Information processing apparatus and information processing method and information processing program

Abstract

To protect confidential information, and achieve a result equal to or better than that obtained when analyzing "non-public information" directly without sending such information, including free-form text, to other service providers.SOLUTION: An information processing apparatus comprises a processor and a memory device that stores a plurality of publicly available sentences, and the processor acquires a sentence to be processed containing non-public information, extracts a similar sentence from the plurality of publicly available sentences that is similar to the sentence to be processed, generates a mixed sentence by combining the sentence to be processed and the similar sentence, and outputs the mixed sentence.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to technology for using "private information" online. [Background technology]

[0002] In recent years, there has been a demand for analysis using machine learning models published online, which requires enormous computational resources. When using these models, companies must input confidential information into the models, such as know-how and technology they have researched and collected, which can bring significant benefits to the company. However, there is a risk that the information input at this time could be used by the service provider that owns the model. Even if a company publishes its model online, there are those who attempt to uncover the confidential information and use it as a reference to create new machine learning models.

[0003] In response to this, intellectual property protection technology has been proposed that verifies whether newly released machine learning models contain non-public information from datasets consisting of pairs of input and output information, or heat maps generated from input information and feature maps (see Non-Patent Document 1 and Patent Document 1). Furthermore, in order to utilize non-public information (customer data) in in-house training for new employees, there is a pseudo-data generation technology that calculates a distribution from structured data such as age and address, automatically generates pseudo-data that conforms to the distribution, and enables analysis without using any non-public information (see Patent Document 2). Furthermore, a privacy protection technology has been proposed that extracts named entities, which are personal information, from non-structured data such as free text, and abstracts the named entities, thereby enabling the analysis of non-structured data while reducing the risk of personal identification (see Patent Document 3). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] International Publication No. 2023 / 135682 [Patent Document 2] Japanese Patent Application Laid-Open No. 2010-277481 [Patent Document 3] Japanese Patent Application Publication No. 2018-200602 [Non-patent literature]

[0005] [Non-Patent Document 1] Lukas, Nils, et al. "SoK: How Robust is Image Classification Deep Neural Network Watermarking?(Extended Version)." arXiv preprint arXiv:2108.04974 (2021). Summary of the Invention [Problem to be solved by the invention]

[0006] However, it is difficult to determine the trend of appearance of unstructured data, such as the distribution of structured data in Patent Document 2. In other words, it is not easy to automatically generate pseudo data. For this reason, Non-Patent Document 1 and Patent Document 1 do not prevent the input of non-public information. Also, in Patent Document 3, non-public information is processed to protect personal information, but this does not change the meaning or structure of the entire sentence, and it does not prevent the input of confidential information that brings great benefits to a company, such as know-how and technology contained in the data.

[0007] The present invention has been made in consideration of the above points, and aims to protect confidential information by achieving effects equal to or greater than those achieved by inputting and analyzing "private information" as is, without transmitting the "private information," including free text, to other service providers. [Means for solving the problem]

[0008] In order to achieve the above object, one aspect of the present invention is an information processing device having a processor and a storage device, wherein the storage device holds a plurality of publicly available sentences, and the processor acquires a processing target sentence that includes private information, extracts similar sentences that are similar to the processing target sentence from the plurality of publicly available sentences, generates a mixed sentence by mixing the processing target sentence and the similar sentence, and outputs the mixed sentence. [Effects of the Invention]

[0009] As described above, according to one aspect of the present invention, in order to protect confidential information, it is possible to obtain effects equal to or greater than those obtained by inputting "private information" as is and analyzing it without transmitting the "private information" including free text to other service providers. Problems, configurations, and effects other than those described above will become clear from the description of the following examples. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a schematic diagram illustrating a system configuration of an information processing apparatus according to an embodiment of the present invention. [Figure 2] FIG. 2 is a block diagram showing an example of a hardware configuration of each device according to the embodiment of the present invention. [Figure 3] 1 is a functional configuration diagram of an information processing apparatus according to an embodiment of the present invention; [Figure 4] FIG. 3 is a sequence diagram showing processing of the information processing device according to the embodiment of the present invention. [Figure 5] FIG. 2 is an explanatory diagram showing an example of a screen image displayed by an information processing apparatus according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, embodiments will be described with reference to the drawings, in which the same or equivalent components and parts are designated by the same reference numerals. [Example]

[0012] FIG. 1 is a schematic diagram of a system configuration of an information processing apparatus according to an embodiment of the present invention.

[0013] The similar sentence mixing device 100 is realized by an information processing device such as a personal computer or a server used by a user. In the first embodiment, the similar sentence mixing device 100 is realized by the configuration of a personal computer 20.

[0014] The similar sentence mixing device 100 communicates with an analysis device 200 managed by another service provider via a communication network 33. The analysis device 200 will be described later.

[0015] FIG. 2 is a block diagram showing an example of the hardware configuration of each device according to the embodiment of the present invention.

[0016] The personal computer 20 comprises a processor 21, a device interface 22, a communication interface 23, a main storage device 24, and an auxiliary storage device 25, which are connected via a path 26.

[0017] 2 includes one of each component, the personal computer 20 may include multiple of the same component. For example, the personal computer 20 may include multiple processors 21.

[0018] In the first embodiment, the processor 21 is a CPU (Central Processing Unit) that executes processing according to a program stored in a storage device such as a hard disk, etc. Alternatively, the processor 21 may be an electronic circuit (such as a GPU) that includes a control device and a computing function of a computer.

[0019] The processor 21 performs arithmetic processing using data and programs input from each device, etc., configured internally of the personal computer 20, and can output the arithmetic results and control signals to each device, etc. The processor 21 may control each component constituting the personal computer 20 by executing the OS (Operating System) of the personal computer 20, applications, etc.

[0020] The device interface 22 is an interface that directly connects to the input device 31 and the display 32 wirelessly or via a wired connection.

[0021] The input device 31 is configured with devices for inputting various operational instructions to the personal computer 20, such as a keyboard, a mouse, buttons, or a touch panel.

[0022] The display 32 is configured as a device such as a display or a touch panel for displaying a GUI (Graphical User Interface) or the like to a user.

[0023] The communication interface 23 is an interface for connecting to a communication network 33 wirelessly or via a wire. The communication interface 23 may be one that conforms to an existing communication standard. The personal computer 20 may exchange information with an external device 34 via the communication network 33 using the communication interface 23. The external device 34 may be an external storage device such as a cloud.

[0024] The main memory device 24 is a storage device that stores instructions executed by the processor 21, various data, etc., and information stored in the main memory device 24 is read by the processor 21. In the first embodiment, the main memory device 24 is configured by a volatile RAM (Random Access Memory) that can store electronic information. Alternatively, the main memory device 24 may be any other electronic component that can store electronic information.

[0025] The auxiliary storage device 25 is a storage device other than the main storage device 24, and in the first embodiment, the auxiliary storage device 25 is configured by a non-volatile (non-temporary) hard disk capable of storing electronic information. Alternatively, the auxiliary storage device 25 may be any other electronic component capable of storing electronic information.

[0026] In the first embodiment, the electronic information of the similar sentence mixing device 100 is held in the auxiliary storage device 25, and during execution, various data of the similar sentence mixing device 100 is written to the main storage device 24. Alternatively, the electronic information of the similar sentence mixing device 100 may be held in, for example, the main storage device 24 or the external device 34. That is, the electronic information of the similar sentence mixing device 100 is realized by an information processing device having a CPU that executes processing according to a program stored in a storage device such as a RAM or a hard disk. In other words, the electronic information of the similar sentence mixing device 100 includes a program, and the processor 21 executes instructions written in this program to control the personal computer 20 and realize the functions of the similar sentence mixing device 100.

[0027] Some or all of the programs and data that realize the processing described below may be stored in the auxiliary storage device 25 in advance, or, if necessary, may be stored in the auxiliary storage device 25 via an I / F (not shown) that the personal computer 20 has, from a non-transitory storage device of another device connected via a network, or from a non-transitory storage medium, via an I / F (not shown).

[0028] FIG. 3 is a functional configuration diagram of the information processing device according to the embodiment of the present invention.

[0029] The similar sentence mixing device 100 has a processing target input program 1, a similar sentence search program 2, a mixed sentence generation program 3 of the processing target and similar sentences, a screen display program 4, a mixed sentence transmission program 5, an analysis result screen display program 6, a public information list 11, and a mixed sentence generation model 12. Of the above, each of the programs from the processing target input program 1 to the analysis result screen display program 6 is included in the electronic information of the similar sentence mixing device 100 shown in FIG.

[0030] The processing target input program 1 is a program for realizing the function of accepting input of a processing target by controlling the input device 31 and the display 32. The processing target input program 1 displays the screen image at the time of input on the display 32 using a programmatic GUI (Graphical User Interface) or the like, and is used to select the processing target and start execution of the similar sentence mixing device 100 by inputting various operation instructions using the input device 31 to specify the address where the processing target of the processing target and similar sentence mixing generation program 3 is stored.

[0031] In the first embodiment, the input data D1 to be processed is stored in the auxiliary storage device 25.

[0032] Alternatively, if necessary, an external storage device such as a USB memory on which input data D1 is stored may be connected to the device interface 22, or the input data D1 may be stored in the auxiliary storage device 25 from a non-transitory storage device of another device connected via a network or from a non-transitory storage medium via an I / F (not shown) provided in the personal computer 20.

[0033] The screen display program 4 and the analysis result screen display program 6 are programs for realizing the screen display function by controlling the input device 31 and the display 32. The program displays the results on the display 32 using a GUI or the like, allowing the user to perceive the processing results. Also, by inputting various operation instructions using the input device 31, it is possible to make corrections to the processing target and then perform the processing again, or to terminate the operation of the similar sentence mixing device 100.

[0034] However, the similar sentence search program 2 and the public information list 11 can also be realized by separate hardware, and may be installed outside the similar sentence mixing device 100, or may be installed so as to be accessible via the communication network 33.

[0035] Furthermore, the mixed sentence generation program 3 for the processing target and similar sentences and the mixed sentence generation model 12 can be realized by separate hardware and may be installed outside the similar sentence mixing device 100. In that case, however, the hardware that realizes the mixed sentence generation program 3 for the processing target and similar sentences and the mixed sentence generation model 12 should be installed so that it cannot be accessed via the communication network 33.

[0036] The public information list 11 holds information such as government guidelines that can be accessed by an unspecified number of people via the Internet or the like.

[0037] In Example 1, the public information list 11 includes guidelines such as the "Occupational Ability Assessment Standards" published by the Ministry of Health, Labor and Welfare, which describe employment-related assessment criteria such as "knowledge," "technology and skills" required to perform a job, and "examples of work behavior that leads to results (job performance ability)," as well as the perspectives for each assessment.

[0038] The above-mentioned "occupational ability evaluation criteria" is an example of information included in the public information list 11, and the public information list 11 may include any information relating to evaluation criteria of abilities based on a person's behavior.

[0039] The mixed sentence generation model 12 holds a model that has been trained based on the public information list 11, and when a sentence is input, outputs one mixed sentence that is related to the public information list 11. Alternatively, it may hold a model that, when two sentences are input, randomly extracts words that make up each sentence, rearranges them, and outputs one mixed sentence.

[0040] In the first embodiment, the mixed sentence generation model 12 holds a model that has been trained based on the public information list 11, and when a sentence is input, outputs a sentence that is relevant to the evaluation axis or evaluation viewpoint in employment.

[0041] The analysis device 200 is configured by one or more computers. The analysis device 200 can receive data to be analyzed via the communication network 33 using the analysis target receiving unit 7.

[0042] The analytical model 13 holds a model that receives data to be analyzed as input and outputs the analysis results.

[0043] In Example 1, the analytical model 13 is a trained sentence generation machine learning model that, when inputted with sentences related to evaluation criteria related to employment and each evaluation perspective, outputs the occupation that suits the person and the reason for that occupation.

[0044] The model is not limited to a sentence generation machine learning model. Other examples include a random forest (RF), a support vector machine (SVM), and a neural network (NN).

[0045] Furthermore, the model may be capable of additional learning by inputting learning data from the outside, in addition to analysis from the outside.

[0046] The analysis device 200 transmits the analysis results output by the analysis model 13 via the analysis result transmission unit 9 to the communication interface 23 that has transmitted the data to be analyzed via the communication network 33.

[0047] FIG. 4 is a sequence diagram showing the processing of the information processing device according to the embodiment of the present invention.

[0048] The operation based on the flowchart in FIG. 4 is as follows.

[0049] Step S101: The user causes the similar sentence mixing device 100 to acquire input data D1 to be processed using the input program 1 to be processed. In the first embodiment, the input data D1 is a support record of a user (i.e., Mr. A) who is a job seeker, such as "Mr. A continued working on (game name) for two years during his job search period." At this time, to protect privacy, named entities may be abstracted from the input data D1. For example, the above "(game name)" may be a proper noun that identifies the game (e.g., a product name), or may be information that identifies the type to which the game belongs. Furthermore, the description of the game in the above input data D1 is an example, and the input data D1 may include any information related to the user's behavior.

[0050] Step S102: The similar sentence mixing device 100 uses the similar sentence search program 2 to extract sentences similar to the input data D1 from the public information list 11. In the first embodiment, similar sentences are determined by finding the appearance tendency of words appearing in the input data D1 and the sentences in the public information list 11, and expressing this as a vector W = {w1, w2, ..., wN}, and determining that the larger the cosine value of the angle formed by the two numerical vectors, the greater the similarity (cosine similarity). However, the above method of determining similarity is just one example, and any other method can be adopted as described later.

[0051] In Example 1, for example, in the "skills and technology" aspect of automobile maintenance, evaluation criteria related to employment in each industry and each evaluation perspective, such as "understands the standard work for the job in question, such as processing, assembly, inspection, and maintenance, and performs the work in the correct way," are determined as similar sentences to the input data D1.

[0052] In step S102, vectorization is not limited to word frequency, as it is sufficient to vectorize a single sentence and search for the most similar one among multiple vectors. Another example is a sentence vectorization machine learning model that outputs a single vector when a single sentence is input. Similar sentences may also be determined using other methods of calculating the similarity between vectors, such as Euclidean distance. Furthermore, when using a sentence vectorization machine learning model, additional learning can be performed based on data approved as similar sentences or data requested for a similar sentence search again upon user confirmation in step S105, making it possible to search for similar sentences that are more suitable for the user. Additional learning will be described later (step S106).

[0053] The sentences included in the public information list 11 may be classified into multiple categories. For example, if the public information list 11 is a vocational ability assessment standard, the sentences may be classified by occupational field. In this case, in step S102, the similar sentence mixing device 100 may extract similar sentences based on the category corresponding to the input data D1.

[0054] For example, the similar sentence mixing device 100 may perform a similarity search only on sentences that belong to a category similar to the input data D1, or may weight the similarity between the input data D1 and a sentence included in the public information list 11 according to the similarity between the input data D1 and the category to which the sentence belongs.

[0055] The similarity between the input data D1 and a category may be determined by, for example, having the user specify the target occupational field when entering the input data D1, and by specifying the corresponding category, or may be determined automatically based on the words contained in the input data D1.

[0056] Step S103: The similar sentence mixing device 100 uses the mixed sentence generation program 3 for the processing target and similar sentences to input the input data D1 and the similar sentences obtained in step S102 into the mixed sentence generation model 12, and obtains one mixed sentence that is related to the output public information list 11.

[0057] In Example 1, a sentence related to an evaluation axis or viewpoint in employment, such as "The person understood the tasks in the game, such as processing and assembly, and performed the tasks in the correct way," is output. At this time, named entities included in the mixed sentence may be abstracted to protect privacy.

[0058] The details of the mixed sentence generation model 12 are not limited, but it may be a model that generates mixed sentences by at least one of replacing words included in the input data D1 with words included in similar sentences and adding words included in one of them to the other. For example, in the above sentence "He understood the in-game tasks, such as processing and assembly, and performed the tasks in the correct way," "processing, assembly, etc." and "in the correct way" originate from the similar sentence, "in the game" originates from the input data D1, and "was performing the tasks" originates from both the input data D1 and the similar sentence. In addition, in this case, if the tasks performed in the game correspond to "processing and assembly" but not to "inspection and maintenance," only "processing" and "assembly," which correspond to the tasks in the game, may be included in the mixed sentence from the similar sentences "processing, assembly, inspection, and maintenance."

[0059] Step S104: The similar sentence mixing device 100 displays the screen image 51 shown in FIG. 5 on the display 32 using the screen display program 4, thereby allowing the user to perceive the search results of the similar sentences and the generation results of the mixed sentence.

[0060] FIG. 5 is an explanatory diagram showing an example of a screen image displayed by the information processing device according to the embodiment of the present invention.

[0061] The screen image 51 includes an input sentence display section 55, a similar sentence display section 56, a mixed sentence display section 57, a similar sentence reexamination button 52, a mixed sentence reexamination button 53, and a mixed sentence send button 54.

[0062] The sentence input as input data D1 in step S101 is displayed in input sentence display section 55. In the example of Fig. 5, "Mr. A continued working at (game name) for two years while he was looking for a job" is displayed.

[0063] The similar sentences found in step S102 are displayed in the similar sentence display section 56. In the example of Fig. 5, "Understand the standard work for the work you are responsible for, such as processing, assembly, inspection, and maintenance, and perform the work in the correct way" is displayed.

[0064] The mixed sentence generated in step S103 is displayed in the mixed sentence display unit 57. In the example of Fig. 5, "The player understood the tasks in the game, such as processing and assembly, and performed the tasks in the correct way" is displayed. This is a mixture of "the tasks in the game" derived from the input sentence and "processing, assembly, etc." and "the player understood the tasks, and performed the tasks in the correct way" derived from similar sentences.

[0065] Referring again to FIG.

[0066] Step S105: If the user is dissatisfied with the search results of the similar sentences perceived in step S104, the user can press the similar sentence reconsider button 52. This causes the processing of the similar sentence mixing device 100 to proceed to step S102, excluding the current similar sentence from the public information list 11. Alternatively, the similar sentence mixing device 100 may display a plurality of similar sentences, from which the user can select the one they prefer.

[0067] Alternatively, if the user is dissatisfied with the mixed sentence generation result perceived in step S104, the user can press the mixed sentence reexamination button 53. This causes the processing of the similar sentence mixing device 100 to proceed to step S103. Alternatively, the similar sentence mixing device 100 may display a plurality of mixed sentences, from which the user can select a preferred one.

[0068] When the user determines that the mixed sentence generated in step S104 may be sent to another service provider, the user presses the mixed sentence send button 54, thereby moving the process to step S107.

[0069] At this time, the similar sentence mixing device 100 may perform additional learning of at least one of the model used for similar sentence search and the mixed sentence generation model 12 in step S106 before proceeding to step S107.

[0070] For example, when the mixed sentence send button 54 is operated, it is considered that the user has judged that the similar sentence displayed in the similar sentence display unit 56 as a result of the similar sentence search is appropriate for the input sentence displayed in the input sentence display unit 55 (i.e., a positive evaluation has been made). For this reason, the similar sentence mixing device 100 may update the model for similar sentence search by adding a combination of the input sentence and the corresponding similar sentence as a desirable output to training data and performing training. Alternatively, if the similar sentence reexamination button 52 was operated before the mixed sentence send button 54 was operated, the similar sentence displayed in the similar sentence display unit 56 at that time may be added to training data as an undesirable output and performing training.

[0071] Furthermore, when the mixed sentence send button 54 is operated, it is considered that the user has judged that the mixed sentence displayed in the mixed sentence display unit 57 as a result of mixed sentence generation is appropriate for the combination of the input sentence displayed in the input sentence display unit 55 and the similar sentence displayed in the similar sentence display unit 56 (i.e., a positive evaluation has been made). Therefore, the similar sentence mixing device 100 may update the mixed sentence generation model 12 by performing training by adding a combination of the input sentence, the similar sentence, and the mixed sentence corresponding to them as a desired output to training data. Alternatively, if the mixed sentence reexamination button 53 is operated before the mixed sentence send button 54 is operated, the mixed sentence displayed in the mixed sentence display unit 57 at that time may be added to training data as an undesirable output, and training may be performed.

[0072] Although performing the above additional learning is not essential, by doing so, it is possible to expect the generation of a highly accurate model that is suited to the user.

[0073] Step S107: The similar sentence mixing device 100 uses the mixed sentence sending program 5 and the communication interface 23 to send the mixed sentence to the analysis target receiving unit 7 via the communication network 33.

[0074] Step S108: The analysis device 200 receives the mixed sentence sent in step S107 via the communication network 33 using the analysis target receiving unit 7 and stores it.

[0075] Step S109: The analysis device 200 inputs the mixed sentence stored in step S108 into the analysis program 8 and outputs the analysis result. In the first embodiment, for example, an occupation suited to the person and the reason for it is output, such as "Occupation: production engineer in the manufacturing industry. Reason: production engineers in the manufacturing industry need to understand in detail the work processes on production lines and in factories and improve efficiency, so the ability to understand the work in games is useful."

[0076] Step S110: The analysis device 200 uses the analysis result transmission unit 9 to transmit the analysis results output in step S109 via the communication network 33 to the communication interface 23 that transmitted the data to be analyzed in step S107.

[0077] Step S111: The similar sentence mixing device 100 receives the analysis results sent via the communication network 33 in step S110 using the communication interface 23 and the analysis result screen display program 6, and stores them.

[0078] Step S112: The similar sentence mixing device 100 displays the analysis result stored in step S111 on the display 32 so that the user can perceive it.

[0079] When the process has been confirmed, various operational instructions are inputted through the input device 31, thereby ending the flow of FIG.

[0080] This mixes the input data D1, which is "private information," with similar sentences, which are "public information," and changes the meaning and structure of the entire sentence from the original "private information" to make it difficult to distinguish it from "public information." This then adds additional information, such as evaluation criteria and evaluation perspectives for employment, and makes it possible to output a suitable occupation for the individual and the reasons for it, without inputting the "private information" into a machine learning model published online. Furthermore, upon confirmation by the user in step S105, additional learning is performed in step S106 based on data approved as similar sentences and data for which a similar sentence search is requested again, leading to more suitable similar sentence search and mixed sentence generation, and occupation suggestions suited to the individual.

[0081] While the above disclosure has been described with respect to exemplary embodiments, those skilled in the art will recognize that various changes and modifications in form and detail are possible without departing from the spirit and scope of the disclosed subject matter. For example, while the above embodiment is intended to provide job suggestions to job seekers, this is merely an example, and the present invention can be applied to other applications, such as when private information, such as personal information or confidential information, is input into a publicly available analytical model (e.g., so-called generative AI) for analysis.

[0082] Furthermore, the system according to the embodiment of the present invention may be configured as follows.

[0083] (1) An information processing device (for example, a personal computer 20 realizing a similar sentence mixing device 100) having a processor (for example, a processor 21) and a memory device (for example, a main memory device 24 and an auxiliary memory device 25), wherein the memory device holds a plurality of publicly available sentences (for example, a public information list 11), and the processor acquires a target sentence including private information (for example, step S101), extracts similar sentences from the publicly available sentences that are similar to the target sentence (for example, step S102), generates a mixed sentence by mixing the target sentence and the similar sentence (for example, step S103), and outputs the mixed sentence.

[0084] This allows the same or better results to be achieved as when the "private information" is entered as is and analyzed without sending the "private information," including free text, to other service providers.

[0085] (2) In the information processing device described in (1) above, the storage device holds a mixed sentence generation model (e.g., mixed sentence generation model 12) that outputs the mixed sentence when the sentence to be processed and the similar sentence are input, and the processor generates the mixed sentence using the mixed sentence generation model (e.g., step S103).

[0086] This allows for proper mixed sentence generation.

[0087] (3) In the information processing device described in (2) above, when information indicating a positive evaluation of the output mixed sentence is input, the processor performs additional learning of the mixed sentence generation model by including the input sentence to be processed, the input similar sentence, and the output mixed sentence in training data (e.g., step S106).

[0088] This can improve the accuracy of the model.

[0089] (4) In the information processing device described in (2) above, the mixed sentence generation model is a model that has been trained to generate the mixed sentence by at least one of replacing a word contained in one sentence with a word contained in the other sentence, and adding a word contained in one sentence to the other sentence, when the sentence to be processed and the similar sentence are input.

[0090] This allows for proper mixed sentence generation.

[0091] (5) In the information processing device described in (1) above, the published sentences are classified into a plurality of categories, and the processor extracts the similar sentences from among the published sentences, a plurality of sentences that are classified into a category corresponding to the sentence to be processed.

[0092] This allows for proper mixed sentence generation.

[0093] (6) An information processing device as described in (1) above, wherein the private information includes information regarding a person's behavior, and the plurality of publicly available sentences each include information regarding evaluation criteria for ability based on the person's behavior.

[0094] This allows a person's capabilities to be analyzed without transmitting private information to other service providers.

[0095] (7) In the information processing device described in (1) above, the processor outputs the mixed sentence to an analysis device (e.g., analysis device 200) having a published analysis model (e.g., analysis model 13) (e.g., step S107), obtains the output of the analysis model in response to the input of the mixed sentence as an analysis result (e.g., step S111), and outputs the analysis result (e.g., step S112).

[0096] This allows analysis to be performed without transmitting private information to other service providers.

[0097] It should be noted that the present invention is not limited to the above-described embodiment, and includes various modifications. For example, the above-described embodiment has been described in detail to provide a better understanding of the present invention, and the present invention is not necessarily limited to an embodiment having all of the configurations described.

[0098] Furthermore, the above-described configurations, functions, processing units, processing means, etc. may be partially or entirely implemented in hardware, for example, by designing them as integrated circuits. The above-described configurations, functions, etc. may also be implemented in software, with a processor interpreting and executing a program that implements each function. Information such as the programs, tables, and files that implement each function can be stored in storage devices such as nonvolatile semiconductor memory, hard disk drives, and solid-state drives (SSDs), or in computer-readable, non-transitory data storage media such as IC cards, SD cards, and DVDs.

[0099] In addition, the control lines and information lines shown are those that are considered necessary for the explanation, and not all control lines and information lines in the product are necessarily shown. In reality, it can be considered that almost all components are interconnected. [Explanation of symbols]

[0100] 1... Input program to be processed 2...Similar sentence search program 3...Mixed sentence generation program for processing target and similar sentences 4...Screen display program 5...Mixed text sending program 6...Analysis result screen display program 7...Analysis target receiver 8...Analysis Program 9...Analysis results transmission section 11...Public Information List 12...Mixed sentence generation model 13...Analysis model 20...Personal computer 21...Processor 22...Device Interface 23...Communication interface 24…Main memory 25…Auxiliary storage device 26...Pass 31...Input device 32...Display 33...Communication Network 34...External device 100…Similar sentence mixing device 200…Analyzer

Claims

1. An information processing device having a processor and a storage device, The storage device holds a plurality of published sentences; The processor: Obtain a sentence to be processed that contains private information; extracting sentences similar to the processing target sentence from the plurality of published sentences; generating a mixed sentence by mixing the target sentence and the similar sentence; An information processing device that outputs the mixed sentence.

2. 2. The information processing device according to claim 1, the storage device holds a mixed sentence generation model that outputs the mixed sentence when the target sentence and the similar sentence are input; The information processing device is characterized in that the processor generates the mixed sentence using the mixed sentence generation model.

3. 3. The information processing device according to claim 2, When information indicating a positive evaluation of the output mixed sentence is input, the processor performs additional learning of the mixed sentence generation model by including the input sentence to be processed, the input similar sentence, and the output mixed sentence in training data.

4. 3. The information processing device according to claim 2, The information processing device is characterized in that the mixed sentence generation model is a model that has been trained to generate the mixed sentence by at least one of replacing words contained in one sentence with words contained in the other sentence when the sentence to be processed and the similar sentence are input, and adding words contained in one sentence to the other sentence.

5. 2. The information processing device according to claim 1, The published sentences are classified into a plurality of categories; The information processing device is characterized in that the processor extracts the similar sentences from among the plurality of publicly available sentences, a plurality of sentences that are classified into a category corresponding to the sentence to be processed.

6. 2. The information processing device according to claim 1, the non-public information includes information about the actions of a person; The information processing device is characterized in that each of the plurality of published sentences includes information regarding an evaluation standard for ability based on a person's behavior.

7. 2. The information processing device according to claim 1, The processor: outputting the mixed sentences to an analysis device having a published analysis model; An information processing apparatus characterized in that it acquires an output of the analysis model for the input of the mixed sentence as an analysis result, and outputs the analysis result.

8. An information processing method executed by an information processing device having a processor and a storage device, The storage device holds a plurality of published sentences; The information processing method includes: The processor obtains a sentence to be processed that includes private information; a step by the processor extracting sentences similar to the sentence to be processed from the plurality of published sentences; a step in which the processor generates a mixed sentence by mixing the target sentence and the similar sentence; and a step of the processor outputting the mixed sentence.

9. An information processing program for controlling an information processing device having a processor and a storage device, The storage device holds a plurality of published sentences; The information processing program obtaining a sentence to be processed that contains non-public information; extracting sentences similar to the target sentence from the plurality of published sentences; a step of generating a mixed sentence by mixing the target sentence and the similar sentence; and a procedure for outputting the mixed sentence.

Images