Information processing method and information processing program

Abstract

To convert unstructured data which is incompatible with a system into structured data which is compatible with the system.SOLUTION: The present invention relates to an information processing method capable of converting unstructured data incompatible with a system into structured data compatible with the system (1). In the method, a computer (2, 3) executes the processes of: receiving the unstructured data incompatible with the system (1); converting the modality of the unstructured data incompatible with the system into modality that a generative AI model can process; and converting the unstructured data whose modality is converted into structured data in a form compatible with the system using the generative AI model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing method and an information processing program. [Background technology]

[0002] BACKGROUND ART Conventionally, a workflow system that systematizes approval applications is known (see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2008-250557 Summary of the Invention [Problem to be solved by the invention]

[0004] As systems like these become outdated or easier-to-use systems appear, users periodically switch to new systems. If the data from the old system could be imported directly into the new system, users could easily switch systems, but when switching to systems from different manufacturers, there is often no data compatibility between the systems, and switching to a new system can be a significant hassle.

[0005] On the other hand, there are few systems that cannot export any system data. For example, in the case of the workflow system mentioned above, it was often possible to output at least data such as an image showing the workflow.

[0006] SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide an information processing method and an information processing program capable of converting unstructured data that is incompatible with a system into structured data that is compatible with the system. [Means for solving the problem]

[0007] One aspect of the present invention is an information processing method capable of converting unstructured data that is incompatible with a system into structured data that is compatible with the system, characterized in that a computer executes the following steps: accepting unstructured data that is incompatible with the system; converting the modality of the unstructured data that is incompatible with the system into a modality that can be processed by a generative AI model; and converting the unstructured data whose modality has been converted into structured data in a format that is compatible with the system by the generative AI model. [Effects of the Invention]

[0008] According to the present invention, unstructured data that is incompatible with the system can be converted into structured data that is compatible with the system. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a block diagram of an approval processing system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram showing unstructured data according to the present embodiment. [Figure 3] 10 is a diagram showing a flow of a process for converting unstructured data into structured data according to the present embodiment. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. <Configuration of the approval system> First, the configuration of an approval system 1 as a system according to the present embodiment will be described. As shown in Fig. 1, the approval system 1 is a workflow system that provides users with workflow services related to approval applications, and includes an approval processing server 2, a data conversion server 3, and a terminal device 5. The approval processing server 2, the data conversion server 3, and the terminal device 5 are communicably connected via a network 6 such as the Internet, and the workflow service related to the approval application is provided as a so-called cloud service when a user accesses the approval processing server 2 via the Internet.

[0011] The terminal device 5 is, for example, a user terminal such as a personal computer or a mobile terminal, and can access the approval request processing server 2 via a web browser or a dedicated application. The terminal device 5 is equipped with a display unit 51 such as a display, and is configured to be able to access the approval request processing server 2 to generate / update a workflow related to an approval request application and to actually submit an approval request application.

[0012] The approval processing server 2 is an approval processing system that can cooperate with the data conversion server 3 to import image data showing a workflow diagram and automatically generate a workflow route related to approval regulations, and is composed of multiple or a single computer equipped with a control unit 10, a storage unit 20, and a communication unit 30. The communication unit 30 is a communication interface for performing encrypted communication with the data conversion server 3 and terminal device 5 via the network 6.

[0013] The storage unit 20 stores information processing programs to be executed by the control unit 10, image data displaying the above-mentioned workflow, etc., and is composed of RAM (Random Access Memory), ROM (Read Only Memory), etc. More specifically, it includes a program storage unit 21, an unstructured data storage unit 22, and a structured data storage unit 23. In addition to the above-mentioned information processing programs, the program storage unit 21 also stores prompt data (described later) for communicating with the data conversion server 3, etc.

[0014] The unstructured data storage unit 22 stores unstructured data that is incompatible with this system. This unstructured data is data in an unimportable format whose contents cannot be read and incorporated by the approval system 1, and in this embodiment, this corresponds to, for example, image data displaying the workflow described above. The image data displaying the workflow is image data displaying the workflow route and conditions, and is exported or captured from the workflow system previously used by the user when switching to the approval system 1 according to this embodiment, for example.

[0015] Fixed form data compatible with the system is stored in the fixed form data storage unit 23. This fixed form data is data in an importable format whose contents can be read and incorporated by the approval system 1, and in this embodiment, for example, this corresponds to fixed form data obtained after image data displaying the above-mentioned workflow is converted by the data conversion server 3. More specifically, this corresponds to YAML data into which the image data displaying the above-mentioned workflow is converted.

[0016] The control unit 10 is composed of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), etc., and functions as a data acquisition unit 11, a workflow generation unit 14, a workflow update unit 15, an approval processing unit 16, and a screen generation unit 17 by executing programs stored in the program storage unit 21.

[0017] The data acquisition unit 11 is configured to acquire various types of data such as the above-mentioned unstructured data. Specifically, the data acquisition unit 11 receives data such as image data displaying a workflow uploaded from the terminal device 5, and stores the data in the unstructured data storage unit 22 of the storage unit 20.

[0018] The workflow generation unit 14 is configured to generate a workflow route. The workflow update unit 15 is configured to update the workflow generated by the workflow generation unit 14. The workflow generation unit 14 and workflow update unit 15 can each generate / update a workflow by having the user manually set branching conditions for the workflow route, but can also automatically generate / update a workflow based on fixed data converted by the data conversion server 3, for example, fixed data converted from image data displaying the above-mentioned workflow.

[0019] The approval processing unit 16 is configured to propose to the user an appropriate route from among the workflow routes stored in the memory unit 20 based on the contents of the approval request, and to circulate the approval request based on the selected workflow route.

[0020] The screen generation unit 17 is configured to generate various screen data to be displayed on the display unit 51 of the terminal device 5.

[0021] The data conversion server 3 is a generation AI server that can be used not only for data conversion but also for general purposes, and is composed of one or more computers, similar to the approval processing server 2. The data conversion server 3 is also equipped with a storage unit composed of RAM (Random Access Memory), ROM (Read Only Memory), etc., and a control unit composed of a CPU (Central Processing Unit), GPU (Graphics Processing Unit), etc., as the computer configuration, and the storage unit stores information processing programs such as a modality conversion program and an AI model. The control unit of the data conversion server 3 also functions as a modality conversion unit 12 and a data conversion unit 13 by executing the information processing program.

[0022] In this embodiment, a generative AI model and a large-scale language model that can be used for general purposes are used as the AI ​​model. However, this is not necessarily limited to this. For example, a dedicated AI model trained for data conversion, or an AI model trained from a generative AI model for data conversion may also be used. For example, GPT (generative pre-transfer)-2, GPT-3, GPT-4, Llama, etc. may be used as the above-mentioned AI model. In this embodiment, modality refers to information or the type (form) of information, and is a component necessary for the system to function. Modality also includes, for example, protocols, information hierarchies, inclusion relationships, and references.

[0023] The modality conversion unit 12 functions by the control unit executing the above-mentioned modality conversion program (modality conversion AI model), and is configured to convert the modality (form) of data acquired by the data acquisition unit 11 into a modality that can be imported by the data conversion unit 13 when the modality is different from a modality that can be processed by the data conversion unit 13. For example, in this embodiment, the generative AI model that functions as the data conversion unit 13 is a generative AI model capable of natural language processing. Therefore, when image data is acquired by the data acquisition unit 11, the modality of the acquired image data is converted into text data by the modality conversion unit 12.

[0024] The data conversion unit 13 functions by the control unit 10 executing the above-mentioned AI model, and is configured to be able to convert unstructured data into structured data. In the present embodiment, the data conversion server 3 is configured as a server separate from the approval processing server 2, but this is not limiting and it may be configured integrally with the approval processing server 2. In addition, the modality conversion unit 12 is configured as a plug-in for the generation AI that performs data conversion, but this is not necessarily limiting and it may be provided on the approval processing server 2 side, for example.

[0025] <How to import non-standard data> Next, an import processing method (information processing method) for unstructured data according to this embodiment will be described with reference to Figures 1 to 3. In the following description, image data displaying the workflow shown in Figure 2 is regarded as unstructured data, and the process of converting this image data into YAML (an example of structured data) and importing it into the approval processing server 2 will be described as an example of unstructured data import processing.

[0026] When switching from a previously used workflow system to the approval system 1 according to this embodiment, there are cases where data from the old system cannot be exported in a format that can be imported into the approval system 1. On the other hand, it is usually possible to export data in some format in any system, and in the case of a workflow system, an image showing the workflow used in the old system can usually be displayed on the screen, so the user can either export the image or, if this is not possible, capture it.

[0027] Therefore, if the migration data cannot be obtained in the form of fixed data, the user obtains image data that displays the workflow in the old system and uploads this image data, which is non-fixed data, as migration data to the approval processing server 2 via the terminal device 5 (step S1 in Figure 3).

[0028] When the image data is uploaded, the control unit 10 of the approval processing server 2 functions as the data acquisition unit 11, receives the uploaded image data, and stores it at a predetermined address in the storage unit 20 (step S2 in FIG. 3, a step of accepting unstructured data). Furthermore, after storing the image data, the control unit 10 transmits a predetermined prompt to the data conversion server 3 instructing it to convert the stored image data into YAML data (step S3 in FIG. 3). More specifically, the following prompt is input to the data conversion server 3:

[0029] Read the workflow from the given image and convert it to YAML. The image shows the workflow for approving an internal request. The YAML that configures the workflow must be output in the following format. Image: URL --- Workflow: Nodes: # The workflow is considered as a directed acyclic graph, and actions such as approvals and branches are represented as nodes. - ID: 1 # ID assigned to specify in Edge, automatically generated even if it does not appear in the image ApproverType: "Type of approver" #Attribute that indicates how to specify the approver, such as affiliation, position, or nomination Approvers: # List approvers if multiple approvers are configured - "Approver" ApproveAtLeast: 1 # Minimum number of approvers Edges: # Objects that connect each step of the flow, treating the workflow as a directed acyclic graph. If there is a conditional branch, it is represented on the edge as a passing condition. - SourceNodeId: "Connection source ID" DestinationNodeId: "Destination ID" Criteria: # Enter the condition when branching. Not required if there is no condition. Target: "Verification content" # Source of the value to be verified, such as application amount, applicant attributes, etc. Threashold: 1,000,000 # Number type condition value Description: "Over 1 million yen" # A brief explanation of the conditions NotNumberThreashold: "Manager" # Used as a non-numeric condition value when the applicant is at a certain position or higher Operator: "GTE" # Comparison methods such as equals, greater than or equal to

[0030] This prompt is an instruction to the generating AI model when converting unstructured data into structured data, and includes information such as the data format of the structured data to be converted, the type of information contained in the workflow of the input image data, and how to handle this information.

[0031] When the data conversion server 3 receives the prompt (step S4 in FIG. 3), it first converts the image data into a modality that can be processed by the generative AI model (step S5 in FIG. 3, modality conversion step). That is, since the generative AI model in this embodiment is a modality that can read text data, the data conversion server 3 converts the modality of the image data into text (step S5 in FIG. 3). More specifically, in the case of the image data in FIG. 2, the data conversion server 3 converts into text data the attributes (job title and name) of each applicant and approver in the workflow displayed on the image, branching conditions (contents and amount of the approval request), order of review, and other content.

[0032] Once the conversion to the above text data is complete, the data conversion server 3 uses the generation AI model to convert the converted text data into YAML format data that can be read by the workflow generation unit 14, in accordance with the contents of the prompt input from the approval processing server 2, and sends it to the approval processing server 2 (step S6 in Figure 3, data conversion process).

[0033] When the approval processing server 2 receives the data (fixed-form data) after the image data displaying the above-mentioned workflow has been converted into YAML format from the data conversion server 3 (step S7 in FIG. 3), it generates a workflow based on the received YAML format data (step S8 in FIG. 3). Then, it sends an image of the generated workflow to the terminal device 5, and the user checks the workflow diagram displayed on the display unit 51 of the terminal device 5 to check whether there are any errors in the generated workflow (step S9 in FIG. 3). If there are any errors in the generated workflow, the user manually corrects the errors and then registers the workflow as an official workflow in the approval processing server 2 (step S9 in FIG. 3).

[0034] In the above example, a new workflow was generated based on an image displaying the workflow. However, the generation AI of the data conversion server 3 can convert any data into fixed-format data as long as it is data that can be processed using natural language. Therefore, for example, instead of an image displaying a workflow, the approval processing server 2 can convert image data of an organizational chart or approval regulations into data in a format that allows the approval processing server 2 to generate a workflow, and the approval processing server 2 can then generate a workflow from that data. Furthermore, the image data of the organizational chart or approval regulations may be in a modality that at least one of the modalities cannot be processed by the generation AI, such as an image. For example, one of the modalities, such as the approval regulations, may originally be text data. Furthermore, these may be data in a format that combines text data and image data.

[0035] Furthermore, the information processing method according to this embodiment can update a workflow that already exists on the approval processing server 2 by converting the non-standard data related to the workflow update into standard data using a similar method and then importing it, even when updating a workflow that already exists on the approval processing server 2.

[0036] <Summary> (Configuration 1) An information processing method capable of converting unstructured data that is incompatible with a system into structured data that is compatible with the system, the method comprising: accepting unstructured data that is not compatible with the system; converting a modality of unstructured data that is incompatible with the system into a modality that can be processed by a generative AI model; converting the modality-converted unstructured data into structured data in a format compatible with the system using the generative AI model; An information processing method comprising:

[0037] In this way, by using a generative AI model to convert unstructured data that is incompatible with the system into structured data that is compatible with the system, it becomes possible to easily convert the unstructured data into structured data and import it into the system. Therefore, even when replacing an incompatible system, for example, data migration becomes possible, reducing the burden on users associated with system changes.

[0038] Furthermore, even if the modality of the unstructured data cannot be processed by the generative AI model, the modality can be changed and then converted to structured data using the generative AI model, thereby increasing the range (type) of unstructured data that can be converted. For example, image data displaying the workflow described above can also be converted to structured data. Note that the image data is not limited to image data output from a computer, and can also be data obtained by capturing an image handwritten by a user.

[0039] (Configuration 2) The non-formatted data that is incompatible with the system includes at least one of image data, audio data, and video data. 2. The information processing method according to configuration 1,

[0040] As described above, since unstructured data is converted into structured data using a generative AI model after converting the modality, not only image data as described in the above embodiment, but also audio data and video data can be converted into structured data using a generative AI after changing the modality. In this case, for example, for image conversion, an AI model that reads images and explains them in natural language is used, for audio, an AI model that transcribes audio is used, and for video, a video reading AI model is used.

[0041] (Configuration 3) The non-standard data that is not compatible with the system is image data, The generative AI model is a generative AI model capable of natural language processing, In the step of converting the image data into a modality processable by the generation AI, the image data is converted into text data. 3. The information processing method according to configuration 2.

[0042] In this way, by changing image data to text data when changing modality, data conversion to standardized data can be performed using a generative AI model capable of natural language processing.

[0043] (Configuration 4) the system is a workflow system capable of generating a workflow, the image data is image data of a workflow diagram, a step of generating a workflow by the computer using the standard data, 4. The information processing method according to any one of configurations 1 to 3.

[0044] Image data of a workflow diagram can be accurately converted to standard data and modality-changing. Therefore, the converted standard data can be used to generate workflows with relatively high accuracy. While the information processing according to this embodiment is suitable for use in workflow generation as described above, it can also be applied to any system that imports unformatted data. For example, it is suitable for importing personnel information or organizational information from image data, which requires the transfer of large amounts of data during system migration, for example, when importing images of documents such as PDF data, or for acquiring information from images that is difficult to scrape in an aggregation system. Furthermore, this information processing method is suitable not only for image information, but also for converting export data in a non-image format that is incompatible with other systems into standard data. Furthermore, because this information processing method can convert various data into a system-importable format, it can also generate test data before system release from easily editable resources such as images and spreadsheets.

[0045] Furthermore, in the above-described embodiment, YAML was used as an example of the format of the fixed data, but the data format of the fixed data is not limited to this and may be any format depending on the system. For example, a GraphQL schema may be provided as an output schema to the generation AI, and input values ​​to the linkage API may be output. Furthermore, table information of an RDBMS may be provided, and SQL such as INSERT / UPDATE may be created. Furthermore, the output format of the fixed data may be HTML, and the HTML may be used to provide a viewer function.

[0046] (Configuration 5) The system is a workflow system capable of generating a workflow related to an approval application, The image data includes image data of at least one of an organizational chart and a request for approval, The computer generates a workflow related to the approval request using the fixed form data. 4. The information processing method according to any one of configurations 1 to 3.

[0047] In the information processing method according to this embodiment, unstructured data is converted into structured data using a generative AI model, and by performing natural language processing on the organizational chart and approval regulations described above, information that is missing for workflow generation can be obtained from the organizational chart and approval regulations data, and the structured data necessary for the system to generate a workflow can be generated.

[0048] (Configuration 6) receiving updated image data by the computer; A step in which the computer converts the updated image data into the fixed data using the generative AI model; and updating the workflow by the computer based on fixed data into which the updated image data is converted. 6. The information processing method according to configuration 4 or 5.

[0049] In this way, not only when creating a new workflow, but also when updating the content of the workflow, the workflow can be easily updated by importing data in which image data has been converted to standard data.

[0050] (Configuration 7) Processing to accept unstructured data that is not compatible with the system, A process of converting unstructured data that is incompatible with the system into structured data in a format that is compatible with the system using a generative AI model; An information processing program characterized by causing a computer to execute the above.

[0051] The inventions described in the above-described embodiments may be combined in any manner. [Explanation of symbols]

[0052] 1: System (approval system), 2: Computer (approval processing server), 3: Computer (data conversion server)

Claims

1. An information processing method capable of converting unstructured data that is incompatible with a system into structured data that is compatible with the system, comprising: accepting unstructured data that is not compatible with the system; converting a modality of unstructured data that is incompatible with the system into a modality that can be processed by a generative AI model; and converting the modality-converted unstructured data into structured data in a format compatible with the system using the generative AI model.

1. An information processing method comprising:

2. The non-formatted data that is not compatible with the system includes at least one of image data, audio data, and video data.

2. The information processing method according to claim 1.

3. The non-standard data that is not compatible with the system is image data, The generative AI model is a generative AI model capable of natural language processing, In the step of converting the image data into a modality processable by the generating AI, the image data is converted into text data.

3. The information processing method according to claim 2.

4. the system is a workflow system capable of generating a workflow, the image data is image data of a workflow diagram, a step of generating a workflow by the computer using the standard data, 4. The information processing method according to claim 3.

5. The system is a workflow system capable of generating a workflow related to an approval application, The image data includes image data of at least one of an organizational chart and a request for approval, The computer generates a workflow related to the approval request using the fixed form data.

4. The information processing method according to claim 3.

6. receiving updated image data by the computer; The computer converts the updated image data into the fixed data using the generative AI model; and updating the workflow by the computer based on fixed data into which the updated image data is converted.

6. The information processing method according to claim 4 or 5.

7. Processing to accept unstructured data that is not compatible with the system, A process of converting unstructured data that is incompatible with the system into structured data in a format that is compatible with the system using a generative AI model; An information processing program characterized by causing a computer to execute the above.

Images