56). After that, control shifts to S60.

In S60, it is determined whether or not the command signal for mobile reception and the data identification signal have been received from the IoT server 8, and if not, the control shifts to S84. On the other hand, when the IoT server 8 determines that there are various sensors 38 of the mobile communication device 3 to be received at the position received by S24, it determines YES by S40, and determines that there is a command signal and data for mobile reception by S62. Sends a specific signal. Upon receiving this, the mobile communication device 3 determines YES in S60 and the control proceeds to S63, and whether or not the data (temperature, humidity, illuminance, ultraviolet rays, etc.) specified by the data identification signal requires the time required for detection. Make a judgment. In the case of illuminance or ultraviolet rays, it can be detected in an instant and the detection time is not required. Therefore, S63 determines NO and the control proceeds to S64, and the specified data is detected by the corresponding sensor 38. The detection data is transmitted to the IoT server 8 (S65).

The IoT server 8 that has received it in S66 searches the IoT device DB 15 in S67 and updates it with the new detection data received. Then, the points are transmitted to the mobile communication device 3 by S68. The mobile communication device 3 that receives it in S69 adds and updates the points stored in the EEPROM 26 (S70). After that, control shifts to S84.

On the other hand, when the data specified by the data identification signal requires the detection time, such as temperature and humidity, the determination of YES is made in S63, the control proceeds to S71, and the detection time (for example, the display unit 30). 30 seconds, etc.) is notified to the user. When cooperating with the transmission of the sensor detection data, the user who sees it stays at the current position for 30 seconds and waits for the detection to be completed, but if he / she is not willing to cooperate, he / she moves without stopping. After the notification by S71, the specified data is detected by the corresponding sensor 38 by S72, and it is determined whether or not the detection is completed by S73. If the detection is not completed yet, the specified data stays in the detection area by S74. It is determined whether or not, and if it remains, the loop returning to S72 is repeatedly executed. If the user's mobile communication device 3 moves out of the detection area before the detection is completed, the control shifts to S84, but the user's mobile communication device 3 stays in the detection area until the detection is completed. In that case, the control proceeds to S75, the detection data is transmitted to the IoT server 8, and then the control shifts to S69.

The IoT server 8 that has received it in S66 searches the IoT device DB 15 in S67 and updates it with the new detection data received. Then, the points are transmitted to the mobile communication device 3 by S68. The mobile communication device 3 that receives it in S69 adds and updates the points stored in the EEPROM 26 (S70). After that, control shifts to S84.

In the control shown in FIGS. 5 to 7, a determination means for determining whether or not there is an IoT device that needs to exchange information between the IoT device group 2 and the IoT server 8 ( For example, S26, S38, S49, etc.) are provided on the IoT server 8 side, but this determination means may be provided on the IoT device group 2 side. As a result, the control load on the IoT server 8 side can be reduced.

Next, processing for the wireless sensor network 4 will be described with reference to FIG. The wireless sensor network 4 collects the detection data of each sensor by S80 and stores them in association with each sensor ID. Next, according to S81, it is determined whether or not each of the detected data should be urgently transmitted. When there is nothing to be urgently transmitted, the control proceeds to S82, and it is determined whether or not it is the time of periodic transmission (for example, every 24 hours). If it is not at the time of periodic transmission, the control returns to S80, and the loop of S80 → S81 → S82 is repeated.

In the middle of this loop, if a clear abnormality that can be determined on the wireless sensor network 4 side is found in each detection data and it is determined by S81 that there is something to be urgently transmitted, control proceeds to S83. .. Further, when it is determined by S82 that the periodic transmission time has come, the control proceeds to S83.

In S83, a process of transmitting a transmission signal is performed. Next, S86 determines whether or not the response signal from the mobile communication device 3 has been received, and if not, returns to S80 and repeats the loop of S80 to S83 → S86 → S80.

In this state, when the user having the mobile communication device 3 passes near the wireless sensor network 4, the transmission signal by S83 is received, YES is determined by S84, and the control proceeds to S85. The mobile communication device 3 transmits a response signal to the wireless sensor network 4 by S85, the wireless sensor network 4 that receives the response signal determines YES by S86, and the control proceeds to S87. The wireless sensor network 4 transmits the detection data collected by S80 to the mobile communication device 3 for each sensor ID (S87). The mobile communication device 3 that has received it in S89 transmits the received detection data and the sensor ID together with the transmission signal to the IoT server 8 in S90.

Then, the IoT server 8 determines YES by S61, the control proceeds to S91, and if there is emergency detection data in the received detection data, an abnormality is notified together with the sensor ID, and then the sensor of the detection data is detected by S92. The wireless sensor DB 16 is searched based on the ID, and the detected data is updated to the new one received. Next, the control returns to S24 after transmitting the point to the mobile communication device 3 by S93. The mobile communication device 3 that has received it in S94 adds and updates the points stored in the EEPROM 26 (S95), then returns and the control shifts to S2.

In the control shown in FIG. 8, a determination means (for example, S82) for determining whether or not there is an IoT device that needs to exchange information between the wireless sensor network 4 and the IoT server 8 is provided. Although the one provided on the wireless sensor network 4 side is shown, this determination means may be provided on the IoT server 8 side.

In the IoT device mediation system described above, the user's mobile communication device (for example, a wearable computer such as a smart glass, a smartphone, a vehicle such as a car having a communication function, etc.) mediates between the IoT device group 2 and the Internet 1. The Internet connection function is not always required on the IoT device group 2 side in order to connect to the device, and the cost can be reduced accordingly. Moreover, when the user performs the above-mentioned mediation by the mobile communication device, the user is given a privilege (for example, points), so that the user can have an incentive for the above-mentioned mediation and promote the user's mediation act. It becomes possible.

Further, various types of IoT devices (mobile communication devices 3) that need to exchange information with the intermediary destination computer (for example, IoT server 8 or network cloud) that the wearable computer 3 mediates with the IoT device group 2. Since the determination means (for example, S26, S38, S49, etc.) for determining whether or not the sensor 38 is present is provided, it is not necessary to make the above determination on the IoT device group 2 side, and the IoT The processing load on the side of the device group 2 can be reduced, and the cost can be further reduced. Moreover, the IoT server 8 having the determination means identifies the IoT device group 2 that needs to exchange information (for example, S26, S38, S49, etc.), and obtains the determined position information of the IoT device 2. If the wearable computer 3 is notified and the user is notified, there is an advantage that the user can be urged to perform the above-mentioned mediation action.

Further, when collecting the detection data by the various sensors 38 at a desired position on the Internet by using the various sensors 38 provided in the wearable computer 3, it takes a certain amount of time to detect the data. Since the wearable computer 3 notifies the user of the required detection time (for example, S71), the user can be encouraged to cooperate in collecting the detection data that requires a long time.

[Service provision system using machine learning]
Next, a service providing system using machine learning will be described with reference to FIGS. 9 to 23.
The current major issue in putting machine learning to practical use is how to collect a large amount of learning data at low cost. This system builds a mechanism in which a large number of users take the initiative in providing learning data to the system side by providing a service using the result of machine learning to the user. In particular, as shown in FIG. 9, the learning data (personal data, etc.) provided by each user to the system side is also used for personalization, and the general model obtained as a result of machine learning is optimized for each user ( Personalized) and then provide personalized services to each user. This prevents the inconvenience that each user entrusts the provision of learning data to another person and enjoys only the service.

First, the outline of the system will be described with reference to FIG.
Each user 70 transmits the around data (video, line-of-sight position, voice, GPS and other position data, etc. that the user 70 is looking at) collected by a mobile communication device 3 such as a wearable computer to the machine learning 72 to the machine. It is used as the learning data of the learning 72, and the data is also used for the personalization 74.

For example, the state in which each user 70 is learning at school or at home is transmitted to the system side, modeled by machine learning (reinforcement learning or regression) 72, and the total time of review repeatedly performed by humans is minimized. Create a general model that becomes. At that time, a general model is generated with elements (memorization ability, etc.) that differ from user to user as constants. In addition, the data transmitted from each user 70 to the system side is also used for personalization 74, and elements (memorization ability, etc.) that differ for each user are calculated for each user and assigned to the constant part of a general model. By doing so, a personalized model is generated for each user. Using the personalized model, each user is presented with a personalized service, for example, the next review time is presented, and guidance is given according to a review plan that minimizes the total review time.

The learning data is provided not only by the user but also by various specialists 71. For example, a sports gym specialist 71 transmits the aggregated results of muscle training (hereinafter referred to as “muscle training”) of a large number of members to the system side. Based on the learning data of this aggregation result, it is modeled by machine learning (reinforcement learning or regression) 72, and a general model that maximizes the effect of muscle training performed repeatedly by humans is created. At that time, a general model is generated with elements (initial load such as barbell, load increase coefficient, etc.) that differ for each user as constants. In addition, the aggregated result of muscle training is transmitted from each user 70 to the system side and used for personalization 74, and elements (initial load, load increase coefficient, etc.) that differ for each user are calculated for each user and are generally used. By substituting into the constant part of a typical model, a personalized model is generated for each user.

Using the personalized model, each user is presented with a personalized service, for example, the time and load of the next muscle training, and guidance is given according to the muscle training plan that maximizes the effect of the muscle training. At that time, an advertisement of 71 specialists (sports gyms, etc.) who provided learning data consisting of the aggregated results of muscle training is presented together with the muscle training schedule. As a result, various specialists can be motivated to provide learning data.

FIG. 10 is a more detailed representation of the overall system of FIG. 9 described above.
With reference to FIG. 10, specialized data is provided to the artificial intelligence server 9 from the PC 10s of various specialist companies. In addition, data collected by a mobile communication device 3 such as a wearable computer of a large number of users (around data such as video, line-of-sight position, voice, GPS and other position data seen by the user 70) is provided to the artificial intelligence server 9. Will be done. The artificial intelligence server 9 converts the provided data into a numerical value / label set 51 to generate learning data. Deep learning is used to generate this training data in order to convert the raw data into a numerical / label set. In addition, "heterogeneous mixture learning" is used to handle raw data in which multiple regularities are mixed. This is a method of generating a model suitable for each pattern when there are multiple regularities in the collected data.

For example, when collecting time-series data on the power consumption of a building and creating a model for predicting the power consumption, the pattern of the power consumption of the building changes depending on the day and time zone. In the past, humans mobilized expertise to determine when patterns would switch, and created a model suitable for each case. In this "heterogeneous mixture learning", the pattern switching itself is found by machine learning, and a model is generated for each pattern. When new data is generated, the compatibility with the generated model is confirmed, and if the model does not match well, the pattern switching judgment is restarted. In other words, the machine repeats "case classification of patterns" and "creation of models" to create the optimum model for each pattern.

In the case of data that cannot be converted into a numerical value / label set even by using deep learning, an artificial intelligence company artificially converts the data into a numerical value / label set.

Although the artificial intelligence server 9 uses a general von Neumann computer, a neural network processor (NNP) may be used. Many "artificial neurons" modeled on real neurons are mounted on the NNP chip, and each neuron cooperates with each other in a network.

Further, a quantum computer adopting the "quantum annealing method" may be used. As a result, the time required for optimizing calculation in machine learning can be significantly reduced. Note that "artificial intelligence" is a broad concept that includes software agents.

The learning data converted into numerical / label sets is sent to the learning algorithm 52 and modeled 53. As the learning algorithm 52, various learning algorithms 52 are prepared, for example, regression and identification as supervised learning, model estimation and data mining as unsupervised learning, reinforcement learning and deep learning as intermediate methods, and the like. In the modeling using the learning algorithm 52, for example, a general model is generated with elements (individual ability differences, preferences, etc.) that differ from user to user as constants. The generated general model is stored in the artificial intelligence DB17. In FIG. 10, as specific examples of the memorized general model, for example, a maintenance inspection model of a building, a memorization learning model, a muscle training model, a dialogue model, a task processing model, and the like are shown. The above-mentioned learning data is also stored in the learning DB 60.

_{Ti = T 0} · i as a memorization learning model is a general model that minimizes the total time of review performed by humans repeatedly, i is the number of repetitions indicating the number of repetitions, and T ₀ is the first learning. The period from to just before forgetting (specifically, the time from the first learning to the memory retention rate reaching 70%), Ti represents the time from (i-1) to i in the number of repetitions. _{T 0} in this general model is an element (difference in individual ability such as memory ability) that differs for each user, and is expressed as a constant. _{The method of generating Ti = T 0} · i as this memorization learning model will be described in detail later.

Fi = F ₀ + (i-1) / a as a muscle training model is a general model in which the effect of repeated muscle training by humans is maximized, and i is the number of repetitions indicating the number of repetitions. , F ₀ is the load in the first muscle training (barbell weight, etc.), and a is the load increase coefficient that increases with the repetition of muscle training.

The dialogue model is a general model used when the user interacts with artificial intelligence, and consists of emotional and emotional change functions such as dialogue templates for initial targets such as male intellectuals, female moe, and famous talents. .. This will be described in detail with reference to FIG.

The task processing model is a general model used when a user processes a task such as work, and consists of knowledge utilization functions of initial objects such as patent attorneys, lawyers, and famous scientists. This will be described in detail with reference to FIG.

Around data sent from the user's mobile communication device 3 is classified for various models by the artificial intelligence server 9, 54, and personalized data for various models is created 55. The personalized data is stored in the user DB 12. In FIG. 10, specific examples for various models are shown for memorization learning, muscle training, dialogue, task processing, and the like.

Personalized data and around data are associated and stored in the user DB 12 for each user ID. The personalized data for memorization learning is TK for the period from the first learning to just before forgetting (specifically, the time from the first learning until the memory retention rate reaches 70%). Human memory information diminishes according to the forgetting curve of psychologist Hermann Ebbinghaus. According to this forgetting curve, the time until the memory retention rate reaches 70% is defined as TK. This TK is the time until the memory retention rate of the actual user whose user ID is "1" reaches 70%, and is represented by an actual numerical value.

Personalized data for muscle training includes initial load FK such as barbell weight in the first muscle training, load increase coefficient az that increases with repetition of muscle training, and supercompensation time CT. These are represented by actual numerical values for the actual user whose user ID is "1". In addition, "super-recovery" refers to a phenomenon in which the muscle strength level rises after training for 48 to 72 hours as compared with that before training. The time required for the super-recovery is the "super-recovery time".

Personalized data for dialogue consists of initial targets selected by the user from initial targets such as male intellectuals, female moe, and famous talents, and personal weights of emotions and sorrows. In the general model, the initial weights of emotions and emotions of each initial target are determined, and the initial weights of emotions and emotions are gradually increased according to the request from the user while interacting with the initial target selected by the user. Finally, the personal weight that is optimal for the user is used to enable dialogue that expresses emotions that match the user's tastes. This will be explained in detail later.

The personalized data for task processing includes an initial target selected by the user from initial targets such as patent attorneys, lawyers, and famous scientists, and personalized weights of each knowledge. In the general model, the initial weight of the knowledge of each initial target is determined, and the initial knowledge to be utilized in response to questions and requests from the user during a work dialogue with the initial target selected by the user. The weights are gradually modified to finally be the optimum personal weight for the user, enabling answers and advice that match the knowledge required by the user at work. This will be explained in detail later.

The artificial intelligence server 9 provides personalized guidance (service) to each user using the personalized data as described above. Specifically, the personalized data for memorization learning is used to provide guidance according to the review plan (review schedule) that minimizes the total review time of the user, and the personalized data for muscle training is used to repeatedly perform muscle training. Guidance is given according to the muscle training schedule that maximizes the effect of Provide answers and advice that match the required knowledge.

Next, with reference to FIG. 11, a flowchart of a subroutine program of the external display process (S2), the display process (S3), and the like process by the SNS server 11 of FIG. 4 will be described.

FIG. 11 shows a process for an example digital menu board of the display 5. The digital menu board is a display for menu display installed in restaurants, etc. By making the menu digital, the time and cost on the store side can be significantly reduced, such as immediate display of sold-out items and menu changes. .. In the present embodiment, as shown in FIG. 1, the digital menu board 5 is connected to the Internet 1 and is configured to be able to communicate with the artificial intelligence server 9 and the SNS server 11. In the flowchart of FIG. 11, the image or video taken by the user with the wearable computer 3 or the like and uploaded to the SNS DB 13 or the user DB 12 or the like is displayed on the digital menu board 5 and viewed by friends who came to the store together. It is designed so that you can eat and drink while having fun.

First, it is determined whether or not the menu has been updated by S100. If the store side performs an operation to update today's menu, the control proceeds to S101 and the menu update process is performed. Next, the control proceeds to S102, and the menu is displayed.

Next, when the user wants to display an image or a moving image taken by the wearable computer 3 or the like and uploaded to the SNS DB 13 or the user DB 12 or the like on the digital menu board 5, the operation of accessing the artificial intelligence server 9 is performed first. .. Then, the determination of YES is made by S104, the control proceeds to S105, the user ID, the access request, and the like are transmitted to the artificial intelligence server 9, and the access process is performed. The artificial intelligence server 9 that has received the access request determines YES in S106. In the mobile communication device 3, S107 and S108 communicate with the artificial intelligence server 9 to select and specify an image or a moving image that the user wants to display on the digital menu board 5. The artificial intelligence server 9 searches the user DB 12 by S108 and identifies the selected image or moving image.

When the image to be displayed is specified, the user (customer) taps the touch screen of the digital menu board 5 to perform a communication operation. Then, the determination of YES is made by S103, and the communication start processing with the wearable computer 3 is performed by S109. In addition, the user (customer) performs a communication operation of his / her wearable computer 3. Then, it is determined as YES in S108, the control proceeds to S110, the communication start process is performed, and the connection for communication is established between the wearable computer 3 and the digital menu board 5.

Next, the digital menu board 5 transmits the display ID assigned to the digital menu board 5 to the wearable computer 3 (S111). The wearable computer 3 that has received it in S112 transmits the display ID to the artificial intelligence server 9 (S113). The artificial intelligence server 9 that received it in S114 transmits the data specified in S108 described above to the display 5 of the received ID (S105). The digital menu board 5 that has received the data makes a YES determination in S116 and displays the received data (S117). As a result, the image or video specified by the user (customer) is displayed on the digital menu board 5. As a result, images and moving images can be viewed on a relatively large display screen called the digital menu board 5.

What should be noted in the control described above is that when the user selects and specifies the data uploaded on the net, he / she directly accesses the artificial intelligence server 9 with his / her own mobile communication terminal 3 without going through the display 5. It is a point that is specified. If the display 5 itself is connected to the net and the data to be displayed on the display 5 is specified, the data on the net is accessed via the display 5 of the display destination and the data is displayed on the display 5. It's quicker. However, the data uploaded by the user on the net is personal information related to the privacy of the user, and when the user's ID or the like is transmitted via the display 5 in order to access the personal information, the user's ID or the like is used. There is a risk that personal information will be leaked. Therefore, in the present embodiment, when the user selects and specifies the data uploaded on the net, the artificial intelligence server 9 is directly accessed and specified by the own mobile communication terminal 3 without going through the display 5. ing

Next, on the digital menu board 5, it is determined by S118 whether or not the data display is completed, and when the data display is completed, the control proceeds to S119, and a process of allowing the user (customer) to access the homepage of the store is performed. Specifically, the URL of the store's homepage is transmitted to the mobile communication terminal 3. Upon receiving this, the mobile communication terminal 3 transmits an access request to the page of the URL to the SNS server 11 by S120. The SNS server 11 that receives it by S121 transmits the homepage of the store to the mobile communication device 3 by S122.

In response to this, the mobile communication device 3 displays the store homepage by S123. On the other hand, on the digital menu board 5, the message "Please tap like" on our homepage "is displayed by S124. If the user (customer who visits the store) who sees it taps the like, a YES judgment is made in S125, the control proceeds to S126, and the like signal is transmitted to the SNS server 11. The SNS server 11 that has received it in S127 performs the rice registration process in S128. As a result, information can be provided from the store to the user (customer) through SNS. For example, it is possible to introduce seasonal new menus, distribute coupons, notify event holdings, etc., which can lead to revisiting the store.

If it returns in the external display process, it proceeds to S4, and if it returns in the display process, it proceeds to S5.

Next, a flowchart of the expert data processing subroutine program shown in S6 of FIG. 4 will be described with reference to FIG. 12 (a). Maintenance data processing is performed by S135, muscle training data processing is performed by S136, other processing is performed by S137, and the process returns to S8. The maintenance data processing receives and processes the data of the maintenance and inspection results of the building sent from the maintenance specialist, which is an example of various specialists. The muscle training data processing receives and processes aggregated data of muscle training results of a large number of members sent from an example sports gym of various specialists.

Next, a flowchart of the above-mentioned maintenance data processing subroutine program will be described with reference to FIG. 12 (b). The maintenance specialist performs the maintenance point of the building based on the detection data transmitted from the wireless sensor network 4 shown in FIG. 8 and stored in the wireless sensor DB 16, and obtains the result (KO, NG, etc.) as the sensor ID. It is input to the PC 10 every time (S140).

Then, the maintenance specialist PC 10 transmits the inspection result input by S141 to the artificial intelligence server 9 for each sensor ID. The artificial intelligence server 9 determines whether or not the inspection result has been received in S145, and if not, the control shifts to S151. On the other hand, when the inspection result is received, the control proceeds to S146, and the inspection result data is additionally stored in the learning DB 60 for each sensor ID. Next, according to S147, the wireless sensor data is read from the wireless sensor DB 16 and the inspection result data is attached, and then data mining is performed to find out the pattern NGP1 at the time of NG. For example, if the strain and the vibration cycle have a predetermined relationship, the pattern NGP1 having a high probability of being an abnormality (NG) inspection result is found.

Next, S148 determines whether NGP1 is a new one different from the one found so far. All the patterns at the time of NG found so far are stored in the storage area of the "maintenance and inspection model of the building" of the artificial intelligence DB17, and it is determined whether or not the NGP1 found this time is already stored in the artificial intelligence DB17. do. If it is determined that the same thing is already stored, the control shifts to S151, but if it is determined that the same thing is not yet stored, the control proceeds to S149, and NGP1 is artificial intelligence as a pattern at the time of NG. It is additionally stored in the storage area of the "maintenance and inspection model of the building" of DB17.

Next, the wireless sensor DB 16 is searched by S150, and the sensor ID having the NGP1 pattern is extracted from the periodically checked by the wireless sensor and notified. The wireless sensor is regularly checked at regular intervals (for example, every 24 hours) (S151, S152), and the past sensor data that has already been periodically checked is rechecked by applying the newly found NGP1. It is.

Next, it is determined by S151 whether or not the time for the periodic check has come, and when the time for the periodic check comes, the wireless sensor DB 16 is searched by S152, and the wireless sensor data additionally stored after the previous periodic check is included. The sensor ID having the NG pattern is extracted from the sensor ID and the NG notification is performed. As described above, new sensor data is additionally stored in the wireless sensor DB 16 at any time, and all NG time patterns stored in the artificial intelligence DB are applied to the new sensor data that has not yet been regularly checked. And check it. After the processing of S152, it returns and the control shifts to S136.

Next, with reference to FIG. 13A, a flowchart of a subroutine program of the around data terminal processing shown in S4 of FIG. 4 and the around data server processing shown in S5 will be described.

User learning terminal processing is executed by S160, user learning server processing is executed by S161, muscle training terminal processing is executed by S162, muscle training server processing is executed by S163, and other processing is executed by S164 and S165. Will be done.

The user learning terminal process and the user learning server process collect around data from each learning user to create a general memorization learning model for the user, and generate personalized data for learning for each user. It is the processing of. The muscle training terminal processing and the muscle training server processing are processes for collecting around data from each user during muscle training and generating personalized data for muscle training for each user.

Next, with reference to FIG. 13B, a flowchart of a subroutine program of the personal service terminal process shown in S7 of FIG. 4 and the personal service server process shown in S8 will be described.

The learning service terminal process is executed by S170, the learning service server process is executed by S171, the muscle training service terminal process is executed by S172, and the muscle training service server process is executed by S173.

The learning service terminal processing guides the user according to the most efficient review plan based on the general memorization learning model for the user and the personalized data for learning. The muscle training service terminal processing and the muscle training service server processing provide guidance to the user according to the most efficient muscle training plan based on the general muscle training model for the user and the personalized data for the muscle training. be.

Next, with reference to FIG. 14, a flowchart of a subroutine program of the user learning terminal process and the user learning server process will be described.

It is determined by S180 whether or not the user is learning. This judgment may be made by accepting the user's explicit manifestation of intention, but the artificial intelligence server 9 autonomously processes the user's voice by natural language processing or analyzes the image in the user's line-of-sight direction. You may judge it. If it is determined that the user is not learning, it returns and the control shifts to S162. Is transmitted to the artificial intelligence server 9. The artificial intelligence server 9 that received it by S182 searches the predetermined DB for useful information such as the past entrance examination question frequency and the question school regarding the matters currently being learned from the received around data by S183, and the user (wearable). It is transmitted to a mobile communication device 3) such as a computer.

The user's mobile communication device 3 that has received it in S184 displays the received useful data in overlay using AR (Augmented Reality) in S185. Since the user can receive guidance according to the most efficient review plan by transmitting the learning around data to the artificial intelligence server 9, the user should take the initiative in transmitting the around data to the artificial intelligence server 9. However, by displaying useful information such as the frequency of questions in the entrance examination, the incentive will be further improved and the around data will be transmitted to the artificial intelligence server 9.

Next, the artificial intelligence server 9 classifies the received data for various models according to S186 and stores the received data in the user DB 12 and the learning DB 60. Next, according to S187, it is determined whether or not the memorization learning model has been completed. If it is not completed yet, the control proceeds to S188, the received data is converted into numerical values and labels to create data for machine learning, and modeled using a learning algorithm (reinforcement learning, regression, etc.) by S189. .. An example of this modeling will be described.

The condition is that the memory retention rate at a predetermined end point such as the entrance examination date can be maintained at a constant value (for example, 90%) for each user by using the enormous amount of past learning around data collected for each user as learning data. Below, we will find a review plan that minimizes the total time of review that is repeated from the first learning to the above end point by reinforcement learning. For example, if the period from the i-th review to the next review is Ti and the time required for the i-th review is THi, the total review time T = ΣTHi. Reinforcement learning is performed on Ti that minimizes T under the condition that the end point memory retention rate ≥ 90%.

In reinforcement learning, as a method of estimating the Q value when the value when performing the action at in the state st is Q (st, at), the knowledge that models the environment, that is, the state transition probability and the probability distribution of the reward If is given, a model-based method may be used, but if the environmental model is unknown, TD (Temporal Difference) learning is used. First, since it is necessary to search the environment, the ε-greedy method is used. In the early stages of exploration, we try various actions and introduce the concept of temperature so that when we calm down, we choose many of the most suitable actions. Let T be the temperature, and choose the action according to the probability expressed by the following formula.

P (a ｜ s) ＝ {exp (Q (s, a) / T)} / {Σexp (Q (s, b) / T} (Note that b ∈ A is described under Σ, and the above equation The description is omitted)
Here, a is an act, Q (s, a) is the value when performing an act a in the state s,

T is called the temperature in annealing, and if it is high, the action is selected with a probability close to equal probability, and if it is low, it is biased to the optimum one. As the learning progresses, the learning result becomes stable by reducing the value of T. The user may be instructed to actually execute the action selected in this way, and the learning data may be actively collected.

An example of the result of the reinforcement learning is shown in FIG. With reference to FIG. 15, the vertical axis of the graph is Ti (review interval), and the horizontal axis is i (number of reviews indicating the number of reviews). What is shown by ◯ is a plot of the result of reinforcement learning in the user with user ID: 1. What is indicated by x is a plot of the result of reinforcement learning in the user with user ID: 2. What is indicated by Δ is a plot of the result of reinforcement learning for the user with user ID: 3.

Next, using the "regression" algorithm, we take the machine learning result data of these multiple users as input data, and consider that these input data output the target based on a certain function, and that function. Ask for. The sought-after function is
Ti ＝ T ₀・ i
It is a linear function of the straight line of. The coefficient T ₀ is an element (memorization ability, etc.) that differs for each user. Specifically, it is the period from the first learning of each user to immediately before forgetting (specifically, the time from the first learning until the memory retention rate reaches 70%). In this way, the elements (memorization ability, etc.) that differ from user to user are represented by a general model as a _{constant T 0.}
Ti ＝ T ₀・ i
Is.

Returning to FIG. 14, the artificial intelligence DB 17 _{stores the memorization learning model (Ti = T 0} · i) by S190. If an old and incomplete memorization learning model is already stored, it is _{updated with a newer complete memorization learning model (Ti = T 0} · i).

On the other hand, when it is determined by S187 that the memorization learning model has not been completed yet, the control proceeds to S191, and a process of creating personalized data based on the classified received data and storing it in the user DB 12 is performed. The "personalized data" in this case is an element (memorization ability, etc.) T ₀ that differs for each user, that is, the time from the initial learning until the memory retention rate reaches 70%. This can be easily derived from the user's learning around data stored in the user DB 12.

Next, with reference to FIG. 16, a flowchart of the subroutine program of the learning service terminal processing shown in S170 of FIG. 13B and the learning service server processing shown in S171 will be described.

The mobile communication device 3 such as a wearable computer determines whether or not there is a review request from the user by S200. If there is no review request, the control shifts to S172 by returning, but if there is a review request, the review request is transmitted to the artificial intelligence server 9 by S201. The artificial intelligence server 9 that receives it in S202 determines whether or not the memorization learning model is completed by S203. If it is not completed, the control shifts to S205, but if it is already completed, the control proceeds to S204, and T ₀ = TK is substituted for the _{general memorization learning model Ti = T 0 · i.} Then, the learning items whose repetitive review time is near are extracted from the around data of the user DB 12. The TK is a period from the first learning of the user to immediately before forgetting (specifically, the time from the first learning until the memory retention rate reaches 70%).

Next, according to S205, the learning items that the user is looking away from during class or the like are extracted from the around data of the user DB 12. This is because the line-of-sight position of the user who is learning is transmitted to the artificial intelligence server 9 as around data and stored in the user DB 12 (S181), so that it is possible to determine whether or not the user is looking away. Next, the learning items extracted in S204 and S205 are transmitted to the mobile communication device 3 by S206. The mobile communication device 3 that has received it in S207 performs a process of having the user review the received learning items in S208.

Further, instead of or in addition to being looking away, control may be performed to allow the user to review the learning items while the user is talking wastefully during class or the like. That is, it is not possible to learn from the collecting means for collecting information that can identify the attitude of the user during learning, the recording means for recording the learning content received by the user, and the information collected by the collecting means. A discriminating means for discriminating an appropriate user's attitude portion, an extracting means for extracting the learning content corresponding to the portion discriminated by the discriminating means from the recording means, and a learning content extracted by the extracting means. Is provided as a transmission means for transmitting the information to the user terminal. The above S205 does not necessarily have to be performed. After the processing of S206, it returns and the control shifts to S173.

In the above control, the personalized data of the user is substituted into the general model to make a personalized model dedicated to the user, and then the personalized service is provided to the user by using the personalized model. A handmade personalized model for the user may be generated and used to provide a personalized service to the user. For example, when a YES judgment is made in S187 of FIG. 14, the results of machine learning of each of the plurality of users (subjects) who contributed to the creation of the general learning model are available (see FIG. 15). Therefore, each user of those subjects can control the result of the machine learning performed on their own learning to be a handmade personalized model, and use the personalized model to provide a personalized service to the user. You may. Furthermore, even if the user is not the subject, if he / she wants to generate a handmade personalized model for himself / herself, he / she provides a large amount of his / her own learning data (review data) to the artificial intelligence server 9 and handcrafts it. You may control to have a personalized model of.

Next, with reference to FIG. 17, a flowchart of the muscle training data processing subroutine program shown in S136 of FIG. 12A will be described.

PC10 of a sports gym, which is an example of a specialist, aggregates muscle training data of a large number of members (subjects) and transmits it to an artificial intelligence server 9 by S210. The muscle training data is, for example, the initial load of the subject, the muscle training time and number of times, the muscle training interval from the previous muscle training to the current muscle training, the load increase coefficient, the measured value data of the muscle thickness, and the like. The artificial intelligence server 9 receives the muscle training data in S211 and classifies the received data for various models by S212 and stores it in the learning DB 60. Next, according to S213, learning data is created by converting the received data into numerical values or label set data.

Next, according to S214, modeling is performed using a learning algorithm such as reinforcement learning or regression based on the learning data. This modeling method is similar to that described with reference to FIG. A common model of muscle training is
Fi ＝ F ₀ ＋ (i-1) / a
Is. Here, i is the number of repetitions indicating the number of repeated muscle trainings, Fi is the load on the i-th muscle training (barbell weight, etc.), and F ₀ is the load on the first muscle training (barbell weight, etc.). , A is a load increase coefficient that increases with the repetition of muscle training. These F ₀ and a are factors (individual differences in muscle strength, etc.) that differ from user to user.

Next, by S215, the above general model (Fi = F ₀ + (i-1) / a) is stored in the storage area of the muscle training learning model of the artificial intelligence DB17. If an old and incomplete kintore learning model is already stored _{, update to a newer complete memorization learning model (Fi = F 0} + (i-1) / a).

Next, with reference to FIG. 18A, a flowchart of the subroutine program of the muscle training terminal processing shown in S162 of FIG. 13A and the muscle training server processing shown in S163 will be described.

The user's mobile communication device 3 determines whether or not the user is in muscle training by S220. If it is not during muscle training, it returns and control shifts to S164. When the muscle training is in progress, the control proceeds to S221, and the muscle training data of the user is aggregated and transmitted to the artificial intelligence server 9. The artificial intelligence server 9 that has received it in S222 stores the received data in the learning DB 60 in S113, calculates the initial load FK of the user, the load increase coefficient az, and the super recovery time CT, and stores it in the user DB 12. After returning from S113, control shifts to S165.

Next, with reference to FIG. 18B, a flowchart of the subroutine program of the muscle training service terminal processing shown in S172 of FIG. 13B and the muscle training service server processing shown in S173 will be described.

The user's mobile communication device 3 transmits a muscle training menu request signal to the artificial intelligence server 9 in response to the user's muscle training menu request by S226. The artificial intelligence server 9 that received it in S227 reads the user's initial load FK, load increase coefficient az, and supercompensation time CT from the user DB12 by S228, and reads a general muscle training model (Fi = F ₀ + (i−). Substitute in 1) / a) to create this muscle training menu. Then, by S229, the muscle training menu and the advertisement of the muscle training specialist (see S210) who provided the muscle training data are transmitted to the user's mobile communication device 3. The mobile communication device 3 that has received it in S230 notifies the user of the received muscle training menu and the advertisement of the muscle training specialist by S231. After returning from S231, the control shifts to S9, and after returning from S229, the control shifts to S10.

FIG. 19A shows details (specific examples) of the data stored in the storage area of the dialogue model in the artificial intelligence DB 17 of FIG. In the storage area of the dialogue model, each data of the initial object, the dialogue template, the emotional initial weight, and the emotional change function is stored. Various selection targets such as male intellectual system, female intellectual system, male wild system, female moe system, talent A, talent B, and talent C are stored in the initial target. The data of the dialogue template, the emotional initial weight, and the emotional change function are different for each type of initial target, and each data is stored in association with each type of initial target. Dialogue templates are a large number of dialogue templates collected to find out the response to a user's dialogue by template matching. For example, in the case of male intellectuals, the proportion of intellectual dialogue templates is high.

The emotional initial weight and the emotional change function are for controlling the emotional change of emotions in response to dialogue. Control to change to "joy" emotions when the user praises, control to change to "anger" emotions when the user deprecates, and change to "sorrow" emotions when the user mourns Control it so that if the user enjoys it, it will change to a "comfortable" emotion.

However, since the weights of emotions and emotions differ depending on the type of the initial object and the mode of change also differs, the initial weights of emotions and emotions and the emotions and emotions change function differ depending on the type of each initial object. The emotional initial weight and the emotional change function are found by machine learning using a large number of dialogue contents collected in dialogues with a large number of users as learning data. A large number of users are classified into male intellectual system, female intellectual system, male wild system, and female moe system, and the average for each system is calculated to match the initial weight and emotions. Create a change function. For talent A, talent B, talent C, etc., a large number of dialogue data collected in the actual dialogue of a certain talent (for example, talent A) is used as learning data, and the initial weights and emotions of emotions and emotions are obtained by machine learning. Find out the sorrow change function. The emotion / emotion change function may be a unified function common to all human beings, and the difference (variation) of each person may be controlled so as to be adjusted by the emotion / emotion initial weight.

As an example of the emotion change function F (x), the following can be considered. Emotions and sorrows are represented by K, D, I, and R, and emotionless states are represented by M. Let the initial weights of emotions be K, D, I, and R, respectively. Also, if the sigmoid function 1 / (1 + e ^-x ) is expressed as S (x) for convenience,
F (x) = KS (x1-K) + DS (x2-D) + IS (x3-I) + RS (x4-R) + M (1-S (x1-K) -S (2x-D) -S (x3) −I) −S (x4-R))
Will be. That is, if any of x1, x2, x3, and x4 exceeds the respective threshold values (initial weights) K, D, I, and R, the emotions that exceed the thresholds (one of joy, anger, sorrow, I, and R) change. However, if none of them exceeds the above threshold value, it becomes emotionless M.

FIG. 19B shows the details of the data stored in the personalized data storage area for dialogue in the user DB 12 of FIG. FIG. 19B shows a case where a user with user ID: 1 selects and specifies “talent A” as the initial target of the dialogue model. And "0.7AK, 1.5AD, 1.3AI, 0.5AR" is memorized as the personal weight of emotions.

The emotional initial weight of "talent A" is "AK, AD, AI, AR" as shown in FIG. 19 (a), but the user has a dialogue based on this emotional initial weight. During the process, the initial weights of emotions and sorrows "AK, AD, AI, AR" were gradually modified according to the user's request, and at the moment it is "0.7AK, 1.5AD, 1.3AI, 0.5AR". .. As a result, the emotional change function for this user's emotional and emotional personal weights "0.7AK, 1.5AD, 1.3AI, 0.5AR" is as follows.
F (x) = KS (x1-0.7AK) + DS (x2-1.5AD) + IS (x3-1.3AI) + RS (x4-0.5AR) + M (1-S (x1-0.7AK) -S (2x-1.5) AD) -S (x3-1.3AI) -S (x4-0.5AR))

As a result, compared to the actual talent A, Ki K and Raku R are more likely to appear, and Anger D and Sad I are less likely to appear.

In this way, it is possible to select and specify an initial target that suits the user's preference, and it is possible to modify the personal weight of emotions and emotions that more reflects the user's preference through further dialogue. In other words, in the case of a human partner, it is almost impossible to change the partner into one's ideal human being, but in the case of artificial intelligence, it is possible to create (nurture) the other party (artificial intelligence) into one's ideal companion. Can be done. Until now, it was an era of searching for an ideal lover, an ideal best friend, and an ideal companion, but in the future world, one's ideal lover, ideal best friend, and ideal companion will be on the network (on the cloud). It will be an era of making a companion.

Next, with reference to FIG. 20, a flowchart of a subroutine program of the dialogue terminal process shown in S9 of FIG. 4 and the dialogue server process shown in S10 will be described.

The mobile communication device 3 such as a wearable computer determines in S235 whether or not the user has started the dialogue, and if not, returns and the control shifts to S11. When the user side starts the dialogue, the control proceeds to S236, and the user ID and the dialogue start signal are transmitted to the artificial intelligence server 9. The artificial intelligence server 9 that has received the result determines YES in S241, determines whether or not the initial target has been received in S242, and if not, the control shifts to S244.

On the other hand, in the mobile communication device 3, it is determined by S237 whether or not the selection designation of the initial target has been made. If the selection is not specified, the control shifts to S239, but if the selection is specified, the specified initial target is transmitted to the artificial intelligence server 9 by S238. Upon receiving this, the artificial intelligence server 9 determines YES in S242 and updates the initial dialogue target corresponding to the user ID in the user DB 12 to the specified one in S243.

Next, the mobile communication device 3 interacts with each other by S239, and the artificial intelligence server 9 interacts with each other by S244. At that time, the artificial intelligence server 9 interacts with the initial dialogue target corresponding to the user ID in the user DB 12, the personal weight of emotions, and the corresponding emotion change function in the artificial intelligence DB 17. The artificial intelligence server 9 is equipped with a natural language processing engine, and in addition to the processing during text mining such as "morphological analysis" and "syntax analysis (processing to determine the dependency relationship between phrases)", it grasps the structure of sentences. It incorporates two techniques, "context analysis" necessary to do so and "semantic analysis" necessary to understand the intent of a sentence. This makes it possible to analyze the user's voice "faster" and "more precisely".

The artificial intelligence server 9 determines whether or not it is necessary to change the personal weight of emotions during the above dialogue (S245), and if necessary, updates the personal weight of emotions by S246. .. For example, the artificial intelligence server 9 recognizes that the user has issued a predetermined cliché such as "Please reduce the weight of anger a little more", and controls to update the personal weight of emotions. In addition to or in addition to the above-mentioned cliché, control may be performed to update the personal weight of emotions and sorrows in response to a user's remark such as "You have happened too much".

Next, the artificial intelligence server 9 determines in S247 whether or not the dialogue content is related to work. If it is not related to work, the process proceeds to S249 to determine whether or not the dialogue is completed, but if it is related to work, the process proceeds to S248 and the task server process is performed.

In the artificial intelligence server 9, when it is determined by S249 that the dialogue has not ended, the control shifts to S244, the steps S244 to S249 are repeatedly executed, and when the dialogue ends, the control returns to S12. Move to. On the other hand, in the mobile communication device 3, when it is determined by S240 that the dialogue has not ended, the control shifts to S237, the steps S237 to S240 are repeatedly executed, and when the dialogue ends, the control returns and controls. Shifts to S11.

FIG. 21A shows details (specific examples) of the data stored in the storage area of the task processing model in the artificial intelligence DB 17 of FIG. In the storage area of the task processing model, the initial target, the initial weight of each knowledge, and each data of the knowledge utilization function are stored. Various selection targets such as patent attorneys, lawyers, physicists, chemists, scientists A, scientists B, and scientists C are stored as initial targets. The initial weight of each knowledge and each data of the knowledge utilization function are different for each type of initial target, and each data is stored in association with each type of each initial target.

The initial weight of each knowledge and the knowledge utilization function are for making it easier for each initial target to use the knowledge that matches the specialized field when performing task processing such as work. The initial weight of each knowledge and the knowledge utilization function are found by machine learning by converting a large number of dialogue contents collected in work dialogues with a large number of users into learning data. A large number of users are classified by profession such as patent attorney, lawyer, physicist, chemist, etc., the average of each profession is calculated, and the initial weight of each knowledge and the knowledge utilization function suitable for each initial target are calculated. create.

For example, a patent attorney user asks a large number of questions and consultations on the artificial intelligence server 9 at work, and a large amount of reactions such as "satisfied" or "not satisfied" with respect to the response from the artificial intelligence server 9. The data is used as training data and modeled by algorithms such as reinforcement learning and regression. For scientist A, scientist B, scientist C, etc., a large number of dialogue data collected in the actual dialogue of a certain scientist is used as learning data, and the initial weight of knowledge and the knowledge utilization function are obtained by machine learning. Find out. The knowledge utilization function may be a unified function common to all human beings, and the difference (variation) of each person may be controlled so as to be adjusted by the initial weight of knowledge.

The following can be considered as an example of the knowledge utilization function G (x). Each knowledge is represented as n1, n2, n3, ..., And the number of positive reactions such as "satisfied" by the user to each knowledge is x1k, x2k, x3k, ... "not satisfied". Let the number of negative reactions such as, etc. be x1h, x2h, x3h, ..., And the initial weights of each knowledge be w ₁ , w ₂ , w ₃ , ..., Respectively. Also, if the sigmoid function 1 / (1 + e ^-x ) is expressed as S (x) for convenience,
G (x) = w ₁ n1S (x1k-x1h) + w ₂ n2S (x2k-x2h) + w ₃ n3S (x3k-x3h) + ...
Will be. That is, knowledge in which the number of positive reactions exceeds the number of negative reactions is regarded as effective knowledge to be used, and the coefficients of each knowledge n1, n2, n3, ... (Weights w ₁ , w). ₂ , w ₃ , ...) Can be controlled so that they are preferentially used from the knowledge of large values.

FIG. 21B shows the details of the data stored in the storage area of the personalized data for task processing in the user DB 12 of FIG. FIG. 21B shows a case where a user with user ID: 1 selects and specifies "patent attorney" as the initial target of the dialogue model. _{Then, "1.2 ・ 1w 1} 1n ₁ , 0.8 ・ 1w ₂ 1n ₂ , 0.7 ・ 1w ₃ 1n ₃ , 1.3 ・ 1w ₄ 1n ₄ " is stored as the personal weight w of each knowledge n.

The initial weight of each knowledge of the "patent attorney" is "1w ₁ 1n ₁ , 1w ₂ 1n ₂ , 1w ₃ 1n ₃ , 1w ₄ 1n ₄ " as shown in FIG. 21 (a). Is conducting a business dialogue based on the initial weight of each knowledge, and according to the request of the user, the initial weight of each knowledge "1w ₁ 1n ₁ , 1w ₂ 1n ₂ , 1w ₃ 1n ₃ , 1w ₄ 1n ₄ ”has been gradually modified to become“ 1.2 ・ 1w ₁ 1n ₁ , 0.8 ・ 1w ₂ 1n ₂ , 0.7 ・ 1w ₃ 1n ₃ , 1.3 ・ 1w ₄ 1n ₄ ”. In this way, in addition to being able to select and specify the initial target that suits the user's profession, it is possible to change it to a personal weight of each knowledge that more reflects the user's questions and consultation contents as the user's questions and consultation contents are repeated. Can be done. In other words, in the future world of the future, it will be an era in which one's ideal work partner (buddy) will be created on the network (on the cloud).

Next, with reference to FIG. 22, a flowchart of the subroutine program of the task terminal process shown in S11 of FIG. 4 and the task server process shown in S12 and S248 of FIG. 20 will be described.

The user's mobile communication device 3 determines in S255 whether or not the user has requested to register the personal weight of himself / herself (user) in the artificial intelligence DB 17 as an initial target. If not, the control shifts to S259, it is determined whether or not there is an update request for the initial target, and if not, the control shifts to S265.

On the other hand, when S255 requests that the personal weight of oneself (user) be registered in the artificial intelligence DB 17 as an initial target, S255 determines YES, the control proceeds to S256, and the user ID and initial target registration request. Is sent to the artificial intelligence server 9.

In the artificial intelligence server 9, S257 determines whether or not the initial target registration request has been received, and if not, the control shifts to S261, determines whether or not the initial request target has been received, and receives the request. If not, control shifts to S267.

If the user ID and the initial target registration request are transmitted from the mobile communication device 3 according to the process of S256, a YES judgment is made by S257, the control proceeds to S258, and the personalized weight of each knowledge corresponding to the user ID is set by the user. A process of reading from the DB 12 and registering it in the artificial intelligence DB 17 together with the user name is performed. As described above, in the present embodiment, the user can register the personalized weight of each knowledge of the user in the artificial intelligence DB 17 and distribute it widely to the general public. For example, when efficiently educating one's work subordinates to become a full-fledged expert, the personalization weight of each work knowledge that the boss has cultivated up to now is transferred to that subordinate. Then, the artificial intelligence server 9 will answer the questions and consultations of the subordinates on your behalf, and the artificial intelligence server 9 will take over the education of the subordinates. When the user transfers the personalized weight of each knowledge of the user to another person, a payment means for settlement with the transferee may be provided and the transfer may be made for a fee.

If the user requests the mobile communication device 3 to update the initial target and specifies a new initial target (request initial target), a YES judgment is made by S259, the control proceeds to S260, and the user ID and request. The initial target is transmitted to the artificial intelligence server 9. Upon receiving this, the artificial intelligence server 9 determines YES in S261, the control proceeds to S262, and determines whether or not the received initial request target has been previously used by the user. If it has not been used yet, S264 reads the requested initial target from the artificial intelligence DB 17 and stores it corresponding to the user ID of the user DB 12, and the control proceeds to S267.

If the received request initial target is previously used by the user, the initial target may already be stored in the user DB 12 and the personal weight may be updated with the use of the initial target. There is (see S271). Therefore, when the received request initial target is the one previously used by the user, the control proceeds to S263 and the personal weight of the previously used knowledge stored in the user DB 12 is used.

Next, referring to FIG. 23, the artificial intelligence server 9 reads out the initial target corresponding to the user ID of the user DB 12 and the personal weight of each knowledge by S267, and the knowledge corresponding to the read initial target by S268. The utilization function is read from the artificial intelligence DB17. Then, when the user engages in a business dialogue using the mobile communication device 3, the dialogue is performed between S265 on the mobile communication device 3 side and S269 on the artificial intelligence server 9 side. In S269, the read knowledge utilization function and the personal weight are used to perform a business dialogue. On the mobile communication device 3 side, the work dialogue control by S265 is continued until it is determined by S266 that the work dialogue is completed. On the artificial intelligence server 9 side, the control of S267 to S274 is repeatedly performed until it is determined by S274 that the work dialogue is completed, and the work dialogue control by S269 is continued.

If it is determined by S270 that the personal weight needs to be changed while the controls of S267 to S274 are being repeatedly performed, the control proceeds to S271 and the personal weight of the user DB 12 is updated. If the personal weight is also registered in the artificial intelligence DB 17 by the user (see S255 to S258), a YES determination is made by S272, and the weight of the artificial intelligence DB 17 is also updated by S273. This judgment by S270 is used on the artificial intelligence server 9 side by, for example, the user saying a predetermined cliché such as "satisfied" or "not satisfied" with respect to the response from the artificial intelligence server 9. Determine the need to change the weight of knowledge you have. In addition to or in addition to the above cliché, the judgment may be made based on the user's normal conversation such as "Thank you for the wonderful answer".

Then, when it is determined by S266 on the mobile communication device 3 side that the work dialogue is completed, the work dialogue end instruction is transmitted to the artificial intelligence server 9 by S273, and the return is returned to S1. Transition. Upon receiving this, the artificial intelligence server 9 makes a YES judgment in S274, returns, and the control shifts to S3.

Modifications and feature points of the embodiments described above are described below.
(1) Even if at least one or all of the IoT server 9 and the device DB 15 for IoT, the artificial intelligence server 9, the artificial intelligence DB 17, the user DB 12 and the learning DB 60, and the SNS server 11 and the SNS DB 13 are configured by the network cloud. good. Further, in the above-described embodiment, the mobile communication device 3 has been mainly described as an example of the user-side terminal, but the user-side terminal may be the user PC 7 or the robot 6. Further, as a specific example of the mobile communication device 3, a wearable computer represented by smart glasses or the like has been shown, but the present invention is not limited thereto, and for example, a mobile phone, a smartphone, a vehicle such as an automobile having a communication function, or the like. Various things can be considered.

(2) The IoT server 8 is provided with indexing means (S24 to S26, S38, S40) for indexing IoT devices that need to exchange information. Then, the geographical position of the IoT device calculated by the indexing means may be transmitted to the wearable computer to be controlled to notify the user, and the user may be encouraged to act as an intermediary. At that time, it may be controlled to notify the user that a high score (points, etc.) will be given to the IoT device, which requires particularly high mediation.

(3) It may be controlled to give a large amount of benefits (points, etc.) only to a limited number of IoT devices, and the mediation act may be promoted by using the same psychology as the lottery. In addition, it may be controlled so that the mediation act of the IoT device laid at the geographical position is performed by using the location game.

(4) The following inventions are disclosed in the embodiments described above.
The present invention relates to, for example, a system that suppresses the cost of IoT devices (sensors, actuators, etc.) laid all over the earth and realizes IoT at low cost.

For example, the following are background technologies related to systems that realize IoT at low cost. A border router (gateway device) is installed between the sensor for IoT and the server on the Internet, the server address is not set for the sensor, and the sensor data information is sent to the border router after its own address is confirmed. Send. The border router relays the sensor information to the sensor data management server to collect data. As a result, the server side can detect the address of the sensor and collect pull-type data addressed to the sensor address (Japanese Unexamined Patent Publication No. 2014-78773).

However, in the case of this background technology, although it is possible to collect pull-type data addressed to the sensor address from the server side, each sensor is provided with an Internet connection function such as a network I / F, or an original standard for connecting to a network. There was a drawback that it was necessary to prepare an adapter for this, which caused an increase in cost in realizing IoT.

On the other hand, some IoT devices laid all over the earth do not necessarily need to collect pull-type data addressed to the sensor address from the server side. For example, in the case of an illuminance sensor installed to aggregate the illuminance throughout the year, the periodic illuminance detection result is stored in the memory, and the illuminance aggregation period is performed once a year on the server side. It suffices to send the stored illuminance detection result to the server. Further, in the case of an IoT device installed in a place where many humans come and go, a method of transmitting the data of the IoT device to the server by an artificial method using the human is also conceivable.

The present invention was conceived in view of such circumstances, and an object of the present invention is to provide a system capable of reducing the cost of realizing IoT.

The present invention describes a wearable computer owned by a user and a wearable computer.
A server that communicates with the wearable computer via the Internet,
Equipped with a group of IoT devices
The IoT device group includes a communication means for communicating with the wearable computer.
The wearable computer is
A device-to-device communication means for communicating with the IoT device group,
A means of communicating with the server for communicating with the server via the Internet,
An IoT system including an IoT device of a communication partner by the device-to-device communication means and an intermediary means that mediates between the communication partner's server by the server-to-server communication means and enables information to be exchanged between the two.

According to such a configuration, the Internet connection function is not always required on the IoT device group side, the cost can be suppressed accordingly, and IoT can be realized at low cost.

Preferably, the server further includes privilege-giving means for granting a predetermined privilege to the wearable computer that has been mediated by the mediator.

According to such a configuration, it is possible to give the user an incentive for mediation, and it is possible to promote the user's mediation act.

More preferably, the server further includes a determination means for determining whether or not there is an IoT device that needs to exchange information.

With such a configuration, it is not necessary for the IoT device group side to have a function of determining whether or not information needs to be exchanged, the processing load on the IoT device group side can be reduced, and the cost is further reduced. be able to.

More preferably, the wearable computer further includes a location information transmitting means for transmitting information that can identify the current position to the server.
The server
A position determining means for determining whether or not the IoT device indexed by the indexing means exists in the vicinity of the wearable computer whose position is specified by the information transmitted by the position information transmitting means.
Further including an intermediary command transmitting means for transmitting a command signal for intermediating the IoT device to the wearable computer when it is determined to be present by the position determining means.
The intermediary means mediates according to the command signal from the intermediary command transmitting means.

According to such a configuration, since the wearable computer side acts on the IoT device that requires mediation, it is not necessary to transmit the mediation request signal from the IoT device group side.

More preferably, the wearable computer
Sensors that take a certain amount of time to detect,
Further including a notification means (for example, S71) for notifying the user of the detection time required by the sensor when performing detection using the sensor at a desired geographical position.
When the sensor is able to detect the required detection time within the detection area of the desired geographical position, the detection data is mediated by the mediation means and transmitted to the server.

According to such a configuration, when it takes a certain amount of time to detect, the wearable computer notifies the user of the time required for detection, so that the user cooperates in collecting the detection data which requires time. Can be urged to.

(5) In the embodiment described above, the creation of a building maintenance / inspection model, a memorization learning model, a muscle training model, an dialogue model, and a task processing model has been shown as specific examples of machine learning, but the present invention is limited to these. For example, a learning model for infants, a model for artificial secretaries, a model for artificial home teachers, and the like can be considered. Further, the personalized data is not limited to memorization learning, muscle training, dialogue, and task processing, and various data such as learning for infants, artificial secretary, and artificial tutor can be considered. These general models and personalized data are controlled to be switched and used properly according to the user's situation. As a result, shortly after birth, humans grow up with artificial intelligence-based personalization services using their own personalized data, and live their lives with artificial intelligence. In other words, in the future world of the future, it will be a world in which humans and artificial intelligence are paired. It is a world in which humans and artificial intelligence are paired to form a single personality.
Further, when using the general model and the personalized data properly, in the present embodiment, the personalized data for dialogue is the conventional one unless the user indicates the intention to switch, but the dialogue is performed when switching between the general model and the personalized data. Personalized data for may also be controlled to automatically switch to other suitable ones. For example, in general daily conversation, the talent A is used as an initial target for dialogue, and in the case of business dialogue, it is controlled to automatically switch to the female intellectual system (see FIG. 19).

(6) Based on the around data collected from the user by the mobile communication device 3 or the like, the recommendation control that recommends the artificial intelligence or the general model that seems to match the user may be performed. In addition, the search results displayed in a list when the user searches for artificial intelligence or a general model are preferentially displayed from those that are considered to match the user based on the around data collected from the user. It may be controlled as follows. Further, it may be controlled to collect and display user reviews for artificial intelligence and general models. At that time, as described above, since humans and artificial intelligence are paired to form one personality, the paired humans and artificial intelligence are grouped together to collect word-of-mouth about it. May be controlled to be displayed.

(7) For infants learned using the above general model for learning for infants and personalized data, a huge amount of data consisting of the provided learning materials (learning information) and learning results is collected and machined. Machine learning (supervised learning, etc.) may be performed on human intellectual growth by using learning data for learning. The same learning material (learning information) provided to the above infants is also provided to artificial intelligence, and as a result, the learning results of artificial intelligence that has achieved intellectual growth and the actual infants are compared, and the differences are minimal. You may perform reinforced learning or regression so that it becomes possible, and create a learning model for artificial intelligence to enable learning close to humans. In particular, in fields where artificial intelligence is not good (for example, common sense, emotions, creativity, etc.), it is considered that machine learning (supervised learning, etc.) that is useful for artificial intelligence will be achieved by comparing the learning results of both.

(8) In the present embodiment, personalization means for personalizing the general model into a model suitable for the user based on the information sent from the user (for example, S191, S223, S246, S271). ) Is provided in the artificial intelligence server 9, but instead of or in addition to it, it may be provided in a terminal on the user side (for example, a mobile communication device 3 such as a wearable computer, a user PC, a robot, etc.).

The following inventions are disclosed in the embodiments described above.
The present invention relates to, for example, a service providing system and a program using machine learning by artificial intelligence. More specifically, the present invention relates to a service providing system that provides a service using machine learning by artificial intelligence, and a program executed by a computer in order for a user to receive a service using machine learning by artificial intelligence.

In order to perform machine learning by artificial intelligence, there is one that creates learning data from each of a plurality of information sources (for example, Japanese Patent Application Laid-Open No. 2011-232997).

However, a huge amount of learning data is required for machine learning, and in order to create this huge amount of learning data, it is necessary to collect a large amount of necessary information from information sources, and machine learning is put into practical use. In order to do so, it is indispensable to reduce the labor and cost required for collecting this information.

Also, for example, a general model modeled by collecting a large amount of information from an unspecified number of people and performing machine learning with a huge amount of learning data is an average model for the entire unspecified number of people. If a service is provided to each user who has individual differences using the general model, there is a risk that the service will not match the users who deviate from the average.

The disclosed invention has been conceived in view of such circumstances, and an object thereof is to reduce the labor and cost required for collecting a huge amount of information necessary for creating learning data. A further objective is to solve the inconvenience that a service using a general model of machine learning results may not match the user.

Next, an example of the correspondence between various means for solving the problem and the embodiment is shown in parentheses below.

The present invention is a service providing system (for example, FIGS. 9 and 10) that provides a service using machine learning by artificial intelligence.
General modeled by machine learning by inputting learning data based on information sent from the user (for example, video, line-of-sight position, voice, GPS position data, and other around data that the user 70 is viewing) Machine learning means (eg, S189, S214, FIG. 15) for generating a typical model (for example, memorization learning model Ti = T ₀ · I, muscle training model Fi = F _{0 + (i-1) / a, etc.)} Etc.) and
Personalization means for personalizing the general model to a model suitable for the user based on the information sent from the user (for example, S191 and S204, S223 and S228, etc.), and
Personalized service to the user using the personalized model (for example, provision of learning items based on the most efficient review plan, provision of muscle training menu based on the most efficient muscle training plan, etc. ) Is provided with a service providing means (for example, S206, S229, etc.).
The information sent from the user is used for both the machine learning and the personalization.

According to such a configuration, by providing a service using the result of machine learning to a user, a large number of users take the initiative in providing information for learning data. Moreover, since the information provided by the user is also used to personalize the general model to the model suitable for the user, each user leaves the information provision for the learning data to others and only the service. The inconvenience of enjoying it can be prevented as much as possible.

Preferably, the machine learning means has constant elements (for example, individual ability differences such as memorization ability, initial load for muscle training, load increase coefficient, preference, etc.) that differ from user to user (for example, T ₀ , F). _{Generate the general model as (0} , a, etc.) (eg, S189, S214, FIG. 15, etc.).
The personalization means
A value deriving means (eg, TK, FK, az, CT, etc.) for deriving the actual value (eg, TK, FK, az, CT, etc.) of the user that applies to the constant part in the general model based on the information sent from the user. S191, S223, etc.) and
Substitution means (for example, S204, 228, etc.) for substituting the actual value derived by the value derivation means into the constant portion in the general model and personalizing the model suitable for the user. include.

More preferably, the machine learning means also inputs learning data based on information sent from a specialist (for example, a muscle training specialist) to generate a general model modeled by machine learning. (For example, S214 etc.).

Another aspect of the present invention is a service providing system that provides a service using machine learning by artificial intelligence.
Storage means for storing a plurality of types of general models (for example, the dialogue model of FIG. 19A, the task processing model of FIG. 21A, etc.) generated by modeling using machine learning (for example, Artificial intelligence DB17 etc.) and
A service (for example, a dialogue service, a response to a business question or a consultation, etc.) is provided to the user by using a general model selected by the user from a plurality of types of general models stored in the storage means. ) (For example, S241 to S244, S247, S248, S261 to S263, S267 to S269, etc.)
A personalization means (for example, S246, S271, etc.) for personalizing the general model into a model suitable for the user according to the reaction of the user to the service provided by the service providing means is provided.
The service providing means provides personal service providing means for providing a service to the user using a model personalized by the personalized means (for example, executing S244, S269, etc. according to an updated personal weight). include.

According to such a configuration, a general model desired by the user is selected from a plurality of types of general models generated by modeling using machine learning, and the service is used by using the general model. Is provided to the user, so that the service can be enjoyed by a general model that reflects the user's wishes. Moreover, since the general model is personalized to the model suitable for the user by the reaction of the user who received the service, it is possible to provide the service matching the user by personalizing the provided service.

Preferably, the personal service providing means is personalized at the time of use when the general model selected by the user has previously been used to provide the service to the user (eg, YES in S262). It has a means (for example, S263, etc.) for providing a service to a user using the model.

More preferably, it further includes model use means (for example, S257, S258, etc.) for allowing another person to use the model personalized by the personalization means.

Another aspect of the disclosed invention is a program executed by a computer in order for a user to receive a service utilizing machine learning by artificial intelligence.
The user desires from a plurality of types of general models (for example, the dialogue model of FIG. 19 (a), the task processing model of FIG. 21 (a), etc.) generated by modeling using machine learning by artificial intelligence. Steps to select and specify what to do (for example, S237 and S238, S259 and S260, etc.) and
Processing steps (for example, S236 to S239, S260, S265, etc.) for the user to receive a service using the general model selected and specified by the selection and specification step, and
A step of providing the artificial intelligence with the reaction of the user in order to personalize the general model into a model suitable for the user according to the reaction of the user serviced by the processing step (eg, S239, S265). Etc.) and,
Let the computer run
The processing step is a personal service enjoyment step for the user to receive a personalized service using the model personalized by the artificial intelligence (for example, a response according to the updated personal weight is sent to S239, S265, etc.). Etc.) including.

According to such a configuration, a general model desired by the user is selected from a plurality of types of general models generated by modeling using machine learning, and the service is used by using the general model. Is provided to the user, so that the service can be enjoyed by a general model that reflects the user's wishes. Moreover, since the general model is personalized to the model suitable for the user by the reaction of the user who received the service, it is possible to provide the service matching the user by personalizing the provided service.

Although the embodiments of the present invention have been described above, it should be considered that the embodiments disclosed this time are examples in all respects and are not restrictive. The scope of the present invention is shown not by the above description but by the scope of claims, and is intended to include the meaning equivalent to the scope of claims and all modifications within the scope.

1 Internet, 2 IoT devices, 3 mobile communication devices, 4 wireless sensor networks, 8 IoT servers, 9 artificial intelligence servers, 10 PCs of various specialists, 12 user DBs, 15 IoT device DBs, 17 artificial intelligence DBs, 60 Learning DB, 38 Various sensors.

Claims (3)

A service provision system that provides services for users and artificial intelligence to interact with each other.
A dialogue model generated by modeling using machine learning, which is a general model used when a user interacts with artificial intelligence, and a task with specialized knowledge in a specific field. A task processing model for processing, a storage means for storing, and
A service providing means for providing a service for a user and artificial intelligence to have a dialogue using the model, and a service providing means.
Through interaction with Ruyu over The by the said interaction model, and a personalized means for personalizing the said interaction model the interaction model suitable for those該Yu over THE,
The dialogue model is a plurality of dialogue models including a specific talent dialogue model generated by using machine learning using the dialogue data of an actual talent as learning data, and can be selected by the user.
The personalization means makes the selected dialogue model suitable for the user through a dialogue between the selected dialogue model selected by the user among the plurality of dialogue models and the user who has selected the selected dialogue model. Personalized to
The service providing means
Using the selected interactive models personalized by the personalization unit, a personal service providing means for providing a service for performing interactive the user who has selected the selected interaction model,
When the user is interacting with the selected dialogue model, the user can take turns interacting with the task processing model according to the content of the dialogue, and can use the knowledge specific to a specific field of the task processing model. A service providing system, including specialized knowledge utilization means. The service providing system according to claim 1, wherein the dialogue model includes an emotion change control means that changes emotions according to a dialogue with a user. A program run by a computer to receive services for users to interact with artificial intelligence.
A dialogue model generated by modeling using machine learning, which is a general model used when a user interacts with artificial intelligence, and a task with specialized knowledge in a specific field. A task processing model for processing, and processing steps for the user to receive a service for interacting with artificial intelligence using
Let the computer run
The dialogue model is a plurality of dialogue models including a specific talent dialogue model generated by using machine learning using the dialogue data of an actual talent as learning data, and can be selected by the user.
The processing step
Through interaction with the user selecting the selected selected interaction model and the selected interaction model by the user among a plurality of said interaction model, personalize the selected interaction model the interaction model suitable for those該Yu over THE and personal services enjoyed step for Kiyu over the receives the service to perform the pre-interaction using the selected interactive model,
When the user is interacting with the selected dialogue model, the user can take turns interacting with the task processing model according to the content of the dialogue, and can use the knowledge specific to a specific field of the task processing model. A program that includes specialized knowledge utilization steps and so on.

Images