Information processing system and method thereof, and program and recording medium storing such program

Abstract

To provide an information processing system and a method thereof, and a program and a recording medium storing the program that can organically acquire information from generative AI and establish knowledge in any field, regardless of whether it is unknown or known.SOLUTION: An information processing system 1 for aggregating knowledge includes: an intra-organization information section 40 filled with one or more nodes having knowledge; a prompter 11 that, to fill the intra-organization information section 40 with nodes, generates a text prompt and also generates the nodes; a generative AI 20 that acquires the text prompt and generates response information; a main processing section 12 having a queue 13 that specifies the execution of the prompt, a node base 14 that stores the nodes generated from the response information by the prompter, and an execution log storage section 16 that stores an execution log; and an application section 30 that manages communication of the nodes between the main processing section 12 and the intra-organization information section 40. The information processing system executes a process of incorporating the nodes into the intra-organization information section 40 until the intra-organization information section 40 is completed.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to, for example, an information processing system, a method therefor, a program, and a recording medium storing the program, and in particular to an information processing system related to generative AI, a method therefor, a program, and a recording medium storing the program. [Background technology]

[0002] Modern society requires dramatic improvements in productivity in traditional industries through innovation, as well as the creation of industries that bring about completely new value.

[0003] Innovation in traditional industries and the creation of new industries both pose a challenge to business organizations into unknown fields, making it essential for them to acquire new knowledge.

[0004] On the other hand, the SECI (Socialization Externalization Combination Internalization) model is a model that represents the process of knowledge acquisition, as described in Non-Patent Document 1. This model shows how social knowledge is transformed and expanded in a spiral through the four quadrants of SECI. [Prior art documents] [Patent documents]

[0005] [Non-Patent Document 1] "The Knowledge-Creating Company" (1991) Summary of the Invention [Problem to be solved by the invention]

[0006] However, whether in an unknown field or an existing field, acquiring knowledge - that is, obtaining and consolidating information that is in line with a particular perspective or value system - requires basic knowledge, perseverance, proficiency, highly skilled and expensive human resources, as well as special support systems, and as a result, there is the problem that it is inevitably a low-productivity task.

[0007] The present invention was created by extracting the above-mentioned problems, and its objective is to provide an information processing system, a method and program for the system, and a recording medium storing the program, that can organically acquire information from generative AI and establish knowledge in any field, regardless of whether it is unknown or known. [Means for solving the problem]

[0008] In solving the above problem, the inventor first focused on the following points. That is, while Figure 20 is a model showing the social expansion of knowledge, it can be said to be a model of the expansion of knowledge in an organization in the sense that explicit knowledge exists. The Internet has developed tremendously, and numerous homepages and SNSs have been born on the Internet space, and a huge amount of information has been generated in the form of blogs, papers, etc. This huge amount of information is mainly in text format and lacks formality, and is knowledge that corresponds to the left side of the SECI 4 quadrant, that is, the area of ​​tacit knowledge. From the above, it is expected that the speed of this expansion can be accelerated if tacit knowledge on the Internet can be supplementarily incorporated into the SECI model, which is a model of knowledge expansion mainly in organizations. However, in order to do this, it is necessary to efficiently incorporate publicly accessible, that is, publicly known, information on the Internet space into the organization, but there was no appropriate method for doing so. The reason for this is that, as mentioned above, since they are text information, they lack connectivity, and there was no means to acquire information scattered across the vast Internet space according to a specific perspective. However, this situation changed dramatically in November 2022 with the birth of ChatGPT (registered trademark), a generative AI. As a result of being able to directly acquire and incorporate tacit knowledge in the Internet space into explicit knowledge in a form that can be converted into explicit knowledge and in line with a specific perspective or theme, it has become possible to accelerate the knowledge spiral in organizations through efficient acquisition of public knowledge. The present invention was conceived based on the above points.

[0009] Therefore, in order to solve the above problem, the information processing system of the first aspect of the present invention is characterized in that it comprises a prompter unit that generates a prompt for eliciting information from a generation AI and provides the prompt to the generation AI, receives response information from the generation AI that has been provided with the prompt, and generates a series of nodes from the response information that satisfy a node model, which is a format defined for each required task or application field, and a main processing unit that determines equivalence between an individual node in the series of nodes and an existing node already existing in the node base, and if the equivalence is confirmed, merges the individual node with the existing node that has equivalence.

[0010] Alternatively, in order to solve the above problem, an information processing system according to a second aspect of the present invention is characterized in that it comprises a prompter unit that generates a prompt for eliciting information from a generation AI and provides the prompt to the generation AI, and a main processing unit that receives response information from the generation AI that has been given the prompt, and based on the response information, generates a series of nodes that satisfy a node model, which is a type defined for each required task or application field, from the response information, determines equivalence between individual nodes in the series of nodes and existing nodes already existing in the node base, and if the equivalence is confirmed, merges the individual node with the existing node that has equivalence.

[0011] As a third aspect of the present invention, in the first or second aspect, the prompter unit may include a framework layer prompter located in a framework layer, which is a part that implements common processing, and having a function of launching the generation AI, and an application layer prompter located in an application layer, which is a part that implements requirements corresponding to individual application fields and / or application problems, and complementing the prompter.

[0012] As a fourth form of the present invention, in the first or second form, the prompter unit may include a definition information unit in an application layer, which is a part that implements requirements corresponding to individual application fields and / or application problems, in which definition information related to the individual application fields or application problems is specified, and an engine, which is a program in a framework layer, which is a part that implements common processing, having a function of operating the generation AI by reading the definition information obtained from the definition information unit.

[0013] As a fifth aspect of the present invention, in the first or second aspect, the main processing unit may include a framework layer main processing unit which is a part that executes main processing in a framework layer which is a part that implements common processing, and an application layer main processing unit which is a part that implements requirements corresponding to individual application fields and / or application problems and complements the framework layer main processing unit in an application layer which is a part that implements requirements corresponding to individual application fields and / or application problems.

[0014] As a sixth form of the present invention, in the first or second form, the main processing unit may include a definition information section in an application layer, which is a section that implements requirements corresponding to individual application fields and / or application problems, in which definition information related to the application fields or application problems is specified, and an engine, which is a program in a framework layer, which is a section that implements common processing, having a function of operating the main processing unit by reading the definition information obtained from the definition information section.

[0015] As a seventh aspect of the present invention, in the first or second aspect, an application unit may be further provided that provides the main processing unit with at least one of the functions of initialization, status acquisition, queue control, and event reception.

[0016] As an eighth aspect of the present invention, in the seventh aspect, the system may further include an internal information unit in which nodes generated by said main processing unit are aggregated via said application unit.

[0017] As a ninth aspect of the present invention, in the first or second aspect, the system may further comprise a generation AI for receiving the text prompt and generating response information.

[0018] As a tenth aspect of the present invention, in the seventh aspect, said application unit may include a history storage unit that stores past nodes in order to display past states on the screen of said interactive processing.

[0019] In order to solve the above problem, an information processing method according to an eleventh aspect of the present invention is characterized in that, in an information processing system having a prompter unit and a main processing unit, the prompter unit generates a prompt for eliciting information from a generation AI and provides the prompt to the generation AI, the prompter unit receives response information from the generation AI to which the prompt has been provided, and generates a series of nodes from the response information that satisfy a node model, which is a type defined for each required task or application field, and the main processing unit determines equivalence between an individual node in the series of nodes and an existing node already existing in a node base, and if the equivalence is confirmed, merges the individual node with the existing node having equivalence.

[0020] In addition, in order to solve the above problem, an information processing method according to a 12th aspect of the present invention is characterized in that, in an information processing system having a prompter unit and a main processing unit, the prompter unit generates a prompt for eliciting information from a generation AI and provides the prompt to the generation AI, a second step in the main processing unit receives response information from the generation AI to which the prompt has been given, and based on the response information, generates a series of nodes from the response information that satisfy a node model, which is a type specified for each required task or application field, and a third step in the main processing unit determines equivalence between an individual node in the series of nodes and an existing node already existing in a node base, and if the equivalence is confirmed, merges the individual node into the existing node having equivalence.

[0021] In order to solve the above problem, the program of the 13th aspect of the present invention is characterized in that it causes a computer to function as a prompter unit that generates a prompt to elicit information from a generation AI and provides the prompt to the generation AI, receives response information from the generation AI that has been provided with the prompt, and generates a series of nodes from the response information that satisfy a node model, which is a format defined for each required task or application field, and a main processing unit that determines equivalence between an individual node in the series of nodes and an existing node already existing in a node base, and if the equivalence is confirmed, merges the individual node with the existing node that has equivalence.

[0022] In addition, in order to solve the above-mentioned problems, a program related to a 14th aspect of the present invention is characterized in that it causes a computer to function as a prompter unit that generates a prompt to elicit information from a generation AI and gives the prompt to the generation AI, and a main processing unit that receives response information from the generation AI that has been given the prompt, and based on the response information, generates a series of nodes that satisfy a node model, which is a type defined for each required task or application field, from the response information, determines equivalence between individual nodes in the series of nodes and existing nodes already existing in the node base, and if the equivalence is confirmed, merges the individual node with the existing node that has equivalence.

[0023] As a fifteenth aspect of the present invention, the program according to the thirteenth or fourteenth aspect may be realized as a computer-readable recording medium storing therein the program. Effect of the Invention

[0024] According to each aspect of the present invention, it is possible to greatly contribute to providing a system that requires proficiency in data handling operations or that can replace knowledgeable personnel in carrying out business activities in unknown fields. [Brief description of the drawings]

[0025] [Figure 1] 1 is a schematic diagram showing an overview of an information processing system according to an embodiment of the present invention. [Diagram 2] FIG. 1 is illustrated by way of object-oriented programming techniques. [Diagram 3] The framework layer in the upper half of Figure 2 represents the part that implements the common processing in the present invention, and the application layer in the lower half is the part that implements the requirements corresponding to individual applications, and this figure shows the correspondence relationship. [Figure 4] 4 is a flowchart showing the operation of the information processing system according to the embodiment of the present invention. [Diagram 5]FIG. 2 is a diagram showing an example of a program for controlling the operation of an information processing system according to an embodiment of the present invention. [Figure 6] 1 is a matrix for business model generation according to one embodiment of the present invention. [Figure 7] FIG. 1 is a diagram showing a node model for business model generation in one embodiment of the present invention. [Figure 8] FIG. 2 is a diagram showing a data structure related to business model generation in one embodiment of the present invention. [Figure 9] FIG. 2 is a diagram illustrating the relationships between celebrities in an undirected graph according to an embodiment of the present invention. [Figure 10] FIG. 2 is a diagram showing the relationships between celebrities in a node model according to an embodiment of the present invention. [Figure 11] FIG. 2 is a diagram illustrating a node model related to celebrity associations in accordance with an embodiment of the present invention. [Figure 12] FIG. 2 is a diagram showing a data structure relating to celebrity associations according to an embodiment of the present invention. [Figure 13] FIG. 2 illustrates parameterization of a prompter according to an embodiment of the present invention. [Figure 14] FIG. 2 illustrates parameterization of a prompter according to an embodiment of the present invention. [Figure 15] FIG. 2 is a diagram illustrating a one-to-many node model according to an embodiment of the present invention. [Figure 16] FIG. 2 is a diagram showing a node model according to a hierarchical structure in one embodiment of the present invention. [Figure 17] FIG. 2 illustrates a node model according to a recursive hierarchical structure in accordance with an embodiment of the present invention. [Figure 18] FIG. 2 is a diagram illustrating a node model for a directed graph according to an embodiment of the present invention. [Figure 19] FIG. 2 is a diagram illustrating a node model for a directed acyclic graph according to an embodiment of the present invention. [Figure 20] FIG. 1 is a diagram showing a schematic configuration of the SECI model that forms the framework of the present invention. [Figure 21]FIG. 2 illustrates an example of a parameterized prompter according to an embodiment of the present invention. [Figure 22] FIG. 2 illustrates an example of a parameterized prompt according to one embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0026] Hereinafter, the embodiment of the present invention will be described with reference to the drawings. Note that the scope necessary for the description to achieve the object of the present invention will be shown in a schematic manner, and the scope necessary for the description of the relevant parts of the present invention will be mainly described, and the parts that are omitted will be described as being related to the publicly known technology.

[0027] <Overall composition> 1 is a conceptual block diagram for explaining an example of the configuration of an information processing system 1 according to one embodiment of the present invention. The information processing system 1 according to one embodiment of the present invention includes one or more computers (not shown), and the computer includes, as a hardware configuration, a CPU (Central Processing Unit), a ROM (Read-only Memory) that stores a control program or the like that runs on the CPU, and a RAM (none of which are shown) for temporarily storing various data (not shown).

[0028] As shown in FIG. 1, the information processing system 1 is an information processing system that aggregates knowledge. The information processing system 1 has, as functional block configuration, a processing unit 10 related to the core of the present invention, a generation AI 20, a prompter 11, a main processing unit 12, an application unit 30, an internal information unit 40, and a history storage unit 50. The main processing unit 12 has a queue 13, a node base 14, and an execution log storage unit 16. The application unit 30 has a display unit 31. The generation AI 20, the prompter 11, the main processing unit 12, the application unit 30, and the internal information unit 40 are communicatively connected by communication means (including wireless or wired). Note that only the processing unit 10 related to the core of the present invention can be regarded as a component of the present invention, and each unit other than the processing unit 10 can be regarded as forming the periphery of the present invention.

[0029] <Description of each component of information processing system 1> The generation AI 20 is a generation AI that receives a text prompt and generates response information, and may be installed in an organization's computer or an external web service. By using the generation AI 20, a user can ask or pose a question to the generation AI 20 using a text called a prompt. In the present invention, the user is not a person but a prompter. The prompter requests information via the API (Application Programming Interface) of the generation AI 20. In response, the generation AI 20 returns the requested information to the prompter in a format such as JSON.

[0030] The prompter 11 is given one specific node that already exists in the node base 14 from the main processing. The prompter 11 obtains necessary information from the given node according to its characteristics and generates text for the generation AI 20. Upon receiving a response from the generation AI 20, the prompter 11 returns to the main processing unit 12 multiple nodes that are connected to the given node and satisfy the conditions of the node model from the format, such as JSON, returned from the generation AI 20. A large number of prompters 11 can be prepared according to the application field of the present invention and easily used as needed.

[0031] As can be seen from Figure 2, a node type, which is a type of node, may be supported by multiple prompters at the same time. In the case of a value chain explained in the graph structure example in "Examples of use classified by node structure", we want to start with a specific node, for example the semiconductor industry, and obtain both the nodes for items and purchasers flowing upstream, and items and buyers flowing downstream. However, synthesizing the text for this purpose using the same prompter places a burden on the generation AI 20, and may prevent the generation AI 20 from understanding the intent of the request. In such a case, it is desirable to prepare two prompters corresponding to the downstream and upstream as mentioned above. For this reason, a prompter is made capable of supporting multiple node types.

[0032] The main processing unit 12 is configured to be capable of executing the main processing for realizing the utility of the present invention. The application programs of the present invention that bring convenience to the user will utilize this main processing unit 12. This main processing unit 12 is capable of supporting various IT (information technology) architectures. For example, the application unit 30 may function as a link with the internal information unit 40 of an organization, or as a server for interactive processing or batch processing. Although it depends on the architecture, the processing that the main processing unit 12 should implement in response to the application unit 30 is, for example, as follows: 1: Node-based initialization. 2: Get node status. 3: Queue control: Change the order of nodes in the queue 13, delete them, etc. 4: Event transmission: Events according to the progress of the execution process are transmitted to the application system.

[0033] The application unit 30 can be used for various purposes and forms, such as batch-like cooperation processing, conversational processing, etc. In particular, a form that utilizes a graphic user interface to display node changes in an easy-to-understand manner and executes the main processing unit 12 in real time in response to screen operations, etc., is highly convenient.

[0034] The internal information unit 40 is a system that already exists within the organization, or a newly developed system, or a web service such as SaaS (Software as a Service). There are various systems such as ERP (Enterprise Resource Planning), various business systems, and BI (Business Intelligence). By linking these systems with the present invention, it is possible to build an internal information unit 40 that actively uses external information in a timely manner. Note that the above-mentioned application unit 30 and this internal information unit 40 or web service, etc. may be conceived and developed simultaneously as one application unit 30.

[0035] <Explanation of components using object technology> The mechanism of the present invention has been devised by utilizing object-oriented programming technology in software engineering. Object-oriented programming technology is a concept in which a system is constructed by combining a large number of relatively small units (objects) that have the properties of both programs and data, rather than separating the entire system into so-called programs and data. Objects consist of programs and data, but data can further be divided into data for programs and data that is classified as knowledge or concepts, such as those managed in so-called databases.

[0036] In this invention, since special emphasis is placed on data that falls under the category of knowledge or concepts, the data is classified into program data and node data, and hereinafter, an object that functions as node data will be simply referred to as a node or a node object.

[0037] FIG. 2 illustrates FIG. 1 in the manner of object-oriented programming techniques.

[0038] The whole is divided into two layers, an upper layer and an lower layer. The upper framework layer 2 shows the parts that function regardless of the application of the present invention. Prompter 2b is an object that implements the functionality of a prompter. AI Adaptor: 2a acts as an intermediary between Prompter 2b and Generator AI 20 so that the prompter can be connected to various Generator AIs. This allows the same Prompter 2b to be connected to different Generator AIs 20.

[0039] NodeType 2e is an object that implements processing for each type of node. Each node of the node 2f is a unit for holding individual information handled by the present invention. NodeQueue 2g implements queues for each component.

[0040] Main 2h is a main program object that performs processing when the present invention is operated simply.

[0041] CellPrompter 3a is an object that implements a prompter function in Business Model Generation, which is a specific example of the information processing system 1. Cell 3b is a node that corresponds to the intersection (cell) of the matrix in business model generation. Influence 3e is the node that corresponds to "influence" in business model generation. Raw 3c is an object (non-node because it is not obtained from the generation AI) that corresponds to a row of the matrix in the business model generation. Column 3d is an object (non-node because it is not obtained from the generation AI) that corresponds to the column of the matrix in business model generation. MatrixMain 3f is an implementation of the application part in business model generation, which is a specific example of the information processing system 1.

[0042] In Figure 2, the framework layer, which is the upper half, represents the part that implements the common processing of the present invention. The application layer, which is the lower half, is the part that implements the requirements corresponding to individual applications. The correspondence is shown in Figure 3. There are two types of implementation methods for this type of framework method. One is a method called an engine, in which the framework layer reads definition information related to the application layer and operates based on that. The other is a method in which both the framework layer and the application layer are implemented and the application layer is injected into the framework layer. The explanation in this document focuses on the latter.

[0043] <Definition of terms related to this invention> A node is a unit of data with an identity, such as an individual person, an individual company, or a country. A node has multiple attributes and attribute values. Nodes are also connected to each other by links.

[0044] The distinctiveness of a node means that different nodes can be distinguished even if they have the same attributes or attribute values. For example, node 1, which has the name "Toyota Motor Corporation", and node 2, which also has the name "Toyota Motor Corporation", are different nodes because they are different nodes.

[0045] Node equivalence is a concept for regarding different nodes as substantially the same. In this method, the concept of equivalence is introduced, so a process for determining equivalence is necessary, and there are several methods for doing so. For example, the following are available: 1: The attribute values ​​considered to be names are exactly the same, such as "Toyota Motor Corporation" and "Toyota Motor Corporation". 2: The attribute values ​​considered to be names are similar, such as "Toyota Motors" and "Toyota." Such judgments can be made into algorithms. 3: If there is no attribute value that corresponds to the name, a comprehensive judgment is made from several attributes. This can also be made into an algorithm. 4: Have an external service, such as a generation AI, determine whether two nodes are equivalent. 5:Other.

[0046] A node model gives typing to nodes and links. In the case of press dies and press parts, the typing of the press dies makes it possible to produce a large number of parts with the same shape. In the same way, typing of a node model makes it possible to generate a large number of nodes with the same structural characteristics. A node model is an expression of typing. For example, the shape of a typical organizational chart is represented by a node model called organizational structure, which can be represented by a class diagram in UML (Unified Modeling Language). A class diagram can represent a structure in which there is only one upper organization and multiple lower organizations. Since typing such as "the hierarchy must be no more than four levels in total" cannot be represented in a diagram, it is represented by a statement called a constraint.

[0047] Node types are categories or concepts such as people, companies, and countries. The prompters described below are set for these node types, not for nodes.

[0048] A node base (NB) 14 is a container in which nodes exist according to the node model. Nodes in a node base 14 must satisfy the typing mentioned above, i.e., they must have the same structural characteristics.

[0049] The prompter 11 takes one node provided by the node base 14 as input, synthesizes a prompt text for the generation AI from it, and launches the generation AI 20. Furthermore, it generates a new series of nodes that satisfy the node model from data in JSON (JavaScript Object Notation) format or the like returned from the generation AI 20, and inputs them into the node base 14.

[0050] Now, let us explain about the queue 13. Each node in the node base becomes an input to the prompter, and the prompter returns multiple related nodes. In other words, when the main processing starts, a wait occurs in the processing of the node. A queue exists to keep the nodes in this waiting state waiting.

[0051] The execution log storage unit 41 stores various information when the main processing unit 12 executes the program. Its main purpose is to record exceptional events such as when a node acquired via a prompter does not satisfy a node model.

[0052] The history storage unit 50 is provided so that the application unit 30 can compare old and new reports when reporting node states at a certain time interval.

[0053] <Description of Operation of Information Processing System 1> The flow of the main processing is shown in Fig. 4, and each stage will be described. First, in step S100, the main processing unit 12 initializes the nodes and queue 13. That is, for example, in the case of business model generation, 27 nodes are prepared. In step S200, the main processing unit 12 processes the top node in the queue 13. In step S250, the main processing unit 12 acquires a plurality of prompters 11 set in the node type. In step S270, the main processing unit 12 executes the prompter 11 to execute the acquired new node.

[0054] In step S271, the main processing section 12 executes the prompter 11. In step S 273 , the main processing section 12 processes one new node returned from the prompter 11 . In step S275, the main processing unit 12 judges whether or not there is a node with the same value. If it is judged that there is a node with the same value, the process proceeds to step S277. If it is judged that there is no node with the same value, the process proceeds to step S279.

[0055] In step S277, the main processing unit 12 merges the new node with the equivalent node. In step S 279 , the main processing unit 12 adds the new node to the queue 13 . In step S281, the main processing unit 12 judges whether or not there is an unprocessed node. If it is judged that there is an unprocessed node, the process returns to step S273. If it is judged that there is no unprocessed node, the process proceeds to step S290.

[0056] In step S290, the main processing unit 12 determines whether or not there is an unexecuted prompter 11. If it is determined that there is an unexecuted prompter 11, the process returns to step S270. If it is determined that there is no unexecuted prompter, the process proceeds to step S300.

[0057] In step S300, the main processing unit 12 determines whether or not the termination condition is satisfied. If it is determined that the termination condition is satisfied, the process ends. If it is determined that the termination condition is not satisfied, the process returns to step S200.

[0058] Examples of termination conditions are shown below. 1: The queue is empty. 2: The limit on the number of tokens (the unit of information exchange with the generating AI 20) has been exceeded. 3: The prompt call limit has been exceeded. 4: The length of queue 13 has exceeded its limit. 5: A special exit condition is met.

[0059] Pseudocode: The pseudocode for the application part to collectively execute the processes of this method is shown in a Python-like pseudo language. It explains how the programs (methods) shown in each class shown in Figure 5 are called.

[0060] <Specific examples of use of information processing system 1> The following are some specific examples of application. Business model generation is a method used when an organization starts a new business or when searching for a new business that can be implemented by the organization. (See Business Model Generation, published by Shoeisha, 2012)

[0061] For its implementation, Business Model Generation recommends the use of a visualization tool called the Business Model Canvas, which lays out nine building blocks: 1:Customer Segments:Customer Segments 2: Value Propositions 3:Channels 4:Customer Relationships: 5: Revenue Streams: 6:Key Resources: Resources 7:Key Activities:Main activities 8:Key Partners:Partners 9: Cost Structure:

[0062] Write down the elements of the business model on each of these building blocks arranged on the Business Model Canvas and discuss them. In order to make the business model strategic, it is recommended to consider the impact of various business model environment factors on each of the above building blocks. Some of the business model environment factors classified as "important trends" are shown below. 1: Technology trends. 2: Regulatory trends. 3: Social and cultural trends.

[0063] With these three "important trends" on the horizontal axis and the above-mentioned building blocks on the vertical axis, the impact of business model environmental factors on the building blocks can be entered into the 27 cells of this matrix. Then, a strategic review of the business model can be performed using a matrix with multiple impacts entered in each cell. In addition, in FIG. 6, it is assumed that a pharmaceutical company is conducting a strategic review of this business model. Using this method, the elements in this cell can be automatically obtained from the generation AI 20. The prompts used at that time will be described later. The "impacts" of "social and cultural trends" on the "cost structure" of pharmaceutical companies obtained by the generation AI 20 (ChatGPT (registered trademark)) are the following four points. 1: Increased profit opportunities due to health consciousness 2: Biotech and personalized medicine will drive revenue growth 3: Digitalization of healthcare and increased revenue through telehealth 4: Revenue constraints due to medical expenses

[0064] Please refer to Figure 7. This is a node model of the above-mentioned method of the Business Model Canvas. Here, "*" represents many, and "1" represents one (similarly in other figures). There are many business environment factors (horizontal axis) 101, and many building blocks (vertical axis) 103, and the intersection is a cell (one). Multiple influences 109 obtained from the generation AI 20 are linked to this cell via a queue 105.

[0065] Please refer to Fig. 8. This is a specific node structure of the node model shown in Fig. 7. Impact 1 and impact 2 are obtained from the generation AI 20 and linked to cell 1 where the cost structure 111 and the technology trend 113 intersect. In addition, impact 3 and impact 4 are obtained from the generation AI 20 and linked to cell 2 where the cost structure 111 and the regulatory trend intersect. In addition, impact 119, impact 121, impact 123, and impact 125 are obtained from the generation AI 20 and linked to cell 2 where the cost structure 111 and the social / cultural trend 117 intersect.

[0066] 8 is passed to the prompter 11 by an instruction (explode) to call the generation AI 20 by the API, and the prompt that is actually sent to the generation AI 20 at that time is as follows: However, in the following, explanations regarding output format instructions such as JSON are omitted. "Describe how upcoming social and cultural trends will impact pharmaceutical companies' revenue streams."

[0067] In order to generate such a prompt for the repetitive processing of the main processing unit 12, the prompter may be parameterized as shown in FIG. TIFF2025021361000002.tif36116

[0068] Since this matrix should be created for one company or industry, input parameter 1 (&1) is fixed to "pharmaceutical company." The reason for parameterizing the company type here is to make the prompter 11 reusable.

[0069] The process ends by expanding all the nodes in the queue 13 as seed nodes. As a result, multiple effects that business environment factors have on the building blocks are associated with all the cells of the matrix. It is not necessary in this example to perform equivalence judgment and judge the equivalence of the node newly entered in the database 14 with the pre-existing node. As explained in <Example of use classified by node structure>, the node model in this example has a one-to-many association structure, so there is no need to judge equivalence regarding "influence". The vertical and horizontal axes passed from the outside in the above explanation can also be imported from the generation AI using the same one-to-many association. In that case, a process equivalent to the main process is used nested in the main processing unit 12.

[0070] Publicly known information such as personal connections between famous people can be obtained from the generation AI 20 and visualized by this method. A typical visualization format is as shown in FIG. 9. As a node model, it is appropriate to use a structure called an undirected graph as shown in FIG. 10. That is, for a person 131, there exist many relationships 133 between people (bars connecting people). On the other hand, when viewed from the relationships 133 between people, two people 131 are connected. And the person 131 is connected to a queue 135.

[0071] As an example of a node-based state, in Fig. 11, person AM exists on the node base, and person AM is further connected to a queue. If person AM is at the head of the queue 145 as shown in Fig. 12, person AM is passed to the prompter 11, and person MS and person YO are entered into the node base 14, and once verification is complete, person MS and person YO are entered into the node base 14. In this example, since it is considered that there is no equivalence between person AM, person MS, and person YO, person MS and person YO are added to the queue 145 in the next cycle. Thereafter, person MS and person YO are passed to the prompter 11 in sequence.

[0072] As an example of the prompt that is synthesized with the specifications of the prompter 11 to which the person AM in Figure 12 is passed, an example of the prompt text actually sent to the generation AI 20 (explanation of output format instructions such as JSON is omitted) is shown below. "Please tell me which celebrities are good friends with celebrity XXX." (XXX represents the name of any celebrity)

[0073] In order to generate such a prompt for the repetitive processing of the main processing unit 12, the prompter may be parameterized as shown in Fig. 13. The parameterized prompt is as shown in Fig. 22 below. TIFF2025021361000003.tif12138When the example in Figure 13 is substituted for each parameter, the prompt passed to the generation AI 20 will be as described above ("Please tell me which celebrities are friends with celebrity XXX."). Figure 14 shows an example of a prompt generated by combining the parameters in Figure 13. The equivalence between a node newly entered into the node base 14 and a pre-existing node is determined by the person's name.

[0074] <Examples of usage classified by node structure> As explained above, when applying the present invention, it is necessary to determine the node structure in the field of application. Any node structure can be adopted, from one that is highly universal to one that is specific to an organization. However, it is difficult to expect a generation AI that uses Internet information as learning material to be able to deal well with a specific node structure that is specific to an organization. Therefore, in the present invention, it is recommended to adopt a universal node structure. Examples of usage categorized using a universal node structure are described below.

[0075] A one-to-many relationship structure is shown in Figure 15. For node A, there are many node Bs. Conversely, for many node Bs, there is one node A. An example of the application of this node model is the aforementioned Business Model Generation. The Business Model Generation example uses a matrix structure, but the matrix structure is not a structure obtained through generative AI, but a structure learned and used from a book. In the Business Model Generation example, what is obtained from generative AI is each cell of the matrix (the relationship of the impact that a certain business model environmental factor has on the business model building block) and multiple actual "influences," which is a one-to-many relationship structure.

[0076] A node model of a hierarchical structure is shown in Figure 16 (a hierarchical structure based on serialized one-to-many associations). Many nodes B exist for one node A. Many nodes C exist for one of the nodes B. On the other hand, a recursive hierarchical structure 161 is shown in Figure 17. The one-to-many structure can be looped to create multiple hierarchical layers. Examples of applications include organizational structures, bills of materials, and WBS (work breakdown structures). This type of configuration information can also be obtained by this method by using an appropriate initial node. Note that the acquisition of nodes that make up a hierarchical structure using this method does not necessarily have to start from the top of the hierarchy. By implementing the concept of equivalence, it is also possible to acquire nodes from the bottom up in the hierarchy.

[0077] A graph structure is a structure that expresses the mutual relationships between multiple things. Below, we will further classify this structure and show examples to which this method can be applied. Graph structures are divided into undirected graphs and directed graphs, and undirected graphs have already been explained in the specific examples in Figures 9 to 14. Examples of application include a network diagram (see Figure 9) and geographical relationships such as between cities.

[0078] As shown in Fig. 18, a directed graph is a structure that can express multiple things including directionality. There are many edges 173 for a vertex 171. Directed graphs are further classified into cyclic directed graph structures and acyclic directed graph structures. A cyclic directed graph is a structure in which it is permissible to follow an edge from a vertex back to the same vertex, whereas an acyclic directed graph structure does not allow this.

[0079] An example of a cyclic directed graph structure is the visual representation of a value chain. A value chain is understood by diagramming the items that flow between organizations or industries. For example, semiconductors sold from the semiconductor industry may be turned into machinery and equipment and flow back into the semiconductor industry, so a representation using an acyclic directed graph is appropriate. In this way, this method can also be used to visualize value chains.

[0080] The structure of an acyclic directed graph is as in the node model in Figure 18 above, but with the added constraint that "vertices connected by edges must not be circular." An example of its application is the activity list, a concept used in project management. Activities have a chronological relationship as work, but because cycles are not possible, the use of an acyclic directed graph is appropriate.

[0081] A composite structure is a combination of the universal node models described above. The node models described above include basic node models such as one-to-many, so a composite structure can be called an arbitrary structure. An example of its application is a hierarchical acyclic directed graph.

[0082] As mentioned above, an acyclic directed graph can be used as a node model for an activity list, but when using a practical project management system, this activity often has a hierarchy in addition to a graph structure. Therefore, when using this method to input an activity list into such a system, it is better to use a node model like that shown in Figure 19, which has the characteristics of hierarchy and acyclic directed graph. However, even if you try to obtain a node that has both the properties of a recursive hierarchy and an acyclic directed graph using a generation AI, acyclicity and hierarchy may not be compatible. In this method, such node states are excluded by checking the node model, and it is possible to avoid a situation where an error occurs when inputting into a project management system.

[0083] Many different composite structures can be devised for this method, including a structure that overlaps a one-to-many structure and a recursive hierarchical structure, as shown in Figure 19. For example, it may be possible to group multiple articles suspected of being fake news and search for the news source from the bottom up in the hierarchy.

[0084] In the business model generation explained as a specific example of a one-to-many related structure, a matrix structure is used as the representation structure of the acquired nodes. This matrix structure is not acquired from the outside by the generation AI 20, but is used as it is as the initial data in the present invention, and is used in the present invention to acquire the "influence" node to be entered in the cell in the matrix structure.

[0085] On the other hand, it is also possible to obtain the matrix structure itself from the generation AI. The matrix structure itself can be obtained by requesting the prompt "Please tell me a matrix structure that can be used for XX analysis. Please tell me all the elements of the vertical axis (row) and the horizontal axis (column) for that purpose." It is also possible to obtain concepts and their elements that can be used for the vertical and horizontal axes in a certain field, and then select N items from them to create an N-ary (dimensional) relationship. In this case, N=2 is the matrix structure. The N-ary relationship constructed in this way can be used in combination with the various universal node structures described above. When N=2 and the universal node structure is a one-to-many structure, it is an example similar to the business model generation described above. When N=3, the matrix becomes a cube. When N>3, it is not possible to visually express the whole, so a projection of a three-dimensional space is used. In addition to the business model generation described above, the application of this method using the matrix structure can also be considered as the so-called SWOT (strengths, weaknesses, opportunities, threats) analysis and its application, cross analysis.

[0086] In the present invention, it is not necessarily a requirement that the internal information section 40 be filled with nodes.

[0087] Although each embodiment has been described above in detail, the present invention is not limited to the specific embodiment, and various modifications and changes are possible within the scope of the claims. In addition, it is also possible to combine all or a plurality of the components of the above-described embodiments. [Explanation of symbols]

[0088] 1. Information Processing Systems 11 Teleprompter 12 Main processing section 13. Queue 14 Node-based 16 Execution log 20 Generation AI 30 Application Section 31 Display section 40 Internal Information Department 50 History memory section

Claims

1. a prompter unit that generates a prompt for extracting information from a generation AI, provides the prompt to the generation AI, receives response information from the generation AI that has been provided with the prompt, and generates a series of nodes that satisfy a node model, which is a format defined for each requested task or application field, from the response information; a main processing unit that determines equivalence between an individual node in the set of nodes and an existing node already in the node base, and if equivalence is confirmed, merges the individual node with the existing node with which it has equivalence; An information processing system comprising:

2. a prompter unit that generates prompts for eliciting information from the generation AI and provides the prompts to the generation AI; a main processing unit that receives response information from the generation AI given the prompt, generates a series of nodes that satisfy a node model, which is a format defined for each required task or application field, from the response information based on the response information, determines equivalence between individual nodes in the series of nodes and existing nodes already existing in a node base, and, if the equivalence is confirmed, merges the individual node with the existing node having equivalence; An information processing system comprising:

3. The prompter unit includes: A framework layer prompter that is in a framework layer that implements common processing and has a function of starting the generation AI; An application layer prompter that is in an application layer that implements requirements corresponding to individual application fields and / or application problems and complements the prompter; 3. The information processing system according to claim 1, further comprising:

4. The prompter unit includes: A definition information section in an application layer that is a section for implementing requirements corresponding to individual application fields and / or application problems, in which definition information relating to the individual application fields or application problems is defined; An engine is a program in a framework layer that implements common processing and has a function of operating the generation AI by reading the definition information obtained from the definition information unit.

3. The information processing system according to claim 1, further comprising:

5. The main processing unit: A framework layer main processing unit that is a part that executes main processes in a framework layer that is a part that implements common processes; An application layer main processing unit that complements the framework layer main processing unit in an application layer that is a part that implements requirements corresponding to individual application fields and / or application problems; 3. The information processing system according to claim 1, further comprising:

6. The main processing unit: A definition information section in an application layer that is a section for implementing requirements corresponding to individual application fields and / or application problems, in which definition information relating to the application fields or application problems is defined; an engine, which is a program in a framework layer that implements common processing and has a function of operating the main processing unit by reading the definition information obtained from the definition information unit; 3. The information processing system according to claim 1, further comprising:

7. 3. The information processing system according to claim 1, further comprising an application section which provides at least one of the functions of initialization, status acquisition, queue control, and event reception to said main processing section.

8. 8. The information processing system according to claim 7, further comprising an internal information section in which the nodes generated by said main processing section are aggregated via said application section.

9. 3. The information processing system according to claim 1, further comprising a generation AI that receives the text prompt and generates response information.

10. 8. The information processing system according to claim 7, wherein said application unit includes a history storage unit that stores past nodes in order to display past states on said interactive processing screen.

11. In an information processing system having a prompter unit and a main processing unit, a first step of generating a prompt for eliciting information from a generating AI and providing the prompt to the generating AI in the prompter unit; a second step of receiving, in the prompter unit, response information from the generation AI given the prompt, and generating, from the response information, a series of nodes that satisfy a node model, which is a format defined for each required task or application field; a third step of determining, in the main processing unit, equivalence between an individual node in the series of nodes and an existing node already existing in the node base, and merging the individual node into the existing node with which it has equivalence if the equivalence is confirmed; An information processing method comprising:

12. In an information processing system having a prompter unit and a main processing unit, a first step of generating a prompt for eliciting information from a generating AI and providing the prompt to the generating AI in the prompter unit; A second step in which, in the main processing unit, response information is received from the generation AI given the prompt, and a series of nodes that satisfy a node model, which is a type defined for each required task or application field, is generated from the response information based on the response information; a third step of determining, in the main processing unit, equivalence between an individual node in the series of nodes and an existing node already existing in the node base, and merging the individual node into the existing node with which it has equivalence if the equivalence is confirmed; An information processing method comprising:

13. Computer, a prompter unit that generates a prompt for extracting information from a generation AI, provides the prompt to the generation AI, receives response information from the generation AI that has been provided with the prompt, and generates a series of nodes that satisfy a node model, which is a format defined for each requested task or application field, from the response information; a main processing unit that determines equivalence between an individual node in the set of nodes and an existing node already in the node base, and if equivalence is confirmed, merges the individual node with the existing node with which it has equivalence; A program characterized by causing the program to function as a

14. Computer, a prompter unit that generates prompts for eliciting information from the generation AI and provides the prompts to the generation AI; a main processing unit that receives response information from the generation AI given the prompt, generates a series of nodes that satisfy a node model, which is a format defined for each required task or application field, from the response information based on the response information, determines equivalence between individual nodes in the series of nodes and existing nodes already existing in a node base, and, if the equivalence is confirmed, merges the individual node with the existing node having equivalence; A program characterized by causing the program to function as a

15. 15. A computer-readable recording medium storing the information processing program according to claim 13 or 14.

Images