Data name presentation device, learning device, data name presentation system, data name presentation method, and learning method

Abstract

To facilitate determination of an appropriate data name by a user.SOLUTION: A data registration server 2 includes a UI processing unit 200, a metadata acquisition unit 201, first learned model storage means, and a data name inference unit 202. The metadata acquisition unit 201 acquires metadata corresponding to registration target data from a file storing the registration target data acquired from an apparatus. The UI processing unit 200 presents the acquired metadata to a user. The first learned model storage means stores a first learned model for inferring a data name corresponding to the registration target data from the metadata determined by the user. The data name inference unit 202 infers a data name based on the metadata determined by the user and the first learned model. The UI processing unit 200 presents the inferred data name to the user.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present disclosure relates to a data name presentation device, a learning device, a data name presentation system, a data name presentation method, and a learning method. [Background technology]

[0002] When registering new data in a database, there are many cases where the user, the operator, decides the data name to be given to the data based on his or her own judgment or the conventions of the department or project to which the user belongs.

[0003] The method of determining data names as described above is not only inefficient, but also often results in similar data being given different data names, or inappropriate data names being assigned, which raises concerns that this may hinder the subsequent use of the data.

[0004] In response to this, a technique has been proposed in which candidates for data names according to the content of the data to be registered are presented to the user (for example, Patent Document 1).

[0005] The software component management device described in Patent Document 1 analyzes registered software components and the software component to be registered when registering a software component, and presents the registrant with the functional names of registered software components that are similar to the software component to be registered. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Laid-Open No. 2012-168598 Summary of the Invention [Problem to be solved by the invention]

[0007] The technology described in Patent Document 1 is effective for registering software components. However, when the data to be registered is data acquired from devices such as home appliances, equipment, and various sensors, the data name is often determined by taking into consideration various information, and presenting only candidate data names presents a problem in that there is insufficient information for the user to rely on to determine the data name.

[0008] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a data name presentation device, a learning device, a data name presentation system, a data name presentation method, and a learning method that allow a user to easily determine an appropriate data name when registering data obtained from a device. [Means for solving the problem]

[0009] In order to achieve the above object, the data name presentation device according to the present disclosure includes: The file storing the registration target data acquired from the device is used to select the registration target data corresponding to the registration target data. The metadata includes at least device information about the device and data value information about values ​​included in the registration target data. a metadata acquisition means for acquiring metadata; a metadata presentation means for presenting the acquired metadata to a user; User-confirmed metadata This is a trained model trained based on training data that includes a dataset with the data name determined by the user as the correct answer data. a first trained model storage means for storing the first trained model; User-confirmed metadata of The first trained model By entering a data name inference means for inferring a data name; and a data name presenting means for presenting the inferred data name to a user. [Effects of the Invention]

[0010] According to the present disclosure, when registering data acquired from a device, the user can easily determine an appropriate data name. [Brief explanation of the drawings]

[0011] [Figure 1]FIG. 1 is a diagram showing the overall configuration of a data registration system according to an embodiment. [Figure 2] FIG. 1 is a block diagram showing the hardware configuration of a data registration server according to an embodiment. [Figure 3] FIG. 1 is a block diagram showing a hardware configuration of a terminal according to an embodiment. [Figure 4] FIG. 1 is a diagram showing the functional configuration of a data registration server according to an embodiment. [Figure 5] FIG. 1 is a diagram for explaining a UI processing unit according to an embodiment. [Figure 6] FIG. 10 is a diagram showing an example of a data registration screen according to an embodiment. [Figure 7] FIG. 1 is a diagram showing an example of data to be registered in an embodiment; [Figure 8] FIG. 1 is a diagram showing an example of data to be registered in an embodiment; [Figure 9] FIG. 1 is a diagram showing an example of data to be registered in an embodiment; [Figure 10] FIG. 10 is a diagram showing an example of a data registration screen according to an embodiment. [Figure 11] FIG. 10 is a diagram showing an example of a data registration screen according to an embodiment. [Figure 12] FIG. 1 is a diagram illustrating a metadata acquisition unit according to an embodiment. [Figure 13] FIG. 1 is a diagram for explaining a data name inference unit according to an embodiment. [Figure 14] FIG. 1 is a diagram illustrating a data registration unit according to an embodiment. [Figure 15] FIG. 10 is a diagram illustrating a metadata learning unit according to an embodiment. [Figure 16] A diagram showing an example of a neural network [Figure 17] FIG. 10 is a diagram illustrating a data name learning unit according to an embodiment. [Figure 18] 1 is a flowchart showing a procedure for data registration processing according to an embodiment. [Figure 19] 1 is a flowchart showing the procedure of a metadata learning process according to an embodiment. [Figure 20]1 is a flowchart showing the procedure of a data name learning process according to an embodiment. [Figure 21] FIG. 10 is a diagram showing the functional configuration of a data registration server according to a modified example of the embodiment. [Figure 22] FIG. 10 is a diagram for explaining a data name inference unit according to a modified example of the embodiment. [Figure 23] 10 is a flowchart showing the procedure of data integration processing according to a modified example of the embodiment. [Figure 24] 10A and 10B are diagrams showing examples of displays ((a) is a graph, and (b) to (e) are symbols) relating to transition trends presented to a user in a modified example of an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings.

[0013] 1 is a diagram showing the overall configuration of a data registration system 1 according to an embodiment of the present disclosure. The data registration system 1 is a system that registers data to be registered in a database through a registration operation by a user, and includes a data registration server 2 and a terminal 3.

[0014] <Data registration server 2> The data registration server 2 is an example of a data name presentation device, a learning device, and a data name presentation system according to the present disclosure. As shown in Fig. 2, the data registration server 2 includes a communication interface 20, a CPU (Central Processing Unit) 21, a ROM (Read Only Memory) 22, a RAM (Random Access Memory) 23, and an auxiliary storage device 24. These components are connected to each other via a bus 25.

[0015] The communication interface 20 is hardware for communicating with other devices, for example, the terminal 3, via a network N such as a LAN (Local Area Network), a CAN (Campus Area Network), a MAN (Metropolitan Area Network), a WAN (Wide Area Network), or the Internet. For example, the communication interface 20 is a wired or wireless LAN interface.

[0016] The CPU 21 performs overall control of the data registration server 2. The functions of the data registration server 2 realized by the CPU 21 will be described in detail later. The ROM 22 stores firmware and data used when the firmware is executed. The RAM 23 is used as a working area for the CPU 21.

[0017] The auxiliary storage device 24 is composed of a readable / writable nonvolatile semiconductor memory, an HDD (Hard Disk Drive), etc. Examples of the readable / writable nonvolatile semiconductor memory include an EEPROM (Electrically Erasable Programmable Read-Only Memory), a flash memory, etc. The auxiliary storage device 24 stores various programs including a program for performing operations related to data registration (hereinafter referred to as a data registration program), and data used when these programs are executed.

[0018] The data registration server 2 can obtain the above-mentioned data registration program from another server via the network N. The data registration program can also be stored and distributed on a computer-readable recording medium such as a CD-ROM (Compact Disc Read Only Memory), a DVD (Digital Versatile Disc), a magneto-optical disk, a USB (Universal Serial Bus) memory, a HDD, an SSD (Solid State Drive), or a memory card. When such a recording medium is attached to the data registration server 2, the data registration server 2 can read and import the data registration program from the recording medium.

[0019] <Terminal 3> The terminal 3 is a computer device that allows a user, who is an operator, to access the data registration server 2 and perform work related to data registration. As shown in Fig. 3, the terminal 3 includes a display 30, an operation reception unit 31, a communication interface 32, a CPU 33, a ROM 34, a RAM 35, and an auxiliary storage device 36. These components are interconnected via a bus 37.

[0020] The display 30 includes a display device such as a liquid crystal display, an organic EL (Electro Luminescence) display, a plasma display, a CRT display, etc. Under the control of the CPU 33, the display 30 displays various screens etc. in response to user operations.

[0021] The operation reception unit 31 is configured to include one or more input devices such as a push button, keyboard, mouse, keypad, touch panel, touchpad, etc., and receives operation input from the user and outputs a signal related to the received operation to the CPU 33.

[0022] The communication interface 32 is hardware for communicating with the data registration server 2 via the network N. For example, the communication interface 32 is a wired or wireless LAN interface.

[0023] The ROM 34 stores firmware and data used when the firmware is executed. The RAM 35 is used as a working area for the CPU 33.

[0024] The auxiliary storage device 36 is composed of a readable / writable nonvolatile semiconductor memory, a HDD, etc. Examples of the readable / writable nonvolatile semiconductor memory include an EEPROM, a flash memory, etc. The auxiliary storage device 36 stores a program (e.g., a web browser, etc.) for communicating with the data registration server 2, a file in which registration target data, which is data to be registered, is stored (hereinafter, a registration data file), etc.

[0025] <Functional configuration of data registration server 2> Fig. 4 is a diagram showing the functional configuration of the data registration server 2. As shown in Fig. 4, the data registration server 2 includes a UI (User Interface) processing unit 200, a metadata acquisition unit 201, a data name inference unit 202, a data registration unit 203, a metadata learning unit 204, and a data name learning unit 205. These functional units are realized by the CPU 21 executing the above-mentioned data registration program stored in the auxiliary storage device 24.

[0026] (1) UI processing unit 200 The UI processing unit 200 is an example of a metadata presentation means and a data name presentation means according to the present disclosure. The UI processing unit 200 executes user interface processing when a user performs data registration via the terminal 3. Specifically, the UI processing unit 200 exchanges data as shown in Fig. 5 with the terminal 3. When a user accesses the data registration server 2 via the terminal 3, the UI processing unit 200 presents a data registration screen as shown in Fig. 6 via the terminal 3.

[0027] The user selects a registration data file stored in the auxiliary storage device 36 of the terminal 3 by operating a selection button 300 provided on the data registration screen of FIG. 6. The registration data file is an example of a file according to the present disclosure, and is a file that stores data to be registered (hereinafter referred to as registration target data) acquired from devices such as home appliances, facility equipment, and various sensors. Examples of the registration target data are shown in FIGS. 7 to 9. FIG. 7 shows an example of the registration target data acquired from an air conditioner, FIG. 8 shows an example of the registration target data acquired from an outside air temperature sensor, and FIG. 9 shows an example of the registration target data acquired from an electric water heater. After selecting a registration data file, when the user operates a confirmation button 301 provided on the data registration screen of FIG. 6, the registration data file is transmitted to the data registration server 2.

[0028] The UI processing unit 200 receives the registration data file transmitted from the terminal 3 and stores the received registration data file in the registration data file storage unit 210. The registration data file storage unit 210 is a memory area provided by the auxiliary storage device 24.

[0029] Furthermore, the UI processing unit 200 transmits the metadata acquired by the metadata acquisition unit 201 to the terminal 3 as metadata candidates, and presents them to the user via a data registration screen (see FIG. 10). As shown in FIG. 10, in this embodiment, the metadata candidates include date and time information indicating the date and time when the registration target data was acquired, location information indicating the location where the registration target data was acquired, device information about the corresponding device, and data value information. The device information includes information identifying the manufacturer of the device (for example, official name, abbreviation, nickname, etc.) and a communication ID (identification). The communication ID is a communication ID assigned to the device in advance. Note that the device information may include other items instead of or in addition to the above items.

[0030] The data value information also includes a data type indicating the type of value contained in the data to be registered, a data value range indicating the range of values ​​contained in the data to be registered, and a transition trend indicating the transition trend of the values ​​contained in the data to be registered.

[0031] The metadata candidates are displayed on the data registration screen in a format that allows the user to edit them using the operation reception unit 31. After the user edits the metadata candidates as necessary, by operating the confirm button 302 provided on the data registration screen of FIG. 10, the metadata candidates currently displayed on the data registration screen are sent to the data registration server 2 as the final metadata results. Note that for metadata candidates having multiple detailed items (e.g., device information, data value information, etc.) on the data registration screen of FIG. 10, the corresponding detailed items may be displayed in a hierarchical fold-out menu format. Furthermore, if all information cannot be displayed on the data registration screen, vertical and horizontal scroll bars are displayed as appropriate, allowing vertical and horizontal scrolling.

[0032] The UI processing unit 200 receives the metadata confirmation result transmitted from the terminal 3, and stores the received metadata confirmation result in the metadata confirmation result storage unit 211. The metadata confirmation result storage unit 211 is a memory area provided by the auxiliary storage device 24.

[0033] Furthermore, the UI processing unit 200 transmits the data name inferred by the data name inference unit 202 to the terminal 3 as a data name candidate, and presents it to the user via the data registration screen (see FIG. 11). The data name candidate is displayed on the data registration screen in a manner that allows the user to edit it using the operation reception unit 31. After the user edits the data name candidate as necessary, when the user operates the confirm button 303 provided on the data registration screen in FIG. 11, the data name candidate currently displayed on the data registration screen is transmitted to the data registration server 2 as the confirmed data name result.

[0034] The UI processing unit 200 receives the data name determination result transmitted from the terminal 3, and stores the received data name determination result in the data name determination result storage unit 212. The data name determination result storage unit 212 is a memory area provided by the auxiliary storage device 24.

[0035] (2) Metadata Acquisition Unit 201 The metadata acquisition unit 201 is an example of a metadata acquisition means according to the present disclosure. The metadata acquisition unit 201 acquires metadata corresponding to registration target data from a registration data file. In detail, as shown in FIG. 12 , the metadata acquisition unit 201 includes a data separation unit 220, an inference unit 221, and a metadata generation unit 222. The data separation unit 220 reads the registration data file from the registration data file storage unit 210, separates the read registration data file into a header and registration target data, stores the registration target data (see FIGS. 7 to 9) in the registration target data storage unit 213, and supplies the data to the inference unit 221, and supplies the header to the metadata generation unit 222.

[0036] The registration target data storage unit 213 is a memory area provided by the auxiliary storage device 24. The header includes date and time information indicating the date and time when the registration target data was acquired, location information indicating the location where the registration target data was acquired, and device information about the corresponding device. The device information includes information identifying the manufacturer of the device (for example, official name, abbreviation, nickname, etc.) and the communication ID of the device.

[0037] The inference unit 221 infers data value information, i.e., the data type, data value range, and transition tendency described above, based on the data to be registered and the trained model for metadata inference stored in the trained model storage unit 214. The trained model storage unit 214 is an example of a second trained model storage means according to the present disclosure. The trained model storage unit 214 is a memory area provided by the auxiliary storage device 24, and stores a trained model for metadata inference, which is a trained model generated by the metadata learning unit 204 described below. The trained model for metadata inference is an example of a second trained model according to the present disclosure. The inference unit 221 infers data value information by inputting the data to be registered into the trained model for metadata inference.

[0038] The metadata generation unit 222 generates metadata corresponding to the data to be registered based on the header supplied from the data separation unit 220 and the data value information inferred by the inference unit 221, and supplies the generated metadata to the UI processing unit 200.

[0039] (3) Data name inference unit 202 The data name inference unit 202 is an example of a data name inference means according to the present disclosure. As shown in FIG. 13 , the data name inference unit 202 infers a data name to be used as the registration name of the data to be registered, based on the metadata confirmation result and the trained model for data name inference stored in the trained model storage unit 215. The trained model storage unit 215 is an example of a first trained model storage means according to the present disclosure. The trained model storage unit 215 is a memory area provided by the auxiliary storage device 24, and stores a trained model for data name inference, which is a trained model generated by the data name learning unit 205 described below. The trained model for data name inference is an example of a first trained model according to the present disclosure.

[0040] The data name inference unit 202 infers a data name by inputting the metadata determination result read from the metadata determination result storage unit 211, i.e., the metadata corresponding to the registration target data determined by the user, into a trained model for data name inference. The data name inference unit 202 supplies the inferred data name to the UI processing unit 200.

[0041] (4) Data registration unit 203 14 , the data registration unit 203 generates registration data based on the data name determination result read from the data name determination result storage unit 212, i.e., the data name corresponding to the registration target data determined by the user, the registration target data read from the registration target data storage unit 213, and the metadata determination result read from the metadata determination result storage unit 211, i.e., the metadata corresponding to the registration target data determined by the user, and registers the generated registration data in the database 216.

[0042] (5) Metadata learning unit 204 As shown in FIG. 15 , the metadata learning unit 204 includes a data acquisition unit 230 and a model generation unit 231. The data acquisition unit 230 is an example of a second learning data acquisition means according to the present disclosure. The data acquisition unit 230 reads and acquires registration target data from the registration target data storage unit 213. The data acquisition unit 230 also reads and acquires data value information (i.e., data type, data value range, and transition tendency) included in the metadata determination result from the metadata determination result storage unit 211. The data acquisition unit 230 uses the acquired registration target data as input data and supplies learning data (an example of second learning data according to the present disclosure) including a dataset in which the acquired data value information is used as correct answer data to the model generation unit 231.

[0043] The model generation unit 231 is an example of a second trained model generation means according to the present disclosure. The model generation unit 231 generates a trained model for inferring data value information based on the training data supplied from the data acquisition unit 230. The model generation unit 231 trains data value information corresponding to the data to be registered by, for example, supervised learning using a neural network.

[0044] A neural network is composed of an input layer to which input data is input, an output layer to which output data is output, and at least one intermediate layer (also called a hidden layer), and each layer is composed of multiple nodes. The number of nodes in the input layer corresponds to the number of input data, and the number of nodes in the output layer corresponds to the number of output data. Figure 16 shows an example of a three-layer neural network. In the example shown in Figure 16, the input layer is composed of nodes X1 to X3, the intermediate layer is composed of nodes Y1 to Y2, and the output layer is composed of nodes Z1 to Z3.

[0045] The model generation unit 231 uses the data set included in the learning data as training data, and performs learning by adjusting the connection weights between each layer (in the example shown in Figure 16, weights w11 to w16 between the input layer and the intermediate layer, and weights w21 to w26 between the intermediate layer and the output layer) so that the output data output from the output layer when input data (i.e., data to be registered) is input to the input layer becomes correct data (i.e., confirmed data value information), thereby generating a learned model.

[0046] The model generation unit 231 stores the generated trained model in the trained model storage unit 214 as a trained model for metadata inference.

[0047] (6) Data name learning unit 205 As shown in FIG. 17 , the data name learning unit 205 includes a data acquisition unit 240 and a model generation unit 241. The data acquisition unit 240 is an example of a first learning data acquisition means according to the present disclosure. The data acquisition unit 240 reads and acquires the metadata determination result, i.e., the metadata determined by the user, from the metadata determination result storage unit 211. The data acquisition unit 240 also reads and acquires the data name determination result, i.e., the data name determined by the user, from the data name determination result storage unit 212. The data acquisition unit 240 uses the acquired metadata as input data and supplies learning data (an example of first learning data according to the present disclosure) including a dataset in which the acquired data name is correct data, to the model generation unit 241.

[0048] The model generation unit 241 is an example of a first trained model generation means according to the present disclosure. The model generation unit 241 generates a trained model for inferring a data name based on the training data supplied from the data acquisition unit 240. The model generation unit 241 learns a data name corresponding to the confirmed metadata, for example, by supervised learning using a neural network. That is, the model generation unit 241 uses a data set included in the training data as training data, and performs training by adjusting the connection weights between layers so that when input data (i.e., confirmed metadata) is input to the input layer, output data output from the output layer becomes correct data (i.e., confirmed data name), thereby generating a trained model.

[0049] The model generation unit 241 stores the generated trained model in the trained model storage unit 215 as a trained model for data name inference.

[0050] <Data registration process> 18 is a flowchart showing the procedure of the data registration process executed by the data registration server 2. The data registration process is executed every time a user accesses the data registration server 2 via the terminal 3 for the purpose of data registration.

[0051] (Step S101) The data registration server 2 presents a data registration screen such as that shown in Fig. 6 to the user via the terminal 3. Thereafter, the process of the data registration server 2 proceeds to step S102.

[0052] (Step S102) The data registration server 2 determines whether or not the user has completed the selection of the registration data file. When the user has selected the registration data file and operated the Confirm button 301, the data registration server 2 determines that the selection of the registration data file has been completed. When the selection of the registration data file has not been completed, the data registration server 2 continues to execute the processing of step S102. On the other hand, when the selection of the registration data file has been completed, the processing of the data registration server 2 transitions to step S103.

[0053] (Step S103) The data registration server 2 instructs the terminal 3 to transmit the registration data file selected by the user, and receives the registration data file transmitted from the terminal 3 in response to the instruction. The data registration server 2 stores the received registration data file in the registration data file storage unit 210. Thereafter, the processing of the data registration server 2 proceeds to step S104.

[0054] (Step S104) The data registration server 2 acquires the registration target data from the registration data file and stores the acquired registration target data in the registration target data storage unit 213. The data registration server 2 also acquires a header from the registration data file. After that, the processing of the data registration server 2 proceeds to step S105.

[0055] (Step S105) The data registration server 2 infers data value information, which is metadata, from the data to be registered. In detail, the data registration server 2 infers data value information (i.e., data type, data value range, and transition tendency) by inputting the data to be registered into a learned model for metadata inference stored in the learned model storage unit 214. Thereafter, the processing of the data registration server 2 proceeds to step S106.

[0056] (Step S106) The data registration server 2 generates metadata corresponding to the data to be registered based on the acquired header and the inferred data value information, and presents the generated metadata as metadata candidates to the user via the terminal 3 (see FIG. 10). Thereafter, the process of the data registration server 2 proceeds to step S107.

[0057] (Step S107) The data registration server 2 determines whether the metadata has been confirmed by the user. If the Confirm button 302 on the data registration screen (see FIG. 10) is operated, the data registration server 2 determines that the metadata has been confirmed. If the metadata has not been confirmed, the data registration server 2 continues to execute the process of step S107. On the other hand, if the metadata has been confirmed, the process of the data registration server 2 proceeds to step S108.

[0058] (Step S108) The data registration server 2 instructs the terminal 3 to transmit the metadata confirmation result, and receives the metadata confirmation result transmitted from the terminal 3 in response to the instruction. The data registration server 2 stores the received metadata confirmation result in the metadata confirmation result storage unit 211. Thereafter, the processing of the data registration server 2 proceeds to step S109.

[0059] (Step S109) The data registration server 2 infers a data name to be the registration name of the data to be registered from the metadata confirmation result. In detail, the data registration server 2 infers the data name by inputting the metadata confirmation result read from the metadata confirmation result storage unit 211, i.e., the metadata corresponding to the data to be registered confirmed by the user, into the trained model for data name inference stored in the trained model storage unit 215. Thereafter, the processing of the data registration server 2 transitions to step S110.

[0060] (Step S110) The data registration server 2 presents the inferred data names as data name candidates to the user via the terminal 3 (see FIG. 11). Thereafter, the process of the data registration server 2 proceeds to step S111.

[0061] (Step S111) The data registration server 2 determines whether the data name has been confirmed by the user. If the user operates the Confirm button 303 on the data registration screen (see FIG. 11), the data registration server 2 determines that the data name has been confirmed. If the data name has not been confirmed, the data registration server 2 continues to execute the processing of step S111. On the other hand, if the data name has been confirmed, the processing of the data registration server 2 transitions to step S112.

[0062] (Step S112) The data registration server 2 instructs the terminal 3 to transmit the data name determination result, and receives the data name determination result transmitted from the terminal 3 in response to the instruction. The data registration server 2 stores the received data name determination result in the data name determination result storage unit 212. Thereafter, the processing of the data registration server 2 proceeds to step S113.

[0063] (Step S113) The data registration server 2 generates registration data based on the data name determination result read from the data name determination result storage unit 212, i.e., the data name corresponding to the registration target data determined by the user, the registration target data read from the registration target data storage unit 213, and the metadata determination result read from the metadata determination result storage unit 211, i.e., the metadata corresponding to the registration target data determined by the user. Thereafter, the processing of the data registration server 2 proceeds to step S114.

[0064] (Step S114) The data registration server 2 registers the generated registration data in the database 216, and ends the data registration process.

[0065] <Metadata learning process> 19 is a flowchart showing the procedure of the metadata learning process executed by the data registration server 2. The metadata learning process is executed immediately after the above-mentioned data registration process is completed. The timing of executing the metadata learning process is a matter of design. For example, the metadata learning process may be executed periodically, or may be executed in response to a user operation that instructs learning.

[0066] (Step S201) The data registration server 2 reads and acquires the registration target data from the registration target data storage unit 213. After that, the process of the data registration server 2 proceeds to step S202.

[0067] (Step S202) The data registration server 2 reads and acquires the data value information (i.e., data type, data value range, and transition tendency) confirmed by the user from the metadata confirmation result storage unit 211. After that, the processing of the data registration server 2 proceeds to step S203.

[0068] (Step S203) The data registration server 2 generates a trained model for inferring data value information based on the training data including a data set in which the data value information confirmed by the user is used as correct answer data, using the data to be registered as input data. After that, the process of the data registration server 2 proceeds to step S204.

[0069] (Step S204) The data registration server 2 stores the generated trained model in the trained model storage unit 214 as a trained model for metadata inference, and ends the metadata learning process.

[0070] <Data name learning process> 20 is a flowchart showing the steps of the data name learning process executed by the data registration server 2. The data name learning process is executed immediately after the above-mentioned data registration process is completed. The timing at which the data name learning process is executed is an arbitrary design matter. For example, the data name learning process may be executed periodically, or may be executed in response to a user operation that instructs learning.

[0071] (Step S301) The data registration server 2 reads and acquires the metadata confirmation result, that is, the metadata confirmed by the user, from the metadata confirmation result storage unit 211. Thereafter, the process of the data registration server 2 proceeds to step S302.

[0072] (Step S302) The data registration server 2 reads and acquires the data name confirmed by the user from the data name confirmation result storage unit 212. Thereafter, the process of the data registration server 2 transitions to step S303.

[0073] (Step S303) The data registration server 2 uses the metadata confirmed by the user as input data and generates a trained model for inferring data names based on training data including a dataset in which the data name confirmed by the user is used as correct answer data. After that, the processing of the data registration server 2 proceeds to step S304.

[0074] (Step S304) The data registration server 2 stores the generated trained model in the trained model storage unit 215 as a trained model for data name inference, and ends the data name learning process.

[0075] As described above, according to the data registration system 1 of the present embodiment, when registering data to be registered acquired from a device, the system presents the user with metadata corresponding to the data to be registered, and also presents the user with a data name inferred based on the metadata determined by the user and the trained model for data name inference obtained by training. This allows the user to easily determine an appropriate data name.

[0076] (Variation 1) The data registration server 2 may acquire data names through multi-stage inference. The functional configuration of the data registration server 2 in this case is shown in Fig. 21. As shown in Fig. 21, the data registration server 2 in this modification includes a data name inference unit 202a instead of the data name inference unit 202, and a data name learning unit 205a instead of the data name learning unit 205.

[0077] 22, the data name inference unit 202a includes a first inference unit 250 and a second inference unit 251. The first inference unit 250 infers a data name to be used as the registration name of the data to be registered, based on the result of determining the metadata and the first trained model for data name inference stored in the trained model storage unit 217.

[0078] The trained model storage unit 217 is an example of a first trained model storage means according to the present disclosure. The trained model storage unit 217 is a memory area provided by the auxiliary storage device 24, and stores a first trained model for data name inference, which is a trained model generated by the data name learning unit 205a (described later). The first trained model for data name inference is an example of a first trained model according to the present disclosure. The first inference unit 250 infers a data name (an example of a first data name according to the present disclosure) by inputting the metadata confirmation result read from the metadata confirmation result storage unit 211, i.e., the metadata corresponding to the registration target data confirmed by the user, into the first trained model for data name inference. The first inference unit 250 supplies the inferred data name to the second inference unit 251.

[0079] The second inference unit 251 infers a data name that will be the registered name of the data to be registered based on the data name inferred by the first inference unit 250 and the second learned model for data name inference stored in the learned model memory unit 218.

[0080] The trained model storage unit 218 is an example of a third trained model storage means according to the present disclosure. The trained model storage unit 218 is a memory area provided by the auxiliary storage device 24, and stores a second trained model for data name inference, which is a trained model generated by the data name learning unit 205a described below. The second trained model for data name inference is an example of a third trained model according to the present disclosure. The second inference unit 251 infers a data name (an example of a second data name according to the present disclosure) by inputting the data name inferred by the first inference unit 250 into the second trained model for data name inference. The second inference unit 251 supplies the inferred data name to the UI processing unit 200.

[0081] The data name learning unit 205a generates a trained model by performing a data name learning process (see Figure 20) similar to that of the data name learning unit 205, and stores the generated trained model in the trained model memory unit 217 as a first trained model for data name inference.

[0082] In addition, the data name learning unit 205a uses the data name inferred by the first inference unit 250 of the data name inference unit 202a as input data, generates a trained model for inferring data names based on training data that includes a dataset in which the data name confirmed by the user is used as correct data, and stores the generated trained model in the trained model storage unit 218 as a second trained model for data name inference.

[0083] By inferring data names in multiple stages in this way, more appropriate data name candidates can be presented to the user.

[0084] (Variation 2) The data registration server 2 may also include a data name unifying unit (an example of a data name unifying means according to the present disclosure) not shown, which searches for registered data whose metadata content is similar to that of the data newly registered in the database 216 but whose data name is different, and if matching data is found, may change the data name of the registered data to the data name of the newly registered data. The data name unifying unit may also periodically or in response to a user's operation via a terminal such as the terminal 3 search for registered data whose metadata content is similar to that of the data, but whose data name is different, and if matching data is found, update the database 216 to unify the data name of the data to the data name of the data most recently registered among the data.

[0085] (Variation 3) The data registration server 2 may also have a function of integrating multiple registered data in response to a user's operation via a terminal such as the terminal 3. Fig. 23 is a flowchart showing the steps of the data integration process executed by the data registration server 2 in this modified example. The data integration process is executed each time a user accesses the data registration server 2 via a terminal such as the terminal 3 for the purpose of data integration.

[0086] (Step S401) The data registration server 2 presents a data integration screen (not shown) to the user via the terminal. After that, the process of the data registration server 2 proceeds to step S402.

[0087] (Step S402) The data registration server 2 determines whether the user has completed the selection of multiple registered data to be integrated. If the selection of multiple data has not been completed, the data registration server 2 continues to execute the process of step S402. On the other hand, if the selection of multiple data has been completed, the process of the data registration server 2 proceeds to step S403.

[0088] (Step S403) The data registration server 2 reads out each piece of data selected by the user from the database 216 and presents the contents of each piece of data and inferred data name candidates to the user via the terminal. After that, the process of the data registration server 2 proceeds to step S404.

[0089] (Step S404) The data registration server 2 determines whether the data name has been confirmed by the user. If the data name has not been confirmed, the data registration server 2 continues to execute the process of step S404. On the other hand, if the data name has been confirmed, the process of the data registration server 2 proceeds to step S405.

[0090] (Step S405) The data registration server 2 instructs the terminal to transmit the result of determining the data name, and receives the result of determining the data name transmitted from the terminal in response to the instruction. After that, the process of the data registration server 2 proceeds to step S406.

[0091] (Step S406) The data registration server 2 assigns the determined data name, integrates the plurality of data, and ends the data integration process.

[0092] In addition, the data registration server 2 executes a process of learning and generating a trained model to be used to infer the data name in step S403 above at a predetermined timing, such as immediately after the above data integration process.

[0093] (Variation 4) The data value information may include at least one of a data type, a data value range, and a transition tendency. The metadata presented to the user may include at least one of date and time information, location information, device information, a data type, a data value range, and a transition tendency.

[0094] (Variation 5) The data registration screen may be provided with operation buttons or the like that allow the user to delete candidate metadata items presented by the data registration server 2 or to add new metadata items.

[0095] (Variation 6) In addition, the contents of the data to be registered (for example, a list of labels and data values) may be displayed on the data registration screen or on a separate screen. Furthermore, to make it easier for the user to understand the transition trends, the data values ​​included in each piece of data to be registered may be displayed in graph form on the data registration screen or on a separate screen (FIG. 24(a)), or the transition trends may be displayed as symbols (FIGS. 24(b) to (e)).

[0096] (Variation 7) Furthermore, the data registration server 2 may store the metadata acquired by the metadata acquisition unit 201 as a result of determining the metadata in the metadata determination result storage unit 211. Similarly, the data name inferred by the data name inference unit 202 may be stored as a result of determining the data name in the data name determination result storage unit 212. By inputting the acquired metadata to the data name inference unit 202 and performing the data registration process using the inferred data name, without presenting the metadata candidates or data name candidates to the user and having the user perform the determination process, the efficiency of the data registration work can be improved and the burden on the user can be reduced.

[0097] (Variation 8) In addition, when the data registration server 2 infers a data name to be the registered name of the data to be registered from the metadata determination result and presents it to the user via the terminal 3, it may present the user with one or more data name candidates and the accuracy corresponding to the data name candidates.

[0098] Specifically, the data name learning unit 205 generates a trained model for data name inference that outputs a predetermined number of highly accurate data name candidates together with their accuracy. Then, the data name inference unit 202 supplies a predetermined number of pairs of data names and accuracy obtained by inputting the metadata determination results into the trained model for data name inference to the UI processing unit 200. The UI processing unit 200 transmits each data name as a data name candidate together with the corresponding accuracy to the terminal 3.

[0099] The UI processing unit 200 may transmit the data name candidates in descending order of accuracy, that is, a ranking based on accuracy, to the terminal 3 instead of the accuracy, or may transmit the data name candidates, the accuracy, and a ranking based on accuracy to the terminal 3. This allows the user to select a data name from multiple data name candidates while taking accuracy into consideration, thereby improving the accuracy of the assigned data name.

[0100] The number of data name candidates, probabilities, and rankings based on the probabilities to be presented is arbitrary, and the number may be determined in advance, or all data name candidates with a probability higher than a predetermined value may be output.

[0101] (Variation 9) The data registration server 2 can also be used to assign data names when transferring data stored in the data lake to a data warehouse. In this case, for example, the data stored in the data lake is the data to be registered, and the data warehouse is the database 216. The trained model for metadata inference is trained using the data stored in the data warehouse and inferred data value information, and the trained model for data name inference is trained using the metadata of the data stored in the data warehouse and the data name determined by the user.

[0102] The data registration server 2 acquires metadata from the data lake selected by the user as the data to be registered, infers a data name, presents it to the user, generates registration data with the data name determined by the user, and registers it in the data warehouse. This reduces the burden on the user who transfers data and improves the accuracy of the assigned data names.

[0103] (Variation 10) Furthermore, there is no limitation on the learning algorithm used by the data registration server 2 when generating trained models (trained models for metadata inference and trained models for data name inference), and in addition to supervised learning, various well-known learning algorithms can be adopted, such as unsupervised learning, reinforcement learning, and deep learning that learns to extract features themselves.

[0104] (Variation 11) The data registration system 1 may also be implemented by a single computer. In this case, the computer has the same hardware configuration as the data registration server 2 (see FIG. 2), as well as an operation reception unit including one or more input devices such as a push button, keyboard, mouse, keypad, touch panel, or touchpad, and a display including a display device such as a liquid crystal display, an organic EL display, a plasma display, or a CRT display. The computer stores the data registration program in the above embodiment, and the CPU of the computer executes the data registration program to implement the same functions as the data registration server 2 (see FIG. 4).

[0105] (Variation 12) The data registration server 2 may also be configured to be composed of multiple separate computer devices. For example, the learning-related functions (metadata learning unit 204, data name learning unit 205) may be separated from the data registration server 2 and implemented by a separate computer device, or the metadata inference trained model, data name inference trained model, database 216, etc. may be stored in a storage device provided in another server such as a file server.

[0106] (Variation 13) In addition, all or part of the functional units (see FIGS. 4, 12, 15, and 17) of the data registration server 2 may be realized by dedicated hardware, such as a single circuit, a composite circuit, a programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof.

[0107] The technical ideas according to the above-described modifications may be realized independently or in appropriate combination.

[0108] The present disclosure is not limited to the above-described embodiments and modifications, and various modifications are of course possible within the scope of the gist of the present disclosure. [Explanation of symbols]

[0109] 1 Data registration system, 2 Data registration server, 3 Terminal, 20, 32 Communication interface, 21, 33 CPU, 22, 34 ROM, 23, 35 RAM, 24, 36 Auxiliary storage device, 25, 37 Bus, 30 Display, 31 Operation reception unit, 200 UI processing unit, 201 Metadata acquisition unit, 202, 202a Data name inference unit, 203 Data registration unit, 204 Metadata learning unit, 205, 205a Data name learning unit, 210 Registration data file storage unit, 211 Metadata confirmation result storage unit, 212 Data name confirmation result storage unit, 213 Registration target data storage unit, 214, 215, 217, 218 Trained model storage unit, 216 Database, 220 Data separation unit, 221 Inference unit, 222 Metadata generation unit, 230, 240 Data acquisition unit, 231, 241 Model generation unit, 250 First inference unit, 251, second inference unit, 300, selection button, 301-303, confirmation buttons

Claims

1. a metadata acquisition means for acquiring, from a file storing registration target data acquired from a device, metadata corresponding to the registration target data, the metadata including at least device information about the device and data value information about values ​​included in the registration target data; a metadata presentation means for presenting the acquired metadata to a user; a first trained model storage means for storing a first trained model, which is a trained model trained based on training data including a dataset in which metadata determined by a user is used as input data and the data name determined by the user is used as correct answer data; a data name inference means for inferring a data name by inputting metadata determined by a user into the first trained model; and a data name presenting means for presenting the inferred data name to a user.

2. The data name presentation device of claim 1, wherein the data value information includes at least one of a data type indicating the type of value contained in the data to be registered, a data value range indicating the range of values ​​contained in the data to be registered, and a transition trend indicating the transition trend of the values ​​contained in the data to be registered.

3. The method further includes a second trained model storage means for storing a second trained model trained based on training data including a dataset in which the registration target data is used as input data and data value information included in metadata confirmed by a user is used as correct answer data, The data name presentation device according to claim 1 or 2, wherein the metadata acquisition means infers and acquires the data value information by inputting the data to be registered into the second trained model.

4. Further comprising a third trained model storage means for storing the third trained model, The data name inference means infers a first data name corresponding to the registration target data by inputting the determined metadata into the first trained model, and infers a second data name corresponding to the registration target data by inputting the first data name into the third trained model; the data name presenting means presents the inferred second data name to a user; The third trained model is a trained model trained based on training data including a data set in which the first data name is input data and the data name confirmed by the user is correct data. The data name presentation device described in any one of claims 1 to 3.

5. A first learning data acquisition means for acquiring first learning data including a data set in which metadata corresponding to the data to be registered acquired from a device is presented to a user from a file in which the data to be registered is stored, the metadata including at least device information about the device and data value information about values ​​included in the data to be registered, and in which the metadata confirmed by the user is used as input data and the data name confirmed by the user is used as correct data; A learning device comprising: a first trained model generation means that generates a first trained model for inferring the data name from metadata confirmed by a user based on the acquired first learning data.

6. a second learning data acquisition means for acquiring second learning data including a data set in which the registration target data is used as input data and data value information included in the metadata confirmed by a user is used as correct answer data; and a second trained model generation means for generating a second trained model for inferring the data value information from the registration target data based on the acquired second training data, The learning device of claim 5, wherein the data value information includes at least one of a data type indicating the type of value included in the registration target data, a data value range indicating the range of values ​​included in the registration target data, and a transition trend indicating the transition trend of the values ​​included in the registration target data.

7. a metadata acquisition means for acquiring, from a file storing registration target data acquired from a device, metadata corresponding to the registration target data, the metadata including at least device information about the device and data value information about values ​​included in the registration target data; a metadata presentation means for presenting the acquired metadata to a user; a first trained model storage means for storing a first trained model, which is a trained model trained based on training data including a dataset in which metadata determined by a user is used as input data and the data name determined by the user is used as correct answer data; a data name inference means for inferring a data name by inputting metadata determined by a user into the first trained model; a data name presentation means for presenting the inferred data name to a user.

8. A metadata acquisition means acquires, from a file in which data to be registered acquired from a device is stored, metadata corresponding to the data to be registered, the metadata including at least device information about the device and data value information about values ​​included in the data to be registered; a metadata presentation means for presenting the acquired metadata to a user; The first trained model storage means stores a first trained model, which is a trained model trained based on training data including a dataset in which metadata determined by a user is used as input data and the data name determined by the user is used as correct answer data; a data name inference means for inferring a data name by inputting metadata determined by a user into the first trained model; A data name presentation method, wherein a data name presentation means presents the inferred data name to a user.

9. A metadata acquisition means acquires, from a file in which the registration target data acquired from a device is stored, metadata corresponding to the registration target data, the metadata including at least device information about the device and data value information about values ​​included in the registration target data; a first learning data acquisition means for acquiring first learning data including a data set in which the metadata confirmed by the user is used as input data by presenting the metadata to the user and the data name confirmed by the user is used as correct answer data; A learning method in which a first trained model generation means generates a first trained model for inferring the data name from metadata confirmed by a user based on the acquired first training data.

Images