Support system

Abstract

To provide a service that matches a user by personalizing a service to be provided.SOLUTION: During a dialogue using a dialogue model for general dialogue (NO in S235-S246, S247, No in S249), the dialogue model is switched to a task processing model in accordance with the content of the dialogue (NO in S247, S248), to allow a user to utilize the knowledge dedicated to a specific field of the task processing model. Through dialogues with individual users using dialogue models, the dialogue models are personalized to dialogue models suitable for the individual users (YES in S245, S246).SELECTED DRAWING: Figure 20

Description

[Technical Field]

[0001] The present invention relates to a support system that uses artificial intelligence to support a user in repeated practice for self-improvement. [Background technology]

[0002] There have been techniques for creating learning data from multiple information sources to perform machine learning using artificial intelligence (for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2011-232997 Summary of the Invention [Problem to be solved by the invention]

[0004] However, machine learning requires a huge amount of training data, and in order to create this vast amount of training data, it is necessary to collect a large amount of necessary information from sources. In order to put machine learning into practical use, it is essential to reduce the effort and cost required to collect this information.

[0005] Furthermore, for example, a general model created by collecting large amounts of information from an unspecified number of people and performing machine learning using a huge amount of learning data tends to be an average model for the entire unspecified number of people, and if that general model is used to provide a service to each user, each of whom is unique, there is a risk that the service will not be a good match for users who deviate from the average.

[0006] The present invention was conceived in light of the above situation, and its purpose is to solve the inconvenience that arises when a service using a general model resulting from machine learning does not match the user.

[0007] Next, an example of the correspondence between various means for solving the problems and embodiments is shown below in parentheses.

[0008] The present invention is a support system (e.g., Figs. 9 and 10 ) that uses artificial intelligence to support a user's repeated practice (e.g., rote learning, muscle training, etc.) for self-improvement, A data collection means (e.g., S182, S211, etc.) for collecting data on a plurality of users who perform the practice; a storage means (e.g., an artificial intelligence DB 17) for storing general models (e.g., a rote learning model Ti=T0·i, a muscle training model Fi=F0+(i−1) / a, etc.) generated by performing machine learning (e.g., reinforcement learning, regression, etc.) based on the data of multiple users collected by the data collection means; and a personalized model generation means (e.g., S191 and S204, S223 and S228, etc.) that generates a personalized model personalized for an individual user based on the general model stored in the storage means and the individual data for one user collected by the data collection means; a practice content specification means (e.g., S204, S228, etc.) that specifies practice content to be executed when the individual user repeatedly performs practice using the personalized model generated by the personalized model generation means; a service providing means (e.g., S206, S229, etc.) that provides a service for the individual user to perform the practice content identified by the practice content identifying means (e.g., providing study items based on the most efficient review plan, providing a muscle training menu based on the most efficient muscle training plan, etc.), The general model is a model that applies to multiple users in common and is generated to improve the efficiency of the repeated practice (e.g., the rote learning model Ti=T0·i in FIG. 15, the muscle training model Fi=F0+(i−1) / a, etc.), The personalized model is a model created exclusively for the individual user in order to streamline the repeated practice of the individual user (for example, S204, S228, etc.).

[0009] This configuration can solve the problem of a service that uses a general model resulting from machine learning not matching the user.

[0010] Preferably, the general model is a model generated using factors that differ for each user (e.g., individual differences in ability, preferences, etc., such as memorization ability, initial load for muscle training, and load increase coefficient) as constants (e.g., T0, F0, a, etc.) (e.g., S189, S214, FIG. 15, etc.), The personalized model generation means A value derivation means (e.g., S191, S223, etc.) for deriving actual values ​​(e.g., TK, FK, az, CT, etc.) of the individual user that fit into the constant part of the general model based on information sent from the individual user; and substitution means (for example, S204, 228, etc.) for substituting the actual value derived by the value derivation means into a constant part in the general model to generate the personalized model.

[0011] Preferably, the data collection means collects data of a plurality of users sent from a professional company (such as the professional company 71 in FIG. 9) that provides the practice to users. [Brief explanation of the drawings]

[0012] [Figure 1] This is a system diagram showing the overall configuration of an IoT device intermediation system and a service provision system that uses machine learning. [Figure 2] FIG. 1(a) is a block diagram showing a control circuit of an IoT device, and FIG. 1(b) is a block diagram showing a control circuit of a mobile communication device. [Figure 3] FIG. 1 is a block diagram showing a control circuit for a PC or a server. [Figure 4]4 is a flowchart showing the main routine program of the mobile communication device and the artificial intelligence server. [Figure 5] 1 is a flowchart showing a subroutine program for IoT processing and a flowchart for an IoT device (sensor) and an IoT server. [Figure 6] 1 is a flowchart showing a subroutine program for IoT processing and a flowchart for an IoT device (actuator) and an IoT server. [Figure 7] 1 is a flowchart showing a subroutine program for IoT processing and a flowchart for an IoT device (actuator) and an IoT server. [Figure 8] 1 is a flowchart showing a subroutine program for IoT processing and a flowchart for a wireless sensor network and an IoT server. [Figure 9] FIG. 1 is an explanatory diagram illustrating an overview of a service providing system that uses machine learning. [Figure 10] FIG. 1 is a functional block diagram showing a service providing system that uses machine learning. [Figure 11] 10 is a flowchart showing a subroutine program for external display processing and display processing, and a flowchart for a digital menu board. [Figure 12] 10A is a flowchart showing a subroutine program for processing expert data, and FIG. 10B is a flowchart showing a subroutine program for processing maintenance data and a flowchart for the maintenance specialist's PC. [Figure 13] 1A is a flowchart showing a subroutine program for around data terminal processing and around data server processing, and FIG. 1B is a flowchart showing a subroutine program for personal service terminal processing and personal service server processing. [Figure 14] 10 is a flowchart showing a subroutine program of a user learning terminal process and a user learning server process. [Figure 15]FIG. 1 is an explanatory diagram illustrating a general model creation method. [Figure 16] 10 is a flowchart showing a subroutine program of learning service terminal processing and learning service server processing. [Figure 17] 10 is a flowchart showing a subroutine program for processing muscle training data and a flowchart of a PC of a muscle training professional. [Figure 18] 10A is a flowchart showing a subroutine program for muscle training terminal processing and muscle training server processing, and FIG. 10B is a flowchart showing a subroutine program for muscle training service terminal processing and muscle training service server processing. [Figure 19] FIG. 1A is a diagram showing a specific example of a dialogue model stored in the artificial intelligence DB, and FIG. 1B is a diagram showing a specific example of personalized data for dialogue stored in the user DB. [Figure 20] 10 is a flowchart showing a subroutine program of an interactive terminal process and an interactive server process. [Figure 21] FIG. 1A is a diagram showing a specific example of a task processing model stored in an artificial intelligence DB, and FIG. 1B is a diagram showing a specific example of personalized data for task processing stored in a user DB. [Figure 22] 10 is a flowchart showing a subroutine program of task terminal processing and task server processing. [Figure 23] 10 is a flowchart showing a subroutine program of task terminal processing and task server processing. DETAILED DESCRIPTION OF THE INVENTION

[0013] [IoT device intermediation system] First, an IoT device intermediation system will be described with reference to Figures 1 to 8. IoT is an abbreviation for Internet of Things, in which things are networked using the Internet Protocol and send signals over the Internet themselves. The IoT device intermediation system is a system in which a user's mobile communication device 3 mediates communication with a group 2 of IoT devices (various IoT sensors, actuators, etc.) that do not have the functionality to connect to the Internet and communicate, and connects the group 2 of IoT devices to the Internet 1 via the mobile communication device 3, enabling communication.

[0014] Referring to the overall system shown in Figure 1, a mobile communication device 3, such as a user's wearable computer, a robot 6 in the user's home, a personal computer (hereinafter referred to as "PC") 7 in the user's home, an IoT server 8, an artificial intelligence server 9, PCs 10 owned by various professional vendors, and an SNS server 11 are connected to the Internet 1 and are configured to communicate with each other. The mobile communication device 3 has the functionality to communicate with a group of IoT devices 2, a wireless sensor network 4, and a group of displays 5, such as digital menu boards. Possible communication methods include Wi-Fi (registered trademark), Bluetooth (registered trademark), Wi-Fi Direct (registered trademark), Zigbee (registered trademark), Z-wave (registered trademark), and Ant+ (registered trademark). The operating system is compatible with iOS, Android (registered trademark), Linux (registered trademark), TIZEN (registered trademark), and other real-time operating systems. IEEE802.15.4 is used as the wireless standard for transmission and reception. IPv6 (Internet Protocol Version 6) is used as the Internet Protocol.

[0015] A wireless sensor network4 is a wireless network in which multiple sensor-equipped wireless terminals are dispersed throughout space, enabling them to collaborate and collect information on the environment and physical conditions. For example, sensor devices are created using energy harvesting, M2M, or batteries. Pressure and gauge sensors constantly monitor, for example, metal fatigue degradation, and report any changes. They are primarily installed on structures such as bridges and tunnels. They typically include multiple sensor nodes and a gateway sensor node. These nodes typically consist of one or more sensors, a wireless chip, a microprocessor, and a power source (e.g., a battery). The hardware configuration of the node's control circuit is the same as that shown in Figure 2(a). Wireless sensor networks typically have ad hoc functionality and a routing algorithm for sending data from each node to a central node. In other words, they have the ability to autonomously reconstruct an alternative communication path if a communication failure occurs between nodes. Because nodes work together as a group, they also have elements of distributed processing. In addition, they can operate for long periods without external power supply, which allows them to save power or generate electricity. In this embodiment, the wireless sensor network 4 is a type of IoT device group 2.

[0016] The IoT server 8 can write and read data to 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 the SNS DB 13.

[0017] Next, with reference to FIG. 2(a), the hardware configuration of the control circuit of the IoT device 2 will be described. The IoT device group 2 is made up of sensors and actuators installed all over the globe, and includes a CPU (Central Processing Unit) 18 for overall control, a ROM (Read Only Memory) 20 storing programs for executing various functions, a RAM (Random Access Memory) 19 serving as a work area for the CPU 18, and a wireless communication interface unit 22 using, for example, Wi-Fi, Bluetooth, Wi-Fi Direct, Zigbee, Zwave, or Ant+. If the IoT device 2 is a sensor, it includes a sensor unit 14. On the other hand, if the IoT device 2 is an actuator, an actuator unit 21 is provided instead of or in addition to the sensor unit 14. The CPU 18, RAM 19, ROM 20, wireless communication interface unit 22, sensor unit 14, and actuator unit 21 are connected to enable signal exchange.

[0018] Next, with reference to Fig. 2(b), a hardware configuration of a control circuit of a wearable computer, which is an example of the mobile communication device 3, will be described. This wearable computer 3 is typified by smart glasses or the like, and has a CPU 23, RAM 24, ROM 25, and EEPROM (Electronically Erasable and Programmable Read Only Memory) 26 connected via a bus 27. The functions of the CPU 23, RAM 24, and ROM 25 are basically the same as those described above for the IoT device 2. Application programs downloaded via the Internet 1 and the like are stored in the EEPROM 26.

[0019] Various devices are connected to the bus 27 via an interface unit 28. For example, the following are connected to the interface unit 28: a digital video camera input unit 29 that captures video of the user's surroundings (video in the user's line of sight, etc.), a display unit 30 that overlays information on the lenses of the smart glasses, a wireless communication processing unit 31 that communicates wirelessly with a base station and performs data communication with a server or the like via the Internet 1, an input operation unit 32 that causes the CPU 23 of the wearable computer 3 to execute a desired function by the user, a voice output unit 33 and a voice input unit 34 that allow the user to make a call by voice, a location information acquisition unit 35 that acquires a current location based on GPS information from a satellite, radio waves from a base station, and radio waves from a wireless LAN access point, a wireless communication interface unit 36 ​​that communicates with the IoT device group 2 using Wi-Fi, Bluetooth, Wi-Fi Direct, Zigbee, Zwave, Ant+, etc., a gaze position detection unit 37 that detects the user's gaze position and identifies the position of the user's gaze in the video captured by the digital video camera input unit 29, various sensors 38, etc.

[0020] The input operation unit 32 not only accepts manual operations by the user, but also accepts instructions from the user using gestures, specific winks, etc. Note that voice instructions from the user are accepted by a voice input unit 34. The various sensors 38 are sensors for temperature, humidity, illuminance, ultraviolet light, etc. attached to the wearable computer 3. Note that the hardware configuration of the robot's control circuit would be the same as the hardware configuration shown in FIG. 2(a) except that the various sensors 38 are removed and a movement drive circuit is added.

[0021] Next, the hardware configuration of the control circuits of the PC and various servers 7 to 11 will be described with reference to Fig. 3. As described above, a CPU 40, RAM 41, and ROM 42 are connected by a bus 43. An interface unit 44 to which the bus 43 is connected is connected to a communication unit 45 for communicating with the Internet 1, etc., a display unit 46 for displaying images and information to the operator, and an input operation unit 47 for receiving operations from the operator.

[0022] Next, a flowchart of the main program of the 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 Figure 4. The mobile communication device 3 performs IoT processing at S1. This is processing for connecting the IoT device group 2 and the Internet 1 via the mobile communication device 3, transmitting detection data of the IoT device (sensor) 2 to the IoT server 8, and transmitting a command signal (actuator control signal) from the IoT server 8 to the IoT device (actuator) 2.

[0023] Next, the mobile communication device 3 performs external display processing at S2, and the artificial intelligence server 9 performs display processing at S3. This is processing for displaying videos, still images, etc. uploaded by the user to the Internet on a display 5 such as a digital menu board. Next, the mobile communication device 3 performs around data terminal processing at S4, and the artificial intelligence server 9 performs around data server processing at S5. This is processing for collecting video and audio around the user using the mobile communication device 3, such as a wearable computer, in the artificial intelligence server 9 to use as information for machine learning.

[0024] Next, the mobile communication device 3 performs personal service terminal processing at S7, and the artificial intelligence server 9 performs personal service server processing at S8. This is processing for providing personalized services to the user using the results of machine learning by the artificial intelligence. Next, the mobile communication device 3 performs dialogue terminal processing at S9, and the artificial intelligence server 9 performs dialogue server processing at S10. This is processing for various dialogues between the user and the artificial intelligence. Next, the mobile communication device 3 performs task terminal processing at S11, and the artificial intelligence server 9 performs task server processing at S12. This is processing for the artificial intelligence, which has completed machine learning, to respond to the user's work-related questions and inquiries.

[0025] Next, a flowchart of the subroutine program for the IoT processing shown in S1 above will be described with reference to Figures 5 to 8. First, with reference to Figure 5, the processing performed by the IoT device (sensor) 2, the mobile communication device 3, and the IoT server 8 will be described.

[0026] 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 in S20, and determines whether or not a transmission command signal has been received in S21. If not, the process returns to S20 and repeatedly cycles through the loop of S20 → 21 → S20.

[0027] Meanwhile, in S22, the CPU 23 of the mobile communication device 3 transmits location information based on GPS information, radio waves from a base station, radio waves from a wireless LAN access point, etc. to the IoT server 8. The IoT server 8 receives this information in S24, searches the IoT device DB 15 in S25, and determines in S26 whether or not the desired IoT device 2 is present in the vicinity of the mobile communication device 3 that transmitted the location information. Ideally, the IoT server 8 periodically collects detection data of IoT devices (sensors) 2, and searches the IoT device DB 15 to identify IoT devices (sensors) 2 for which the periodic reception time has arrived. If the desired IoT device 2 is present, a YES determination is made in S26, and control proceeds to S27, where a command signal for reception is transmitted to the mobile communication device 3.

[0028] On the other hand, if there is no IoT device 2 to receive the command signal, a NO determination is made in S26 and control proceeds to S38, where it is determined whether or not there is an IoT device (actuator) 2 nearby to which it wants to transmit a command signal (actuator control signal). If there is, control proceeds to S39, where it transmits a command signal to be transmitted and transmission information (information for controlling the actuator unit 21) to the mobile communication device 3. Upon receiving a command signal in S27 or S39, the mobile communication device 3 makes a YES determination in S23 and control proceeds to S28, where it determines whether or not the received command signal is for reception. If the command signal is for transmission transmitted in S38, control proceeds to S47, but if the command signal is for reception transmitted in S27, control proceeds to S29, where it transmits a transmission command signal to a nearby IoT device (sensor) 2.

[0029] The mobile communication device 3 receives the data in S21, and in S30 transmits the detection data and device ID stored in RAM to the mobile communication device 3, and returns to S20. The mobile communication device 3 receives the data in S31, and transmits the received data to the IoT server 8 in S32. The mobile communication device 3 receives the data in S33, and in S34 updates the stored data of the IoT device corresponding to the received device ID, among the stored data in the IoT device DB 15, with the received data.

[0030] Next, the IoT server 8 transmits points, an example of a benefit, to the mobile communication device 3 in S35, and then control proceeds to S61. The mobile communication device 3 receives the points in S36, updates the points stored in the EEPROM 26 (S37), and then control proceeds to S60. The points stored in the EEPROM 26 can be used to receive various services from the company that operates the IoT server 8. For example, the points can be used to pay highway tolls or to purchase JR or private railway tickets or commuter passes. The points may also be redeemable for cash, which can provide an incentive for users to broker IoT devices.

[0031] On the other hand, if the determination in S38 is NO, control proceeds to S40, where it is determined whether there are any of the various sensors 38 of the mobile communication device 3 that are desired to receive data. That is, it is determined whether there are any of the various sensors 38 provided in the mobile communication device 3 located at the location information received in S24 that are desired to transmit detection data at that geographical location. If there are no such sensors, control proceeds to S24, but if there are such sensors, control proceeds to S62.

[0032] Next, processing for the IoT device (actuator) 2 will be described with reference to FIG. 6. The IoT device (actuator) 2 performs operation control processing based on the transmission information transmitted from the IoT server 8 and stored in RAM 19 (S45). Next, in S46, it is determined whether or not a reception command signal has been received from the mobile communication device 3. If not, control returns to S45, and the loop of S45 → S46 → S45 is repeatedly performed. In this state, the mobile communication device 3 receives the transmission command signal and transmission information, and transmits the reception command signal and transmission information to the IoT device (actuator) 2 in S47. Upon receiving the signal, the IoT device (actuator) 2 determines YES in S46, and control proceeds to S48, where the received transmission information (information for controlling the actuator unit 21) is stored in RAM 19, and a reception completion signal and a device ID are transmitted to the mobile communication device 3 in S49.

[0033] The mobile communication device 3 receives the completion signal in S50 and transmits it to the IoT server 8 in S51. The IoT server 8 receives the completion signal in S52 and searches the IoT device DB based on the device ID in S53, and updates the data corresponding to the device ID. For example, the data may be updated to "XX:XX △X transmission information transmitted." Then, points are transmitted to the mobile communication device 3 in S54, and control proceeds to S61. The mobile communication device 3 receives the completion signal in S55 and updates the points stored in the EEPROM 26 (S56). Then, control proceeds to S60.

[0034] In S60, it is determined whether a command signal for mobile reception and a data identification signal have been received from the IoT server 8. If not, control proceeds to S84. On the other hand, if the IoT server 8 determines in S24 that the desired sensor 38 of the mobile communication device 3 is present at the location received, it determines YES in S40 and transmits the command signal for mobile reception and the data identification signal in S62. Upon receiving the signal, the mobile communication device 3 determines YES in S60 and proceeds to S63, where it determines whether the data identified by the data identification signal (temperature, humidity, illuminance, ultraviolet light, etc.) requires a required detection time. In the case of illuminance or ultraviolet light, these can be detected instantly and do not require a required detection time. Therefore, it determines NO in S63 and proceeds to S64, where the identified data is detected by the corresponding sensor 38 and the detected data is transmitted to the IoT server 8 (S65).

[0035] The IoT server 8 receives the data in S66, searches the IoT device DB 15, and updates it with the new detection data received in S67. Then, in S68, it transmits the points to the mobile communication device 3. The mobile communication device 3 receives the data in S69, and adds or updates the points stored in the EEPROM 26 (S70). Then, control proceeds to S84.

[0036] On the other hand, if the data identified by the data identification signal, such as temperature or humidity, requires a detection time, a YES determination is made in S63, and control proceeds to S71, where the display unit 30 notifies the user of the detection time (e.g., 30 seconds). If the user sees this and cooperates with the transmission of sensor-detected data, they stay at their current location for 30 seconds and wait for detection to be completed. If they do not intend to cooperate, they move on without staying. After the notification in S71, in S72, the identified data is detected by the corresponding sensor 38. In S73, it is determined whether detection is complete. If not, in S74, it is determined whether the device remains within the detection area. If it remains, a loop is repeatedly executed returning to S72. If the user's mobile communication device 3 moves out of the detection area before detection is completed, control proceeds to S84. However, if the user's mobile communication device 3 remains within the detection area until detection is completed, control proceeds to S75, where the detection data is transmitted to the IoT server 8, and then control proceeds to S69.

[0037] The IoT server 8 receives the data in S66, searches the IoT device DB 15, and updates it with the new detection data received in S67. Then, in S68, it transmits the points to the mobile communication device 3. The mobile communication device 3 receives the data in S69, and adds or updates the points stored in the EEPROM 26 (S70). Then, control proceeds to S84.

[0038] 5 to 7, the determination means (for example, S26, S38, S49, etc.) 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 is provided on the IoT server 8 side, but this determination means may also be provided on the IoT device group 2 side. This can reduce the control burden on the IoT server 8 side.

[0039] Next, processing for the wireless sensor network 4 will be described with reference to Figure 8. In S80, the wireless sensor network 4 collects detection data from each sensor and stores the data in association with each sensor ID. Next, in S81, it is determined whether or not any of the detection data requires urgent transmission. If there is no data requiring urgent transmission, control proceeds to S82, where it is determined whether or not it is time for regular transmission (for example, every 24 hours). If it is not time for regular transmission, control returns to S80, and the loop of S80 → S81 → S82 is repeated.

[0040] If an obvious abnormality that can be judged by the wireless sensor network 4 is found in each detected data during the loop and it is judged in S81 that there is data that should be urgently transmitted, control proceeds to S83. Also, if it is judged in S82 that it is time for regular transmission, control proceeds to S83.

[0041] In S83, a process of transmitting a transmission signal is performed. Next, in S86, it is determined whether or not a response signal has been received from the mobile communication device 3. If not, the process returns to S80, and the loop of S80 to S83 → S86 → S80 is repeated.

[0042] In this state, when a user carrying the mobile communication device 3 passes near the wireless sensor network 4, the transmission signal in S83 is received, a determination of YES is made in S84, and control proceeds to S85. The mobile communication device 3 transmits a response signal to the wireless sensor network 4 in S85, and the wireless sensor network 4 receiving the response signal determines YES in S86, and control proceeds to S87. The wireless sensor network 4 transmits the detection data collected in S80 to the mobile communication device 3 for each sensor ID (S87). The mobile communication device 3 receives the data in S89 and transmits the received detection data and sensor ID together with a transmission signal to the IoT server 8 in S90.

[0043] Then, the IoT server 8 makes a YES determination in S61, and control proceeds to S91. If there is emergency detection data in the received detection data, it notifies the abnormality along with the sensor ID, and then in S92, it searches the wireless sensor DB 16 based on the sensor ID of the detection data and updates the detection data with the new one received. Next, in S93, it transmits points to the mobile communication device 3, and then control returns to S24. The mobile communication device 3 receives the points in S94, updates them by adding points stored in EEPROM 26 (S95), and then returns, and control proceeds to S2.

[0044] In the control shown in Figure 8, a determination means (e.g., S82) for determining whether there is an IoT device that needs to exchange information between the wireless sensor network 4 and the IoT server 8 is provided on the wireless sensor network 4 side, but this determination means may also be provided on the IoT server 8 side.

[0045] In the IoT device intermediation system described above, a user's mobile communication device (for example, a wearable computer such as smart glasses, a smartphone, or a vehicle such as an automobile equipped with communication capabilities) acts as an intermediary to connect the IoT device group 2 to the Internet 1, so the IoT device group 2 does not necessarily need to have an Internet connection function, which makes it possible to keep costs low. Moreover, when a user performs the above intermediation using a mobile communication device, the user is given a benefit (for example, points), which gives the user an incentive to perform the above intermediation and makes it possible to encourage users to act as intermediaries.

[0046] Furthermore, since a determination means (e.g., S26, S38, S49, etc.) is provided on the side of an intermediary computer (e.g., IoT server 8, network cloud, etc.) through which the wearable computer 3 acts as an intermediary for the IoT device group 2, for determining whether there is an IoT device (including various sensors 38 of the mobile communication device 3) with which information needs to be exchanged, the above determination does not need to be made on the IoT device group 2 side, reducing the processing burden on the IoT device group 2 side and further reducing costs. Moreover, if the IoT server 8 having the determination means identifies the IoT device group 2 with which information needs to be exchanged (e.g., S26, S38, S49, etc.) and notifies the wearable computer 3 of the location information of the identified IoT device 2 to notify the user, there is an advantage that the user can be encouraged to act as an intermediary.

[0047] Furthermore, when various sensors 38 provided on the wearable computer 3 are used to collect detection data from the various sensors 38 at desired locations onto the Internet, if a certain amount of time is required for detection, the wearable computer 3 notifies the user of the required detection time (e.g., S71), thereby encouraging the user to cooperate in collecting detection data that requires a certain amount of time.

[0048] [Service provision system using machine learning] Next, a service providing system using machine learning will be described with reference to FIGS. A major challenge in the practical application of machine learning is how to inexpensively collect large amounts of training data. This system provides users with services that utilize the results of machine learning, thereby building a mechanism in which many users proactively provide their own training data to the system. In particular, as shown in Figure 9, the training data (data from everyday life, etc.) provided by each user to the system is also used for personalization, and a general model obtained as a result of machine learning is optimized (personalized) for each user, and personalized services are then provided to each user. This prevents the inconvenience of each user relying on others to provide the training data and simply enjoying the service.

[0049] First, the outline of the system will be explained with reference to FIG. Each user 70 transmits around data (such as the image the user 70 is viewing, the position of their gaze, audio, and location data such as GPS) collected by a mobile communication device 3, such as a wearable computer, to machine learning 72, which uses the data as learning data for machine learning 72, and the data is also used for personalization 74.

[0050] For example, the state of each user 70 studying at school, at home, etc. is sent to the system, and modeled using machine learning (reinforcement learning and regression) 72 to create a general model that minimizes the total time spent on repeated review by humans. In this case, factors that vary from user to user (such as memorization ability) are used as constants to create the general model. The data sent from each user 70 to the system is also used for personalization 74, and factors that vary from user to user (such as memorization ability) are calculated for each user and substituted into the constant portion of the general model to create a personalized model for each user. This personalized model is used to provide personalized services to each user, such as presenting the next review date and providing instruction according to a review plan that minimizes the total review time.

[0051] The training data is provided not only by users 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 many members to the system. Based on the training data of these aggregated results, a model is created using machine learning (reinforcement learning and regression) 72 to create a general model that maximizes the effect of muscle training performed repeatedly by humans. In this case, the general model is generated using constants that represent factors that differ from user to user (initial load such as a barbell and a load increase coefficient). The aggregated results of muscle training from each user 70 are also transmitted to the system and used for personalization 74. Factors that differ from user to user (initial load, load increase coefficient, etc.) are calculated for each user and substituted into the constant portion of the general model to generate a personalized model for each user.

[0052] This personalization model is used to provide personalized services to each user, such as presenting the next muscle training schedule and load and providing guidance according to a muscle training plan that maximizes the benefits of muscle training. At the same time, advertisements for 71 specialized businesses (sports gyms, etc.) that provided learning data consisting of aggregated muscle training results are presented along with the muscle training schedule. This can motivate various specialized businesses to provide learning data.

[0053] FIG. 10 shows the overall system of FIG. 9 explained above in more detail. Referring to FIG. 10, specialized data is provided to the artificial intelligence server 9 from PCs 10 of various specialized businesses. Data collected from mobile communication devices 3, such as wearable computers, of numerous users (around data, such as the video viewed by the user 70, the position of the user's gaze, audio, and location data, such as GPS) is also provided to the artificial intelligence server 9. The artificial intelligence server 9 converts the provided data into a set of numerical values ​​and labels (data conversion 51) to generate training data. Deep learning is used to generate this training data, as raw data is converted into a set of numerical values ​​and labels. Furthermore, "heterogeneous mixture learning" is used to handle raw data containing a mixture of multiple patterns. This is a method for generating a model appropriate for each pattern when multiple patterns exist in the collected data.

[0054] For example, when collecting time-series data on a building's power consumption to create a model that predicts power consumption, the building's power consumption patterns change depending on the day of the week and time of day. Traditionally, humans would use their specialized knowledge to distinguish between cases to determine when the pattern changes, and then create a model that is appropriate for each case. With this "heterogeneous mixture learning" method, machine learning is used to identify the pattern changes themselves, and a model is generated for each pattern. When new data is generated, the compatibility with the generated model is checked, and if the model does not match well, the system starts over by determining whether the pattern has changed. In other words, the machine repeatedly performs "pattern distinction" and "model creation" to create the optimal model for each pattern.

[0055] In addition, in cases where data cannot be converted into a set of numbers and labels even using deep learning, artificial intelligence companies will manually convert the data into a set of numbers and labels.

[0056] This artificial intelligence server 9 uses a general von Neumann-type computer, but it can also use a neural net processor (NNP). The NNP chip is equipped with a large number of "artificial neurons" modeled after real neurons, and each neuron connects with each other in a network.

[0057] Quantum computers that employ the "quantum annealing method" can also be used. This can significantly reduce the time required for optimization calculations in machine learning. Note that "artificial intelligence" is a broad concept that includes software agents.

[0058] The training data converted into a set of numerical values ​​and labels is sent to a learning algorithm 52 and modeled 53. Various learning algorithms are available for the learning algorithm 52, including supervised learning methods such as regression and classification, unsupervised learning methods such as model estimation and data mining, and intermediate methods such as reinforcement learning and deep learning. Modeling using the learning algorithm 52 generates a general model using, for example, factors that vary from user to user (such as individual ability differences and preferences) as constants. The generated general model is stored in an artificial intelligence database 17. FIG. 10 shows specific examples of stored general models, such as a building maintenance and inspection model, a rote learning model, a muscle training model, a dialogue model, and a task processing model. The aforementioned training data is also stored in a learning database 60.

[0059] The rote learning model Ti=T0·i is a general model that minimizes the total time spent by humans on repeated review, where i is the number of repetitions, T0 is the period from initial learning to just before forgetting (specifically, the time from initial learning until memory retention rate reaches 70%), and Ti is the time from repetition number (i-1) to i. T0 in this general model is a factor that differs for each user (individual differences in ability, such as memory), and is expressed as a constant. The method for generating this rote learning model Ti=T0·i will be explained in detail later.

[0060] The muscle training model Fi=F0+(i-1) / a is a general model that maximizes the effect of repeated muscle training performed by humans, where i is the number of repetitions, F0 is the load (such as the weight of a barbell) during the first muscle training session, and a is the load increase coefficient that increases with each repetition of the muscle training session.

[0061] The dialogue model is a general model used when a user dialogues with an artificial intelligence, and includes dialogue templates of initial targets such as intellectual males, cute females, famous celebrities, and change functions of emotions, etc. This will be explained in detail with reference to FIG.

[0062] The task processing model is a general model used when a user processes a task such as work, and is composed of knowledge utilization functions of initial targets such as patent attorneys, lawyers, famous scientists, etc. This will be explained in detail with reference to FIG.

[0063] The around data sent from the user's mobile communication device 3 is classified for various models in the artificial intelligence server 9 54, and personalized data for the various models is created 55. The personalized data is stored in the user DB 12. Figure 10 shows specific examples of the various models, such as for rote learning, muscle training, dialogue, and task processing.

[0064] The user DB 12 stores personalized data and around data associated with each user ID. The personalized data for rote learning is TK, the period from initial learning to just before forgetting (specifically, the time from initial learning until memory retention reaches 70%). Human memory information decreases according to the forgetting curve of psychologist Hermann Ebbinghaus. TK is the time it takes for memory retention to reach 70% according to this forgetting curve. This TK is the time it takes for an actual user with user ID "1" to reach 70% memory retention, and is expressed as an actual number.

[0065] Personalized data for muscle training includes the initial load FK, such as the weight of the barbell used in the first muscle training session, the load increase coefficient az, which increases with each repetition of muscle training, and the supercompensation time CT. These are expressed as actual values ​​for an actual user with user ID "1." Note that "supercompensation" refers to the phenomenon in which muscle strength levels rise above pre-training levels after 48 to 72 hours of rest after training. The time required for this supercompensation is the "supercompensation time."

[0066] The personalized data for the dialogue includes an initial target selected by the user from initial targets such as intellectual males, cute females, or famous celebrities, as well as personal weights for emotions. In a general model, the initial weights for emotions are determined for each initial target, and while the user is interacting with the initial target selected by the user, the initial weights for emotions are gradually changed in response to requests from the user, until the personal weights are finally optimized for the user, enabling dialogue that expresses emotions that match the user's preferences. This will be explained in detail later.

[0067] The personalized data for task processing includes an initial object selected by the user from initial objects such as patent attorneys, lawyers, and famous scientists, as well as personalized weights for each piece of knowledge. In a general model, the initial weights for the knowledge of each initial object are determined, and the initial weights for the knowledge to be utilized are gradually modified in response to questions and requests from the user during a work-related dialogue with the initial object selected by the user, ultimately resulting in a personalized weight that is optimal for the user, enabling answers and advice that match the knowledge required by the user for work. This will be explained in more detail later.

[0068] The artificial intelligence server 9 uses the personalized data described above to provide personalized instruction (services) to each user. Specifically, it uses the personalized data for rote learning to provide instruction according to a review plan (review schedule) that minimizes the user's total review time, uses the personalized data for muscle training to provide instruction according to a muscle training schedule that maximizes the effect of repeated muscle training, uses the personalized data for dialogue to provide dialogue that expresses emotions that match the user's preferences, and uses the personalized data for task processing to provide answers and advice that match the knowledge the user requires for work.

[0069] Next, with reference to FIG. 11, a flowchart of the subroutine program of the external display process (S2), the display process (S3), and the like process by the SNS server 11 in FIG. 4 will be described.

[0070] FIG. 11 shows processing for a digital menu board, which is an example of a display 5. A digital menu board is a display installed in restaurants and the like to display menus. By displaying menus digitally, the restaurant can significantly reduce time and costs by instantly displaying sold-out items and changing the menu. In this 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. The flowchart in FIG. 11 shows how images and videos taken by a user using a wearable computer 3 or the like and uploaded to the SNS DB 13, user DB 12, etc. can be displayed on the digital menu board 5, allowing friends who have visited the restaurant together to view and enjoy their meal.

[0071] First, in S100, it is determined whether the menu has been updated. If the restaurant performs an operation to update today's menu, control proceeds to S101, where the menu is updated. Next, control proceeds to S102, where the menu is displayed.

[0072] Next, when the user wishes to display on the digital menu board 5 an image or video that has been taken with the wearable computer 3 or the like and uploaded to the SNS DB 13, user DB 12, or the like, the user first performs an operation to access the artificial intelligence server 9. Then, a YES determination is made in S104, and control proceeds to S105, where the user ID, access request, etc. are transmitted to the artificial intelligence server 9 for access processing. Upon receiving the access request, the artificial intelligence server 9 makes a YES determination in S106. The mobile communication device 3 communicates with the artificial intelligence server 9 in S107 and S108, and the user selects and specifies the image or video that he or she wishes to display on the digital menu board 5. The artificial intelligence server 9 searches the user DB 12 in S108 to identify the selected image or video.

[0073] Once the user (customer) has identified the image they want to display, they tap the touch screen of the digital menu board 5 to perform a communication operation. YES is determined in S103, and communication start processing with the wearable computer 3 is performed in S109. The user (customer) also performs a communication operation on their own wearable computer 3. YES is determined in S108, and control proceeds to S110, where communication start processing is performed and a connection for communication is established between the wearable computer 3 and the digital menu board 5.

[0074] Next, the digital menu board 5 transmits the display ID assigned to that digital menu board 5 to the wearable computer 3 (S111). The wearable computer 3 receives this in S112 and transmits the display ID to the artificial intelligence server 9 (S113). The artificial intelligence server 9 receives this in S114 and transmits the data identified in S108 to the display 5 with the received ID (S105). The digital menu board 5, which receives this, determines YES 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, the image or video can be viewed on the relatively large display screen of the digital menu board 5.

[0075] A noteworthy point about the control described above is that when a user selects and specifies data uploaded to the Internet, the AI ​​server 9 is accessed and specified directly with the user's mobile communication terminal 3 without going through the display 5. If the display 5 itself is connected to the Internet and data to be displayed on the display 5 is specified, it is quicker to access the data on the Internet via the display 5 at the display destination and display the data on the display 5. However, the data uploaded to the Internet by the user is personal information related to the user's privacy, and if the user's ID or other information is transmitted via the display 5 to access that personal information, there is a risk that the user's ID and personal information will be leaked. Therefore, in this embodiment, when a user selects and specifies data uploaded to the Internet, the AI ​​server 9 is accessed and specified directly with the user's mobile communication terminal 3 without going through the display 5.

[0076] Next, the digital menu board 5 determines in S118 whether the data display has finished, and if so, control proceeds to S119, where processing is performed to allow the user (customer) to access the store's homepage. Specifically, the URL of the store's homepage is sent to the mobile communication device 3. Upon receiving this, the mobile communication terminal 3 sends an access request to the page of that URL to the SNS server 11 in S120. The SNS server 11, having received this in S121, sends the store's homepage to the mobile communication device 3 in S122.

[0077] Upon receiving the signal, the mobile communication device 3 displays the store's homepage in S123. Meanwhile, the digital menu board 5 displays a message saying, "Please tap on the like button on our store's homepage" in S124. If a user (customer) sees the message and taps on the like button, a YES determination is made in S125, control proceeds to S126, and a like signal is sent to the SNS server 11. The SNS server 11 receives the signal in S127 and performs a like registration process in S128. As a result, the store can provide information to the user (customer) via the SNS. For example, it is possible to introduce new seasonal menu items, distribute coupons, notify users of upcoming events, and so on, which can encourage repeat visits.

[0078] If a return is made in the external display processing, the process proceeds to S4, and if a return is made in the display processing, the process proceeds to S5.

[0079] Next, with reference to Figure 12(a), a flowchart of the expert data processing subroutine program shown in S6 of Figure 4 will be described. Maintenance data processing is performed in S135, strength training data processing is performed in S136, other processing is performed in S137, and then the program returns to S8. Maintenance data processing involves accepting and processing data on the results of building maintenance inspections sent from a maintenance specialist, which is an example of a variety of specialists. Strength training data processing involves accepting and processing aggregate data on the strength training results of a large number of members sent from a sports gym, which is an example of a variety of specialists.

[0080] Next, a flowchart of the subroutine program for the maintenance data processing will be described with reference to Fig. 12(b). The maintenance specialist performs maintenance on 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 inputs the results (KO, NG, etc.) for each sensor ID into the PC 10 (S140).

[0081] Then, in S141, the maintenance specialist's PC 10 transmits the input inspection results to the artificial intelligence server 9 for each sensor ID. The artificial intelligence server 9 determines in S145 whether or not it has received the inspection results, and if not, control proceeds to S151. On the other hand, if it has received the inspection results, control proceeds to S146, where the inspection result data is additionally stored in the learning DB 60 for each sensor ID. Next, in S147, the wireless sensor data is read from the wireless sensor DB 16, and the inspection result data is attached, followed by data mining to find a pattern NGP1 for NG results. For example, a pattern NGP1 is found, such as one in which there is a high probability of an abnormal (NG) inspection result if the distortion and vibration period have a predetermined relationship.

[0082] Next, in S148, it is determined whether NGP1 is a new pattern different from those found so far. All NG patterns found so far are stored in the storage area of ​​the "building maintenance and inspection model" of the artificial intelligence DB 17, and it is determined whether the NGP1 found this time is already stored in the artificial intelligence DB 17. If it is determined that the same pattern is already stored, control proceeds to S151, but if it is determined that it has not yet been stored, control proceeds to S149, where NGP1 is additionally stored as an NG pattern in the storage area of ​​the "building maintenance and inspection model" of the artificial intelligence DB 17.

[0083] Next, in S150, the wireless sensor DB 16 is searched to extract and report the sensor IDs having the NGP1 pattern from among the wireless sensors that have been periodically checked. Wireless sensors are periodically checked at regular intervals (e.g., every 24 hours) (S151, S152), and the newly found NGP1 is applied to past sensor data that has already been periodically checked to perform a recheck.

[0084] Next, in S151, it is determined whether or not it is time for a regular check. If it is time for a regular check, in S152, the wireless sensor DB 16 is searched to extract a sensor ID having an NG pattern from the wireless sensor data that has been added and stored since the previous regular check, and an NG notification is issued. As described above, new sensor data is added and stored in the wireless sensor DB 16 from time to time, and new sensor data that has not yet been periodically checked is checked by applying all NG patterns stored in the artificial intelligence DB. After processing in S152, the process returns, and control proceeds to S136.

[0085] Next, with reference to FIG. 13(a), a flowchart of the subroutine program of the around data terminal processing shown in S4 and the around data server processing shown in S5 of FIG. 4 will be described.

[0086] S160 executes user learning terminal processing, S161 executes user learning server processing, S162 executes muscle training terminal processing, S163 executes muscle training server processing, and S164 and S165 execute other processing.

[0087] The user learning terminal processing and user learning server processing are processing for collecting around data from each user who is studying to create a general rote learning model for the user, and for generating personalized data for studying for each user. The muscle training terminal processing and muscle training server processing are processing for collecting around data from each user who is studying to generate personalized data for muscle training for each user.

[0088] Next, with reference to FIG. 13(b), a flowchart of the subroutine program of the personal service terminal processing shown in S7 and the personal service server processing shown in S8 of FIG. 4 will be described.

[0089] Learning service terminal processing is executed in S170, learning service server processing is executed in S171, muscle training service terminal processing is executed in S172, and muscle training service server processing is executed in S173.

[0090] The learning service terminal processing provides instruction to the user according to the most efficient review plan based on the user's general rote learning model and personalized data for learning. The muscle training service terminal processing and muscle training service server processing provide instruction to the user according to the most efficient muscle training plan based on the user's general muscle training model and personalized data for muscle training.

[0091] Next, a flowchart of the subroutine programs for the user learning terminal processing and the user learning server processing will be described with reference to FIG.

[0092] In S180, it is determined whether the user is studying. This determination may be made by accepting an explicit expression of intent from the user, or may be made autonomously by the artificial intelligence server 9 by processing the user's voice using natural language processing or analyzing video in the user's line of sight. If it is determined that the user is not studying, the process returns and control shifts to S162. However, if it is determined that the user is studying, control proceeds to S181, where around data such as video in the user's line of sight, line of sight position, and voice is transmitted to the artificial intelligence server 9. Having received this in S182, the artificial intelligence server 9 searches a predetermined database from the received around data for useful information such as frequency of questions in past entrance exams and schools that offered the questions, related to the subject currently being studied, in S183, and transmits this information to the user (mobile communication device 3 such as a wearable computer).

[0093] The user's mobile communication device 3 receives the useful data in S184 and displays the received useful data as an overlay using AR (Augmented Reality) in S185. By sending the around data during study to the artificial intelligence server 9, the user can receive guidance according to the most efficient review plan, so the user will take the initiative to send the around data to the artificial intelligence server 9, and the display of useful information such as the frequency of questions appearing in the entrance exam will further increase the user's incentive to send the around data to the artificial intelligence server 9.

[0094] Next, in S186, the artificial intelligence server 9 classifies the received data for various models and stores them in the user DB 12 and the learning DB 60. Next, in S187, it is determined whether the rote learning model is complete. If it is not yet complete, control proceeds to S188, where the received data is converted into numerical values ​​and labels to create data for machine learning, and in S189, it is modeled using a learning algorithm (reinforcement learning, regression, etc.). An example of this modeling will be described.

[0095] Using a huge amount of past learning data collected for each user as learning data, reinforcement learning is used to find a review plan that minimizes the total time spent on repeated review from the initial learning to the specified end point, such as the entrance exam date, under the condition that a certain memory retention rate (e.g., 90%) can be maintained for each user at the specified end point. 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, then the total review time is T = ΣTHi. Reinforcement learning is used to find Ti that minimizes the above T under the condition that the end point memory retention rate is ≥ 90%.

[0096] In reinforcement learning, when the value of performing action at in state st is Q(st,at), a model-based method can be used to estimate the Q value if knowledge that models the environment, i.e., the probability distribution of state transition probabilities and rewards, is given, but if the environmental model is unknown, TD (Temporal Difference) learning is used. First, since it is necessary to explore the environment, the ε-greedy method is used. In the early stages of exploration, various actions are tried, and as the exploration settles down, the concept of temperature is introduced so that the most optimal actions are selected. With temperature as T, an action is selected according to the probability expressed in the following formula.

[0097] P(a|s)={exp(Q(s,a) / T)} / {Σexp(Q(s,b) / T} (Note that b∈A is written under Σ, but is omitted in the above formula) Here, a is an action, Q(s, a) is the value of performing action a in state s,

[0098] T is called the annealing temperature; if it is high, actions will be selected with a probability close to equal, and if it is low, the action will be biased towards the optimal one. As learning progresses, the learning results can be stabilized by decreasing the value of T. It is also possible to actively collect learning data by instructing the user to actually perform the actions selected in this way.

[0099] An example of the results of this reinforcement learning is shown in Figure 15. Referring to Figure 15, the vertical axis of the graph is Ti (review interval), and the horizontal axis is i (the number of reviews indicating how many times the review has been). The circled circles represent plots of the results of reinforcement learning for user ID: 1. The crossed circles represent plots of the results of reinforcement learning for user ID: 2. The triangled circles represent plots of the results of reinforcement learning for user ID: 3.

[0100] Next, using a "regression" algorithm, the machine learning results data of these multiple users are used as input data, and this input data is considered to output a target based on a certain function, and that function is calculated. The calculated function is Ti=T0·i The coefficient T0 is a factor that varies from user to user (memorization ability, etc.). Specifically, it is the period from the first learning to just before forgetting for each user (specifically, the time from the first learning until the memory retention rate reaches 70%). In this way, the factor that varies from user to user (memorization ability, etc.) is expressed as a constant T0 in a general model as follows: Ti=T0·i is.

[0101] 14, in step S190, the rote learning model (Ti=T0·i) is stored in the artificial intelligence DB 17. If an old and incomplete rote learning model is already stored, it is updated to a newer, more complete rote learning model (Ti=T0·i).

[0102] On the other hand, if it is determined in S187 that the rote learning model is complete, control proceeds to S191, where personalized data is created based on the classified received data and stored in the user DB 12. In this case, the "personalized data" is the element (memory ability, etc.) T0 that differs for each user, i.e., the time from initial learning until the memory retention rate reaches 70%. This can be easily derived from the user's around data during learning stored in the user DB 12.

[0103] Next, with reference to FIG. 16, a flowchart of the subroutine program of the learning service terminal processing shown in S170 and the learning service server processing shown in S171 in FIG. 13(b) will be described.

[0104] In S200, a mobile communication device 3 such as a wearable computer determines whether a user has requested a review. If no review request has been received, the process returns and control proceeds to S172. If a review request has been received, the process sends the review request to the artificial intelligence server 9 in S201. The artificial intelligence server 9 receives the request in S202 and determines in S203 whether the rote learning model is complete. If not, control proceeds to S205. If the rote learning model is complete, control proceeds to S204, where T0 = TK is substituted into the general rote learning model Ti = T0·i, and learning items due for repetitive review are extracted from the around data in the user DB 12. Note that TK refers to the period from the user's first learning to just before forgetting (specifically, the time from first learning until memory retention reaches 70%).

[0105] Next, in S205, the learning items when the user is looking away during class or the like are extracted from the around data in the user DB 12. This is because the gaze position of the user while studying is transmitted as around data to the artificial intelligence server 9 (S181) and stored in the user DB 12, making it possible to determine whether the user is looking away. Next, in S206, the learning items extracted in S204 and S205 are transmitted to the mobile communication device 3. The mobile communication device 3 receives the learning items in S207, and then in S208 performs a process of having the user review the received learning items.

[0106] Alternatively, instead of or in addition to detecting that the user is looking away, the system may control the user to review the learning material while chatting during class, etc. That is, the system includes a collection means for collecting information that can identify the user's behavior during learning, a recording means for recording the learning content the user is receiving, a discrimination means for discriminating the user's behavior that is inappropriate for learning from the information collected by the collection means, an extraction means for extracting from the recording means the learning content corresponding to the portion discriminated by the discrimination means, and a transmission means for transmitting the learning content extracted by the extraction means to the user terminal. Note that S205 above does not necessarily have to be performed. After processing S206, the system returns and control proceeds to S173.

[0107] In the above control, the user's personalized data is substituted into the general model to create a personalized model dedicated to the user, and then the personalized service is provided to the user using the personalized model. However, a handmade personalized model may be generated for the user and used to provide the personalized service to the user. For example, when a YES determination is made in S187 of FIG. 14, the machine learning results of each of the multiple users (subjects) who contributed to the creation of the general learning model are available (see FIG. 15). Therefore, the results of the machine learning performed on each of the subject users' own learning can become a handmade personalized model, and the personalized service may be provided to the user using the personalized model. Furthermore, even if a user other than the subject users wishes to create a handmade personalized model for themselves, the user may be controlled to provide a large amount of their own learning data (review data) to the artificial intelligence server 9 to create a handmade personalized model.

[0108] Next, with reference to FIG. 17, a flowchart of the subroutine program for processing muscle training data shown in S136 of FIG. 12(a) will be described.

[0109] In S210, PC 10 at a sports gym, an example of a professional gym, compiles muscle training data from numerous members (subjects) and sends it to AI server 9. The muscle training data includes, for example, the subject's initial load, the duration and number of repetitions of the muscle training, the muscle training interval between the previous and current muscle training sessions, the load increase coefficient, and muscle thickness measurements. In S211, AI server 9 receives the muscle training data, and in S212 classifies the received data for various models and stores it in learning DB 60. Next, in S213, the received data is converted into data of numerical values ​​and label sets to create learning data.

[0110] Next, in S214, a model is created based on the learning data using a learning algorithm such as reinforcement learning or regression. This modeling method is the same as that explained with reference to FIG. 15. A general model for muscle training is as follows: Fi=F0+(i-1) / a Here, i is the number of repetitions indicating the number of repetitions of the muscle training, Fi is the load (weight of the barbell, etc.) in the i-th muscle training, F0 is the load (weight of the barbell, etc.) in the first muscle training, and a is the load increase coefficient that increases with each repetition of the muscle training. F0 and a are factors that differ for each user (individual differences in muscle strength, etc.).

[0111] Next, in S215, the general model (Fi=F0+(i-1) / a) is stored in the muscle training learning model storage area of ​​the artificial intelligence DB 17. If an old and incomplete muscle training learning model is already stored, it is updated to a newer, more complete rote learning model (Fi=F0+(i-1) / a).

[0112] Next, with reference to FIG. 18(a), a flowchart of the subroutine program of the muscle training terminal process shown in S162 and the muscle training server process shown in S163 in FIG. 13(a) will be described.

[0113] In S220, the user's mobile communication device 3 determines whether the user is currently doing muscle training. If the user is not currently doing muscle training, the process returns and control proceeds to S164. If the user is currently doing muscle training, the process proceeds to S221, where the user's muscle training data is compiled and sent to the artificial intelligence server 9. The artificial intelligence server 9 receives this data in S222 and stores the received data in the learning DB 60 in S113, while also calculating the user's initial load FK, load increase coefficient az, and super recovery time CT and storing them in the user DB 12. After S113, the process returns and control proceeds to S165.

[0114] Next, with reference to FIG. 18(b), a flowchart of the muscle training service terminal process shown in S172 of FIG. 13(b) and the muscle training service server process shown in S173 will be described.

[0115] In S226, 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. The artificial intelligence server 9 receives this in S227 and reads the user's initial load FK, load increase coefficient az, and super recovery time CT from the user DB 12 in S228, and substitutes them into the general muscle training model (Fi = F0 + (i-1) / a) to create the muscle training menu for this session. Then, in S229, the artificial intelligence server 9 transmits the muscle training menu and an advertisement for the muscle training professional (see S210) that provided the muscle training data to the user's mobile communication device 3. The mobile communication device 3 receives this in S230 and notifies the user of the received muscle training menu and advertisement for the muscle training professional in S231. After S231, the process returns and control proceeds to S9. After S229, the process returns and control proceeds to S10.

[0116] FIG. 19(a) shows details (specific examples) of data stored in the dialogue model storage area in the artificial intelligence DB 17 of FIG. 10. The dialogue model storage area stores data on the initial target, dialogue template, initial weights for emotions, and emotion change functions. Various selectable targets are stored as the initial target, such as intellectual males, intellectual females, wild males, cute females, Talent A, Talent B, and Talent C. The data on the dialogue template, initial weights for emotions, and emotion change functions differ for each type of initial target, and each data is stored in association with each type of initial target. The dialogue templates are collected in large quantities to find responses to user dialogues through template matching. For example, for intellectual males, the proportion of intellectual dialogue templates is high.

[0117] The initial weights of emotions and the emotion change function are used to control changes in emotions according to the conversation. If the user praises, the emotion changes to "joy", if the user criticizes, the emotion changes to "anger", if the user is sad, the emotion changes to "sad", and if the user is happy, the emotion changes to "happy".

[0118] However, since the weights of emotions and the manner of change vary depending on the type of initial object, different initial weights and change functions for emotions are used depending on the type of initial object. The initial weights and change functions for emotions are found by machine learning using a large number of dialogue contents collected from dialogues with a large number of users as training data. The large number of users are classified into male intellectual, female intellectual, male wild, and female cute types, and the average for each type is calculated to create initial weights and change functions for emotions appropriate for each initial object. Note that for talent A, talent B, talent C, etc., the initial weights and change functions for emotions are found by machine learning using a large number of dialogue data collected from the dialogues of an actual talent (e.g., talent A). Note that the change functions for emotions may be a unified function common to all people, and individual differences (variations) may be adjusted using the initial weights for emotions.

[0119] An example of the joy, anger, sadness, and happiness change function F(x) is as follows. Let K, D, I, and R represent joy, anger, sadness, and happiness, respectively, and M represent an emotionless state. Let x1 be the number of times the user praised, x2 be the number of times they criticized, x3 be the number of times they were sad, and x4 be the number of times they enjoyed. Let K, D, I, and R be the initial weights for joy, anger, sadness, and happiness, respectively. Also, let the sigmoid function 1 / (1+e -x ) is conveniently expressed as S(x), 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)) In other words, if any of x1, x2, x3, and x4 exceeds the respective thresholds (initial weights) K, D, I, and R, the emotion will change to the emotion that exceeded it (happiness K, anger D, sadness I, or pleasure R), but if none of them exceed the above thresholds, the emotion will be neutral M.

[0120] Figure 19(b) shows details of the data stored in the storage area for personalized data for dialogue in the user DB 12 of Figure 10. Figure 19(b) shows a case where a user with user ID: 1 has selected and specified "Talent A" as the initial target for the dialogue model. "0.7AK, 1.5AD, 1.3AI, 0.5AR" are stored as personal weights for joy, anger, sadness, and happiness.

[0121] The initial emotion weights for "Talent A" are "AK, AD, AI, AR" as shown in Figure 19(a), but while the user is engaged in a conversation based on these initial emotion weights, the initial emotion weights "AK, AD, AI, AR" are gradually modified according to the user's request, and are currently "0.7AK, 1.5AD, 1.3AI, 0.5AR." As a result, the emotion change function for this user's personal emotion 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-1.5AD)-S(x3-1.3AI)-S(x4-0.5AR))

[0122] As a result, compared to the actual talent A, happiness K and happiness R are more likely to appear, and anger D and sadness I are less likely to appear.

[0123] In this way, not only can the user select and specify an initial target that matches their preferences, but as they continue to interact with each other, they can also modify the personal weights of emotions to better reflect the user's preferences. In other words, while it is nearly impossible to turn a human into the person of one's dreams, with AI it is possible to create (raise) the other person (AI) into one's ideal partner. Until now, we have lived in an age where we searched for and found our ideal lover, best friend, or partner, but in the coming future, we will be able to create our ideal lover, best friend, or partner on a network (cloud).

[0124] Next, with reference to FIG. 20, a flowchart of the subroutine program of the dialogue terminal process shown in S9 and the dialogue server process shown in S10 in FIG. 4 will be described.

[0125] In S235, the mobile communication device 3 such as a wearable computer determines whether the user has started a dialogue, and if not, returns and control proceeds to S11. If the user has started a dialogue, control proceeds to S236, where the user ID and a dialogue start signal are transmitted to the artificial intelligence server 9. Upon receiving this, the artificial intelligence server 9 determines YES in S241, and determines whether an initial target has been received in S242, and if not, control proceeds to S244.

[0126] Meanwhile, in the mobile communication device 3, in S237, it is determined whether or not an initial target has been selected. If no selection has been made, control proceeds to S239. If a selection has been made, however, in S238, the designated initial target is transmitted to the artificial intelligence server 9. Upon receiving the transmission, the artificial intelligence server 9 determines YES in S242 and performs processing in S243 to update the dialogue initial target corresponding to the user ID in the user DB 12 to the designated one.

[0127] Next, the mobile communication device 3 and the artificial intelligence server 9 communicate with each other at S239 and S244, respectively. The AI ​​server 9 then communicates based on the initial dialogue target and personal weights of emotions corresponding to the user ID in the user DB 12, and the corresponding emotion change functions in the AI ​​DB 17. The AI ​​server 9 is equipped with a natural language processing engine, and in addition to the text mining processes of "morphological analysis" and "syntactic analysis (processing to determine the dependency relationships between phrases)," it incorporates two technologies: "context analysis" required to grasp the structure of a sentence and "semantic analysis" required to understand the intent of a sentence. This enables faster and more precise analysis of the user's voice.

[0128] The artificial intelligence server 9 determines whether or not it is necessary to change the personal weights of joy, anger, sadness, and happiness during the dialogue (S245), and if it is necessary, updates the personal weights of joy, anger, sadness, and happiness in S246. For example, the artificial intelligence server 9 recognizes that the user has uttered a predetermined set phrase such as "Please reduce the weight of anger a little more," and performs control to update the personal weights of joy, anger, sadness, and happiness. Note that instead of or in addition to the set phrase, the personal weights of anger, sadness, and happiness may be updated in response to a user's utterance such as "You get upset too much."

[0129] Next, in S247, the artificial intelligence server 9 determines whether the dialogue content is work-related. If it is not work-related, the process proceeds to S249 to determine whether the dialogue has ended, but if it is work-related, the process proceeds to S248 to perform task server processing.

[0130] In the artificial intelligence server 9, if it is determined in S249 that the dialogue has not ended, control proceeds to S244, where steps S244 to S249 are repeatedly executed, and when the dialogue has ended, control returns and proceeds to S12. On the other hand, in the mobile communication device 3, if it is determined in S240 that the dialogue has not ended, control proceeds to S237, where steps S237 to S240 are repeatedly executed, and when the dialogue has ended, control returns and proceeds to S11.

[0131] Fig. 21(a) shows details (specific example) of data stored in the storage area of ​​the task processing model in the artificial intelligence DB 17 of Fig. 10. The storage area of ​​the task processing model stores data on the initial object, the initial weight of each piece of knowledge, and the knowledge utilization function. The initial objects stored include various selection objects such as patent attorney, lawyer, physicist, chemist, scientist A, scientist B, and scientist C. The data on the initial weight of each piece of knowledge and the knowledge utilization function differs for each type of initial object, and each data is stored in association with each type of initial object.

[0132] The initial weight and knowledge utilization function for each piece of knowledge are intended to make it easier for each initial subject to utilize knowledge that matches their field of expertise when processing tasks such as work. The initial weight and knowledge utilization function for each piece of knowledge are found through machine learning using a large number of conversational contents collected from work-related conversations with a large number of users as learning data. The large number of users are classified by profession, such as patent attorneys, lawyers, physicists, and chemists, and the average for each profession is calculated to create an initial weight and knowledge utilization function for each piece of knowledge that matches each initial subject.

[0133] For example, a patent attorney user asks the AI ​​server 9 many questions and consultations about his work, and a large amount of data consisting of responses such as "satisfied" or "not satisfied" to the responses from the AI ​​server 9 is used as learning data and modeled using algorithms such as reinforcement learning and regression. Note that for scientist A, scientist B, scientist C, etc., a large amount of dialogue data collected from actual dialogues between a certain scientist is used as learning data, and the initial knowledge weights and knowledge utilization functions are found by machine learning. Note that the knowledge utilization function may be a unified function common to all people, and individual differences (variations) may be controlled to be adjusted by the initial knowledge weights.

[0134] An example of the knowledge utilization function G(x) is as follows. Each piece of knowledge is represented as n1, n2, n3, ..., the number of positive responses by users to each piece of knowledge, such as "satisfied", is represented as x1k, x2k, x3k, ..., the number of negative responses, such as "not satisfied", is represented as x1h, x2h, x3h, ..., and the initial weights of each piece of knowledge are represented as w1, w2, w3, .... Also, the sigmoid function 1 / (1+e -x ) is conveniently expressed as S(x), G(x)=w1n1S(x1k-x1h)+w2n2S(x2k-x2h)+w3n3S(x3k-x3h)+... In other words, knowledge for which the number of positive responses exceeds the number of negative responses is deemed to be effective knowledge to use, and it is possible to control the coefficients (weights w1, w2, w3, ...) of each piece of knowledge n1, n2, n3, ... so that knowledge with larger values ​​is used first.

[0135] Figure 21(b) shows details of the data stored in the storage area for personalized data for task processing in the user DB 12 of Figure 10. Figure 21(b) shows a case where a user with user ID: 1 has selected and specified "Patent Attorney" as the initial target of the dialogue model. The personal weights w of each knowledge n are stored as "1.2·1w11n1, 0.8·1w21n2, 0.7·1w31n3, 1.3·1w41n4."

[0136] As shown in Figure 21(a), the initial weights for each piece of knowledge for a patent attorney are "1w11n1, 1w21n2, 1w31n3, 1w41n4." However, as the user engages in business-related conversations based on these initial weights, the initial weights for each piece of knowledge ("1w11n1, 1w21n2, 1w31n3, 1w41n4") are gradually modified according to the user's requests, and are currently "1.2·1w11n1, 0.8·1w21n2, 0.7·1w31n3, 1.3·1w41n4." In this way, the user can select and specify the initial target that best suits their profession. Furthermore, as business conversations continue, the weights for each piece of knowledge can be modified to better reflect the user's questions and concerns. In other words, in the future, we will be able to create our ideal business partner (partner) on the network (cloud).

[0137] Next, with reference to FIG. 22, a flowchart of the task terminal processing shown in S11 of FIG. 4 and the task server processing subroutine program shown in S12 and S248 of FIG. 20 will be described.

[0138] In S255, the user's mobile communication device 3 determines whether or not the user has requested that the user's personal weight be registered as an initial target in the AI ​​DB 17. If not, the control proceeds to S259, where it is determined whether or not a request to update the initial target has been made, and if not, the control proceeds to S265.

[0139] On the other hand, if a request is made in S255 to register the user's (the user's) personal weight as the initial target in the artificial intelligence DB 17, S255 judges YES, control proceeds to S256, and the user ID and the initial target registration request are sent to the artificial intelligence server 9.

[0140] In the artificial intelligence server 9, S257 determines whether an initial target registration request has been received, and if not, control proceeds to S261, where it is determined whether a requested initial target has been received, and if not, control proceeds to S267.

[0141] If the user ID and the initial target registration request are transmitted from the mobile communication device 3 in accordance with the processing of S256, a YES determination is made in S257, and control proceeds to S258, where the personalized weights of each piece of knowledge corresponding to the user ID are read from the user DB 12 and registered together with the user name in the AI ​​DB 17. In this manner, in this embodiment, the user can register the personalized weights of each piece of knowledge in the AI ​​DB 17 and distribute them widely to the general public. For example, when efficiently training a subordinate at work to turn them into a fully-fledged expert, the supervisor can transfer the personalized weights of each piece of work-related knowledge that the supervisor has cultivated to the subordinate. The AI ​​server 9 will respond to the subordinate's questions and inquiries on the user's behalf, and the AI ​​server 9 will take over the training of the subordinate. When the user transfers the personalized weights of each piece of knowledge to another person, a payment method for settlement with the transferee may be provided, allowing the transfer to be made for a fee.

[0142] If the user requests the mobile communication device 3 to update the initial object and specifies a new initial object (requested initial object), a YES determination is made in S259 and control proceeds to S260, where the user ID and requested initial object are sent to the artificial intelligence server 9. Upon receiving the request, the artificial intelligence server 9 makes a YES determination in S261 and control proceeds to S262, where it determines whether the received requested initial object is one that has been used previously by the user. If the requested initial object has not been used before, a process is performed in S264 to read the requested initial object from the artificial intelligence DB 17 and store it in the user DB 12 in association with the user ID, and control proceeds to S267.

[0143] If the received requested initial object is one that has been used previously by the user, the initial object may already be stored in the user DB 12, and the personal weight may have been updated with the use of the initial object (see S271). Therefore, if the received requested initial object is one that has been used previously by the user, control proceeds to S263, and the personal weight of the previously used knowledge stored in the user DB 12 is used.

[0144] 23, the artificial intelligence server 9 reads out in S267 an initial object corresponding to the user ID from the user DB 12 and the personal weight of each piece of knowledge, and then reads out in S268 a knowledge utilization function corresponding to the read initial object from the artificial intelligence DB 17. When the user then uses the mobile communication device 3 to engage in a business-related dialogue, the dialogue is carried out between S265 on the mobile communication device 3 side and S269 on the artificial intelligence server 9 side. In S269, the business-related dialogue is carried out using the read knowledge utilization function and personal weight. On the mobile communication device 3 side, the business-related dialogue control in S265 continues until it is determined in S266 that the business-related dialogue has ended. On the artificial intelligence server 9 side, the controls of S267 to S274 are repeated until it is determined in S274 that the business-related dialogue has ended, and the business-related dialogue control in S269 continues.

[0145] If it is determined in S270 that the personal weight needs to be changed while the controls of S267 to S274 are being repeated, control proceeds to S271, where the personal weight in the user database 12 is updated. If the personal weight has also been registered in the AI ​​database 17 by the user (see S255 to S258), a YES determination is made in S272, and the weight in the AI ​​database 17 is also updated in S273. This determination in S270 determines the need to change the knowledge weight used by the AI ​​server 9, for example, by the user uttering a predetermined set phrase such as "satisfied" or "not satisfied" in response to a response from the AI ​​server 9. Note that instead of or in addition to the set phrase, the determination may be made based on the user's normal conversation, such as "Thank you for the great answer."

[0146] If it is determined in S266 on the mobile communication device 3 side that the business dialogue has ended, an instruction to end the business dialogue is sent to the artificial intelligence server 9 in S273, and the process returns and the control shifts to S1. The artificial intelligence server 9, having received this instruction, judges YES in S274, returns, and the control shifts to S3.

[0147] Modifications and features of the above-described embodiment will be described below. (1) At least one or all of the IoT server 9 and IoT device DB 15, the artificial intelligence server 9, the artificial intelligence DB 17, the user DB 12 and learning DB 60, and the SNS server 11 and SNS DB 13 may be configured as a network cloud. In the above-described embodiment, the mobile communication device 3 has been mainly used as an example of a user terminal, but the user terminal may also be a user PC 7 or a robot 6. While a wearable computer, such as smart glasses, has been shown as a specific example of the mobile communication device 3, the mobile communication device 3 is not limited to this, and various devices, such as a mobile phone, a smartphone, or a vehicle, such as an automobile, that has communication capabilities, may be used.

[0148] (2) The IoT server 8 is provided with an identification means (S24 to S26, S38, S40) for identifying IoT devices with which information needs to be exchanged. The geographical locations of the IoT devices identified by the identification means may be transmitted to the wearable computer and notified to the user, encouraging the user to act as an intermediary. In this case, the user may be notified that high scores (e.g., points) will be awarded to IoT devices with a particularly high need for intermediation.

[0149] (3) It is also possible to control the system so that only a limited number of IoT devices are given high-value rewards (points, etc.), and to promote intermediation activities by utilizing lottery-like psychology. It is also possible to control the system so that intermediation activities of IoT devices installed in geographical locations are performed using location games.

[0150] (4) The above-described embodiments disclose the following inventions. The present invention relates to a system that reduces the cost of IoT devices (sensors, actuators, etc.) that are installed all over the Earth, thereby realizing IoT at low cost.

[0151] The following is an example of background technology related to systems that realize IoT inexpensively. A border router (gateway device) is introduced between an IoT sensor and a server on the Internet, and the server address is not set in the sensor; instead, sensor data information is sent to the border router after its own address is determined. The border router relays the sensor information to a sensor data management server to collect data. This allows the server to detect the sensor address and perform pull-type data collection addressed to the sensor address (Japanese Patent Application Laid-Open No. 2014-78773).

[0152] However, in the case of this background technology, although pull-type data collection from the server to the sensor address is possible, it is necessary to provide each sensor with an internet connection function such as a network I / F, or to prepare a proprietary adapter for connecting to the network, which has the disadvantage of increasing the costs involved in realizing IoT.

[0153] On the other hand, some IoT devices, which are installed all over the world, do not necessarily require pull-type data collection from the server to the sensor address. For example, in the case of an illuminance sensor installed to tally illuminance throughout the year, it is sufficient to store periodic illuminance detection results in memory and send the stored illuminance detection results to the server by the time the server tallys the illuminance throughout the year once a year. Also, for IoT devices installed in places with a lot of human traffic, it is possible to consider a method of sending IoT device data to the server using a human-assisted method.

[0154] The present invention was conceived in view of the above circumstances, and its purpose is to provide a system that can reduce costs in realizing IoT.

[0155] The present invention relates to a wearable computer carried by a user, a server communicating with the wearable computer via the Internet; Equipped with a group of IoT devices, the IoT devices include a communication means for communicating with the wearable computer; The wearable computer includes: a device communication means for communicating with the IoT device group; a server communication means for communicating with the server via the Internet; an intermediary means that mediates between the IoT device that is the communication partner of the device-to-device communication means and the server that is the communication partner of the server-to-server communication means, enabling the exchange of information between them.

[0156] With this configuration, the IoT devices do not necessarily need to have Internet connection capabilities, which reduces costs and allows IoT to be realized at low cost.

[0157] Preferably, the server further includes a reward granting unit for granting a predetermined reward to the wearable computer that has performed the mediation through the mediation unit.

[0158] According to this configuration, it is possible to provide users with an incentive to act as intermediaries, thereby encouraging users to act as intermediaries.

[0159] More preferably, the server further includes a determination means for determining whether or not there is an IoT device with which information needs to be exchanged.

[0160] With this configuration, the IoT device group does not need to have the function to determine whether or not information exchange is necessary, which reduces the processing burden on the IoT device group and further reduces costs.

[0161] More preferably, the wearable computer further includes a location information transmitting means for transmitting information capable of identifying a current location to the server; The server a location determination means for determining whether the IoT device identified by the identification means is present in the vicinity of the wearable computer whose location is identified by the information transmitted by the location information transmission means; an intermediation command sending means for sending a command signal to the wearable computer to cause the intermediation means to mediate the IoT device when the location determining means determines that the IoT device is present; The mediation means performs mediation in accordance with the command signal from the mediation command transmission means.

[0162] With this configuration, the wearable computer side requests the IoT device that needs to act as an intermediary, so there is no need for the IoT device group side to transmit an intermediation request signal.

[0163] More preferably, the wearable computer comprises: a sensor that requires a predetermined time for detection; and a notification means (e.g., S71) for notifying a user of a time required for detection by the sensor when performing detection using the sensor at a desired geographical position, If the sensor can perform the detection for the required detection time within the detection area of ​​the desired geographical location, the detection data is transmitted to the server via the intermediary means.

[0164] With this configuration, if a certain amount of time is required for detection, the wearable computer notifies the user of the required detection time, thereby encouraging the user to cooperate in collecting detection data that requires time.

[0165] (5) In the above-described embodiments, specific examples of machine learning have been described, including the creation of a building maintenance and inspection model, a memorization learning model, a muscle training model, a dialogue model, and a task processing model. However, these examples are not limited to these, and various other models are conceivable, such as a learning model for young children, a model for an artificial secretary, and a model for an artificial tutor. Furthermore, the personalized data is not limited to those for memorization learning, muscle training, dialogue, and task processing, and various other models are conceivable, such as those for young children, a model for an artificial secretary, and a model for an artificial tutor. These general models and personalized data are controlled so that they can be switched and used appropriately depending on the user's situation. As a result, shortly after birth, humans grow while receiving personalized services based on artificial intelligence using their own personalized data, and will live their entire lives together with artificial intelligence. In other words, in the future, humans and artificial intelligence will be paired together. This will be a world in which humans and artificial intelligence are paired together to form a single personality. Furthermore, in this embodiment, when switching between the general model and personalized data, the previous personalized data for dialogue is used unless the user indicates an intention to switch, but the personalized data for dialogue may also be controlled to automatically switch to another appropriate one when switching between the general model and personalized data. For example, talent A may be used as the initial subject for dialogue during general everyday conversation, and automatically switched to a female intellectual type (see FIG. 19) during business-related conversation.

[0166] (6) Recommendation control may be performed to recommend artificial intelligence or general models that are thought to match the user based on around data collected from the user by the mobile communication device 3 or the like. Furthermore, when a user searches for an artificial intelligence or general model, the search results displayed as a list may be controlled to display those that are thought to match the user based on around data collected from the user. Furthermore, user reviews of the artificial intelligence or general model may be collected and displayed. In this case, since a person and an artificial intelligence are paired to form a personality, as described above, control may be performed to collect and display reviews about the paired person and artificial intelligence.

[0167] (7) For infants trained using the general model and personalized data for infant learning, a large amount of data consisting of the learning materials (learning information) and learning results can be collected and used as learning data for machine learning to model human intellectual growth (e.g., supervised learning). The same learning materials (learning information) provided to the infant can also be provided to an artificial intelligence, and the learning results of the resulting intellectually developed artificial intelligence can be compared with those of an actual infant. Reinforcement learning or regression can be performed to minimize the differences, creating a learning model for artificial intelligence that enables learning closer to that of a human. Comparing the learning results of both can be particularly beneficial for artificial intelligence in areas where artificial intelligence is weak (e.g., common sense, emotions, creativity, etc.).

[0168] (8) In this embodiment, the personalization means (e.g., S191, S223, S246, S271) for personalizing the general model to a model suitable for the user based on information sent from the user is provided in the artificial intelligence server 9, but instead of or in addition to this, it may be provided in the user's terminal (e.g., a mobile communication device 3 such as a wearable computer, a user PC, a robot, etc.).

[0169] The above-described embodiments disclose the following inventions. The present disclosure relates to a service provision system and a program that utilizes machine learning through artificial intelligence, for example. More specifically, the present disclosure relates to a service provision system that provides a service that utilizes machine learning through artificial intelligence, and a program that is executed by a computer to allow a user to receive a service that utilizes machine learning through artificial intelligence.

[0170] There have been techniques for creating learning data from each of a plurality of information sources in order to perform machine learning using artificial intelligence (for example, Japanese Patent Application Laid-Open No. 2011-232997).

[0171] However, machine learning requires a huge amount of training data, and in order to create this vast amount of training data, it is necessary to collect a large amount of necessary information from sources. In order to put machine learning into practical use, it is essential to reduce the effort and cost required to collect this information.

[0172] Furthermore, for example, a general model created by collecting large amounts of information from an unspecified number of people and performing machine learning using a huge amount of learning data tends to be an average model for the entire unspecified number of people, and if that general model is used to provide a service to each user, each of whom is unique, there is a risk that the service will not be a good match for users who deviate from the average.

[0173] The present disclosure was conceived in light of this situation, and its purpose is to reduce the effort and cost required to collect the vast amount of information necessary to create training data. Another purpose is to solve the inconvenience that occurs when a service using a general model resulting from machine learning does not match the user.

[0174] Next, an example of the correspondence between various means for solving the problems and embodiments is shown below in parentheses.

[0175] The present disclosure relates to a service provision system (e.g., FIGS. 9 and 10 ) that provides services using machine learning by artificial intelligence, A machine learning means (e.g., S189, S214, FIG. 15, etc.) for generating a general model (e.g., a rote learning model Ti=T0·I, a muscle training model Fi=F0+(i-1) / a, etc.) modeled by machine learning using learning data based on information sent from the user (e.g., around data such as the video and gaze position, voice, and GPS location data viewed by the user 70); personalization means (e.g., S191 and S204, S223 and S228, etc.) for personalizing the general model to a model suitable for the user based on information sent by the user; and a service providing means (e.g., S206, S229, etc.) for providing the user with personalized services (e.g., providing study items based on the most efficient review plan, providing muscle training menus based on the most efficient muscle training plan, etc.) using the personalized model; Information sent by the user is used for both the machine learning and the personalization.

[0176] With this configuration, by providing users with a service that utilizes the results of machine learning, many users will proactively provide information for training data. Moreover, since the information provided by users is also used to personalize a general model to a model suitable for that user, it is possible to minimize the inconvenience of each user relying on others to provide information for training data and only enjoying the service.

[0177] Preferably, the machine learning means generates the general model using factors that differ for each user (e.g., individual ability differences such as memorization ability, initial load for muscle training, load increase coefficient, preferences, etc.) as constants (e.g., T0, F0, a, etc.) (e.g., S189, S214, FIG. 15, etc.), The personalization means comprises: A value derivation means (e.g., S191, S223, etc.) for deriving the actual values ​​(e.g., TK, FK, az, CT, etc.) of the user that fit into the constant part of the general model based on information sent from the user; and substitution means (e.g., S204, 228, etc.) for substituting the actual value derived by the value derivation means into the constant part of the general model to personalize the model to suit the user.

[0178] More preferably, the machine learning means also inputs learning data based on information sent from a professional (e.g., a muscle training professional) and generates a general model modeled by machine learning (e.g., S214, etc.).

[0179] Another aspect of the present invention is a service providing system that provides a service using machine learning by artificial intelligence, A storage means (e.g., an artificial intelligence DB 17) for storing a plurality of types of general models (e.g., the dialogue model of FIG. 19(a), the task processing model of FIG. 21(a), etc.) generated by modeling using machine learning; a service providing means (e.g., S241 to S244, S247, S248, S261 to S263, S267 to S269, etc.) for providing a service (e.g., dialogue service, response to business-related questions or consultations, etc.) to a user by utilizing a general model selected by the user from a plurality of types of general models stored in the storage means; personalizing means (e.g., S246, S271, etc.) for personalizing the general model into a model suitable for the user in accordance with the user's response to the service provided by the service providing means; The service providing means includes a personal service providing means (for example, executing S244, S269, etc. according to the updated personal weights) for providing a service to the user using the model personalized by the personalization means.

[0180] With this configuration, a user can select a desired general model from among multiple types of general models generated by modeling using machine learning, and a service is provided to the user using that general model, allowing the user to enjoy a service based on a general model that reflects the user's preferences. Moreover, the general model is personalized to a model suitable for the user based on the user's reaction to the service, making it possible to provide a service that matches the user's preferences through personalization of the service provided.

[0181] Preferably, the personal service providing means has means (e.g., S263, etc.) for providing a service to the user using the personalized model at the time of use when the general model selected by the user has previously been used to provide a service to the user (e.g., YES in S262).

[0182] More preferably, the system further comprises a model using means (for example, S257, S258, etc.) for allowing other people to use the model personalized by the personalizing means.

[0183] Another aspect of the present disclosure is a program executed by a computer to allow a user to receive a service utilizing machine learning by artificial intelligence, the program comprising: A step (e.g., S237 and S238, S259 and S260, etc.) in which the user selects and specifies a desired model from among multiple types of general models (e.g., 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; processing steps (e.g., S236 to S239, S260, S265, etc.) for the user to receive a service using the general model selected and designated by the selecting and designating step; A step (e.g., S239, S265, etc.) of providing the user's reaction to the artificial intelligence to personalize the general model into a model suitable for the user according to the reaction of the user who received the service by the processing step; causing the computer to execute The processing steps include a personal service receiving step (e.g., receiving a response according to the updated personal weights in S239, S265, etc.) in which the user receives personalized services using the model personalized by the artificial intelligence.

[0184] With this configuration, a user can select a desired general model from among multiple types of general models generated by modeling using machine learning, and a service is provided to the user using that general model, allowing the user to enjoy a service based on a general model that reflects the user's preferences. Moreover, the general model is personalized to a model suitable for the user based on the user's reaction to the service, making it possible to provide a service that matches the user's preferences through personalization of the service provided.

[0185] Although the embodiments of the present invention have been described above, the embodiments disclosed herein are illustrative in all respects and should not be considered limiting. The scope of the present invention is defined by the claims rather than the above description, and it is intended to include meanings equivalent to the claims and all modifications within the scope of the claims. [Explanation of symbols]

[0186] 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 databases, 15 IoT device databases, 17 Artificial intelligence databases, 60 Learning databases, 38 Various sensors.

Claims

1. A support system that uses artificial intelligence to support a user's repeated practice for self-improvement, data collection means for collecting data on a plurality of users who perform the practice; a storage means for storing a general model generated by performing machine learning based on data of a plurality of users collected by the data collection means; a personalized model generation means for generating a personalized model personalized for an individual user based on the general model stored in the storage means and the individual data for one user collected by the data collection means; a practice content specification means for specifying practice content to be executed when the individual user repeatedly performs practice using the personalized model generated by the personalized model generation means; a service providing means for providing a service for the individual user to perform the practice content identified by the practice content identifying means, the general model is a model that applies commonly to multiple users and is generated to streamline the repeated practice; The personalized model is a model created specifically for an individual user to streamline the repeated practice of the individual user, according to an assistance system.

2. The general model is a model generated using factors that vary from user to user as constants, The personalized model generation means a value derivation means for deriving an actual value of the individual user that fits into the constant part of the general model based on information sent from the individual user; 2. The assistance system according to claim 1, further comprising: substitution means for substituting the actual value derived by said value derivation means into a constant part in said general model to generate said personalized model.

3. 3. The support system according to claim 1, wherein the data collection means collects data on a plurality of users sent from a professional company that provides the practice to users.

Images