Service providing 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 Machine learning a service provision system that enables Provide services using a model To the extent of .

Background Art

[0002] For example, as a voice control system, by recognizing uttered commands and associated words (e.g., "call mom at home") and causing the selected application (e.g., phone dialer) to execute the command, the data processing system such as a smartphone is made to execute an operation based on the command (e.g., look up moms phone number at home and dial it to establish a telehone call). There have been such things (for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] An object of the present invention is to enable a model personalized for a specific person. Specific examples of means for solving the problems and their effects Utilization

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

[0008] The invention according to claim 1 is Machine learning a service provision system that enables Provide services using a model (for example, After optimizing (personalizing) the general model obtained as a result of machine learning for each user, provide a personalized service to each user. etc.), Means for storing a plurality of types of models for task processing as an average general model modeled by performing machine learning on learning data collected from a large number of people (e.g., FIG. 21(a), etc.) as an initial target, and storing the initial weights of each piece of knowledge in the model for task processing in association with each model for task processing. The initial target storage means (e.g., the artificial intelligence DB17 in FIG. 10), By using the model for task processing selected by the user from among the various models for task processing stored as the initial target in the initial target storage means (e.g., NO in S260, S261, YES in S264, S265, S269), the weight update means for updating and personalizing the weights of each piece of knowledge in the model for task processing so that the user is satisfied (e.g., YES in S270, S271), Personalized weight storage means for storing the personalized weights of each piece of knowledge updated and personalized by the weight update means (e.g., FIG. 21(b)) in association with the user (e.g., the user DB12 in FIG. 10), Means for storing the personalized weights of each piece of knowledge stored in the personalized weight storage means in the initial target storage means as the user's own initial target (e.g., YES in S255, S256 to S258). Make the personalization weights of each piece of knowledge of the user himself / herself stored in the initial target storage means generally available, and make the personalization weights of each piece of knowledge available as a substitute for the user himself / herself (for example, the user can register the personalization weights of his / her own knowledge in the artificial intelligence DB17 and make them widely available for general circulation. For example, when efficiently educating subordinates at work to train them into full-fledged experts, if the user, who is the supervisor, transfers the personalization weights of each piece of work-related knowledge cultivated so far to the subordinates, the artificial intelligence server 9 can respond to the subordinates' questions and consultations on behalf of the user, and the artificial intelligence server 9 can take over the education of the subordinates, etc.).

[0010] The invention according to claim 2 is A service providing system capable of providing a service using a machine learning model (for example, after optimizing (personalizing) a general model obtained as a result of machine learning for each user, providing a personalized service to each user, etc.), a large number of humanModeled by performing machine learning on learning data collected from average The first means for generating a general model ( For example, Classify a large number of users into each professional occupation such as patent attorneys, lawyers, physicists, chemists, etc., calculate the average of each professional occupation, and create an initial weight of each knowledge and a knowledge utilization function suitable for each initial target. As an example of the knowledge utilization function G(x), for example, the following can be considered. Represent each knowledge as n1, n2, n3, ···, and for each knowledge, the number of positive reactions such as "satisfied" by users is x1k, x2k, x3k, ···, and the number of negative reactions such as "not satisfied" is x1h, x2h, x3h, ···, and the initial weights of each knowledge are w1, w2, w3, ···, respectively. Also, if the sigmoid function 1 / (1 + e -x ) is expressed as S(x) for convenience, G(x) = w1n1S(x1k - x1h) + w2n2S(x2k - x2h) + w3n3S(x3k - x3h) + ··· That is, knowledge in which the number of positive reactions exceeds the number of negative reactions is regarded as effective knowledge for utilization, and it can be controlled so that knowledge with large coefficients (weights w1, w2, w3, ···) of each knowledge n1, n2, n3, ··· is preferentially utilized.) and, The second means for specifying the general model generated by the first means for a specific person personalize it to a personalized model (for example, S255~S258, S270~S273 Regarding the determination by this S270, for example, based on predetermined set phrases such as "satisfied" or "not satisfied" uttered by the user in response to the response from the artificial intelligence server 9, the artificial intelligence server 9 determines the necessity of changing the weights of the knowledge being used. Note that instead of or in addition to the above set phrases, the determination may be made based on the user's normal conversation such as "Thank you for the wonderful answer". ) and, comprising Make the personalized model generally available, and make the personalized model available as a substitute for the specific person (for example, the user can register the personalization weights of his / her own knowledge in the artificial intelligence DB17 and make them widely available for general circulation. For example, when efficiently educating subordinates at work to train them into full-fledged experts, if the user, who is the supervisor, transfers the personalization weights of each piece of work-related knowledge cultivated so far to the subordinates, the artificial intelligence server 9 can respond to the subordinates' questions and consultations on behalf of the user, and the artificial intelligence server 9 can take over the education of the subordinates, etc.).

[0011] The invention according to claim 3 is the claim 2 described in In the service providing system, the first means subtracts the total number of negative responses x1h, x2h, x3h,... from the total number of positive responses x1k, x2k, x3k,... by the user for each of the pieces of knowledge n1, n2, n3,... in the model to be generated, and for each of these values, applies the sigmoid function 1 / (1 + e -x ) is substituted into, and by multiplying the values of the sigmoid functions corresponding to the respective knowledges by the respective knowledges n1, n2, n3, ··· corresponding thereto, a general model is formed in which the priorities in utilizing the respective knowledges n1, n2, n3, ··· are determined. The second means is such that when the user, who is the specific person, uses the general model, the weights of the respective knowledges in the general model are updated so as to satisfy the user, and a personalized model having the personalized weights of the updated and personalized knowledges is formed (for example, S255 to S258, S270 to S273. The determination in this S270 is made, for example, by the user saying predetermined set phrases such as "satisfied" or "not satisfied" in response to the response from the artificial intelligence server 9, and the artificial intelligence server 9 determines the necessity of changing the weights of the knowledges used on its side. Note that instead of or in addition to the above set phrases, it may be determined based on the user's normal conversation such as "Thank you for the wonderful answer").

Brief Description of Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Figure 23

Embodiments for Carrying Out the Invention

[0013] [Device Mediation System for IoT] First, a device mediation system for IoT will be described based on FIGS. 1 to 8. IoT is an abbreviation for Internet of Things, in which things are networked using Internet protocols and things transmit signals onto the Internet by themselves. The device mediation system for IoT is a system that mediates communication for a group 2 of IoT devices (various sensors and actuators for IoT, etc.) that do not have a function for connecting to the Internet and communicating, and enables the group 2 of IoT devices to be connected to the Internet 1 via a mobile communication device 3 of a user and communicate with each other.

[0014] Referring to the overall system in FIG. 1, a mobile communication device 3 represented by a user's wearable computer or the like, a robot 6 at the user's home, a personal computer (hereinafter referred to as "PC") 7 at the user's home, an IoT server 8, an artificial intelligence server 9, a PC 10 of various specialists, and an SNS server 11 are connected to the Internet 1 and configured to be able to communicate with each other. The mobile communication device 3 has a function for communicating with the group 2 of IoT devices, a wireless sensor network 4, and a display group 5 such as a digital menu board. As communication methods, for example, Wi-Fi (registered trademark), Bluetooth (registered trademark), Wi-Fi Direct (registered trademark), Zigbee (registered trademark), Z-wave (registered trademark), Ant+ (registered trademark), etc. are assumed. Also, it is compatible with iOS, Android (registered trademark), Linux (registered trademark), TIZEN (registered trademark), and other real-time operating systems. The wireless standard for transmission and reception adopts IEEE802.15.4. Also, IPv6 (Internet Protocol Version 6) is adopted as the Internet Protocol.

[0015] The wireless sensor network 4 is a wireless network that scatters a plurality of wireless terminals with sensors in space, enabling them to cooperate to collect environmental and physical conditions. For example, a sensor device is made using energy harvesting, M2M, or a battery, etc. For example, deterioration such as metal fatigue is constantly monitored with a pressure sensor or a gauge sensor, and when there is a change, it is notified. It is mainly installed in structures such as bridges and tunnels. Generally, it includes a plurality of sensor nodes and a gateway sensor node. These nodes are usually composed of one or more sensors, a wireless chip, a microprocessor, and a power source (such as a battery). The hardware configuration of the control circuit of the node is the same as that shown in Fig. 2(a). The wireless sensor network usually has an ad hoc function and a routing algorithm for sending data from each node to the central node. That is, it has a function of autonomously reconstructing another communication path when there is an obstacle in communication between nodes. There is also an element of distributed processing for nodes to cooperate as a group. In addition, it has a function of operating for a long time without receiving external power supply, and for this purpose, it has a power-saving function or a self-power generation function. 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 and from the IoT device database (hereinafter referred to as "DB"), the IoT device DB 15. The artificial intelligence server 9 can write and read data to and from the artificial intelligence DB 17, the user DB 12, and the learning DB 60. The SNS server can write and read data to and from the SNSDB 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 includes sensors and actuators installed all over the earth, a CPU (Central Processing Unit) 18 for controlling the whole, a ROM (Read Only Member) 20 storing programs for executing various functions, a RAM (Random Access Memory) 19 which is the work area of the CPU 18, and a wireless communication interface unit 22 using, for example, Wi-Fi, Bluetooth, Wi-Fi Direct, Zigbee, Zwave, Ant+ and the like. When the IoT device 2 is a sensor, it includes a sensor unit 14. On the other hand, when 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, the RAM 19, the ROM 20, the wireless communication interface unit 22, the sensor unit 14, and the actuator unit 21 are connected so that signal exchange is possible.

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

[0019] Bus 27 has various devices connected via interface unit 28. For example, a digital video camera input unit 29 that captures video around the user (such as video in the user's line of sight direction), a display unit 30 that overlays information and the like on the lens unit of the smart glasses, a wireless communication processing unit 31 that performs wireless communication with a base station and conducts data communication with a server etc. via the Internet 1, an input operation unit 32 for causing the CPU 23 of the wearable computer 3 to execute functions desired by the user, a voice output unit 33 and a voice input unit 34 for the user to make a call by voice, a position information acquisition unit 35 for acquiring the current position based on GPS information from a satellite, radio waves from a base station, and wireless radio waves from a wireless LAN access point, a wireless communication interface unit 36 for communicating with the IoT device group 2 using Wi-Fi, Bluetooth, Wi-Fi Direct, Zigbee, Zwave, Ant+ etc., a line-of-sight position detection unit 37 for detecting the user's line-of-sight position and specifying the position of the user's line of sight in the video captured by the digital video camera input unit 29, various sensors 38 etc. are connected to the interface unit 28.

[0020] The input operation unit 32 not only accepts manual operations by the user, but also accepts instructions by the user's gestures and specific winks etc. Note that instructions by the user's voice are accepted by the voice input unit 34. The various sensors 38 are sensors such as temperature, humidity, illuminance, ultraviolet rays etc. attached to the wearable computer 3. Note that for the hardware configuration shown in this Figure 2(a), removing the various sensors 38 and adding a drive circuit for movement results in the hardware configuration of the control circuit of the robot.

[0021] Next, based on Figure 3, the hardware configuration of the control circuits of a PC and various servers 7 - 11 will be described. Similar to the above, a CPU 40, a RAM 41, and a ROM 42 are connected by a bus 43. To the interface unit 44 to which the bus 43 is connected, a communication unit 45 with the Internet 1 etc., a display unit 46 for displaying video and information to an operator, and an input operation unit 47 for accepting operations from the operator are connected.

[0022] Next, based on FIG. 4, a flowchart of the main program of the control process executed by the CPU 23 of the mobile communication device 3 and the CPU 40 of the artificial intelligence server 9 will be described. The mobile communication device 3 performs IoT processing by S1. This is a process for the mobile communication device 3 to connect the IoT device group 2 and the Internet 1 through mediation, transmit the detection data of the IoT device (sensor) 2 to the IoT server 8, and transmit the 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 by S2, and the artificial intelligence server 9 performs display processing by S3. This is a process for displaying videos, still images, etc. uploaded by the user on the Internet on the display 5 such as a digital menu board. Next, the mobile communication device 3 performs around data terminal processing by S4, and the artificial intelligence server 9 performs around data server processing by S5. This is a process for collecting videos and sounds around the user by the mobile communication device 3 such as a wearable computer and using them as information for machine learning by the artificial intelligence server 9.

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

[0025] Next, based on FIGS. 5 to 8, the flowchart of the subroutine program for the IoT processing shown in S1 above will be described. First, referring to FIG. 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 from the sensor unit 14 in the RAM 19 in S20, determines in S21 whether a transmission command signal has been received, and if not, returns to S20 and repeatedly loops through the loop of S20→21→S20.

[0027] On the other hand, the CPU 23 of the mobile communication device 3 transmits, in S22, position 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 that has received this in S24 searches the IoT device DB 15 in S25 and determines in S26 whether there is an IoT device 2 to be received in the vicinity of the mobile communication device 3 that has transmitted the position information. It is ideal for the IoT server 8 to periodically collect the detection data of the IoT device (sensor) 2, and it determines the IoT device (sensor) 2 for which the periodic reception time has come by searching the IoT device DB 15. If the IoT device 2 to be received exists, a YES determination is made in S26 and the control proceeds to S27, and 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 be received, a NO determination is made at S26 and the control proceeds to S38, where it is determined whether there is an IoT device (actuator) 2 for which a command signal (actuator control signal) is to be transmitted in the vicinity. If there is, the control proceeds to S39, and the command signal for transmission and the transmission information (information for controlling the actuator unit 21) are transmitted to the mobile communication device 3. The mobile communication device 3 that has received the command signal at S27 or S39 makes a YES determination at S23 and the control proceeds to S28, where it is determined whether the received command signal is for reception. In the case of the transmission command signal transmitted at S38, the control shifts to S47, but in the case of the reception command signal transmitted at S27, the control proceeds to S29, and the transmission command signal is transmitted to the nearby IoT device (sensor) 2.

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

[0030] Next, the IoT server 8 transmits, at S35, a point as an example of a privilege to the mobile communication device 3, and then the control shifts to S61. The mobile communication device 3 that has received it at S36 updates the points stored in the EEPROM 26 by addition (S37), and then the control shifts to S60. The points stored in this EEPROM 26 can be used to receive various services from the operator operating the IoT server 8. For example, the points can be used to pay highway tolls or purchase train tickets or commuter passes for JR or private railways. Also, the points may be exchangeable for money. This can give an incentive to the user for the mediation of IoT devices.

[0031] On the other hand, if it is determined as NO in S38, the control proceeds to S40, and it is determined whether or not there are various sensors 38 in the mobile communication device 3 that wants to receive. That is, it is determined whether or not there is any sensor 38 provided in the mobile communication device 3 located at the position information received in S24 for which it is desired to receive detection data at the geographical position. If not, the control shifts to S24, but if there is, the control shifts to S62.

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

[0033] The mobile communication device 3 that has received that in S50 transmits a completion signal to the IoT server 8 in S51. The IoT server 8 that has received that in S52 searches the IoT device DB based on the device ID in S53 and updates the data corresponding to the device ID. For example, it is updated such as "At ○:○△× transmission information has been transmitted". Then, after transmitting points to the mobile communication device 3 in S54, the control shifts to S61. The mobile communication device 3 that has received that in S55 adds and updates the points stored in the EEPROM 26 (S56). After that, the control shifts to S60.

[0034] In S60, it is determined whether a command signal for mobile reception and a data identification signal are received from the IoT server 8. If not received, the control proceeds to S84. On the other hand, when the IoT server 8 determines that there are various sensors 38 of the mobile communication device 3 to be received at the position received by S24, it determines YES by S40 and transmits a command signal for mobile reception and a data identification signal by S62. The mobile communication device 3 that has received it determines YES by S60 and the control proceeds to S63, and determines whether the data (such as temperature, humidity, illuminance, ultraviolet rays, etc.) specified by the data identification signal requires a detection time. In the case of illuminance and ultraviolet rays, since they can be detected instantaneously and do not require a detection time, it is determined NO by S63 and the control proceeds to S64, and performs a process of detecting the specified data with the corresponding sensor 38 and transmits the detection data to the IoT server 8 (S65).

[0035] The IoT server 8 that has received it at S66 updates the IoT device DB15 by searching and updating it with the newly received detection data by S67. Then, it transmits points to the mobile communication device 3 by S68. The mobile communication device 3 that has received it at S69 adds and updates the points stored in the EEPROM 26 (S70). After that, the control proceeds to S84.

[0036] On the other hand, when the data specified by the data specifying signal, such as temperature or humidity, requires a detection time, a YES determination is made in S63, and the control proceeds to S71, where the display unit 30 notifies the user of the detection time (e.g., 30 seconds, etc.). A user who sees this will wait for 30 seconds at the current position for the detection to complete if they cooperate with the transmission of the sensor detection data, but will move without stopping if they are not willing to cooperate. After the notification in S71, in S72, the specified data is detected by the corresponding sensor 38, and it is determined in S73 whether the detection is complete. If it is not yet complete, it is determined in S74 whether the user is staying within the detection area, and if so, the loop that returns to S72 is repeatedly executed. If the user's mobile communication device 3 moves outside the detection area before the detection is complete, the control shifts to S84, but if the user's mobile communication device 3 stays within the detection area until the detection is complete, the control proceeds to S75. After transmitting the detection data to the IoT server 8, the control shifts to S69.

[0037] The IoT server 8 that receives it in S66 updates the IoT device DB15 by searching for it in S67 and updating it with the newly received detection data. Then, in S68, points are transmitted to the mobile communication device 3. The mobile communication device 3 that receives it in S69 adds and updates the points stored in the EEPROM 26 (S70). After that, the control shifts to S84.

[0038] In addition, in the control shown in FIGS. 5 to 7, a determination means (e.g., S26, S38, S49, etc.) for determining whether there is an IoT device that requires information exchange between the IoT device group 2 and the IoT server 8 is provided on the IoT server 8 side, but this determination means may be provided on the IoT device group 2 side. This can reduce the control burden on the IoT server 8 side.

[0039] Next, the processing for the wireless sensor network 4 will be described with reference to FIG. 8. The wireless sensor network 4 collects the detection data of each sensor by S80 and stores it in association with each sensor ID. Next, S81 determines whether there is any data that should be urgently transmitted among the respective detection data. When there is no data that should be urgently transmitted, the control proceeds to S82, and it is determined whether it is time for periodic transmission (for example, every 24 hours). If it is not time for periodic transmission, the control returns to S80, and the loop of S80→S81→S82 is repeatedly traversed.

[0040] During the traversal of this loop, if an obvious abnormality that can be determined on the side of the wireless sensor network 4 is found in each detection data and it is determined by S81 that there is data that should be urgently transmitted, the control proceeds to S83. Also, when it is determined by S82 that it is time for periodic transmission, the control also proceeds to S83.

[0041] In S83, a process of transmitting a transmission signal is performed. Next, S86 determines whether a response signal from the mobile communication device 3 has been received. If not, the control returns to S80, and the loop of S80~S83→S86→S80 is repeatedly traversed.

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

[0043] Then, the IoT server 8 determines YES in S61, and the control proceeds to S91. If there is emergency detection data in the received detection data, it will notify the abnormality along with the sensor ID. Then, in S92, it searches the wireless sensor DB 16 based on the sensor ID of the detection data and updates it to the newly received detection data. Next, in S93, it transmits the points to the mobile communication device 3, and then the control returns to S24. The mobile communication device 3 that receives this in S94 adds and updates the points stored in the EEPROM 26 (S95), and then returns and the control shifts to S2.

[0044] Note that in the control shown in FIG. 8, a determination means (for example, S82) for determining whether there is an IoT device that requires information exchange between the wireless sensor network 4 and the IoT server 8 is provided on the wireless sensor network 4 side. However, this determination means may be provided on the IoT server 8 side.

[0045] In the IoT device mediation system described above, since a user's mobile communication device (for example, a wearable computer such as smart glasses, a smartphone, a vehicle such as an automobile equipped with a communication function, etc.) mediates and connects the IoT device group 2 and the Internet 1, an Internet connection function is not necessarily required on the IoT device group 2 side, and thus the cost can be suppressed at a low level. Moreover, since the user is given a privilege (for example, points, etc.) by performing the above mediation with the mobile communication device, an incentive for the user's above mediation can be provided, and it becomes possible to promote the user's mediation behavior.

[0046] Furthermore, in order to provide determination means (e.g., S26, S38, S49, etc.) for determining whether there are IoT devices (including various sensors 38 of the mobile communication device 3) that require information exchange on the side of the mediation destination computer (e.g., IoT server 8, network cloud, etc.) mediated by the wearable computer 3 for the IoT device group 2, it is not necessary to perform the above determination on the side of the IoT device group 2. This can reduce the processing load on the side of the IoT device group 2 and further suppress costs. Moreover, the IoT server 8 having the determination means identifies the IoT device group 2 that requires information exchange (e.g., S26, S38, S49, etc.) and notifies the wearable computer 3 of the location information of the identified IoT devices 2 to inform the user, which has the advantage of prompting the user to perform the above mediation action.

[0047] Also, when collecting detection data by various sensors 38 at a desired position using the various sensors 38 provided in the wearable computer 3 on the Internet, if it takes a certain amount of time to detect, the wearable computer 3 notifies the user of the detection time required (e.g., S71). Therefore, it is possible to prompt the user to cooperate in collecting detection data that requires time.

[0048] [Service Provision System Utilizing Machine Learning] Next, a service provision system utilizing machine learning will be described with reference to FIGS. 9 to 23. A major current challenge in putting machine learning into practical use is how to collect a large amount of learning data at low cost. This system constructs a mechanism in which a large number of users proactively provide learning data to the system side by providing users with services that utilize the results of machine learning. In particular, as shown in Fig. 9, the learning data (data around oneself, etc.) provided by each user to the system side is also used for personalization, and after optimizing (personalizing) the general model obtained as a result of machine learning for each user, a personalized service is provided to each user. This prevents the inconvenience that each user leaves the provision of learning data to others and only enjoys the service.

[0049] First, the outline of the system will be described with reference to Fig. 9. Each user 70 transmits around data (video the user 70 is viewing, position of the line of sight, voice, position data such as GPS, etc.) collected by a mobile communication device 3 such as a wearable computer to machine learning 72 and uses it as learning data for machine learning 72, and those data are also used for personalization 74.

[0050] For example, the state in which each user 70 is learning at school, home, etc. is transmitted to the system side, modeled by machine learning (reinforcement learning or regression) 72, and a general model is created in which the total time of review repeatedly performed by humans is minimized. At that time, elements with differences for each user (such as memorization ability) are used as constants to generate a general model. In addition, the data transmitted from each user 70 to the system side is also used for personalization 74, elements with differences for each user (such as memorization ability) are calculated for each user, and by substituting them into the constant part of the general model, a personalized model is generated for each user. Using the personalized model, a personalized service is provided to each user, for example, the next review time is presented and guidance is provided according to a review plan in which the total review time is minimized.

[0051] Note that the learning data is provided not only by users but also by various professionals 71. For example, a sports gym professional 71 sends the system the aggregated results of the muscle training (hereinafter referred to as "weight training") of a large number of members. Based on this aggregated result learning data, a general model that maximizes the effect of weight training performed repeatedly by humans is created through machine learning (reinforcement learning or regression) 72. At that time, elements with differences for each user (such as the initial load like a barbell and the load increase coefficient) are set as constants to generate a general model. Also, the aggregated results of weight training are sent from each user 70 to the system side and used for personalization 74. By calculating elements with differences for each user (such as the initial load and the load increase coefficient) for each user and substituting them into the constant part of the general model, a personalized model for each user is generated.

[0052] Using the personalized model, a service personalized for each user, for example, presenting the next weight training time and load, and providing guidance according to a weight training plan that maximizes the effect of weight training. At that time, an advertisement of the professional (such as a sports gym) 71 that provided the learning data consisting of the aggregated results of weight training is presented together with the weight training schedule. This can motivate various professionals to provide learning data.

[0053] Figure 10 shows the overall system of Figure 9 described above in more detail. Referring to FIG. 10, specialized data is provided from the PCs 10 of various professionals to the artificial intelligence server 9. Also, data collected by mobile communication devices 3 such as wearable computers of a large number of users (surrounding data such as images viewed by the user 70, positions of lines of sight, voices, position data such as GPS) is provided to the artificial intelligence server 9. In the artificial intelligence server 9, the provided data is converted into numerical-label sets 51 to generate learning data. The generation of this learning data uses deep learning (deep neural learning) in order to convert raw data into numerical-label sets. Also, in order to handle raw data in which a plurality of regularities are mixed, "heterogeneous mixed learning" is used. This is a method of generating a model suitable for each pattern when there are a plurality of regularities in the collected data.

[0054] For example, when collecting time-series data on the power consumption of a building and creating a model for predicting power consumption, the power consumption of the building changes in pattern for each day of the week and time zone. Conventionally, humans mobilize specialized knowledge to make case-by-case divisions of when the pattern switches and create models suitable for each case. In this "heterogeneous mixed learning", the switching of the pattern itself is found by machine learning, and a model is generated for each pattern. When new data occurs, its compatibility with the generated model is confirmed, and if the model does not fit well, it starts over from the determination of the pattern switch. That is, by repeating "pattern case-by-case division" and "model creation" by a machine, an optimal model for each pattern is created.

[0055] In the case of data for which data conversion into numerical-label sets cannot be performed even using deep learning, an artificial intelligence professional artificially performs data conversion into numerical-label sets.

[0056] This artificial intelligence server 9 uses a general Neumann-type computer, but a neural network processor (NNP) may also be used. A large number of "artificial neurons" modeled after real neurons are mounted on the chip of the NNP, and each neuron cooperates with each other in the network.

[0057] Alternatively, a quantum computer adopting the "quantum annealing method" may be used. Thereby, the time required for optimization calculation in machine learning can be significantly reduced. Note that "artificial intelligence" is a broad concept including software agents.

[0058] The learning data converted into numerical and label sets is sent to the learning algorithm 52 and modeled 53. As the learning algorithm 52, various algorithms are prepared, such as regression and discrimination as supervised learning, model estimation and data mining as unsupervised learning, and reinforcement learning and deep learning as intermediate methods. The modeling using the learning algorithm 52 generates a general model by setting elements with differences for each user (such as individual ability differences and preferences) as constants. The generated general model is stored in the artificial intelligence DB 17. In FIG. 10, as specific examples of the stored general models, for example, a building maintenance inspection model, a rote learning model, a muscle training model, a dialogue model, a task processing model, etc. are shown. The above-mentioned learning data is also stored in the learning DB 60.

[0059] Ti = T0·i as a rote learning model is a general model in which the total time of review repeatedly performed by a human is minimized. i represents the number of repetitions indicating which repetition it is, T0 represents the period from initial learning until just before forgetting (specifically, the time until the memory retention rate reaches 70% from initial learning), and Ti represents the time from the (i - 1)-th to the i-th repetition. T0 in this general model is an element with differences for each user (such as individual ability differences like memory ability) and is represented as a constant. The generation method of Ti = T0·i as this rote learning model will be described in detail later.

[0060] Fi = F0+(i - 1) / a as a muscle training model is a general model in which the effect of muscle training repeatedly performed by a human is maximized. i represents the number of repetitions indicating which repetition it is, F0 represents the load in the first muscle training (such as the weight of the barbell), and a is a load increase coefficient that increases with the repetition of muscle training.

[0061] The dialogue model is a general model used when a user interacts with artificial intelligence, and consists of initial target dialogue templates such as male intellectual type, female moe type, and famous talents, as well as change functions of emotions such as joy, anger, sorrow, and happiness. This will be described in detail based on FIG. 19.

[0062] The task processing model is a general model used when a user processes tasks such as work, and consists of knowledge utilization functions for initial targets such as lawyers, attorneys, and famous scientists. This will be described in detail based on FIG. 21.

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

[0064] In the user DB 12, personalized data and around data are stored in association with each user ID. The personalized data for memorization learning is the period from initial learning until just before forgetting (specifically, the time until the memory retention rate reaches 70% from initial learning), TK. Human memory information decreases according to the forgetting curve of psychologist Hermann Ebbinghaus. The time until the memory retention rate reaches 70% according to this forgetting curve is defined as TK. This TK is the time until the memory retention rate of the actual user with user ID "1" reaches 70%, and is represented by an actual numerical value.

[0065] The personalized data for muscle training includes the initial load FK such as the weight of the barbell in the first muscle training, the load increase coefficient az that increases with the repetition of muscle training, the super-recovery time CT, etc. These are represented by actual numerical values for the actual user with user ID "1". Note that "super-recovery" refers to the phenomenon where the muscle strength level increases compared to before training by taking a rest for 48 to 72 hours after training. The time required for that super-recovery is the "super-recovery time".

[0066] As the personalized data for conversation, it consists of the initial target selected by the user from among the initial targets such as male intellectual types, female cute types, and famous talents, and the personal weights of emotions such as joy, anger, sorrow, and happiness. In a general model, the initial weights of emotions for each initial target are determined. During the conversation with the initial target selected by the user, the initial weights of emotions are gradually modified according to requests from the user, etc., and finally, the optimal personal weights for the user are obtained to enable a conversation that represents emotions matching the user's preferences. This will be explained in detail later.

[0067] As the personalized data for task processing, it consists of the initial target selected by the user from among the initial targets such as lawyers, attorneys, and famous scientists, and the personalized weights of each piece of knowledge, etc. In a general model, the initial weights of knowledge for each initial target are determined. During the conversation about work with the initial target selected by the user, the initial weights of knowledge to be utilized are gradually modified according to questions and requests from the user, etc., and finally, the optimal personal weights for the user are obtained to enable answers and advice that match the knowledge required by the user in their work. This will be explained in detail later.

[0068] The artificial intelligence server 9 performs personalized guidance (services) for each user using the above - mentioned personalized data. Specifically, it conducts guidance according to a review plan (review schedule) that minimizes the total review time of the user using the personalized data for memorization learning, conducts guidance according to a muscle training schedule that maximizes the effect of repeated muscle training using the personalized data for muscle training, conducts a conversation that represents emotions matching the user's preferences using the personalized data for conversation, and provides answers and advice that match the knowledge required by the user in their work using the personalized data for task processing.

[0069] Next, referring to FIG. 11, a flowchart of a sub - routine program of the external display process (S2) in FIG. 4, the display process (S3), and the like - button process by the SNS server 11 will be explained.

[0070] In FIG. 11, the processing for a digital menu board as an example of the display 5 is shown. A digital menu board is a display for menu display installed in restaurants and the like. By digitally displaying the menu, the time and cost on the store side can be significantly reduced, such as immediate display of out-of-stock items and menu changes. In the present embodiment, as shown in FIG. 1, the digital menu board 5 is connected to the Internet 1 and configured to be able to communicate with the artificial intelligence server 9 and the SNS server 11. In the flowchart of FIG. 11, images and videos taken by the user with the wearable computer 3 or the like and uploaded to the SNSDB 13, the user DB 12, etc. are displayed on the digital menu board 5 so that friends who come to the store together can view them and enjoy eating and drinking.

[0071] First, at S100, it is determined whether there is a menu update. If the store side performs an operation to update the menu for today, the control proceeds to S101 and the menu update process is performed. Next, the control proceeds to S102 and the menu is displayed.

[0072] Next, when the user wants to display images and videos taken with the wearable computer 3 or the like and uploaded to the SNSDB 13, the user DB 12, etc. on the digital menu board 5, first an operation to access the artificial intelligence server 9 is performed. Then, at S104, a determination of YES is made and the control proceeds to S105, where the user ID, access request, etc. are transmitted to the artificial intelligence server 9 and an access process is performed. The artificial intelligence server 9 that has received the access request makes a determination of YES at S106. In the mobile communication device 3, at S107 and S108, communication is performed with the artificial intelligence server 9 to select and specify the images and videos that the user wants to display on the digital menu board 5. The artificial intelligence server 9 searches the user DB 12 at S108 to identify the selected and specified images and videos.

[0073] When the user (customer) can identify the image to be displayed, they tap the touch screen of the digital menu board 5 to perform a communication operation. Then, a YES determination is made in S103, and communication start processing with the wearable computer 3 is performed in S109. Also, the user (customer) performs a communication operation on their own wearable computer 3. Then, a YES determination is made in S108 and the 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 the digital menu board 5 to the wearable computer 3 (S111). The wearable computer 3 that receives it in S112 transmits the display ID to the artificial intelligence server 9 (S113). The artificial intelligence server 9 that receives it in S114 transmits the data identified in the aforementioned S108 to the display 5 of the received ID (S105). The digital menu board 5 that receives it makes a YES determination in S116 and displays the received data (S117). As a result, the image or video specified by the user (customer) is displayed on the digital menu board 5. As a result, images and videos can be viewed on the relatively large display screen of the digital menu board 5.

[0075] The point to note in the above-described control is that when the user selects and specifies the data being uploaded onto the network, the user directly accesses and specifies the artificial intelligence server 9 using their own mobile communication terminal 3 without going through the display 5. If the display 5 itself is connected to the network and the data to be displayed on the display 5 is to be specified, it is quicker to access the data on the network via the display 5 at the display destination and display the data on the display 5. However, the data that the user uploads onto the network is personal information related to the user's privacy. If the user ID, etc. is transmitted via the display 5 in order to access the personal information, there is a risk of leakage of the user ID and personal information. Therefore, in the present embodiment, when the user selects and specifies the data being uploaded onto the network, the user directly accesses and specifies the artificial intelligence server 9 using their own mobile communication terminal 3 without going through the display 5.

[0076] Next, in the digital menu board 5, it is determined at S118 whether the data display has ended. If it has ended, the control proceeds to S119, and a process is performed to allow the user (customer in the store) to access the store's homepage. Specifically, the URL of the store's homepage is transmitted to the mobile communication terminal 3. The received mobile communication terminal 3 transmits a request to access the page of the URL to the SNS server 11 at S120. The SNS server 11 that has received it at S121 transmits the store's homepage to the mobile communication device 3 at S122.

[0077] Upon receiving it, the mobile communication device 3 displays the store's homepage according to S123. On the other hand, on the digital menu board 5, a message "Please tap the like button for the store's homepage" is displayed according to S124. If a user (a customer who visits the store) who sees this taps the like button, a YES determination is made at S125 and the control proceeds to S126, and the like signal is transmitted to the SNS server 11. The SNS server 11 that receives it at S127 performs the like registration process at S128. As a result, it becomes possible to provide information from the store to the user (customer who visits the store) through the SNS. For example, it becomes possible to introduce season-limited new menus, distribute coupons, notify of event held, etc., and it can lead to repeat visits.

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

[0079] Next, referring to FIG. 12(a), the flowchart of the subroutine program of the expert data process shown in S6 of FIG. 4 will be described. Maintenance data processing is performed by S135, muscle training data processing is performed by S136, other processing is performed by S137, and then the process returns and proceeds to S8. The maintenance data processing is to receive and process the data of the building maintenance inspection results sent from a maintenance specialist, which is an example of various specialists. The muscle training data processing is to receive and process the aggregated data of the muscle training results of a large number of members sent from a sports gym, which is an example of various specialists.

[0080] Next, referring to FIG. 12(b), the flowchart of the subroutine program of the above-described maintenance data processing will be described. The maintenance specialist performs maintenance points of the building based on the detection data transmitted from the wireless sensor network 4 shown in FIG. 8 and stored in the wireless sensor DB 16, and inputs the results (KO, NG, etc.) to the PC 10 for each sensor ID (S140).

[0081] Then, the maintenance specialist's PC 10 transmits the input inspection results to the artificial intelligence server 9 for each sensor ID according to S141. The artificial intelligence server 9 determines whether it has received the inspection results according to S145. If not, the control proceeds to S151. On the other hand, if the inspection results are received, the control proceeds to S146, and the inspection result data is additionally stored in the learning DB60 for each sensor ID. Next, according to S147, wireless sensor data is read from the wireless sensor DB16, inspection result data is attached, and data mining is performed to find the NG pattern NGP1 at the time of NG. For example, a pattern NGP1 such as a high probability of an abnormal (NG) inspection result if the strain and vibration period are in a predetermined relationship is found.

[0082] Next, according to S148, it is determined whether NGP1 is a new one different from those found so far. All the NG patterns found so far are stored in the storage area of the "building maintenance inspection model" in the artificial intelligence DB17, and it is determined whether the newly found NGP1 this time is already stored in the artificial intelligence DB17. If it is determined that the same one is already stored, the control proceeds to S151. However, if it is determined that it has not been stored yet, the control proceeds to S149, and NGP1 is additionally stored in the storage area of the "building maintenance inspection model" in the artificial intelligence DB17 as the NG pattern.

[0083] Next, according to S150, the wireless sensor DB16 is searched, and the sensor ID having the NGP1 pattern is extracted from the sensors that have been regularly checked by the wireless sensor and notified. The wireless sensor is regularly checked at regular intervals (for example, every 24 hours) (S151, S152), and the newly found NGP1 is applied to the past sensor data that has already been regularly checked for re-checking.

[0084] Next, it is determined by S151 whether it is time for a periodic check. If it is time for a periodic check, then by S152, by searching the wireless sensor DB16, a process is performed to extract sensor IDs having an NG pattern from the wireless sensor data additionally stored after the previous periodic check and issue an NG notification. As described above, new sensor data is stored in the wireless sensor DB16 at any time, and for new sensor data that has not yet been periodically checked, all NG patterns stored in the artificial intelligence DB are applied for checking. After the process of S152, the program returns and the control shifts to S136.

[0085] Next, referring to Fig. 13(a), a flowchart of a subroutine program for the around data terminal process shown in S4 of Fig. 4 and the around data server process shown in S5 will be described.

[0086] The user learning terminal process is executed by S160, the user learning server process is executed by S161, the muscle training terminal process is executed by S162, the muscle training server process is executed by S163, and other processes are executed by S164 and S165.

[0087] The user learning terminal process and the user learning server process are processes for collecting around data from each user during learning to create a general memorization learning model for the user and generating personalized data for learning for each user. The muscle training terminal process and the muscle training server process are processes for collecting around data from each user during muscle training and generating personalized data for muscle training for each user.

[0088] Next, referring to Fig. 13(b), a flowchart of a subroutine program for the personal service terminal process shown in S7 of Fig. 4 and the personal service server process shown in S8 will be described.

[0089] The learning service terminal process is executed by S170, the learning service server process is executed by S171, the muscle training service terminal process is executed by S172, and the muscle training service server process is executed by S173.

[0090] The learning service terminal process provides guidance to the user according to the most efficient review plan based on a general memorization learning model for the user and personalized data for learning. The muscle training service terminal process and the muscle training service server process provide guidance to the user according to the most efficient muscle training plan based on a general muscle training model for the user and personalized data for muscle training.

[0091] Next, referring to FIG. 14, a flowchart of a subroutine program of the user learning terminal process and the user learning server process will be described.

[0092] It is determined by S180 whether the user is learning. This determination may be made by receiving an explicit indication of the user's intention, or the artificial intelligence server 9 may autonomously determine by processing the user's voice through natural language processing or analyzing the video of the user's line of sight direction. If it is determined that the user is not learning, the process returns and the control transfers to S162. If it is determined that the user is learning, the control proceeds to S181, and the surrounding data such as the video of the user's line of sight direction, the line of sight position, and the voice is transmitted to the artificial intelligence server 9. The artificial intelligence server 9 that receives it by S182 searches a predetermined DB for useful information such as the past entrance examination question frequencies and the schools that set the questions related to the matter currently being learned from the received surrounding data, and transmits it to the user (the mobile communication device 3 such as a wearable computer).

[0093] The mobile communication device 3 of the user that received it through S184, through S185, overlays and displays the received useful data using AR (Augmented Reality). Since the user can receive guidance according to the most efficient review plan by transmitting the surrounding data during learning to the artificial intelligence server 9, the user will take the initiative to transmit the surrounding data to the artificial intelligence server 9. However, the incentive will be further improved by the display of useful information such as the entrance examination question frequency, and the user will transmit the surrounding data to the artificial intelligence server 9.

[0094] Next, the artificial intelligence server 9, through S186, performs a process of classifying the received data for various models and storing it in the user DB12 and the learning DB60. Next, through S187, it is determined whether the rote learning model is completed. If it is not yet completed, the control proceeds to S188, where the received data is converted into numerical values or labels to create data for machine learning, and through S189, it is modeled using a learning algorithm (such as reinforcement learning or regression). An example of this modeling will be described.

[0095] Using the vast amount of past surrounding data during learning collected for each user as learning data, under the condition that the memory retention rate at a predetermined end point such as the entrance examination date can be maintained at a certain value (for example, 90%), find a review plan with the least total review time from the initial learning to the above end point using reinforcement learning. For example, when the period from the i-th review to the next review is Ti and the time required for the i-th review is THi, the total review time T = ΣTHi. Reinforce Ti when T is minimized under the condition that the end point memory retention rate ≥ 90%.

[0096] In reinforcement learning, as a method for estimating the Q-value when the value of performing action \(a_t\) in state \(s_t\) is \(Q(s_t, a_t)\), if knowledge for modeling the environment, that is, the state transition probability and the probability distribution of rewards are given, a model-based approach can be used. However, when the environment model is unknown, TD (Temporal Difference) learning is used. First, since exploration of the environment is necessary, the ε-greedy method is used. At the initial stage of exploration, various actions are tried, and as it settles down, the concept of temperature is introduced so that the optimal action is selected more often. Let the temperature be \(T\), and select an action according to the probability represented by the following formula.

[0097] \(P(a|s)=\frac{\exp(Q(s,a) / T)}{\sum_{b\in A}\exp(Q(s,b) / T)}\) (Note that \(b\in A\) is written under \(\sum\), and this notation is omitted in the above formula) Here, \(a\) is an action, \(Q(s,a)\) is the value when performing action \(a\) in state \(s\),

[0098] Let \(T\) be the temperature in annealing. If it is high, select an action with a probability close to equal probability. If it is low, bias it towards the optimal one. As learning progresses, by reducing the value of \(T\), the learning result becomes stable. The action selected in this way may be instructed to the user to actually execute, and learning data may be actively collected.

[0099] An example of the result of that reinforcement learning is shown in FIG. 15. Referring to FIG. 15, the vertical axis of the graph is \(T_i\) (review interval), and the horizontal axis is \(i\) (the number of reviews indicating which review it is). What is indicated by ○ is the result of plotting the reinforcement learning of the user with user ID: 1. What is indicated by × is the result of plotting the reinforcement learning of the user with user ID: 2. What is indicated by △ is the result of plotting the reinforcement learning of the user with user ID: 3.

[0100] Next, using the "regression" algorithm, taking the data of the machine learning results of these multiple users as input data, assuming that these input data output a target based on a certain function, and finding that function. The obtained function is Ti = T0·i is a linear function of a straight line. The coefficient T0 is an element that varies from user to user (such as memorization ability). Specifically, it is the period from initial learning to just before forgetting for each user (specifically, the time until the memory retention rate reaches 70% from initial learning). In this way, what represents a general model with an element (such as memorization ability) that varies from user to user as the constant T0 is Ti = T0·i is.

[0101] Returning to FIG. 14, by S190, the memorization learning model (Ti = T0·i) is memorized in the artificial intelligence DB17. If an old and incomplete memorization learning model is already memorized, it is updated to a newer and complete memorization learning model (Ti = T0·i).

[0102] On the other hand, when it is determined by S187 that the memorization learning model is completed, the control proceeds to S191, and processing is performed to create personalized data based on the classified received data and memorize it in the user DB12. The "personalized data" in this case is the element (such as memorization ability) T0 that varies from user to user, that is, the time until the memory retention rate reaches 70% from initial learning. This can be easily derived from the surrounding data during the learning of the user memorized in the user DB12.

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

[0104] The mobile communication device 3 such as a wearable computer determines, by S200, whether there is a review request from the user. If there is no review request, it returns and the control shifts to S172. If there is a review request, it transmits the review request to the artificial intelligence server 9 by S201. The artificial intelligence server 9 that receives it at S202 determines, by S203, whether the memorization learning model is completed. If it is not completed, the control shifts to S205. If it is already completed, the control proceeds to S204, substitutes T0 = TK into the general memorization learning model Ti = T0·i, and extracts learning items for which the time for repeated review is approaching from the surrounding data in the user DB12. Note that the above TK refers to the period from the user's initial learning until just before forgetting (specifically, the time until the memory retention rate reaches 70% from the initial learning).

[0105] Next, by S205, the mobile communication device extracts learning items that the user is looking at during class or the like from the surrounding data in the user DB12. Since the line-of-sight position of the user during learning is transmitted as surrounding data to the artificial intelligence server 9 (S181) and stored in the user DB12, it becomes possible to determine whether the user is looking around. Next, by S206, the learning items extracted in S204 and S205 are transmitted to the mobile communication device 3. The mobile communication device 3 that receives it at S207 performs a process of having the user review the received learning items by S208.

[0106] Alternatively, instead of or in addition to being in a side view, control may be performed to cause the user to review the learning items while the user is chatting idly during class or the like. That is, a collection means for collecting information capable of identifying the attitude of the user during learning, a recording means for recording the learning content received by the user, a discrimination means for discriminating an inappropriate attitude part of the user from the information collected by the collection means, an extraction means for extracting the learning content corresponding to the part discriminated by the discrimination means from the recording means, and a transmission means for transmitting the learning content extracted by the extraction means to the user terminal. Note that the above S205 does not necessarily have to be performed. After the process of S206, the control returns and shifts to S173.

[0107] In the above control, the personalization data of the user is substituted into a general model to create a user-specific personalization model, and the personalization service is provided to the user using the personalization model. However, a handmade personalization model for the user may be generated and the personalization service may be provided to the user using the model. For example, when a YES determination is made in S187 of FIG. 14, the machine learning results of each of a plurality of users (subjects) who contributed to the creation of the general learning model are available (see FIG. 15). Therefore, for each of these subject users, the machine learning result performed on their own learning can become a handmade personalization model, and control may be performed to provide the personalization service to the user using the personalization model. Further, even for users other than the above subjects, if they wish to generate a handmade personalization model for themselves, control may be performed to provide a large amount of their own learning data (review data) to the artificial intelligence server 9 to create a handmade personalization model.

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

[0109] A sports gym PC10, an example of a professional, aggregates the muscle training data of a large number of members (subjects) by S210 and transmits it to the artificial intelligence server 9. The muscle training data is, for example, the initial load of the subject, the muscle training time and number of times, the muscle training interval from the previous muscle training to the current muscle training, the load increase coefficient, the measured value data of the muscle thickness, and the like. The artificial intelligence server 9 receives the muscle training data at S211 and classifies the received data for various models and stores it in the learning DB60 by S212. Next, at S213, the received data is made into data of numerical values and label sets to create learning data.

[0110] Next, at S214, based on the learning data, it is modeled using learning algorithms such as reinforcement learning and regression. This modeling method is the same as the one described based on FIG. 15. A general model of muscle training is Fi = F0 + (i - 1) / a Here, i is the number of repetitions indicating which repetition of muscle training, Fi is the load (such as the weight of the barbell) at the i-th muscle training, F0 is the load (such as the weight of the barbell) at the first muscle training, and a is the load increase coefficient that increases with the repetition of muscle training. These F0 and a are elements with differences for each user (such as individual differences in muscle strength).

[0111] Next, at S215, the above general model (Fi = F0 + (i - 1) / a) is stored in the storage area of the muscle training learning model in the artificial intelligence DB17. If an old and incomplete muscle training learning model is already stored, it is updated to a newer and complete memorization learning model (Fi = F0 + (i - 1) / a).

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

[0113] The user's mobile communication device 3 determines whether the user is exercising muscle strength at S220. If not during muscle training, it returns and the control shifts to S164. If during muscle training, the control proceeds to S221, aggregates the user's muscle training data, and transmits it to the artificial intelligence server 9. The artificial intelligence server 9 that receives it at S222 stores the received data in the learning DB60 at S113, and calculates the user's initial load FK, load increase coefficient az, and super recovery time CT, and stores them in the user DB12. After S113, it returns and the control shifts to S165.

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

[0115] The user's mobile communication device 3 transmits a muscle training menu request signal to the artificial intelligence server 9 in response to a request for the user's muscle training menu at S226. The artificial intelligence server 9 that receives it at S227 reads out the user's initial load FK, load increase coefficient az, and super recovery time CT from the user DB12 at S228, substitutes them into the general muscle training model (Fi = F0+(i - 1) / a), and creates the current muscle training menu. Then, at S229, it transmits the muscle training menu and the advertisement of the muscle training specialist who provided the muscle training data (see S210) to the user's mobile communication device 3. The mobile communication device 3 that receives it at S230 notifies the user of the received muscle training menu and the advertisement of the muscle training specialist at S231. After S231, it returns and the control shifts to S9, and after S229, it returns and the control shifts to S10.

[0116] FIG. 19(a) shows the details (specific examples) of the data stored in the memory area of the dialogue model in the artificial intelligence DB 17 of FIG. 10. In the memory area of the dialogue model, data for each of the initial target, dialogue template, initial weights of emotions (joy, anger, sorrow, and pleasure), and emotion change functions are stored. In the initial target, various selectable targets such as male intellectual type, female intellectual type, male wild type, female moe type, talent A, talent B, and talent C are stored. The data for the dialogue template, initial weights of emotions, and emotion change functions differ for each type of the initial target, and each data is stored in association with each type of the initial target. The dialogue template is a large number of dialogue templates collected to find a response to the user's dialogue by template matching. For example, in the case of the male intellectual type, the ratio of intellectual dialogue templates is high.

[0117] The initial weights of emotions and the emotion change functions are for controlling the change of emotions of joy, anger, sorrow, and pleasure in response to the dialogue. When the user praises, it is controlled to change to the emotion of "joy", when the user criticizes, it is controlled to change to the emotion of "anger", when the user saddens, it is controlled to change to the emotion of "sorrow", and when the user enjoys, it is controlled to change to the emotion of "pleasure".

[0118] However, since the weights of emotions (joy, anger, sorrow, and pleasure) and the modes of change vary depending on the type of initial target, different initial emotion weights and emotion change functions are set for each type of initial target. The initial emotion weights and emotion change functions are found through machine learning using, as learning data, a large number of conversation contents collected from interactions with a large number of users. A large number of users are classified into male intellectual type, female intellectual type, male wild type, and female moe type, and the average for each type is calculated to create the initial emotion weights and emotion change functions corresponding to each initial target. For talents such as Talent A, Talent B, and Talent C, the initial emotion weights and emotion change functions are found through machine learning using, as learning data, a large number of conversation data collected from the interactions of a specific talent (for example, Talent A). Note that the emotion change function may be a unified function common to all humans, and the differences (variances) among individuals may be controlled by adjusting with the initial emotion weights.

[0119] As an example of the emotion change function F(x), for instance, the following can be considered. Represent joy, anger, sorrow, and pleasure as K, D, I, and R respectively, and represent the emotionless state as M. Let the number of times the user praises be x1, the number of times the user scolds be x2, the number of times the user feels sad be x3, and the number of times the user has fun be x4, and let the initial emotion weights be K, D, I, and R respectively. Also, if the sigmoid function 1 / (1 + e -x ) is expressed as S(x) for convenience, F(x) = KS(x1 - K) + DS(x2 - D) + IS(x3 - I) + RS(x4 - R) + M(1 - S(x1 - K) - S(2x - D) - S(x3 - I) - S(x4 - R)) That is, if any of x1, x2, x3, or x4 exceeds the respective thresholds (initial weights) K, D, I, or R, it changes to the corresponding emotion (either joy K, anger D, sorrow I, or pleasure R), but if none of them exceeds the above thresholds, it becomes the emotionless state M.

[0120] Figure 19(b) shows the details of the data stored in the storage area for interactive personalization data in the user DB12 of Figure 10. Figure 19(b) shows the case where the user with user ID: 1 selects and designates "Talent A" as the initial target of the dialogue model. And "0.7AK, 1.5AD, 1.3AI, 0.5AR" is stored as the personal weights of joy, anger, sorrow, and joy.

[0121] The initial weights of joy, anger, sorrow, and joy of "Talent A" are "AK, AD, AI, AR" as described in Figure 19(a). However, while the user is having a dialogue based on this initial weight of joy, anger, sorrow, and joy, the initial weight of joy, anger, sorrow, and joy "AK, AD, AI, AR" is gradually modified according to the user's request, and at the current time, it has become "0.7AK, 1.5AD, 1.3AI, 0.5AR". As a result, the function of the change in joy, anger, sorrow, and joy with the personal weight of joy, anger, sorrow, and joy of this user "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 with the actual Talent A, joy K and joy R are more likely to appear, while anger D and sorrow I are less likely to appear.

[0123] In this way, it is possible to select and designate an initial target that matches the user's preferences, and furthermore, it can be modified to a personal weight of joy, anger, sorrow, and joy that more reflects the user's preferences as the dialogue continues. That is, in the case of a human partner, it is almost impossible to change the partner into one's ideal human, but in the case of artificial intelligence, it is possible to create (nurture) the partner (artificial intelligence) into one's ideal partner. Until now, it has been an era of searching for and meeting ideal lovers, ideal friends, and ideal partners, but in the future world, it will be an era of creating ideal lovers, ideal friends, and ideal partners on the network (in the cloud).

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

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

[0126] On the other hand, in the mobile communication device 3, it is determined by S237 whether there is a selection specification of the initial target. If there is no selection specification, the control shifts to S239, but if there is a selection specification, the specified initial target is transmitted to the artificial intelligence server 9 by S238. The received artificial intelligence server 9 makes a YES determination by S242 and performs a process of updating the dialogue initial target corresponding to the user ID in the user DB12 to the specified one by S243.

[0127] Next, the mobile communication device 3 and the artificial intelligence server 9 perform a dialogue with each other by S239 and S244, respectively. At this time, in the artificial intelligence server 9, the dialogue is performed according to the dialogue initial target corresponding to the user ID in the user DB12, the personal weight of emotions, and the corresponding emotion change function in the artificial intelligence DB17. The artificial intelligence server 9 is equipped with a natural language processing engine, and in addition to the processes at the time of text mining such as "morphological analysis" and "syntactic analysis (a process of determining the dependency relationship between clauses)", two technologies, "context analysis" necessary for grasping the structure of the sentence and "semantic analysis" necessary for understanding the intention of the sentence, are incorporated. As a result, it becomes possible to analyze the user's voice "faster" and "more precisely".

[0128] In the artificial intelligence server 9, during the above conversation, it is determined whether it is necessary to change the personal weights of joy, anger, sorrow, and pleasure (S245). If necessary, the personal weights of joy, anger, sorrow, and pleasure are updated according to S246. For example, when 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", control is performed to update the personal weights of joy, anger, sorrow, and pleasure. In addition to or instead of the above set phrase, control may be performed to update the personal weights of anger, joy, and sorrow according to the user's utterance such as "You are overreacting".

[0129] Next, in the artificial intelligence server 9, it is determined according to S247 whether the conversation content is related to work. If it is not related to work, the process proceeds to S249 to determine whether the conversation has ended. If it is related to work, the process proceeds to S248 and task server processing is performed.

[0130] In the artificial intelligence server 9, if it is determined according to S249 that the conversation has not ended, the control shifts to S244, and the steps of S244 to S249 are repeatedly executed. When the conversation ends, the process returns and the control shifts to S12. On the other hand, in the mobile communication device 3, if it is determined according to S240 that the conversation has not ended, the control shifts to S237, and the steps of S237 to S240 are repeatedly executed. When the conversation ends, the process returns and the control shifts to S11.

[0131] FIG. 21(a) shows the details (specific examples) of the data stored in the storage area of the task processing model in the artificial intelligence DB17 of FIG. 10. In the storage area of the task processing model, the initial target, the initial weights of each knowledge, and each data of the knowledge utilization function are stored. In the initial target, various selection targets such as attorney, lawyer, physicist, chemist, scientist A, scientist B, and scientist C are stored. The initial weights of each knowledge and each data of the knowledge utilization function are different for each type of the initial target, and each data is stored in association with each type of the initial target.

[0132] The initial weights of each piece of knowledge and the knowledge utilization function are for facilitating the use of knowledge whose initial target matches its respective specialized field when performing task processing such as work. The initial weights of each piece of knowledge and the knowledge utilization function are found through machine learning using, as learning data, a large number of conversation contents collected from work-related conversations with a large number of users. A large number of users are classified by professional type such as lawyers, attorneys, physicists, chemists, etc., and the average of each professional type is calculated to create the initial weights of each piece of knowledge and the knowledge utilization function corresponding to each initial target.

[0133] For example, a lawyer user makes a large number of work-related questions and consultations to the artificial intelligence server 9, and uses, as learning data, a large amount of data consisting of reactions such as "satisfied" or "not satisfied" to the responses from the artificial intelligence server 9, and models it using algorithms such as reinforcement learning and regression. For scientists A, B, C, etc., the initial weights of knowledge and the knowledge utilization function are found through machine learning using, as learning data, a large number of conversation data collected from the conversations of an actual scientist. Note that the knowledge utilization function may be a unified function common to all humans, and the differences (variances) between individuals may be controlled to be adjusted by the initial weights of knowledge.

[0134] As an example of the knowledge utilization function G(x), for example, the following can be considered. Each piece of knowledge is represented as n1, n2, n3, ··· respectively, the number of positive reactions such as "satisfied" by the user for each piece of knowledge is x1k, x2k, x3k, ···, and the number of negative reactions such as "not satisfied" is x1h, x2h, x3h, ···, and the initial weights of each piece of knowledge are w1, w2, w3, ··· respectively. Also, if the sigmoid function 1 / (1 + e -x ) is expressed as S(x) for convenience, G(x) = w1n1S(x1k - x1h) + w2n2S(x2k - x2h) + w3n3S(x3k - x3h) + ··· That is, the knowledge for which the number of positive reactions exceeds the number of negative reactions is regarded as effective knowledge for utilization, and it can be controlled so that the knowledge with large coefficient (weights w1, w2, w3, ···) of each piece of knowledge n1, n2, n3, ··· is preferentially utilized.

[0135] Figure 21(b) shows the details of the data stored in the storage area for personalized data for task processing in the user DB12 of FIG. 10. FIG. 21(b) shows the case where the user with user ID: 1 selects and designates "attorney" as the initial target of the dialogue model. And as the personal weight w of each knowledge n, "1.2·1w11n1, 0.8·1w21n2, 0.7·1w31n3, 1.3·1w41n4" is stored.

[0136] The initial weights of each knowledge of "attorney" are "1w11n1, 1w21n2, 1w31n3, 1w41n4" as described in FIG. 21(a). However, while the user is having a work-related dialogue based on the initial weights of each knowledge, the initial weights of each knowledge "1w11n1, 1w21n2, 1w31n3, 1w41n4" are gradually modified according to the requests of the user, and at the current time, they have become "1.2·1w11n1, 0.8·1w21n2, 0.7·1w31n3, 1.3·1w41n4". In this way, it is possible to select and designate an initial target suitable for the user's profession, and further modify the personal weights of each knowledge to more reflect the questions and consultations of the user as the work-related dialogue progresses. That is, in the future world, it will be an era of creating one's ideal work partner on the network (in the cloud).

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

[0138] The user's mobile communication device 3 determines, by S255, whether there is a request from the user to register the user's (user's) personal weight as the initial target in the artificial intelligence DB17. If not, the control transfers to S259, and it is determined whether there is a request to update the initial target. If not, the control transfers 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 makes a request to update the initial target for the mobile communication device 3 and designates a new initial target (requested initial target), a YES determination is made in S259 and the control proceeds to S260, where the user ID and the requested initial target are transmitted to the artificial intelligence server 9. Upon receiving this, the artificial intelligence server 9 makes a YES determination in S261 and the control proceeds to S262, where it determines whether the received requested initial target has been previously used by the user. If it has not been used yet, in S264, the requested initial target is read from the artificial intelligence DB 17 and stored corresponding to the user ID in the user DB 12, and the control proceeds to S267.

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

[0144] Next, referring to FIG. 23, the artificial intelligence server 9 reads out the initial target corresponding to the user ID in the user DB 12 and the personal weight of each piece of knowledge in S267, and reads out the knowledge utilization function corresponding to the read initial target from the artificial intelligence DB 17 in S268. Then, if the user conducts a work-related conversation using the mobile communication device 3, a conversation is carried out between S265 on the mobile communication device 3 side and S269 on the artificial intelligence server 9 side. In S269, processing is performed to conduct a work-related conversation using the read knowledge utilization function and personal weight. On the mobile communication device 3 side, the work-related conversation control by S265 continues until it is determined in S266 that the work-related conversation has ended. On the artificial intelligence server 9 side, the control from S267 to S274 is repeatedly performed until it is determined in S274 that the work-related conversation has ended, and the work-related conversation control by S269 continues.

[0145] When it is determined during the repeated execution of the control of S267 to S274 that there is a need to change the personal weight by S270, the control proceeds to S271, and the personal weight in the user DB 12 is updated. If that personal weight is also registered in the artificial intelligence DB 17 by the user (see S255 to S258), a YES determination is made by S272, and the weight in the artificial intelligence DB 17 is also updated by S273. This determination by S270 is made, for example, when the user says predetermined set phrases such as "satisfied" or "not satisfied" in response to the response from the artificial intelligence server 9, to determine the necessity of changing the weight of the knowledge used on the artificial intelligence server 9 side. Note that instead of or in addition to the above set phrases, it may be determined based on the user's normal conversation such as "Thank you for the wonderful answer".

[0146] And when it is determined by S266 on the mobile communication device 3 side that the work-related dialogue has ended, a work-related dialogue end instruction is transmitted to the artificial intelligence server 9 by S273, and the control returns and shifts to S1. The artificial intelligence server 9 that has received this makes a YES determination at S274, returns, and the control shifts to S3.

[0147] Modifications and features of the embodiments described above are described below. ((ID=10)) At least one or all of the IoT server 9 and the IoT device DB 15, the artificial intelligence server 9, the artificial intelligence DB 17, the user DB 12, and the learning DB 60, and the SNS server 11 and the SNSDB 13 may be configured by a network cloud. Also, in the embodiments described above, the mobile communication device 3 was mainly used as an example of the user-side terminal for explanation, but as the user-side terminal, a user PC 7 or a robot 6 may also be used. Further, as a specific example of the mobile communication device 3, a wearable computer represented by smart glasses or the like was shown, but it is not limited thereto, and for example, various things such as mobile phones, smartphones, and vehicles such as automobiles having a communication function can be considered.

[0148] (2) The IoT server 8 is provided with a detecting means (S24 to S26, S38, S40) for detecting IoT devices that require information exchange. Then, it may be controlled to transmit the geographical location of the detected IoT devices to the wearable computer to notify the user, and prompt the user to perform an intermediary act. At this time, it may also be controlled to notify the user that a high score (points, etc.) is given to IoT devices that particularly require intermediation.

[0149] (3) It may be controlled to give high-value privileges (points, etc.) only to a limited number of IoT devices, and utilize a psychology similar to that of a lottery to promote the intermediary act. Also, it may be controlled to perform the intermediary act of IoT devices laid at the geographical location using a location game.

[0150] (4) The following inventions are disclosed in the embodiments described above. This invention relates to a system for realizing IoT at low cost by suppressing the cost of IoT devices (sensors, actuators, etc.) that are laid all over the earth without limitation.

[0151] As background art related to a system for realizing IoT at low cost, for example, the following existed. A border router (gateway device) was introduced between an IoT sensor and a server on the Internet. The sensor was not set with the server address, and after its own address was determined, the sensor data information was transmitted to the border router. The border router relayed the sensor information to the sensor data management server to perform data collection. Thereby, the server side can detect the address of the sensor and also 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 art, although it is possible to perform pull-type data collection addressed to the sensor from the server side, it is necessary to provide each sensor with an Internet connection function such as a network I / F, or to prepare an adapter with a proprietary standard for connecting to the network, which has the drawback of increasing the cost of realizing the IoT.

[0153] On the other hand, among the IoT devices laid out without limit on the earth, there are some that do not necessarily need to perform pull-type data collection addressed to the sensor from the server side. For example, in the case of an illuminance sensor installed to aggregate the illuminance throughout the year, the periodic illuminance detection results are stored in the memory, and it is sufficient to send the stored illuminance detection results to the server by the time of aggregating the illuminance throughout the year, which is performed once a year on the server side. Also, in the case of an IoT device installed in a place where there is a lot of human traffic, a method of sending the data of the IoT device to the server by an artificial method using that human can also be considered.

[0154] This invention has been conceived in view of such circumstances, and its object is to provide a system capable of suppressing the cost of realizing the IoT.

[0155] This invention includes a wearable computer owned by a user, a server that communicates with the wearable computer via the Internet, and a group of IoT devices, wherein the group of IoT devices includes communication means for communicating with the wearable computer, the wearable computer includes pairwise device communication means for communicating with the group of IoT devices, pairwise server communication means for communicating with the server via the Internet, and mediation means for enabling information exchange between the IoT device that is the communication partner by the pairwise device communication means and the server that is the communication partner by the pairwise server communication means. An IoT system.

[0156] According to such a configuration, the IoT device group side does not necessarily require an Internet connection function, and thus the cost can be reduced accordingly, enabling the realization of IoT at low cost.

[0157] Preferably, the server further includes privilege granting means for granting a predetermined privilege to the wearable computer that has been mediated by the mediation means.

[0158] According to such a configuration, an incentive for mediation can be given to the user, and it becomes possible to promote the user's mediation behavior.

[0159] More preferably, the server further includes determination means for determining whether there is an IoT device that needs to exchange information.

[0160] According to such a configuration, it is not necessary for the IoT device group side to have a function of determining whether there is a need to exchange information, the processing burden on the IoT device group side can be reduced, and the cost can be further suppressed.

[0161] More preferably, the wearable computer further includes position information transmission means for transmitting information capable of specifying the current position to the server, The server, position determination means for determining whether the IoT device determined by the determination means exists near the wearable computer whose position is specified by the information transmitted by the position information transmission means; and mediation command transmission means for transmitting a command signal for performing mediation by the mediation means to the wearable computer when it is determined by the position determination means that the IoT device exists. The mediation means performs mediation according to the command signal from the mediation command transmission means.

[0162] According to such a configuration, since the wearable computer prompts mediation for IoT devices that require mediation, it is not necessary for the IoT device group side to transmit a mediation request signal.

[0163] More preferably, the wearable computer includes a sensor that takes a predetermined time to detect, and further includes notification means (for example, S71) for notifying the user of the detection required time by the sensor when performing detection using the sensor at a desired geographical location, and when the sensor can perform the detection required time detection within the detection area of the desired geographical location, the detection data is mediated by the mediation means and transmitted to the server.

[0164] According to such a configuration, since the detection required time is notified to the user by the wearable computer when it takes a certain amount of time to detect, it is possible to prompt the user to cooperate in collecting detection data that takes time.

[0165] (5) In the embodiments described above, as specific examples of machine learning, the creation of a building maintenance inspection model, a rote learning model, a muscle training model, a dialogue model, and a task processing model was shown, but it is not limited thereto. For example, various models such as a learning model for infants, a model for an artificial secretary, and a model for an artificial tutor are conceivable. Also, the personalized data is not limited to that for rote learning, muscle training, dialogue, and task processing. For example, various data such as for infant learning, for artificial secretary, and for artificial tutor are conceivable. These general models and personalized data are controlled to be switched and properly used according to the user's situation. As a result, a person will grow while receiving a personalized service based on artificial intelligence using personalized data dedicated to oneself soon after birth, and will spend a lifetime with artificial intelligence. That is, in the future world, it will be a world where a person and artificial intelligence are paired. It is a world where a person and artificial intelligence are paired to form one personality. Also, when differentiating between the general model and personalized data, in this embodiment, for the personalized data for conversation, the previous data was used unless the user indicated a switching intention. However, when switching between the general model and personalized data, the personalized data for conversation may also be automatically controlled to switch to other appropriate data. For example, during general daily conversations, Talent A is used as the initial target for conversation, and during work-related conversations, it is controlled to automatically switch to a female intellectual type (see Fig. 19).

[0166] (6) Recommendation control may be performed to recommend an artificial intelligence or general model that is considered to match the user based on the around data and the like collected from the user by the mobile communication device 3 or the like. Also, when the user searches for an artificial intelligence or general model, the search results displayed in a list may be controlled to be preferentially displayed from those that are considered to match the user based on the around data and the like collected from the user. Furthermore, control may be performed to collect and display the user's reviews of the artificial intelligence or general model. At this time, as described above, since a human and an artificial intelligence are paired to form one personality, control may be performed to collect and display the reviews for the paired human and artificial intelligence as a set.

[0167] (7) Regarding the infants learned using the general model and personalized data for infant learning, a vast amount of data consisting of the provided learning materials (learning information) and learning results may be collected and used as learning data for machine learning to machine-learn the intellectual growth of humans (such as supervised learning). The same learning materials (learning information) provided to the infants may also be provided to the artificial intelligence, and the learning results of the artificial intelligence that has achieved intellectual growth and the actual infants may be compared, and reinforcement learning or regression may be performed to minimize the differences, and a learning model for the artificial intelligence to enable learning closer to that of humans may be created. In particular, in fields where the artificial intelligence is weak (such as common sense, emotions, creativity, etc.), it is considered that by comparing the learning results of both, it will become beneficial machine learning (such as supervised learning) for the artificial intelligence.

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

[0169] The following inventions are disclosed in the embodiments described above. The present disclosed invention relates to, for example, a service providing system and a program using machine learning by artificial intelligence. Specifically, it relates to a service providing system that provides a service using machine learning by artificial intelligence, and a program executed by a computer for a user to receive a service using machine learning by artificial intelligence.

[0170] There is one that creates learning data from each of a plurality of information sources in order to perform machine learning by artificial intelligence (for example, Japanese Patent Application Laid-Open No. 2011-232997).

[0171] However, a huge amount of learning data is required for machine learning. In order to create this huge amount of learning data, a large amount of necessary information must be collected from information sources. In order to put machine learning into practical use, it is essential to suppress the labor and cost required for this information collection.

[0172] Also, for example, a general model that is modeled by performing machine learning with a huge amount of learning data by collecting a large amount of information from an unspecified large number of people tends to be an average model for the entire unspecified large number of people. When providing services to each user with individual differences using this general model, there is a risk that the service will not match users who deviate from the average.

[0173] The present disclosed invention has been conceived in view of such circumstances, and its object is to reduce the labor and cost required for collecting a huge amount of information necessary for creating learning data. A further object is to solve the inconvenience that may occur when a service using a general model of the result of machine learning does not match the user.

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

[0175] The present disclosed invention is a service providing system (for example, FIGS. 9 and 10, etc.) that provides a service using machine learning by artificial intelligence, Machine learning means (for example, S189, S214, FIG. 15, etc.) for inputting learning data based on information sent from a user (for example, around data such as video being viewed by user 70, position of line of sight, voice, GPS position data, etc.) and modeling it by machine learning to generate a general model (for example, rote learning model Ti = T0·I, muscle training model Fi = F0+(i-1) / a, etc.), Personalization means (for example, S191 and S204, S223 and S228, etc.) for personalizing the general model into a model suitable for the user based on information sent from the user, Service providing means (for example, S206, S229, etc.) for providing a service personalized for the user (for example, providing learning items based on the most efficient review plan, providing a muscle training menu based on the most efficient muscle training plan, etc.) using the personalized model, and The information sent from the user is used for both the machine learning and the personalization.

[0176] According to such a configuration, by providing a service that utilizes the results of machine learning to users, a large number of users will proactively provide information for learning data on their own. Moreover, since the information provided by users is also used to personalize a general model into a model suitable for the user, it is possible to prevent as much as possible the inconvenience that each user leaves the information provision for learning data to others and only enjoys the service.

[0177] Preferably, the machine learning means generates the general model by using elements with differences for each user (for example, individual ability differences such as memorization ability, initial load for muscle training, and load increase coefficient, preferences, etc.) as constants (for example, T0, F0, a, etc.) (for example, S189, S214, FIG. 15, etc.). The personalization means value derivation means (for example, S191, S223, etc.) for deriving the actual values (for example, TK, FK, az, CT, etc.) of the user that apply to the constant part in the general model based on the information sent from the user, and substitution means (for example, S204, 228, etc.) for substituting the actual values derived by the value derivation means into the constant part in the general model to personalize it into a model suitable for the user, and includes.

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

[0179] Another aspect of the present invention is a service providing system that provides a service using machine learning by artificial intelligence, storage means (for example, artificial intelligence DB17, etc.) for storing a plurality of types of general models (for example, the dialogue model in FIG. 19(a), the task processing model in FIG. 21(a), etc.) generated by modeling using machine learning, and Service providing means (e.g., S241 - S244, S247, S248, S261 - S263, S267 - S269, etc.) for providing a service (e.g., an interactive service, a response to a work - related question or consultation, etc.) to a user by using a general model selected by the user from among a plurality of types of general models stored in the storage means. Personalization means (e.g., S246, S271, etc.) for personalizing the general model into a model suitable for the user according to the user's reaction to the service provided by the service providing means. The service providing means includes personalized service providing means (e.g., executing S244, S269, etc. according to the updated personal weights) for providing a service to the user by using the model personalized by the personalization means.

[0180] According to such a configuration, since a general model desired by the user is selected from among a plurality of types of general models generated by modeling using machine learning and the service is provided to the user by using the general model, the user can enjoy the service by the general model that reflects the user's desire. Moreover, since the general model is personalized into a model suitable for the user according to the reaction of the user who has received the service, it becomes possible to provide a service that matches the user by personalizing the provided service.

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

[0182] More preferably, it further includes model using means (e.g., S257, S258, etc.) for allowing others to use the model personalized by the personalization means.

[0183] Another aspect of the disclosed invention is a program executed by a computer for a user to receive services utilizing machine learning by artificial intelligence, a step of selecting and specifying (e.g., S237 and S238, S259 and S260, etc.) what the user desires from among a plurality of types of general models (e.g., the dialogue model in FIG. 19(a), the task processing model in FIG. 21(a), etc.) generated by modeling using machine learning by artificial intelligence, a processing step (e.g., S236 to S239, S260, S265, etc.) for the user to receive a service using the general model selected and specified in the selecting and specifying step, a step of providing the reaction of the user to the artificial intelligence in order to personalize the general model into a model suitable for the user in response to the reaction of the user who has received the service in the processing step (e.g., S239, S265, etc.), causing the computer to execute, the processing step includes a personal service enjoyment step (e.g., receiving a response according to the updated personal weight in S239, S265, etc.) for the user to receive a personalized service using the model personalized by the artificial intelligence.

[0184] According to such a configuration, since a general model desired by the user is selected from among a plurality of types of general models generated by modeling using machine learning and the service is provided to the user using the general model, the user can enjoy the service by the general model reflecting the user's desire. Moreover, since the general model is personalized into a model suitable for the user according to the reaction of the user who has received the service, it becomes possible to provide a service that matches the user by personalizing the provided service.

[0185] Although the embodiments of the present invention have been described as above, it should be considered that the embodiments disclosed this time are illustrative in all respects and not restrictive. The scope of this invention is shown not by the above description but by the claims, and it is intended that all changes within the meaning and scope equivalent to the claims are included.

Explanation of Reference Numerals

[0186] 1 Internet, 2 IoT device, 3 Mobile communication device, 4 Wireless sensor network, 8 IoT server, 9 Artificial intelligence server, 10 PC of various specialists, 12 User DB, 15 IoT device DB, 17 Artificial intelligence DB, 60 Learning DB, 38 Various sensors.

Claims

A service providing system capable of providing a service using a machine learning model, comprising: means for storing a plurality of types of models for task processing as average general models modeled by performing machine learning on learning data collected from a large number of people, and storing the initial weights of each piece of knowledge in the model for task processing in association with each model for task processing; an initial target storage means; weight update means for updating the weights of each piece of knowledge in the model for task processing so as to satisfy the user by using the model for task processing selected by the user from among the models for various task processing stored as the initial target in the initial target storage means, and personalizing the model; personalized weight storage means for storing the personalized weights of each piece of knowledge updated and personalized by the weight update means in association with the user; means for storing the personalized weights of each piece of knowledge stored in the personalized weight storage means in the initial target storage means as the initial target of the user himself / herself; and A service providing system that makes the personalized weights of each piece of knowledge in the user himself / herself stored in the initial target storage means generally available and makes the personalized weights of each piece of knowledge available as a substitute for the user himself / herself. A service providing system capable of providing a service using a machine learning model, comprising: a first means for generating an average general model modeled by performing machine learning on learning data collected from a large number of people; a second means for personalizing the general model generated by the first means into a personalized model for a specific person; and A service providing system that makes the personalized model generally available and makes the personalized model available as a substitute for the specific person.

3. The first means is to subtract the total number of negative responses x1h, x2h, x3h,... from the total number of positive responses x1k, x2k, x3k,... for each of the respective knowledges n1, n2, n3,... in the model to be generated, substitute the resulting values into the sigmoid function 1 / (1 + e^-x), and multiply the values of the sigmoid functions corresponding to the respective knowledges by the corresponding knowledges n1, n2, n3,... to obtain a general model that determines the priority order in utilizing the knowledges n1, n2, n3,... The second means is to update the weights of the respective knowledges in the general model so that the user, who is the specific person, is satisfied by using the general model, and to obtain a personalized model having the personalized weights of the updated and personalized knowledges. The service providing system according to claim 2.

Images