Method for generating machine learning data, and device for generating machine learning data

Abstract

To enable a sufficient amount of machine learning data to be prepared even when third-party rights are required.SOLUTION: A method includes: adding, in advance, to target data (21) on a server, permission-required portion information (23) indicating portions requiring permission (permission-required portions (24)) and replacement data for replacing the target data in a permission-required portion for which permission is not obtained; creating, in advance, permission information (30) indicating whether there is permission for the permission-required portions; loading, when generating machine learning data, the target data and the permission information, and replacing the target data with the replacement data for the permission-required portion for which permission is not obtained, to generate the machine learning data. The machine learning data can be generated even from the target data for which permission is not obtained, thereby easily preparing a sufficient amount of machine learning data.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a technique for generating machine learning data used to generate an estimation model by machine learning. [Background technology]

[0002] Recent technological advances in machine learning have been remarkable. One such remarkable advance is the practical application of a technology that creates an estimation model with high estimation accuracy by performing preliminary learning using a large amount of data related to general content, and then learning data of individual content corresponding to the field of use, purpose of use, etc. In the following, data of individual content corresponding to the field of use, purpose of use, etc. will be referred to as "individual data", and learning using individual data will be referred to as "individual learning". Data related to general content will be referred to as "general data", and preliminary learning using general data will be referred to as "preliminary learning".

[0003] In general, to create an estimation model with high estimation accuracy, it is necessary to train a large amount of data, but it is difficult to prepare a large amount of individual data. Therefore, by performing preliminary training using a large amount of general data that is easily available, and then performing individual training using a smaller amount of individual data than the general data, it becomes possible to relatively easily create an estimation model with high estimation accuracy.

[0004] Even if an estimation model with high estimation accuracy is obtained, the estimation accuracy may gradually decrease due to subsequent changes in environmental conditions. Therefore, various techniques have been proposed to maintain the accuracy of an estimation model by continuously collecting new individual data and performing individual learning even after the estimation model is obtained (for example, Patent Document 1).

[0005] Furthermore, continuous individual learning may actually result in a decrease in the accuracy of the estimation model. Therefore, a technique has been proposed in which data learned through continuous individual learning (i.e., individual data) is stored, and when the accuracy of the estimation model decreases, the learned individual data can be verified (Patent Document 2).

[0006] Here, these individual data are more difficult to obtain than the general data used in preliminary learning, and in addition, compared to general data, they are more likely to include content that is subject to third-party rights (such as copyright, privacy rights, trade secrets, etc.). If the learned individual data includes content that is subject to third-party rights, it becomes necessary to obtain the third party's permission after the fact, and if permission cannot be obtained, the estimated model must be discarded. For this reason, permission must be obtained before learning for individual data that includes content that is subject to third-party rights. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] JP 2023-086053 A [Patent Document 2] JP 2022-150778 A Summary of the Invention [Problem to be solved by the invention]

[0008] However, for machine learning data that requires third-party consent, such as individual data, consent cannot always be obtained, making it difficult to secure a sufficient amount of machine learning data. If a sufficient amount of machine learning data cannot be secured, it becomes difficult to obtain an estimation model with sufficient accuracy.

[0009] This invention has been made to solve the above-mentioned problems associated with conventional technology, and aims to provide technology that makes it possible to easily secure a sufficient amount of data for machine learning, even for data that requires third-party rights. [Means for solving the problem]

[0010] In order to solve the above-mentioned problems, the method for generating machine learning data of the present invention employs the following configuration. A method for generating machine learning data by a computer, the machine learning data being used to generate an estimation model by machine learning, comprising the steps of: a target data reading step of reading target data stored in a server; a permission information reading step of reading permission information indicating the presence or absence of permission to use the target data for the machine learning; a generating step of generating the machine learning data based on the target data and the permission information; It is equipped with The target data reading step includes: permission-required portion information indicating a permission-required portion, which is a portion in the target data where the content requiring permission is described or displayed; When the permission for the permission-required portion has not been obtained, replacement data is substituted for the target data of the permission-required portion; and A step of reading the target data to which The generating step is a step of generating the machine learning data by replacing the target data in the permission-required portion with the replacement data in accordance with the permission information. It is characterized by:

[0011] The method for generating data for machine learning according to the present invention can also be understood as an apparatus for generating data for machine learning. A machine learning data generation device that generates machine learning data for generating an estimation model by machine learning, a target data reading unit that reads target data stored in a server; a permission information reading unit that reads permission information indicating whether or not permission is given to use the target data for the machine learning; a machine learning data generation unit that generates the machine learning data based on the target data and the permission information; It is equipped with The target data reading unit is permission-required portion information indicating a permission-required portion, which is a portion in the target data where the content requiring permission is described or displayed; When the permission for the permission-required portion has not been obtained, replacement data is substituted for the target data of the permission-required portion; and The target data to which the following has been added is read, The machine learning data generation unit generates the machine learning data by replacing the target data in the permission-required portion with the replacement data in accordance with the permission information. It is characterized by:

[0012] In the method and device for generating data for machine learning according to the present invention, permission-required portion information and replacement data are added to target data in advance. The permission-required portion information is information indicating a permission-required portion in the target data, which is a portion where content requiring permission (such as a word, a sentence, a figure, or a graph) is described or displayed. The replacement data is data to be substituted for the target data of the permission-required portion when permission for the permission-required portion has not been obtained. When generating data for machine learning, the target data to which the permission-required portion information and replacement data have been added and the permission information are read. The permission information is information indicating the presence or absence of permission to use the target data for machine learning. Then, based on the permission information, it is determined whether permission has been obtained for the permission-required portion of the permission-required portion information, and when permission has not been obtained, the target data of the permission-required portion is replaced with the replacement data to generate data for machine learning.

[0013] In this way, even if permission has not been obtained for the license-required portion, replacement data that has been modified to a content that does not infringe on the rights of a third party or a content for which permission has been obtained is prepared, and the target data of the license-required portion is replaced with the replacement data, so that machine learning data can be generated. Also, even if the target data includes a license-required portion for which permission has been obtained and a license-required portion for which permission has not been obtained, machine learning data can be generated even from target data for which third-party rights are required by replacing the target data of the license-required portion for which permission has not been obtained with the replacement data. Furthermore, instead of concealing the target data of the license-required portion for which permission has not been obtained by blacking it out, the target data can be replaced with appropriate replacement data prepared in advance, so that machine learning data that can be sufficiently used for machine learning can be generated even from target data that includes a license-required portion for which permission has not been obtained. As a result, it is possible to easily secure a sufficient amount of machine learning data.

[0014] In addition, in the above-mentioned method for generating data for machine learning and device for generating data for machine learning of the present invention, information on the portion requiring permission may be added to the target data by describing the information on the portion requiring permission as a separate layer of the target data.

[0015] In this way, the parts of the target data that require third-party permission can be easily identified, and permission-required part information can be easily added to the target data. In addition, the parts that require third-party permission can be clearly identified without ambiguity, making it easy to create replacement data. Furthermore, since there is no need to modify the target data itself, no copyright problems arise from modifications.

[0016] In addition, in the above-mentioned method for generating data for machine learning and device for generating data for machine learning of the present invention, when generating data for machine learning, target data to which replacement data has been added for each portion requiring permission, and permission information indicating whether or not permission has been given for each portion requiring permission may be read.

[0017] In this way, it is possible to generate appropriate machine learning data even from target data that contains multiple parts that require permission.

[0018] Furthermore, in the above-described method and device for generating data for machine learning according to the present invention, when generating data for machine learning, permission information including partial permission and target data to which replacement data corresponding to the partial permission has been added may be read. Here, partial permission is permission that imposes restrictions on the content of the target data in the permission-required portion, the party receiving permission, the use of the estimation model generated by machine learning, etc. Then, for the permission-required portion for which partial permission has been obtained, the target data may be replaced with replacement data for partial permission to generate data for machine learning.

[0019] In this way, even if unlimited permission is not available, it is possible to generate data for machine learning from target data for which partial permission is available.

[0020] In addition, in the above-mentioned method for generating data for machine learning and device for generating data for machine learning of the present invention, when generating or after generating the data for machine learning, replacement information including information capable of identifying the target data, information on the parts requiring permission, and replacement data replaced for each part requiring permission may be generated, and the replacement information 40 may be stored in a distributed ledger in blockchain format.

[0021] This makes it easy to prove at a later date that the machine learning data was generated with the permission of a third party, which makes it possible to avoid a situation in which an inference model generated by machine learning is suspected of having learned machine learning data without the permission of a third party, and the inference model has to be discarded. [Brief description of the drawings]

[0022] [Figure 1] FIG. 1 is an explanatory diagram showing an overview of a machine learning data generation device 10 according to an embodiment of the present invention. [Diagram 2]FIG. 1 is an explanatory diagram showing an overview of machine learning that performs preliminary learning and individual learning. [Diagram 3] 1 is an explanatory diagram showing the data structure of a target dataset 20 to be read into a machine learning data generation device 10 of the present embodiment. FIG. [Figure 4] 1 is an explanatory diagram conceptually showing the relationship between target data 21 in a target data set 20 and permission-requiring portion information 22.

[0033] FIG. [Diagram 5] 2 is an explanatory diagram conceptually showing replacement data in the target data set 20. FIG. [Figure 6] FIG. 1 is an explanatory diagram of permission information 30 that is read to generate machine learning data. [Figure 7] 11 is a flowchart showing the first half of a process for generating data for machine learning. [Figure 8] 13 is a flowchart showing the latter half of the process of generating data for machine learning. [Figure 9] 13 is a flowchart showing a process of storing replacement information 40 in a block chain format. [Figure 10] FIG. 2 is an explanatory diagram showing the data structure of replacement information 40. [Figure 11] FIG. 11 is an explanatory diagram showing how replacement information 40 is stored in a block chain format. [Figure 12] FIG. 1 is an explanatory diagram showing the reason why a sufficient amount of individual data can be easily prepared by using the machine learning data generation device 10 of the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] FIG. 1 is an explanatory diagram showing an overview of a machine learning data generation device 10 of this embodiment. As shown in the figure, the machine learning data generation device 10 of this embodiment reads target data stored in a server 50 and preset permission information and applies a predetermined process to generate machine learning data. Here, the target data is data that is the subject of machine learning, and includes document data, image data, and data that combines these. In addition, as described later, the machine learning data generation device 10 of this embodiment reads the target data with permission-required portion information and replacement data added thereto, rather than the target data alone.

[0024] The permission information is information indicating whether or not permission has been obtained to use the target data for machine learning. In other words, the target data may contain content that is subject to third-party rights (such as privacy rights and copyrights), and if the target data is learned without the third-party's permission, in the worst case scenario, the estimated model obtained by machine learning may have to be discarded. Therefore, information indicating whether or not the target data contains content that requires permission, and, if such content is included, whether or not permission has been obtained, is compiled in advance as permission information. The permission information will be explained in detail later.

[0025] The machine learning data generation device 10 also includes a target data reading unit 11, a permission information reading unit 12, a machine learning data generation unit 13, a replacement information storage unit 14, and the like. Note that these "units" are abstract concepts that represent functions that the machine learning data generation device 10 has for generating machine learning data, and do not indicate the existence of components that correspond to these "units." These "units" can be realized as software programs executed by a computer, or as hardware such as an LSI or IC. Furthermore, they may be realized by combining a software program with hardware.

[0026] The target data reading unit 11 reads target data and supplies it to the machine learning data generation unit 13. In addition, the permission information reading unit 12 reads permission information and outputs it to the machine learning data generation unit 13.

[0027] When the machine learning data generation unit 13 receives the target data and the permission information, it checks whether the necessary permission has been obtained by referring to the permission information. If permission has been obtained, the machine learning data is generated using the target data as is. On the other hand, if permission has not been obtained, the machine learning data is generated by replacing the part of the target data that requires permission with replacement data that has been set in advance. The process of generating the machine learning data will be described in detail later.

[0028] The replacement information storage unit 14 generates replacement information by acquiring information indicating the portion of the target data replaced with replacement data and information indicating the replaced replacement data from the machine learning data generation unit 13, and stores the replacement information in a distributed ledger on the blockchain network 60. In the blockchain network 60, a plurality of computer nodes (hereinafter, nodes) are connected to each other so as to be able to communicate with each other, and when data is stored in one of the plurality of nodes, the same data is also stored in the other nodes, thereby forming a distributed ledger. The replacement information storage unit 14 generates replacement information in a blockchain format and transmits it to one node n1 of the blockchain network 60, thereby storing the replacement information in a distributed ledger on the blockchain network 60. A method for generating replacement information in a blockchain format will be described in detail later.

[0029] Here, the machine learning data generated by the above-mentioned machine learning data generation device 10 is machine learning data (individual data) used in a learning stage called "individual learning" in machine learning. That is, in recent machine learning, it is common to first perform preliminary learning and then perform individual learning, and the machine learning data generated by the machine learning data generation device 10 of this embodiment is particularly suitable as machine learning data for individual learning. Therefore, as a preparation for the explanation, an overview of machine learning that performs preliminary learning and individual learning will be explained.

[0030] FIG. 2 is an explanatory diagram showing an overview of machine learning that performs preliminary learning and individual learning. In machine learning, which has become mainstream in recent years, a large amount of data related to general content (general data) is learned in the preliminary learning, which is the first learning stage. Machine learning for generating so-called "large-scale language models" also falls under the category of preliminary learning. For example, to generate a large-scale language model for Japanese, it is necessary to learn a large amount of Japanese document data, but the content of the document data can be general. In this way, a large amount of data is machine-learned in preliminary learning, but since the data to be learned is data with general content, even large amounts of data can be obtained relatively easily.

[0031] After the preliminary learning is completed, the second learning stage, individual learning, is performed. In individual learning, data corresponding to the estimated model to be generated is prepared and machine-learned. For example, if an estimated model to be used in the financial field is to be generated, literature on economics and finance, including as much recent literature as possible, is machine-learned. Also, if an estimated model to be used in the security field is to be generated, literature on computers, communications, security, etc., including as much recent literature as possible, is machine-learned. Furthermore, if it is necessary to adapt the estimated model to the circumstances of the field where it is used, materials and data used in the field are also machine-learned. In this way, in individual learning, an estimated model is generated by machine-learning data (individual data) of individual contents corresponding to the field of use, purpose of use, etc. In order to generate an estimated model with high estimation accuracy, a large amount of individual data must be machine-learned, but a small amount of data compared to general data is acceptable.

[0032] In addition, even after an estimation model is obtained, the estimation accuracy may gradually decrease due to subsequent changes in environmental conditions. Therefore, estimation accuracy is maintained by continuously collecting new individual data and updating the estimation model through individual learning.

[0033] Here, since the individual data used in the individual learning or continuous individual learning is data with individual contents according to the purpose or use of the estimation model, it is difficult to prepare a large amount of data like general data. Furthermore, since the individual data is data with individual contents, it may contain contents accompanied by third party rights (copyright, privacy right, trade secret, etc.). In addition, in order to obtain an estimation model with high estimation accuracy, it is necessary to learn including individual data with the latest contents as much as possible, but the latest individual data is more likely to contain contents accompanied by third party rights. For these reasons, it is difficult to prepare a sufficient amount of data for the individual data. In contrast, if the above-mentioned machine learning data generation device 10 of this embodiment is used, it becomes possible to prepare a sufficient amount of individual data. The reason for this will be explained below.

[0034] 3 is an explanatory diagram showing the data structure of a target dataset 20 to be read into the machine learning data generation device 10 of this embodiment. The target dataset 20 has a data structure in which permission-required portion information 22 is added after target data 21 that is the subject of machine learning, and a replacement record set 23 is added after the permission-required portion information 22. Here, the permission-required portion information 22 is information that indicates a portion in the target data 21 where content requiring permission is described or displayed (hereinafter, the permission-required portion).

[0035] 4 is an explanatory diagram conceptually showing the relationship between the target data 21 of the target data set 20 and the permission-requiring portion information 22. As shown in the figure, the permission-requiring portion information 22 is added to the target data 21 as a separate layer, and is data that specifies the permission-requiring portion 24 on the separate layer. As shown in the figure, by displaying the permission-requiring portion 24 on a separate layer of the target data 21, even if there are multiple descriptions or indications requiring permission in the target data 21, they can be easily specified and recognized. Note that each permission-requiring portion 24 is assigned a serial number called a "permission-requiring number."

[0036] FIG. 5 is an explanatory diagram conceptually showing a replacement record set 23 included in the target data set 20. As shown in the figure, the replacement record set 23 is a collection of multiple replacement data records 25, and the replacement data record 25 is formed by arranging the license number, the assumed license status, and replacement data in this order. Here, the assumed license status indicates what license status the replacement data described later is assumed to be created under. For example, the second replacement data record 25 from the top in FIG. 5 is a replacement data record 25 for the license-requiring portion 24 with the license number 2, and indicates that this replacement data is created assuming the license status is "none" (i.e., a situation in which license is not obtained).

[0037] 5, the replacement data record 25 displayed at the top has a blank column for replacement data. The reason for this is that for the permission-required portion 24 with permission number 1, the permission status is "Yes" (i.e., permission has been obtained), so there is no need to create replacement data.

[0038] 5 both have a permission required number of 4. That is, two replacement data are set for the permission required portion 24 with the permission required number 4. The reason for this is that two situations are assumed for the content described in the permission required portion 24 with the permission required number 4: when permission has not been obtained and when partial permission has been obtained, and replacement data is required for each of the cases when permission has not been obtained (assumed permission situation is [none]) and when partial permission has been obtained (assumed permission situation is [partial]).

[0039] In the above example, the explanation was given assuming that there are two options, "a situation where permission is not obtained" and "a situation where partial permission is obtained". In contrast, in the case where there are two options, "a situation where permission is obtained" and "a situation where partial permission is obtained", it is sufficient to set a replacement data record 25 with an assumed permission status of "Yes" and a replacement data record 25 with an assumed permission status of "Partial". Also, as will be described in detail later, if there is a replacement data record 25 with the same permission required number, it is not necessarily necessary to set a replacement data record 25 with an assumed permission status of "Yes". Therefore, in the case where there are two options, "a situation where permission is obtained" and "a situation where partial permission is obtained", it is sufficient to set only a replacement data record 25 with an assumed permission status of "Partial".

[0040] Furthermore, the replacement data is recorded as is in the second, fourth, and fifth replacement data records 25 from the top in Fig. 5. However, as in the third replacement data record 25 from the top, it is also possible to set the URI (Uniform Resource Identifier) ​​where the replacement data is saved.

[0041] 6 is an explanatory diagram of the permission information 30 that the machine learning data generation device 10 of this embodiment reads in order to generate machine learning data. As shown in the figure, the permission information 30 is a collection of multiple permission status records 31, and the permission status record 31 is formed by arranging a license-required number and a permission status in this order. Here, the permission status is a status that indicates whether or not permission has been obtained for the license-required part 24 corresponding to the license-required number.

[0042] The machine learning data generation device 10 of this embodiment generates machine learning data by reading the target dataset 20 described above using Figures 3 to 5 and the permission information 30 described above using Figure 6, and executing the machine learning data generation process described below.

[0043] 7 and 8 are flowcharts of the machine learning data generation process. In the machine learning data generation process, the target dataset 20 created in advance and stored in the server 50 and the permission information 30 created in advance are read (STEP 10). Then, after initializing the license number to No. 1 (STEP 11), the permission status corresponding to the current license number is obtained by referring to the license information 30 (STEP 12). For example, immediately after the license number is initialized in STEP 11, the license number is No. 1, so the permission status of "Yes" set in the permission status record 31 for the license number No. 1 in the license information 30 illustrated in FIG. 6 is obtained.

[0044] Next, it is determined whether the permission status acquired in STEP 12 is "Yes" (STEP 13). As a result, if the permission status is "Yes" (STEP 13: yes), there is no need to replace the contents of the target data 21 with replacement data for the permission-required portion 24 of the current license-required number. In this case, it is determined whether the license-required number has reached the last license-required number (STEP 18 in FIG. 8), and if the last license-required number has not been reached (STEP 18: no), the license-required number is incremented by one (STEP 19), and the process returns to STEP 12 in FIG. 7 to acquire the permission status in which the license-required number is set in the permission status record 31 of the new license-required number from the permission information 30. Then, it is determined whether the obtained permission status is "yes" or not (STEP 13), and if it is "yes" (STEP 13), it is determined again whether the last license number has been reached, and if the last license number has not been reached, the license number is incremented by one (STEP 18 and STEP 19 in Figure 8), and then the process returns to STEP 12 in Figure 7 to repeat the same operations.

[0045] On the other hand, if the permission status acquired in STEP 12 is not "Yes" (STEP 13: no), then the replacement record set 23 of the target data set 20 is referenced to acquire the assumed permission status corresponding to the license-required number (STEP 14). That is, as described above with reference to Figures 3 and 5, the replacement data record 25 in the replacement record set 23 has the assumed permission status and replacement data set in association with the license-required number. Therefore, in STEP 14, the assumed permission status set in association with the current license-required number is acquired. At this time, if multiple types of assumed permission statuses are set for the same license-required number, all of the assumed permission statuses are acquired (see Figure 5).

[0046] Then, it is determined whether or not there is an assumed permission situation that matches the permission situation obtained in STEP 12 among the assumed permission situations obtained in STEP 14. For example, for the license-required number 4 in the license information 30 shown in FIG. 6, a permission situation of [partial], which indicates that partial permission has been obtained, is set. On the other hand, for the license-required number 4 in the target data set 20 shown in FIG. 5, an assumed permission situation of [none] and an assumed permission situation of [partial] are set. Therefore, the assumed permission situation of [partial] that matches the permission situation of [partial] is set. Then, if it is determined that there is an assumed permission situation that matches the assumed situation (STEP 15: yes), the content of the target data 21 described in the permission-required portion 24 is changed to replacement data that corresponds to the assumed permission situation that matches the assumed situation (STEP 16).

[0047] Once the contents of the target data 21 for the license-required portion 24 of the license-required number have been replaced with replacement data in the manner described above, it is determined whether the last license-required number has been reached (STEP 18 in FIG. 8). If the last license-required number has not been reached (STEP 18: no), the license-required number is incremented by one (STEP 19), and the process returns to STEP 12 in FIG. 7 to start the series of operations described above.

[0048] On the other hand, if the answer is "no" in STEP 15, that is, if there is no assumed permission situation among the assumed permission situations acquired in STEP 14 that matches the permission situation acquired in STEP 12, a predetermined alarm is output (STEP 17) and the machine learning data generation process is terminated. For example, if the permission situation acquired in STEP 12 is [None], which indicates that permission has not been obtained, but there is no replacement data record 25 with an assumed permission situation of [None] in the replacement record set 23 referenced in STEP 14, the content of the permission-requiring portion 24 in the target data 21 cannot be replaced. Therefore, an alarm indicating this is output, and the process is terminated without generating machine learning data.

[0049] Also, if it is determined in STEP 18 of FIG. 8 that the license number has reached the last number (STEP 18: yes), the description in the target data 21 for which no license has been obtained is replaced with replacement data, thereby obtaining machine learning data. Then, the data name of the obtained machine learning data is obtained (STEP 20). The data name of the machine learning data may be set when the target data set 20 and the license information 30 are read into the machine learning data generation device 10 in STEP 10 of FIG. 7, but input of the data name may be requested in STEP 20, and the input data name may be read. After that, the machine learning data is output with the obtained data name (STEP 21).

[0050] After the machine learning data is output in this way, a replacement information storage process (STEP 30) is started as described below in order to store the replacement information in a distributed ledger on the blockchain network 60. FIG. 9 is a flowchart of the replacement information storage process (STEP 30). The replacement information 40 is information indicating the content replaced in the target data 21 to generate the machine learning data, and is information obtained by adding the URI (Uniform Resource Identifier) ​​of the target data 21 used, the permission-required portion information 22 in the target dataset 20, the replacement record set 23, and the permission information 30 to the data name of the machine learning data, as shown in FIG. 10.

[0051] As shown in FIG. 9, in the replacement information storage process, the URI of the target data 21 used to generate the machine learning data, the permission-required portion information 22 and the replacement record set 23 in the target data set 20, and the permission information 30 are added to the data name of the machine learning data to generate replacement information 40 (STEP 31). Here, the reason why the URI is used for the target data 21 as the data to be added to the data name of the machine learning data, instead of the data itself, is that the target data 21 has a large data size. In addition, since the target data 21 is not created independently to generate machine learning data, but is a document stored in another institution (in most cases, a public institution), it is sufficient to know the URI. In contrast, the permission-required portion information 22 and the replacement record set 23 in the target data set 20, and the permission information 30 are created to generate machine learning data, so the data itself needs to be used. When generating replacement information 40 in STEP 31, information indicating the date and time when the machine learning data was output in STEP 21 of Fig. 8 may be obtained, and replacement information 40 including the date and time information may be generated. Furthermore, if the data size of the target data 21 is small, replacement information 40 may be generated using the target data 21 itself (instead of the URI).

[0052] Furthermore, among third parties who have rights associated with the target data 21, there may be a right holder who does not mind machine learning the target data 21 but does not want to disclose the contents of the target data 21 at least for the time being. In this case, there is no problem because the contents of the target data 21 cannot be known even if the estimated model obtained by machine learning is analyzed. However, if the URI of the target data 21 (or the target data 21 itself) is included in the replacement information 40, the contents of the target data 21 may be disclosed, and the right holder may be reluctant to do so. In such a case, the URI of the target data 21 (or the target data 21 itself) may be encrypted or concealed using secret calculation. Furthermore, the URI of the target data 21 may be displayed as usual, but the target data 21 itself identified by the URI may be encrypted or concealed using secret calculation. In this way, the contents of the target data 21 will not be disclosed, so it is possible to obtain permission to machine learn the target data 21 even from a right holder who is reluctant to have the contents of the target data 21 disclosed.

[0053] Next, block data 41 is created by adding replacement information 40 obtained in STEP 31 to the stored hash value (STEP 32). Here, the stored hash value is a value obtained by applying a predetermined hash function to previously created block data 41. The hash function converts data of any size into a hash value of the same data length, and if even a part of the data before conversion is different, the converted hash value becomes a completely different value, and furthermore, it is impossible to restore the data before conversion from the hash value. Therefore, since a one-to-one relationship is always established between the hash value and the original data, the hash value can be considered to uniquely represent the data before conversion.

[0054] Once the block data 41 is obtained in this manner, a new hash value is calculated by applying a hash function to the block data 41 (STEP 33), and the stored hash value is changed to the new hash value (STEP 34). This hash value will be used the next time block data 41 is created.

[0055] Then, after transmitting the obtained block data 41 to node n1 of the blockchain network 60 (STEP 35), the replacement information storage process of Figure 9 is terminated, and the process returns to the machine learning data generation process of Figures 7 and 8, and then terminates the machine learning data generation process. Upon receiving the block data 41, node n1 also transmits the block data 41 to the other nodes n2 to n4 of the blockchain network 60, and as a result, the block data 41 is stored in the distributed ledger of the blockchain network 60.

[0056] 11 is an explanatory diagram showing how block data 41 is stored in a distributed ledger of a blockchain network 60. Since the block data 41 includes a hash value indicating the previous block data 41, the block data 41 is stored in a format in which multiple block data 41 are linked in a row (blockchain format). It is known that storing the block data in this manner in the blockchain format makes it extremely difficult to tamper with the replacement information in one block data 41, since it becomes necessary to change the hash values ​​of all block data 41 linked after the block data 41.

[0057] The above is a detailed description of the method for generating machine learning data using the machine learning data generation device 10 of this embodiment. Generating machine learning data using such a method provides a number of advantages, as described below.

[0058] First, the machine learning data (individual data) used in the learning stage of machine learning, known as individual learning, is usually not easy to obtain because there is less of it compared to data with general content. In addition, because it is data about individual content, third-party permission may be required for use in machine learning, and if permission cannot be obtained, the data cannot be used for machine learning. Furthermore, if permission from multiple people is required, the data cannot be used for machine learning unless permission from all of them is obtained. For this reason, in addition to the small amount of data to begin with, the need for third-party permission makes it even more difficult to obtain.

[0059] In contrast, in the method of the present embodiment described above, the parts (license-requiring parts 24) that require permission from a third party are identified in the data (target data 21) to be machine-learned, and permission can be obtained for each permission-requiring part 24. Then, the permission-requiring parts 24 for which permission cannot be obtained can be replaced with content (replacement data) that does not require permission from a third party. Therefore, even if there is a permission-requiring part 24 for which permission cannot be obtained, it is possible to use the target data 21 for machine learning. Also, if permission can be obtained with a little modification of the content, it is possible to use the target data 21 for machine learning by replacing it with replacement data that has been modified as necessary.

[0060] Furthermore, since replacement data can be prepared for each permission-required portion 24, even if the target data 21 of the permission-required portion 24 for which permission has not been obtained is replaced with replacement data, the meaning of the text before and after the portion including the replacement data does not change significantly. If the target data 21 of the permission-required portion 24 for which permission has not been obtained is concealed by blacking out the data or replacing it with meaningless symbols, the meaning of the text before and after the portion including the replacement data will not be understood. Such learning data is noise that adversely affects machine learning and cannot be used as machine learning data. In contrast, in the method of the present embodiment described above, for the permission-required portion 24 for which permission has not been obtained, the target data 21 of the portion can be replaced with replacement data prepared in advance, making it possible to generate machine learning data that is sufficiently usable for machine learning.

[0061] Furthermore, since the permission-requiring portion 24 can be specified on a different layer of the target data 21 (see FIG. 4), the permission-requiring portion 24 in the target data 21 can be easily and unambiguously specified. In addition, since the permission-requiring portion 24 in the target data 21 can be precisely specified, replacement data for replacing that portion can also be easily prepared. In addition, since there is no need to modify the target data 21 itself, the permission-requiring portion information 22 can be easily added to the target data 21.

[0062] Furthermore, even if there are multiple third parties who require permission, it is not necessarily the case that all of them have the rights for all of the permission-required portions 24. Therefore, even if it is difficult to obtain permission from some of the third parties, it is possible to use the target data 21 as is for machine learning for the permission-required portions 24 for which permission has been obtained.

[0063] Even if permission from a third party cannot be obtained, permission may be obtained depending on the party from whom permission is sought or the purpose or use of the machine learning, or permission may be obtained by modifying the content only slightly. In the method of the present embodiment described above, permission status for each permission-requiring portion 24 can be set in the permission information 30, so that such cases can be flexibly dealt with. This point will be described with reference to FIG. 12.

[0064] FIG. 12 is an explanatory diagram showing a state where multiple types of machine learning data are generated from target data 21. In the illustrated example, a case is shown where machine learning data used in a university B, machine learning data used in a related company C, and machine learning data used in a general company D are generated from one target data 21. Even if the same target data 21 is used for research at the university B, it is considered that permission can be obtained for many permission-required parts 24 as long as the content does not include trade secrets. Furthermore, when the target data 21 is used for product development at the related company C, it is considered that permission can be obtained even for the permission-required parts 24 that include content including trade secrets. Furthermore, when the target data 21 is used by the general company D, it is considered that the number of permission-required parts 24 for which permission can be obtained is reduced. Therefore, permission information 30 for the university (permission information b), permission information 30 for the related company (permission information c), and permission information 30 for the general company (permission information d) are created in advance. Then, for university B, permission information 30 for the university (permission information b) is applied to the target dataset 20 of the target data 21 to generate machine learning data for the university (machine learning data Ab). For company C, permission information 30 for the related company (permission information c) is applied to generate machine learning data for the related company (machine learning data Ac). Furthermore, for company D, permission information 30 for general companies (permission information d) is applied to generate machine learning data for general companies (machine learning data Ad). In this way, multiple types of machine learning data can be generated from one target data 21.

[0065] For the reasons described above, by using the method of this embodiment, even if the data for machine learning (individual data) is used in individual learning, a sufficient amount of individual data can be easily prepared, and as a result, an estimation model with high estimation accuracy can be obtained through machine learning.

[0066] In addition, each time machine learning data is generated, replacement information 40 of the machine learning data is stored in a blockchain format in the distributed ledger of the blockchain network 60. As is well known, data stored in a blockchain format in the distributed ledger of the blockchain network 60 is virtually impossible to tamper with, so that even at a later date, it can be easily proven that the machine learning data was generated with the permission of a third party. Furthermore, if the replacement information 40 contains information on the date and time when the machine learning data was output, the date and time when the machine learning data was generated can also be identified, making the proof even easier. This makes it possible to avoid a situation in which an estimation model generated by machine learning must be discarded.

[0067] The above describes the machine learning data generating device 10 of this embodiment, but the present invention is not limited to the above embodiment and can be embodied in various forms without departing from the gist of the present invention. [Explanation of symbols]

[0068] 10...machine learning data generation device, 11...target data reading unit, 12 ... permission information reading unit, 13 ... machine learning data generation unit, 14... Replacement information storage unit; 20... Target data set; 21... Target data; 22... Information on parts requiring permission, 23... Replacement record set, 24... Parts requiring permission, 25... Replacement data record, 30... Permission information, 31... Permission status record, 40...Replacement information, 41...Block data, 50...Server, 60…blockchain network, n1~n4…nodes.

Claims

1. A method for generating machine learning data by a computer, the machine learning data being used to generate an estimation model by machine learning, comprising the steps of: a target data reading step of reading target data stored in a server; a permission information reading step of reading permission information indicating the presence or absence of permission to use the target data for the machine learning; a generating step of generating the machine learning data based on the target data and the permission information; It is equipped with The target data reading step includes: permission-required portion information indicating a permission-required portion, which is a portion in the target data where the content requiring permission is described or displayed; When the permission for the permission-required portion has not been obtained, replacement data is substituted for the target data of the permission-required portion; and A step of reading the target data to which The generating step is a step of generating the machine learning data by replacing the target data in the permission-required portion with the replacement data in accordance with the permission information. A method for generating data for machine learning, comprising:

2. The method for generating machine learning data according to claim 1, The target data reading step is a step of reading the target data in which the permission-required portion information is described in a separate layer from the target data. A method for generating data for machine learning, comprising:

3. The method for generating machine learning data according to claim 2, the target data reading step is a step of reading the target data to which the replacement data has been added for each of the permission-requiring portions, The permission information reading step is a step of reading the permission information indicating the presence or absence of the permission for each of the permission-requiring portions. A method for generating data for machine learning, comprising:

4. The method for generating machine learning data according to any one of claims 1 to 3, the permission information reading step is a step of reading the permission information including a partial permission, The target data reading step is a step of reading the target data to which the replacement data for the case where the partial permission has been obtained has been added, for the portion requiring permission for which the partial permission has been obtained. A method for generating data for machine learning, comprising:

5. The method for generating machine learning data according to any one of claims 1 to 3, A storage step of generating replacement information including information capable of identifying the target data, information on the permission-required portion, and the replacement data replaced for each permission-required portion when or after the machine learning data is generated, and storing the replacement information in a distributed ledger in a blockchain format. A method for generating data for machine learning, comprising:

6. A machine learning data generation device that generates machine learning data for generating an estimation model by machine learning, a target data reading unit that reads target data stored in a server; a permission information reading unit that reads permission information indicating whether or not permission is given to use the target data for the machine learning; a machine learning data generation unit that generates the machine learning data based on the target data and the permission information; It is equipped with The target data reading unit is permission-required portion information indicating a permission-required portion, which is a portion in the target data where the content requiring permission is described or displayed; When the permission for the permission-required portion has not been obtained, replacement data is substituted for the target data of the permission-required portion; and The target data to which the following has been added is read, The machine learning data generation unit generates the machine learning data by replacing the target data in the permission-required portion with the replacement data in accordance with the permission information. A machine learning data generation device comprising:

Images