Information processing systems, information processing methods

Abstract

To provide a novel information processing system that is superior in convenience, usefulness, or reliability. [Solution] This is an information processing system composed of three components. The first component receives documents and sends them to the third component, and receives and provides messages. The second component receives instructions, generates descriptions and evaluations, and processes them using a large-scale language model. The third component receives and shares documents, descriptions, and evaluations, creates instructions, and sends them to the second component. It also receives messages and sends them to the first component. Furthermore, it includes subcomponents that perform processing using a database and management system, subcomponents that create instructions, and subcomponents that create messages based on evaluations.

Description

Technical Field

[0001] One aspect of the present invention relates to an information processing system, an information processing method, or a semiconductor device.

[0002] Note that one aspect of the present invention is not limited to the above technical field. The technical field of one aspect of the invention disclosed in this specification or the like relates to an article, a method, or a manufacturing method. Alternatively, one aspect of the present invention relates to a process, a machine, a manufacture, or a composition of matter. Therefore, more specifically, as the technical field of one aspect of the present invention disclosed in this specification, an information processing device, a semiconductor device, a storage device, a driving method thereof, or a manufacturing method thereof can be cited as an example.

Background Art

[0003] In recent years, the development of language models using neural networks has been actively carried out, and in particular, large language models (LLMs) have attracted attention. A large language model is a natural language processing model learned using a large amount of data. With a large language model, for example, a dialogue model that answers user instructions can be realized. In Non-Patent Document 1, GPT-4 (Generative Pre-trained Transformer 4) (registered trademark) is disclosed as a large language model, and ChatGPT is disclosed as a dialogue model.

[0004] By using a large language model, the capabilities of natural language processing models have been significantly improved. On the other hand, due to the increase in the size of language models, it is difficult to incorporate and operate a language model by oneself in terms of equipment and cost. Therefore, using an external service that provides a language model has become one form of using a language model.

Prior Art Documents

Non-Patent Documents

[0005] [Non-Patent Document 1] Summary of ChatGPT / GPT-4 Research and Perspective Towards the Future of Large Language Models, Yiheng Liu et al. (Submitted on 4 Apr 2023), [online], Internet<URL:https: / / arxiv.org / abs / 2304.01852> [Overview of the project] [Problems that the invention aims to solve]

[0006] One aspect of the present invention aims to provide a novel information processing system that is superior in convenience, usefulness, or reliability. Alternatively, it aims to provide a novel information processing method that is superior in convenience, usefulness, or reliability. Alternatively, it aims to provide a novel information processing system, a novel information processing method, or a novel semiconductor device.

[0007] Furthermore, the description of these problems does not preclude the existence of other problems. Moreover, one aspect of the present invention does not need to solve all of these problems. Other problems will naturally become apparent from the description in the specification, drawings, and claims, and it is possible to extract other problems from the description in the specification, drawings, and claims. [Means for solving the problem]

[0008] (1) One aspect of the present invention is an information processing system having a first component, a second component, and a third component.

[0009] The first component has the functionality to receive documents and send them to the third component, and the functionality to receive and provide messages.

[0010] The second component includes the function of receiving the first instruction and the second instruction and sending the description and evaluation to the third component, and the function of processing using a large-scale language model.

[0011] The third component includes the function of receiving documents and descriptions and sending the first instruction and the second instruction to the second component, and the function of receiving evaluations and sending messages to the first component.

[0012] The first instruction document includes the first instruction and document, and the first instruction includes a procedure for extracting factual statements from the document. The second instruction document includes the second instruction, description and public information, and the second instruction includes a procedure for generating an evaluation. The evaluation expresses the consistency between the public information and the description.

[0013] (2) One aspect of the present invention is an information processing system having a first component, a second component, and a third component.

[0014] The first component has the functionality to receive a first document and send it to the third component, and the functionality to receive and provide messages.

[0015] The second component includes the function of receiving the first instruction and the second instruction and sending the first description and the first evaluation to the third component, and the function of performing processing using a large-scale language model. The large-scale language model includes the function of generating the first description according to the first instruction and the function of generating the first evaluation according to the second instruction.

[0016] The third component includes the functions of receiving and sharing the first document, the first description, and the first evaluation within the third component, creating the first instruction and the second instruction and sending them to the second component, and creating a message and sending it to the first component. The third component also includes the first subcomponent, the second subcomponent, and the third subcomponent.

[0017] The first subcomponent includes functions for processing using a database and management system, and functions for retrieving public information and sharing it within the third component. The database has the function of storing public information sets, and the management system has the function of retrieving public information from the public information sets according to the first query.

[0018] The second subcomponent provides the functionality to generate a first directive, a first query, and a second directive. The first directive includes a first instruction and a first document, the first instruction including a procedure to extract a first statement, which is identified as a fact, from the first document. The first query includes a request to collect information containing the first statement from a set of publicly available information. The second directive includes a second instruction, a first statement, and publicly available information, the second instruction including a procedure to generate a first evaluation, the first evaluation expressing the consistency between the publicly available information and the first statement.

[0019] The third subcomponent has the functionality to create a message based on the first evaluation.

[0020] As a result, it is possible to evaluate the consistency between the first description disclosed as a fact in the first document and the public information. Also, for example, when the first evaluation is lower than the set evaluation, it is possible to provide a message prompting the user of the information processing system to pay attention. Also, it is possible to find out that the information processing system user or the corporation to which the information processing system user belongs is not a fact, has not been publicly disclosed, and has been disclosed by others. As a result, it is possible to provide a new information processing system excellent in convenience, usefulness, or reliability.

[0021] (3) Also, in one aspect of the present invention, the second component has a function of receiving a third instruction sentence and transmitting a second evaluation to a third component, and the large language model has a function of generating a second evaluation according to the third instruction sentence. It is the above information processing system.

[0022] The third component has a function of receiving the second evaluation and sharing it inside the third component, and a function of creating a third instruction sentence and transmitting it to the second component. The database has a function of storing a group of business information, and the management system has a function of acquiring business information from the group of business information according to the second query and sharing it inside the third component.

[0023] The second sub-component has a function of creating a second query and a third instruction sentence. The second query includes a request to collect information including the first description from the group of business information. The third instruction sentence includes a third instruction, the first description, and business information. The third instruction includes a procedure for generating a second evaluation, and the second evaluation represents the consistency between the business information and the first description.

[0024] The third sub-component has a function of creating a message based on the second evaluation.

[0025] As a result, it is possible to evaluate the consistency between the first description disclosed as a fact in the first document and the business information. As a result, it is possible to provide a novel information processing system that is excellent in convenience, usefulness, or reliability.

[0026] (4) Further, in one aspect of the present invention, when the first evaluation is lower than the set evaluation and the second evaluation is higher than the set evaluation, the message indicates that the information included in the business information group is included in the first description even though it is not included in the public information group. It is the above information processing system.

[0027] As a result, it is possible to evaluate whether the first document discloses unpublished information as a fact. In addition, it is possible to find a first document that discloses unpublished information as a fact and convey a message, for example, to a user of the information processing system. As a result, it is possible to provide a novel information processing system that is excellent in convenience, usefulness, or reliability.

[0028] (5) Further, in one aspect of the present invention, the second component has a function of receiving the fourth instruction text and transmitting the type to the third component, and the large language model has a function of classifying the first description into the type according to the fourth instruction text. It is the above information processing system.

[0029] The third component has a function of transmitting the fourth instruction text to the second component and a function of receiving the type and sharing it inside the third component.

[0030] The second sub-component has a function of creating the fourth instruction text. The fourth instruction text includes the fourth instruction and the first description, and the fourth instruction includes a procedure for classifying the first description into a type.

[0031] As a result, the first description disclosed as a fact in the first document can be classified into its type. As a result, it is possible to provide a novel information processing system that is excellent in convenience, usefulness, or reliability.

[0032] (6) Another aspect of the present invention is the above-described information processing system, wherein the second component has a function to receive a fifth instruction and transmit a first source to the third component, and the large-scale language model has a function to identify the first source according to the fifth instruction.

[0033] The third component has the function of sending the fifth instruction to the second component, and the function of receiving the first source and sharing it within the third component.

[0034] The second subcomponent has the function of creating a fifth instruction. The fifth instruction includes a fifth instruction, a first description, and a first document, and the fifth instruction includes a procedure for extracting the first source relating to the first description from the first document.

[0035] This allows the first source of the first statement, which is presented as a fact in the first document, to be extracted from the first document. As a result, a novel information processing system with superior convenience, usefulness, or reliability can be provided.

[0036] (7) Another aspect of the present invention is the above-described information processing system, wherein the database has a function for storing a knowledge graph, and the knowledge graph includes a first node and a second node.

[0037] The first subcomponent has the functionality to create a first node and a second node. Both the first node and the second node have a first label, which is assigned to the node that stores the description identified as fact.

[0038] The first node comprises a first attribute value pair capable of storing the second description, a second attribute value pair capable of storing a type, a third attribute value pair capable of storing information identifying the second document, and a fourth attribute value pair capable of storing information identifying the second source. The second description is a description stated as fact in the second document included in the publicly available information group, and the second source is a source related to the second description that is described in the second document.

[0039] The second node comprises a first label, a first attribute value pair capable of storing a first description, a second attribute value pair capable of storing a type, a third attribute value pair capable of storing information identifying a first document, and a fourth attribute value pair capable of storing information identifying a first source.

[0040] This allows a knowledge graph to represent a second statement identified as a fact in a second document included in a publicly available information set, and a first statement identified as a fact in another first document. Furthermore, for example, when a second statement identified as a fact is found in a second document included in a publicly available information set, its type, information identifying the second document, and information identifying the second source of the second statement extracted from the second document can be associated and registered in the knowledge graph. Similarly, when a first statement identified as a fact is found in a first document, its type, information identifying the first document, and information identifying the first source related to the first statement can be associated and registered in the knowledge graph. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0041] (8) Another aspect of the present invention is the above-described information processing system, wherein the first component has the function of receiving and providing a third node.

[0042] The third component has the ability to send the third node to the first component, and the management system has the ability to retrieve the third node from the knowledge graph according to the third query and share it within the third component.

[0043] The second subcomponent has the capability to create a third query. The third query includes a request to collect nodes classified into types from the knowledge graph to create a result set, and a request to collect from the result set third nodes whose fourth attribute-value pairs do not contain information that identifies documents in the result set.

[0044] This allows descriptions classified into categories to be collected in a result set. Furthermore, descriptions classified into a category in a second document included in the publicly available information set, and descriptions classified into the same category in the first document, can be collected in the result set. Additionally, using the information stored in the fourth attribute value pair, a third node that does not originate from a document containing descriptions of the same category can be found. Descriptions that do not originate from a document containing descriptions of the same category can also be found. Furthermore, documents containing such descriptions can be found. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0045] (9) Another aspect of the present invention is the above-described information processing system, wherein the first component has the function of receiving and providing hearsay paths. The hearsay path includes a chain of nodes, starting from a third node and connecting to a source node and a destination node.

[0046] The third component has the functionality to send hearsay paths to the first component. The management system has the functionality to retrieve hearsay paths from the knowledge graph according to the fourth query and share them within the third component.

[0047] The second subcomponent has the functionality to create a fourth query. The fourth query includes a request to collect a chain of source and destination nodes from the result set, starting from the third node, and the fourth attribute value pair of the destination node stores information that identifies the document stored in the third attribute value pair of the source node.

[0048] This allows us to identify hearsay paths originating from a third node within the result set. Furthermore, it enables us to identify the extent of hearsay paths originating from documents containing descriptions that do not originate from documents containing descriptions of the same type. As a result, we can provide a novel information processing system that is superior in terms of convenience, usefulness, and reliability.

[0049] (10) Another aspect of the present invention is the above-described information processing system, wherein a large-scale language model has the function of generating a third description in accordance with a first instruction.

[0050] The first instruction includes a procedure for extracting a third statement expressed as an opinion or commentary from the first document.

[0051] The knowledge graph includes a fourth node, and the first subcomponent has the function of creating the fourth node. The fourth node has a second label, which is assigned to the node that stores a description expressed as an opinion or commentary.

[0052] The fourth node comprises a fifth attribute value pair capable of storing a third description, a second attribute value pair capable of storing a type, and a third attribute value pair capable of storing information that identifies the first document.

[0053] This allows a third statement, expressed as an opinion or commentary, to be represented in a knowledge graph. Furthermore, for example, if a first document contains a third statement expressed as an opinion or commentary, its type and information identifying the first document can be associated and registered in the knowledge graph. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0054] (11) Another aspect of the present invention is the above-described information processing system, wherein the first component has the function of receiving and providing a third node.

[0055] The third component has the ability to send the third node to the first component, and the management system has the ability to retrieve the third node from the knowledge graph according to the fifth query and share it within the third component.

[0056] The second subcomponent has the functionality to create a fifth query. The fifth query includes a request to collect nodes classified into types from the knowledge graph and create a result set, and a request to collect a third node from the result set.

[0057] The third node has a first label, and the fourth node stores the information stored in the third attribute value pair of the fourth node, which has a second label.

[0058] This allows descriptions classified into categories to be collected in a result set. For example, if a first description, presented as a fact, and a third description, expressed as an opinion or commentary, are both classified into categories, both the first and third descriptions can be collected in the result set. Furthermore, a third node can be found where a document containing the third description, which is presented as an opinion or commentary but is classified into the same category, is used as the source. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0059] (12) One aspect of the present invention is an information processing method having steps 1 to 19.

[0060] In the first step, the first component receives the document and sends it to the second component.

[0061] In the second step, the second component receives the document and sends the first instruction to the third component.

[0062] The second component comprises the first subcomponent and the second subcomponent. The first subcomponent has the function of performing processing using a database and a management system, and the database has the function of storing public information and business information. The second subcomponent has the function of creating the first instruction statement. The first instruction statement includes the first instruction and a document, and the first instruction includes a procedure for extracting descriptions identified as facts from the document.

[0063] In the third step, the third component receives the first instruction and generates a description using a large-scale language model.

[0064] In the fourth step, the third component sends the description to the second component.

[0065] In the fifth step, the second component receives the description and shares it within the second component.

[0066] In the sixth step, the second subcomponent creates and shares the first query within the second component. The first query includes a request to collect information, including descriptions, from the publicly available information set.

[0067] In the seventh step, the first subcomponent uses a management system to retrieve public information from the database according to the first query and shares it within the second component.

[0068] In step eighth, the second subcomponent creates a second instruction and sends it to the third component. The second instruction includes a second instruction, a description, and public information, and the second instruction includes a procedure for generating a first evaluation. The first evaluation expresses the consistency between the public information and the description.

[0069] In the ninth step, the third component receives the second instruction and uses a large-scale language model to generate the first evaluation.

[0070] In the tenth step, the third component sends the first evaluation to the second component.

[0071] In step 11, the second component receives the first evaluation and shares it internally.

[0072] In step 12, the second subcomponent creates and shares a second query within the second component. The second query includes a request to collect information, including descriptions, from a set of business information.

[0073] In step 13, the first subcomponent uses a management system to retrieve business information from the database according to the second query and shares it within the second component.

[0074] In step 14, the second subcomponent creates a third instruction and sends it to the third component. The third instruction includes a third instruction, description, and business information, and the third instruction includes a procedure for generating a second evaluation. The second evaluation expresses the consistency between the business information and the description.

[0075] In step 15, the third component receives a third instruction and uses a large-scale language model to generate a second evaluation.

[0076] In step 16, the third component sends the second evaluation to the second component.

[0077] In step 17, the second component receives the second evaluation and shares it internally.

[0078] Furthermore, the second component comprises a third subcomponent, and the third subcomponent has the function of creating a message based on the first and second evaluations.

[0079] In step 18, the third subcomponent creates a message and sends it to the first component. If the first evaluation is lower than the set evaluation and the second evaluation is higher than the set evaluation, the message indicates that the description contains information that is included in the business information group, even though it is not included in the public information group.

[0080] In step 19, the first component receives and provides a message.

[0081] This allows us to evaluate whether a document presents unpublished information as fact. Furthermore, it enables us to identify documents that present unpublished information as fact and communicate the message, for example, to users of an information processing system. As a result, it becomes possible to provide a novel information processing method that is superior in terms of convenience, usefulness, or reliability.

[0082] (13) Another aspect of the present invention is the above-described information processing method, which has steps 20 to 28 following step 19.

[0083] In step 20, the second subcomponent creates a fourth instruction and sends it to the third component. The fourth instruction includes a fourth instruction and a description, the fourth instruction including a procedure for classifying the description into categories.

[0084] In step 21, the third component receives the fourth instruction and generates a typology using a large-scale language model.

[0085] In step 22, the third component sends the type to the second component.

[0086] In step 23, the second component accepts the type and shares it within the second component.

[0087] In step 24, the second component creates the fifth instruction and sends it to the third component. The fifth instruction includes the fifth instruction, description, and document, and the fifth instruction includes a procedure for extracting the source of the description from the document.

[0088] In step 25, the third component receives the fifth directive and uses a large-scale language model to identify the source.

[0089] In step 26, the third component submits the source to the second component.

[0090] In step 27, the second component accepts sources and shares them within the second component.

[0091] In step 28, the first subcomponent creates nodes. The database has the functionality to store the knowledge graph, and the knowledge graph contains nodes.

[0092] A node has a label, which is assigned to a node that stores a description identified as a fact. A node also has a label, a first attribute value pair that can store a description, a second attribute value pair that can store a type, a third attribute value pair that can store information that identifies the document, and a fourth attribute value pair that can store information that identifies the source.

[0093] This allows for the representation of factual statements in a document on a knowledge graph. Furthermore, for example, when a document contains factual statements, the type of statement, information identifying the document, and information identifying the source of the statement can be associated and registered in the knowledge graph. As a result, a novel information processing method with superior convenience, usefulness, and reliability can be provided. [Effects of the Invention]

[0094] One aspect of the present invention can provide a novel information processing system that is superior in convenience, usefulness, or reliability. Alternatively, it can provide a novel information processing method that is superior in convenience, usefulness, or reliability. Alternatively, it can provide a novel information processing system, a novel information processing method, or a novel semiconductor device.

[0095] Furthermore, the description of these effects does not preclude the existence of other effects. Moreover, one aspect of the present invention does not necessarily have to possess all of these effects. Other effects will naturally become apparent from the description in the specification, drawings, and claims, and it is possible to extract other effects from the description in the specification, drawings, and claims. [Brief explanation of the drawing]

[0096] [Figure 1] Figure 1 is a diagram illustrating the configuration of an information processing system according to an embodiment. [Figure 2] Figure 2 is a diagram illustrating the configuration of components used in the information processing system according to the embodiment. [Figure 3]Figures 3(A) to 3(C) illustrate the structure of instruction statements used in the information processing system according to the embodiment. [Figure 4] Figures 4(A) and 4(B) illustrate the structure of instruction statements used in the information processing system according to the embodiment. [Figure 5] Figures 5(A) to 5(C) illustrate the configuration of a knowledge graph used in an information processing system according to an embodiment. [Figure 6] Figures 6(A) to 6(C) illustrate the configuration of a knowledge graph used in an information processing system according to an embodiment. [Figure 7] Figures 7(A) to 7(D) illustrate the configuration of a knowledge graph used in an information processing system according to an embodiment. [Figure 8] Figures 8(A) to 8(C) illustrate the configuration of a knowledge graph used in an information processing system according to an embodiment. [Figure 9] Figure 9 is a diagram illustrating the configuration of an information processing device used in an information processing system according to an embodiment. [Figure 10] Figure 10 is a diagram illustrating an information processing method according to an embodiment. [Modes for carrying out the invention]

[0097] An information processing system according to one aspect of the present invention comprises a first component, a second component, and a third component.

[0098] The first component has the functionality to receive a first document and send it to the third component, and the functionality to receive and provide messages.

[0099] The second component includes the function of receiving the first instruction and the second instruction and sending the first description and the first evaluation to the third component, and the function of performing processing using a large-scale language model. The large-scale language model includes the function of generating the first description according to the first instruction and the function of generating the first evaluation according to the second instruction.

[0100] The third component includes the functions of receiving and sharing the first document, the first description, and the first evaluation within the third component, creating the first instruction and the second instruction and sending them to the second component, and creating a message and sending it to the first component. The third component also includes the first subcomponent, the second subcomponent, and the third subcomponent.

[0101] The first subcomponent includes functions for processing using a database and management system, and functions for retrieving public information and sharing it within the third component. The database has the function of storing the public information set, and the management system has the function of retrieving public information from the public information set according to the first query.

[0102] The second subcomponent has the functionality to create a first directive, a first query, and a second directive. The first directive includes a first instruction and a first document, and the first instruction includes a procedure to extract a first statement, which is identified as a fact, from the first document. The first query includes a request to collect information containing the first statement from a set of publicly available information. The second directive includes a second instruction, a first statement, and publicly available information, and the second instruction includes a procedure to generate a first evaluation. The first evaluation expresses the consistency between the publicly available information and the first statement.

[0103] The third subcomponent has the functionality to create a message based on the first evaluation.

[0104] This allows for an evaluation of the consistency between the first statement, which is presented as a fact in the first document, and publicly available information. Furthermore, for example, if the first evaluation is lower than a set evaluation, a cautionary message can be provided to users of the information processing system. As a result, a novel information processing system with superior convenience, usefulness, or reliability can be provided.

[0105] Embodiments will be described in detail with reference to the drawings. However, it will be readily apparent to those skilled in the art that the present invention is not limited to the following description, and that its form and details can be modified in various ways without departing from the spirit and scope of the present invention. Accordingly, the present invention is not to be interpreted as being limited to the contents of the embodiments shown below. In the configuration of the invention described below, the same reference numerals are used in common across different drawings for the same parts or parts having similar functions, and repeated descriptions are omitted.

[0106] In this specification, ordinal numbers such as "first," "second," etc., are used to avoid confusion of components and do not limit the number of components or the order of components (e.g., process order or layering order). Furthermore, even if a term does not have an ordinal number in this specification, an ordinal number may be added in the claims to avoid confusion of components. Even if a term has an ordinal number in this specification, a different ordinal number may be added in the claims. Even if a term has an ordinal number in this specification, an ordinal number may be omitted in the claims.

[0107] In the drawings attached to this specification, components are classified by function and shown as independent blocks in block diagrams. However, in reality, it is difficult to completely separate components by function, and a single component may be involved in multiple functions.

[0108] (Embodiment 1) In this embodiment, an information processing system according to one aspect of the present invention will be described with reference to Figures 1 to 9.

[0109] Figure 1 is a diagram illustrating the configuration of an information processing system according to one embodiment of the present invention.

[0110] Figure 2 is a diagram illustrating the configuration of components used in an information processing system according to one embodiment of the present invention.

[0111] Figures 3(A) to 3(C) illustrate the structure of instruction statements transmitted and received within an information processing system according to one embodiment of the present invention.

[0112] Figures 4(A) and 4(B) illustrate the structure of instruction statements transmitted and received within an information processing system according to one embodiment of the present invention.

[0113] Figure 5(A) is a diagram illustrating the configuration of a knowledge graph that can be used in an information processing system according to one embodiment of the present invention, and Figures 5(B) and 5(C) are diagrams illustrating the configuration of nodes stored in the knowledge graph shown in Figure 5(A).

[0114] Figure 6(A) is a diagram illustrating the configuration of a knowledge graph that can be used in an information processing system according to one embodiment of the present invention, and Figures 6(B) and 6(C) are diagrams illustrating the configuration of nodes stored in the knowledge graph shown in Figure 6(A).

[0115] Figure 7(A) is a diagram illustrating the configuration of a knowledge graph that can be used in an information processing system according to one embodiment of the present invention, and Figures 7(B) to 7(D) are diagrams illustrating the configuration of nodes stored in the knowledge graph shown in Figure 7(A).

[0116] Figure 8(A) is a diagram illustrating the configuration of a knowledge graph that can be used in an information processing system according to one embodiment of the present invention, and Figures 8(B) and 8(C) are diagrams illustrating the configuration of nodes stored in the knowledge graph shown in Figure 8(A).

[0117] Figure 9 is a block diagram illustrating the configuration of an information processing device that can be used in an information processing system according to one embodiment of the present invention.

[0118] <Example of an information processing system configuration 1> An information processing system according to one aspect of the present invention includes component 110, component 130, and component 120 (see Figure 1).

[0119] For example, the information processing devices that perform the functions of component 110, component 130, and component 120 each include a computing device and a communication device. Furthermore, these communication devices can be connected to each other via the network 51 to constitute an information processing system according to one embodiment of the present invention.

[0120] <Component 110 Configuration Example 1> Component 110 has the function of receiving document D(1) and sending it to component 120. It also has the function of receiving message msg and providing it, for example, to the user 99 of the information processing system. Specifically, it provides it to the user 99 of the information processing system using output devices such as display devices, speakers, printers, facsimile machines, and storage devices.

[0121] For example, a user 99 of the information processing system inputs document D(1) into component 110. Alternatively, for example, a user inputs a command to select and send document D(1) stored in a storage device into component 110. Specifically, the user 99 of the information processing system inputs into component 110 using an input device such as a keyboard, mouse, facsimile machine, or eye-tracking device.

[0122] For example, newspaper articles or blog posts about user 99 of the information processing system or the corporation to which user 99 belongs can be used in document D(1). Specifically, documents published by other individuals or corporations in the media, mass media (newspapers, television, magazines, etc.), internet media (web media (primary media, secondary media), social media), etc. can be used in document D(1).

[0123] <Component 130 Configuration Example 1> Component 130 has the function of receiving instruction statements Pt0 and Pt11(X1). It also has the function of sending description DF(X1) and evaluation Ev1(X1) to component 120, and the function of performing processing using the large-scale language model LLM.

[0124] 《Example Configuration of a Large-Scale Language Model (LLM) 1》 The large-scale language model LLM has the function of generating a description DF(X1) according to instruction Pt0. It also has the function of generating an evaluation Ev1(X1) according to instruction Pt11(X1).

[0125] Furthermore, large-scale language models such as GPT-3(registered trademark), GPT-3.5, GPT-4(registered trademark), LaMDA, Llama2, or Llama3 can be used in the large-scale language model LLM.

[0126] <Component 120 Configuration Example 1> Component 120 has the function of receiving document D(1), description DF(X1), and evaluation Ev1(X1) and sharing them internally within component 120. It also has the function of creating instruction statements Pt0 and Pt11(X1) and sending them to component 130, and the function of creating message msg and sending it to component 110.

[0127] Furthermore, component 120 includes subcomponents 120A, 120B, and 120C (see Figure 2). For the purposes of this specification, a configuration having one or more functions is referred to as a component or subcomponent.

[0128] 《Example Configuration of Subcomponent 120A》 Subcomponent 120A includes functions for processing using a database DB and a management system DBMS, and functions for sharing public information PI(X1) within component 120.

[0129] The database (DB) has the function of storing public information groups (PI), and the management system (DBMS) has the function of retrieving public information (PI(X1)) from the public information groups (PI) according to query Qu11(X1). For example, documents published by users 99 of the information processing system or the corporation to which users 99 belong can be used as public information (PI(X1)). In addition, archives of documents containing public information (PI(X1)) can be used as public information groups (PI).

[0130] 《Example Configuration of Subcomponent 120B》 Subcomponent 120B has the functionality to create directive Pt0, query Qu11(X1), and directive Pt11(X1).

[0131] [Example 1 of the structure of instruction statement Pt0] Instruction Pt0 includes instruction g0 and document D(1) (see Figure 3(A)). Instruction g0 includes a procedure for extracting a statement DF(X1) identified as a fact from document D(1). In this specification, a statement DF(X1) identified as a fact refers to a statement that identifies some matter as a fact.

[0132] For example, the document in the following paragraph can be used as instruction Pt0.

[0133] " Please extract only the sections from {Document D(1)} that describe facts. "

[0134] [Example of query Qu11(X1) configuration] Query Qu11(X1) includes a request to collect information containing the description DF(X1) from the public information group PI. For example, relevant information can be collected from the public information group PI using a full-text search engine, the index of the public information group PI, and the description DF(X1).

[0135] [Example of the structure of instruction statement Pt11(X1)] Instruction Pt11(X1) includes instruction g11, description DF(X1), and public information PI(X1) (see Figure 3(B)). Instruction g11 includes a procedure for generating evaluation Ev1(X1). Evaluation Ev1(X1) expresses the consistency between public information PI(X1) and description DF(X1). For example, if description DF(X1) quotes public information PI(X1) without modification, a high evaluation value can be used for evaluation Ev1(X1). Also, for example, if description DF(X1) contains information not found in public information PI(X1), a low evaluation value can be used for evaluation Ev1(X1). If description DF(X1) contains a description with an unclear basis, this can be noted in message msg. Also, if description DF(X1) contains an incorrect description, this can be noted in message msg.

[0136] For example, the document in the following paragraph can be used as instruction Pt11(X1).

[0137] " Please indicate whether {Description DF(X1)} is included in {Public Information PI(X1)}. Your answer should be: fully included (3 points), partially included (2 points), or not included at all (1 point). "

[0138] 《Example Configuration of Subcomponent 120C》 Subcomponent 120C has the function of creating a message msg based on evaluation Ev1(X1). For example, a message msg can be used to indicate that the description DF(X1) accurately conveys the public information PI(X1). Alternatively, a message msg can be used to indicate that the description DF(X1) is inaccurate.

[0139] This allows for the evaluation of the consistency between the description DF(X1) identified as a fact in document D(1) and the publicly available information PI(X1). Furthermore, for example, if the evaluation Ev1(X1) is lower than the set evaluation, a cautionary message (msg) can be provided to the user of the information processing system. Additionally, it is possible to detect instances where false or undisclosed information has been disclosed by others regarding the user of the information processing system or the corporation to which the user belongs. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0140] <Component 130 Configuration Example 2> Component 130 has the function of receiving instruction Pt12(X1). It also has the function of sending evaluation Ev2(X1) to component 120.

[0141] 《Example Configuration of a Large-Scale Language Model (LLM) 2》 The large-scale language model LLM has the function of generating evaluation Ev2(X1) according to the instruction Pt12(X1).

[0142] <Component 120 Configuration Example 2> Component 120 has the function to receive evaluation Ev2(X1) and share it internally. It also has the function to create instruction statement Pt12(X1) and send it to component 130.

[0143] 《Example Configuration of Subcomponent 120A 2》 Subcomponent 120A has the function of sharing business information BI(X1) within component 120.

[0144] The database (DB) has the function of storing business information groups (BI), and the management system (DBMS) has the function of retrieving business information BI(X1) from the business information group (BI) according to query Qu12(X1). For example, users 99 of the information processing system or the corporation to which users 99 belong can use documents created in their work as business information BI(X1). In addition, archives of documents containing business information BI(X1) can be used as business information group (BI). Note that business information group (BI) includes not only public information groups (PI) but also non-public documents.

[0145] 《Example Configuration of Subcomponent 120B 2》 Subcomponent 120B has the functionality to create queries Qu12(X1) and directives Pt12(X1).

[0146] [Example of Query Qu12 (X1) Configuration] Query Qu12(X1) includes a request to collect information containing the descriptive DF(X1) from the business information group BI. For example, relevant information can be collected from the business information group BI using a full-text search engine, the business information group BI index, and the descriptive DF(X1).

[0147] [Example of the structure of instruction statement Pt12(X1)] Instruction Pt12(X1) includes instruction g12, description DF(X1), and business information BI(X1) (see Figure 3(C)). Instruction g12 includes a procedure for generating evaluation Ev2(X1). Evaluation Ev2(X1) expresses the consistency between business information BI(X1) and description DF(X1). For example, if description DF(X1) directly quotes business information BI(X1) without modification, a high evaluation value can be used for evaluation Ev2(X1). Also, for example, if description DF(X1) contains information not present in business information BI(X1), a low evaluation value can be used for evaluation Ev2(X1). If description DF(X1) contains a description with an unclear basis, this can be noted in the message msg. Also, if description DF(X1) contains an incorrect description, this can be noted in the message msg.

[0148] For example, the document in the following paragraph can be used in instruction Pt12(X1).

[0149] " Please indicate whether {Description DF(X1)} is included in {Business Information BI(X1)}. Your answer should be: fully included (3 points), partially included (2 points), or not included at all (1 point). "

[0150] 《Example Configuration of Subcomponent 120C 2》 Subcomponent 120C has the function of creating a message msg based on evaluation Ev2(X1). For example, a message msg can be used to indicate that the description DF(X1) accurately conveys the business information BI(X1). Alternatively, a message msg can be used to indicate that the description DF(X1) is inaccurate.

[0151] This allows for the evaluation of the consistency between the description DF(X1) identified as fact in document D(1) and the business information BI(X1). As a result, a novel information processing system with superior convenience, usefulness, or reliability can be provided.

[0152] 《Example Configuration of Subcomponent 120C 3》 When evaluation Ev1(X1) is lower than the set evaluation and evaluation Ev2(X1) is higher than the set evaluation, subcomponent 120C can create a special message.

[0153] For example, message msg indicates that information included in the business information group BI is included in description DF(X1), even though it is not included in the public information group PI.

[0154] This allows us to evaluate whether document D(1) presents unpublished information as fact. Furthermore, it is possible to identify document D(1) that presents unpublished information as fact and communicate a message (msg) to, for example, the user of the information processing system. As a result, it is possible to provide a novel information processing system that is superior in terms of convenience, usefulness, and reliability.

[0155] <Component 130 Configuration Example 3> Component 130 has the function of receiving instruction Pt2(X1) and sending type Typ(i) to component 120.

[0156] 《Example 3 of the configuration of a large-scale language model (LLM)》 The large-scale language model (LLM) has the function of classifying a description DF(X1) into a type Typ(i) according to an instruction Pt2(X1). For example, multiple types can be provided, and the model can select the appropriate type.

[0157] Specifically, a user 99 of the information processing system can set multiple topics and use each topic as a category. If a single description contains multiple topics, that description will be classified into multiple categories. This type of classification is sometimes called multi-label classification. For example, in the case of an article about a business that conducts mail-order sales, the business's name, address, telephone number, sales price of goods or services, payment method and timing, delivery time of goods, rules regarding returns or exchanges, and other information necessary for each product can be used as categories to classify the description.

[0158] Furthermore, users 99 of the information processing system can also set the number of categories to be classified. Each description is classified into one of the categories. This type of classification is sometimes called multi-class classification.

[0159] <Component 120 Configuration Example 3> Component 120 has the function of sending instruction statement Pt2(X1) to component 130 and the function of receiving type Typ(i) and sharing it internally within component 120.

[0160] 《Example Configuration of Subcomponent 120B 3》 Subcomponent 120B has the function of creating instruction statement Pt2(X1).

[0161] [Example of the structure of instruction statement Pt2(X1)] Instruction Pt2(X1) includes instruction g2 and description DF(X1) (see Figure 4(A)). Instruction g2 includes a procedure for classifying description DF(X1) into type Typ(i).

[0162] For example, the document in the following paragraph can be used as instruction Pt2(X1).

[0163] " Please classify {Description DF(X1)} into one of the {Types}. "

[0164] This allows the description DF(X1) identified as a fact in document D(1) to be classified into its type Typ(i). As a result, a novel information processing system with superior convenience, usefulness, or reliability can be provided.

[0165] <Component 130 Configuration Example 4> Component 130 has the function of receiving instruction Pt3(X1) and sending source R(1) to component 120.

[0166] 《Example 4 of the configuration of a large-scale language model (LLM)》 The large-scale language model LLM has the function of identifying the source R(1) according to the instruction Pt3(X1).

[0167] <Component 120 Configuration Example 4> Component 120 has the function of sending instruction Pt3(X1) to component 130 and the function of receiving source R(1) and sharing it internally within component 120.

[0168] 《Example Configuration of Subcomponent 120B 4》 Subcomponent 120B has the function of creating instruction statement Pt3(X1).

[0169] [Example of the structure of instruction statement Pt3(X1)] Instruction Pt3(X1) includes instruction g3, description DF(X1), and document D(1) (see Figure 4(B)). Instruction g3 includes a procedure for extracting the source R(1) related to description DF(X1) from document D(1). For example, the source R(1) related to description DF(X1) is extracted from the description or footnotes in document D(1).

[0170] For example, the document in the following paragraph can be used as instruction Pt3(X1).

[0171] " Please extract the sources related to description DF(X1) from document D(1). "

[0172] This allows the source R(1) of the statement DF(X1) identified as a fact in document D(1) to be extracted from document D(1). As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0173] <Example of an information processing system configuration 2> An information processing system according to one aspect of the present invention includes a function for managing descriptions identified as facts using a knowledge graph (KG).

[0174] <Component 120 Configuration Example 5> Component 120 has a function for managing information transmitted and received within an information processing system according to one embodiment of the present invention.

[0175] 《Example Configuration of Subcomponent 120A 3》 The database DB of subcomponent 120A has the function of storing the knowledge graph KG. The knowledge graph KG includes nodes Nd(X0) and Nd(X1) (see Figure 5(A)).

[0176] Furthermore, subcomponent 120A has the function of creating nodes Nd(X0) and Nd(X1). Both nodes Nd(X0) and Nd(X1) have a label LDF (see Figures 5(B) and 5(C)). Note that the label LDF is assigned to the node that stores the description identified as fact.

[0177] [Example configuration of node Nd(X0)] Node Nd(X0) comprises an attribute-value pair AVP[DF] that can store a description DF(X0), an attribute-value pair AVP[Typ] that can store a type Typ(i), an attribute-value pair AVP[D] that can store information identifying document D(0), and an attribute-value pair AVP[R] that can store information identifying source R(0). Each attribute-value pair contains an attribute and its value. In addition, various types and structures of data, such as numbers, strings, and arrays, can be used as attribute values.

[0178] Description DF(X0) is a description identified as a fact in document D(0), which is included in the public information group PI.

[0179] Source R(0) is the source for description DF(X0) found in document D(0).

[0180] [Example configuration of node Nd(X1)] Node Nd(X1) comprises a label LDF, an attribute value pair AVP[DF] that can store a description DF(X1), an attribute value pair AVP[Typ] that can store a type Typ(i), an attribute value pair AVP[D] that can store information identifying document D(1), and an attribute value pair AVP[R] that can store information identifying source R(1).

[0181] This allows for the representation of a description DF(X0) identified as a fact in document D(0) included in the public information group PI, and a description DF(X1) identified as a fact in another document D(1), in the knowledge graph KG. Furthermore, for example, when document D(0) included in the public information group PI contains a description DF(X0) identified as a fact, its type Typ(i), information identifying document D(0), and information identifying the source R(0) of the description DF(X0) extracted from document D(0) can be associated and registered in the knowledge graph KG. Similarly, when document D(1) contains a description DF(X1) identified as a fact, its type Typ(i), information identifying document D(1), and information identifying the source R(1) related to the description DF(X1) can be associated and registered in the knowledge graph KG. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0182] <Component 110 Configuration Example 2> Component 110 has the function of receiving node Nd(Er0) and providing it, for example, to the user 99 of the information processing system (see Figure 1).

[0183] <Component 120 Configuration Example 6> Component 120 has the function of sending node Nd(Er0) to component 110.

[0184] 《Example Configuration of Subcomponent 120A 4》 The management system DBMS of subcomponent 120A has the functionality to retrieve node Nd(Er0) from the knowledge graph KG according to query Qu2 and share it within component 120 (see Figure 2).

[0185] 《Example Configuration of Subcomponent 120B 5》 Subcomponent 120B has the functionality to create query Qu2.

[0186] [Example of Query Qu2 configuration] Query Qu2 includes a request to collect nodes classified as type Typ(i) from the knowledge graph KG and create result set RS(i). In other words, query Qu2 can be used to collect nodes that store descriptions classified as type Typ(i) from multiple documents into result set RS(i). It can also collect information identifying the source of descriptions classified as type Typ(i) into result set RS(i). For example, nodes classified as type Typ(i) and nodes classified as type Typ(ii) can be illustrated as shown in Figure 6(A), and all nodes classified as type Typ(i) in Figure 6(A) are collected into result set RS(i).

[0187] Furthermore, query Qu2 includes a request to retrieve from result set RS(i) nodes Nd(Er0) whose attribute value pair AVP[R] does not contain information that identifies a document within result set RS(i). In other words, query Qu2 can be used to find nodes that do not originate from documents within result set RS(i) among the nodes that store descriptions classified under type Typ(i). For example, nodes Nd(X0), Nd(X1), and Nd(X2) all originate from nodes within result set RS(i) (nodes classified under type Typ(i)) (see Figures 5(B), 5(C), and 6(B)). In node Nd(X2), the attribute-value pair AVP[DF] stores the description DF(X2), the attribute-value pair AVP[Typ] stores the type Typ(i), the attribute-value pair AVP[D] stores information identifying document D(2), and the attribute-value pair AVP[R] stores information identifying document D(0). On the other hand, node Nd(Er0) uses a node that is not classified under type Typ(i) as its source (see Figures 6(A) and 6(C)). In node Nd(Er0), the attribute-value pair AVP[DF] stores the description DF(X10), the attribute-value pair AVP[Typ] stores the type Typ(i), the attribute-value pair AVP[D] stores information identifying document D(10), and the attribute-value pair AVP[R] stores information identifying source R(10). Furthermore, source R(10) is a document that does not contain any descriptions classified as type Typ(i).

[0188] This allows descriptions classified as type Typ(i) to be collected in result set RS(i). Furthermore, descriptions classified as type Typ(i) within document D(0) and descriptions classified as the same type Typ(i) within document D(1) can be collected in result set RS(i). Additionally, using the information stored in attribute value pair AVP[R], nodes Nd(Er0) that do not originate from documents containing descriptions of the same type Typ(i) can be found. Descriptions that do not originate from documents containing descriptions of the same type Typ(i) can also be found. Furthermore, documents containing such descriptions can be found. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0189] <Component 110 Configuration Example 3> Component 110 has the function of receiving a hearsay path HSP(X) and providing it, for example, to a user 99 of the information processing system (see Figure 1). The hearsay path HSP(X) starts at node Nd(Er0) and includes a chain of originating and destination nodes (see Figure 7(A)).

[0190] <Component 120 Configuration Example 7> Component 120 has the function of sending the hearsay path HSP(X) to component 110 (see Figure 1).

[0191] 《Example Configuration of Subcomponent 120A 5》 The management system DBMS of subcomponent 120A has the function of retrieving the hearsay path HSP(X) from the knowledge graph KG according to query Qu3 and sharing it within component 120 (see Figure 2).

[0192] 《Example Configuration of Subcomponent 120B 6》 Subcomponent 120B has the functionality to create query Qu3.

[0193] [Example of Query Qu3 configuration] Query Qu3 includes a request to collect the chain of source and destination nodes, starting from node Nd(Er0), from result set RS(i). The attribute value pair AVP[R] of the destination node stores information that identifies the document stored in the attribute value pair AVP[D] of the source node.

[0194] For example, in the case of a hearsay path HSP(X) starting from node Nd(Er0), the attribute value of the attribute value pair AVP[R] of the destination node Nd(Er1) is information that identifies document D(10), and the attribute value of the attribute value pair AVP[D] of the originating node Nd(Er0) is information that identifies document D(10) (see Figures 7(A), 7(B), and 7(C)). Note that the attribute value pair AVP[DF] of node Nd(Er1) stores the description DF(X11), the attribute value pair AVP[Typ] stores the type Typ(i), and the attribute value pair AVP[D] stores information that identifies document D(11). Furthermore, the attribute value of the attribute value pair AVP[R] of the connected node Nd(Er2) is information that identifies document D(11), and the attribute value of the attribute value pair AVP[D] of the connecting node Nd(Er1) is information that identifies document D(11) (see Figures 7(A), 7(C), and 7(D)). Note that the attribute value pair AVP[DF] of node Nd(Er2) stores the description DF(X12), the attribute value pair AVP[Typ] stores the type Typ(i), and the attribute value pair AVP[D] stores information that identifies document D(12).

[0195] This allows us to find the hearsay path HSP(X) starting from node Nd(Er0) in the result set RS(i). Furthermore, we can identify the extent of the hearsay path HSP(X) starting from documents containing descriptions that are not sourced from documents containing descriptions classified under the same type Typ(i). As a result, we can provide a novel information processing system that is superior in terms of convenience, usefulness, and reliability.

[0196] <Example of Information Processing System Configuration 3> An information processing system according to one aspect of the present invention includes a function that uses a knowledge graph (KG) to manage not only descriptions that are presented as facts, but also descriptions that are expressed as opinions or commentaries.

[0197] <Component 130 Configuration Example 5> Component 130 has the function to receive instruction statement Pt0. It also has the function to send description DO(Y1) to component 120 (see Figure 1).

[0198] 《Example Configuration of a Large-Scale Language Model (LLM) 1》 The large-scale language model (LLM) has the function of generating a description DO(Y1) according to the instruction Pt0.

[0199] <Component 120 Configuration Example 8> Component 120 has the function to receive a description DO(Y1) and share it internally. It also has the function to create an instruction statement Pt0 and send it to component 130.

[0200] 《Example Configuration of Subcomponent 120B 7》 Subcomponent 120B has the function of creating instruction statement Pt0.

[0201] [Example 2 of the structure of instruction statement Pt0] Instruction Pt0 includes instruction g0 and document D(1) (see Figure 3(A)). Instruction g0 includes a procedure for extracting from document D(1) a statement DF(X1) that is identified as a fact and a statement DO(Y1) that is expressed as an opinion or comment. In this specification, a statement DO(Y1) that is expressed as an opinion or comment refers to a statement that expresses some kind of opinion or comment.

[0202] For example, the document in the following paragraph can be used as instruction Pt0.

[0203] " Extract the sections from {Document D(1)} that describe facts and the sections that describe opinions or commentaries. "

[0204] 《Example Configuration of Subcomponent 120A 6》 The database DB of subcomponent 120A has the function of storing the knowledge graph KG. The knowledge graph KG includes node Nd(Y1) (see Figure 8(A)). In the figure, white circles represent nodes that store descriptions that are presented as facts, and hatched circles represent nodes that store descriptions that are expressed as opinions or commentaries.

[0205] Furthermore, subcomponent 120A has the function of creating node Nd(Y1). Node Nd(Y1) has a label LDO. Note that the label LDO is assigned to nodes that store descriptions expressed as opinions or commentaries.

[0206] [Example configuration of node Nd(Y1)] Node Nd(Y1) comprises an attribute-value pair AVP[DO] that can store the description DO(Y1), an attribute-value pair AVP[Typ] that can store the type Typ(i), and an attribute-value pair AVP[D] that can store information identifying document D(1) (see Figure 8(B)).

[0207] This allows descriptions DO(Y1) expressed as opinions or commentaries to be represented in the knowledge graph KG. Furthermore, for example, if document D(1) contains a description DO(Y1) expressed as an opinion or commentary, its type Typ(i) and information identifying document D(1) can be associated and registered in the knowledge graph KG. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0208] <Component 110 Configuration Example 4> Component 110 has the function of receiving node Nd(Er0) and providing it, for example, to the user 99 of the information processing system (see Figure 1).

[0209] <Component 120 Configuration Example 9> Component 120 has the function of sending node Nd(Er0) to component 110.

[0210] The management system (DBMS) has the functionality to retrieve node Nd(Er0) from the knowledge graph KG according to query Qu4 and share it within component 120 (see Figure 2).

[0211] 《Example Configuration of Subcomponent 120B 8》 Subcomponent 120B has the functionality to create query Qu4.

[0212] [Example of Query Qu4 configuration] Query Qu4 includes a request to collect nodes classified as type Typ(i) from the knowledge graph KG and create a result set RS(i). In other words, query Qu4 can be used to collect nodes containing descriptions classified as type Typ(i) from multiple documents into the result set RS(i). It can also collect information identifying the source of the descriptions classified as type Typ(i) into the result set RS(i).

[0213] Furthermore, query Qu4 includes a request to collect node Nd(Er0) from result set RS(i). Note that node Nd(Er0) has the label LDF and the attribute value pair AVP[R] stores the information stored in the attribute value pair AVP[D] of node Nd(Y1) which has the label LDO (see Figure 8(C)). In other words, query Qu4 can be used to find nodes that cite documents containing descriptions expressed as opinions or commentaries as sources from among the nodes that store descriptions classified as type Typ(i).

[0214] This allows descriptions classified under type Typ(i) to be collected in result set RS(i). For example, when both a description DF(X1) presented as a fact and a description DO(Y1) expressed as an opinion or commentary are classified under type Typ(i), both descriptions DF(X1) and DO(Y1) can be collected in result set RS(i). Furthermore, it is possible to find node Nd(Er0) where a document containing a description DO(Y1) that is also classified under type Typ(i) and expressed as an opinion or commentary, despite being a description DF(X1) presented as a fact, is used as the source. As a result, a novel information processing system with superior convenience, usefulness, and reliability can be provided.

[0215] <Example of an information processing system configuration 4> An information processing device 20 that can be used in an information processing system according to one embodiment of the present invention includes, for example, an input unit 21, a storage unit 22, a processing unit 23, an output unit 24, and a transmission line 25 (see Figure 9).

[0216] In the drawings attached to this specification, the components are classified by function and shown as independent blocks in the block diagram. However, in reality, it is difficult to completely separate the components by function, and one component may be involved in multiple functions. For example, a part of the processing unit 23 may function as the input unit 21. Also, one function may be involved in multiple components. For example, the processing performed by the processing unit 23 may be executed by different information processing devices depending on the processing.

[0217] Input section 21 The input unit 21 can receive data from outside the information processing device. For example, the input unit 21 can receive data via the network 51. Specifically, a device such as a personal computer equipped with a communication port or communication function can be used.

[0218] The input unit 21 supplies the received data to either or both of the storage unit 22 and the processing unit 23 via the transmission line 25.

[0219] 《Storage section 22》 The memory unit 22 has the function of storing the program executed by the processing unit 23. The memory unit 22 may also have the function of storing data generated by the processing unit 23 (for example, calculation results, analysis results, inference results), data received by the input unit 21, etc.

[0220] The storage unit 22 may have a database. The information processing device may also have a database separate from the storage unit 22. The information processing device may have the function to retrieve data from a database located outside the storage unit 22, outside the information processing device itself, or outside the information processing system. Furthermore, the information processing device may have the function to retrieve data from both its own database and an external database.

[0221] Either or both of the storage and / or file server can be used in the storage unit 22. Furthermore, a database recording the paths of files stored on the file server can be used in the storage unit 22.

[0222] The storage unit 22 includes at least one of volatile memory and non-volatile memory. Examples of volatile memory include DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory). Examples of non-volatile memory include ReRAM (Resistive Random Access Memory), PRAM (Phase Change Random Access Memory), FeRAM (Ferroelectric Random Access Memory), MRAM (Magnetoresistive Random Access Memory), and flash memory. The storage unit 22 may also include at least one of NOSRAM (registered trademark) and DOSRAM (registered trademark). The storage unit 22 may also include a recording media drive. Examples of recording media drives include hard disk drives (HDD) and solid state drives (SSD).

[0223] NOSRAM is an abbreviation for "Nonvolatile Oxide Semiconductor Random Access Memory (RAM)". NOSRAM is a type of memory where the memory cell is a 2-transistor (2T) or 3-transistor (3T) gain cell, and the transistors are transistors that use metal oxide in the channel formation region (also called OS transistors). OS transistors have an extremely small current flowing between the source and drain when off, i.e., a leakage current. By utilizing the characteristic of extremely low leakage current, NOSRAM can be used as a non-volatile memory by holding charge corresponding to the data within the memory cell. In particular, NOSRAM can read the stored data without destroying it (non-destructive read), making it suitable for computational processing that involves repeating data read operations a large amount. Because the data capacity of NOSRAM can be increased by stacking it, it can be used as a large-scale cache memory, main memory, or storage memory to improve the performance of semiconductor devices.

[0224] DOSRAM is an abbreviation for "Dynamic Oxide Semiconductor RAM," and refers to RAM with a 1T (transistor) 1C (capacitance) type memory cell. DOSRAM is a type of DRAM formed using OS transistors, and it is a memory that temporarily stores information sent from an external source. DOSRAM is a memory that takes advantage of the low off-current of OS transistors.

[0225] In this specification, "metal oxide" refers to an oxide of a metal in a broad sense. Metal oxides are classified into oxide insulators, oxide conductors (including transparent oxide conductors), oxide semiconductors (also called oxide semiconductors or simply OS), etc. For example, when a metal oxide is used in the semiconductor layer of a transistor, that metal oxide may be referred to as an oxide semiconductor.

[0226] The metal oxide in the channel-forming region preferably contains indium (In). When the metal oxide in the channel-forming region contains indium, the carrier mobility (electron mobility) of the OS transistor is increased. For example, indium oxide (InOx) or indium gallium zinc oxide (In-Ga-Zn oxide, also written as "IGZO") can be used in the channel-forming region. Furthermore, the metal oxide in the channel-forming region is preferably an oxide semiconductor containing element M. Element M is preferably at least one of aluminum (Al), gallium (Ga), and tin (Sn). Other elements applicable to element M include boron (B), silicon (Si), titanium (Ti), iron (Fe), nickel (Ni), germanium (Ge), yttrium (Y), zirconium (Zr), molybdenum (Mo), lanthanum (La), cerium (Ce), neodymium (Nd), hafnium (Hf), tantalum (Ta), and tungsten (W). However, in some cases, element M may be a combination of multiple elements as mentioned above. Element M is, for example, an element with a high bond energy with oxygen. For example, an element with a higher bond energy with oxygen than indium. Furthermore, the metal oxide containing the channel-forming region is preferably a metal oxide containing zinc (Zn). Metal oxides containing zinc may be more prone to crystallization.

[0227] The metal oxides present in the channel-forming regions are not limited to indium-containing metal oxides. For example, the metal oxides present in the channel-forming regions may be zinc-tin oxides, gallium-tin oxides, or other metal oxides that do not contain indium but contain zinc, gallium, or tin.

[0228] Processing Unit 23 The processing unit 23 has the function of performing calculations, analyses, and inferences using data supplied from either or both of the input unit 21 and the storage unit 22. The processing unit 23 can supply the generated data (e.g., calculation results, analysis results, inference results) to either or both of the storage unit 22 and the output unit 24.

[0229] The processing unit 23 has the function of acquiring data from the storage unit 22. The processing unit 23 may also have the function of recording or registering data in the storage unit 22.

[0230] The processing unit 23 may, for example, have an arithmetic circuit. The processing unit 23 may, for example, have a central processing unit (CPU). The processing unit 23 may also have a graphics processing unit (GPU). The processing unit 23 may also have a neural processing unit (NPU / neural network processing unit).

[0231] The processing unit 23 may have a microprocessor such as a DSP (Digital Signal Processor). The microprocessor can be implemented using a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array) or FPAA (Field Programmable Analog Array). The processing unit 23 may also have a quantum processor. The processing unit 23 can perform various data processing and program control by interpreting and executing instructions from various programs via the processor. Programs that can be executed by the processor are stored in at least one of the processor's memory area and the storage unit 22.

[0232] The processing unit 23 may have main memory. The main memory may include at least one of volatile memory such as RAM and non-volatile memory such as ROM (Read Only Memory). Furthermore, the main memory may include at least one of the above-mentioned NOSRAM and DOSRAM.

[0233] For RAM, for example, DRAM or SRAM is used, and a virtual memory space is allocated and used as the workspace for the processing unit 23. The operating system, application programs, program modules, program data, and lookup tables stored in the storage unit 22 are loaded into RAM for execution. These data, programs, and program modules loaded into RAM are each directly accessed and manipulated by the processing unit 23.

[0234] ROM can store BIOS (Basic Input / Output System) and firmware, etc., which do not require rewriting. Examples of ROM include mask ROM, OTPROM (One Time Programmable Read Only Memory), and EPROM (Erasable Programmable Read Only Memory). Examples of EPROM include UV-EPROM (Ultra-Violet Erasable Programmable Read Only Memory), which allows data to be erased by ultraviolet irradiation, EEPROM (Electrically Erasable Programmable Read Only Memory), and flash memory.

[0235] The processing unit 23 may have either or both an OS transistor and a transistor having silicon in its channel formation region (Si transistor).

[0236] The processing unit 23 preferably has an OS transistor. Because the OS transistor has an extremely small off-current, using the OS transistor as a switch to hold the charge (data) that has flowed into a capacitive element that functions as a memory element ensures that the data can be retained for a long period of time. By using this characteristic in at least one of the registers and cache memory of the processing unit, the processing unit can be operated only when necessary, and at other times the information of the previous processing is saved to the memory element, thereby turning off the processing unit. In other words, normally-off computing becomes possible, and the power consumption of the information processing system can be reduced.

[0237] It is preferable for information processing devices to use AI for at least some of their processing.

[0238] Information processing devices preferably utilize artificial neural networks (ANNs, also simply referred to as neural networks). Neural networks are implemented using circuits (hardware) or programs (software).

[0239] In this specification, the term "neural network" refers to any model that mimics the neural network of living organisms, determines the strength of connections between neurons through learning, and possesses problem-solving capabilities. A neural network has an input layer, an intermediate layer (hidden layer), and an output layer.

[0240] In this specification and other documents, when discussing neural networks, the process of determining the connection strength (also called weight coefficient) between neurons from existing information is sometimes referred to as "learning."

[0241] In this specification and other documents, the process of constructing a neural network using connection strengths obtained through learning and deriving new conclusions from it may be referred to as "inference."

[0242] Output section 24 The output unit 24 can output at least one of the calculation results, analysis results, and inference results from the processing unit 23 to the outside of the information processing device. For example, the output unit 24 can transmit data via the network 51. Specifically, a device such as a personal computer equipped with a communication port or communication function can be used. Alternatively, a device equipped with a communication function may be used for both the input unit 21 and the output unit 24.

[0243] Transmission line 25 The transmission line 25 has the function of transmitting data. Data can be transmitted and received between the input unit 21, the storage unit 22, the processing unit 23, and the output unit 24 via the transmission line 25. Specifically, an external bus, LAN, or the internet can be used as the transmission line 25.

[0244] This embodiment can be appropriately combined with other embodiments shown in this specification.

[0245] (Embodiment 2) In this embodiment, an information processing method according to one aspect of the present invention will be described with reference to Figure 10.

[0246] Figure 10 is a flowchart illustrating an information processing method according to one embodiment of the present invention.

[0247] <Example of information processing method 1> One aspect of the present invention is an information processing method having steps S1 to S19 (see Figure 10).

[0248] <Step S1> In step S1, component 110 receives document D(1) and sends it to component 120. For example, user 99 of the information processing system inputs document D(1) into component 110.

[0249] <Step S2> In step S2, component 120 receives document D(1) and sends instruction Pt0 to component 130. Component 120 includes subcomponent 120A and subcomponent 120B.

[0250] Subcomponent 120A has the functionality to perform processing using a database DB and a management system DBMS. The database DB also has the functionality to store public information groups PI and business information groups BI.

[0251] Subcomponent 120B has the function of creating instruction statement Pt0. Instruction statement Pt0 includes instruction g0 and document D(1), and instruction g0 includes a procedure for extracting the description DF(X1) identified as fact from document D(1).

[0252] <Step S3> In step S3, component 130 receives instruction Pt0 and generates description DF(X1) using the large-scale language model LLM.

[0253] <Step S4> In step S4, component 130 sends the description DF(X1) to component 120.

[0254] <Step S5> In step S5, component 120 receives the description DF(X1) and shares it internally within component 120.

[0255] <Step S6> In step S6, subcomponent 120B creates query Qu11(X1) and shares it internally within component 120. Query Qu11(X1) includes a request to collect information containing description DF(X1) from the public information set PI.

[0256] <Step S7> In step S7, the sub-component 120A uses the management system DBMS to obtain the public information PI(X1) from the database DB according to the query Qu11(X1) and share it internally within the component 120.

[0257] 〈Step S8〉 In step S8, the sub-component 120B creates the instruction Pt11(X1) and sends it to the component 130. The instruction Pt11(X1) includes the instruction g11, the description DF(X1), and the public information PI(X1). The instruction g11 includes the procedure for generating the evaluation Ev1(X1). Also, the evaluation Ev1(X1) represents the consistency between the public information PI(X1) and the description DF(X1).

[0258] 〈Step S9〉 In step S9, the component 130 receives the instruction Pt11(X1) and uses the large language model LLM to generate the evaluation Ev1(X1).

[0259] [[ID=!5]] 〈Step S10〉 In step S10, the component 130 sends the evaluation Ev1(X1) to the component 120.

[0260] 〈Step S11〉 In step S11, the component 120 receives the evaluation Ev1(X1) and shares it internally within the component 120.

[0261] 〈Step S12〉 In step S12, the sub-component 120B creates the query Qu12(X1) and shares it internally within the component 120. The query Qu12(X1) includes the request to collect the information including the description DF(X1) from the business information group BI.

[0262] 〈Step S13〉 In step S13, subcomponent 120A uses the management system DBMS to retrieve business information BI(X1) from the database DB according to query Qu12(X1) and shares it within component 120.

[0263] <Step S14> In step S14, subcomponent 120B creates instruction statement Pt12(X1) and sends it to component 130. Instruction statement Pt12(X1) includes instruction g12, description DF(X1), and business information BI(X1), and instruction g12 includes a procedure for generating evaluation Ev2(X1). Evaluation Ev2(X1) expresses the consistency between business information BI(X1) and description DF(X1).

[0264] <Step S15> In step S15, component 130 receives instruction Pt12(X1) and generates evaluation Ev2(X1) using the large-scale language model LLM.

[0265] <Step S16> In step S16, component 130 sends evaluation Ev2(X1) to component 120.

[0266] <Step S17> In step S17, component 120 receives evaluation Ev2(X1) and shares it internally. Component 120 also includes subcomponent 120C, which has the function of creating a message msg based on evaluation Ev1(X1) and evaluation Ev2(X1).

[0267] <Step S18> In step S18, subcomponent 120C creates message msg and sends it to component 110. If evaluation Ev1(X1) is lower than the set evaluation and evaluation Ev2(X1) is higher than the set evaluation, message msg indicates that information included in business information group BI is included in description DF(X1) even though it is not included in public information group PI.

[0268] <Step S19> In step S19, component 110 receives a message msg and provides it, for example, to a user of the information processing system.

[0269] This allows us to evaluate whether document D(1) presents unpublished information as fact. Furthermore, it is possible to identify document D(1) that presents unpublished information as fact and communicate a message (msg) to, for example, the user of an information processing system. As a result, a novel information processing method with superior convenience, usefulness, and reliability can be provided.

[0270] <Example of information processing method 2> An information processing method according to one aspect of the present invention is an information processing method having steps S20 to S28 following the above-described information processing method (see Figure 10).

[0271] <Step S20> In step S20, subcomponent 120B creates instruction statement Pt2(X1) and sends it to component 130. Instruction statement Pt2(X1) includes instruction g2 and description DF(X1), and instruction g2 includes a procedure for classifying description DF(X1).

[0272] <Step S21> In step S21, component 130 receives instruction Pt2(X1) and generates type Typ(i) using the large-scale language model LLM.

[0273] <Step S22> In step S22, component 130 sends type Typ(i) to component 120.

[0274] 〈Step S23〉 In step S23, component 120 receives type Typ(i) and shares it internally within component 120.

[0275] 〈Step S24〉 In step S24, sub-component 120B creates instruction Pt3(X1) and sends it to component 130. Note that instruction Pt3(X1) includes instruction g3, description DF(X1), and document D(1).

[0276] 〈Step S25〉 In step S25, component 130 receives instruction Pt3(X1) and uses the large language model LLM to identify source R(1).

[0277] 〈Step S26〉 In step S26, component 130 sends source R(1) to component 120.

[0278] 〈Step S27〉 In step S27, component 120 receives source R(1) and shares it internally within component 120.

[0279] 〈Step S28〉 In step S28, sub-component 120A creates node Nd(X1). Note that database DB has the function of storing knowledge graph KG.

[0280] Knowledge graph KG includes node Nd(X1), and node Nd(X1) has label LDF. Label LDF is assigned to nodes that store descriptions disclosed as facts.

[0281] Node Nd(X1) comprises a label LDF, an attribute value pair AVP[DF] that can store a description DF(X1), an attribute value pair AVP[Typ] that can store a type Typ(i), an attribute value pair AVP[D] that can store information identifying document D(1), and an attribute value pair AVP[R] that can store information identifying source R(1).

[0282] This allows the description DF(X1) identified as a fact in document D(1) to be represented in the knowledge graph KG. Furthermore, for example, when document D(1) contains the description DF(X1) identified as a fact, its type Typ(i), information identifying document D(1), and information identifying the source R(1) related to the description DF(X1) can be associated and registered in the knowledge graph KG. As a result, a novel information processing method with superior convenience, usefulness, and reliability can be provided.

[0283] This embodiment can be appropriately combined with other embodiments shown in this specification. [Explanation of Symbols]

[0284] AVP[D] Attribute Value Pair AVP[DF] Attribute Value Pair AVP[DO] Attribute Value Pair AVP[R] Attribute Value Pair AVP[Typ] Attribute-value pair BI Business Information Group Document D DB Database DBMS Management System g11 instructions g12 instructions HSP(X) Hearsay Channel KG Knowledge Graph LDF Label LDO label Large-Scale Language Model (LLM) msg message PI public information group RS(i) Result Set Typ(i) 20 Information Processing Devices 21 Input section 22 Memory section 23 Processing Unit 24 Output section 25 Transmission lines 51 Network 99 User 110 components 120 components 120A Subcomponent 120B Subcomponent 120C Subcomponents 130 components

Claims

1. The first component and The second component, It has a third component, The first component has the function of receiving documents and sending them to the third component, and the function of receiving and providing messages. The second component has the function of receiving the first instruction and the second instruction, and transmitting the description and evaluation to the third component. It has the capability to perform processing using a large-scale language model, The third component receives the document and the description, A function to send the first instruction and the second instruction to the second component, The system includes a function to receive the evaluation and send the message to the first component, The first instruction statement includes the first instruction and the document, The first instruction includes a procedure for extracting the statements identified as facts from the document, The second instruction statement includes the second instruction, the description and the public information, The second instruction includes a procedure for generating the evaluation, The aforementioned evaluation is an information processing system that expresses the consistency between the aforementioned publicly available information and the aforementioned description.

2. The first component and The second component, It has a third component, The first component includes a function to receive a first document and send it to the third component, and a function to receive and provide a message. The second component has the function of receiving the first instruction and the second instruction, and transmitting the first description and the first evaluation to the third component, It has the capability to perform processing using a large-scale language model, The large-scale language model comprises a function to generate a first description according to a first instruction, and a function to generate a first evaluation according to a second instruction, The third component has the function of receiving the first document, the first description, and the first evaluation, and sharing them within the third component, A function to create the first instruction and the second instruction and send them to the second component, The system includes a function for creating the aforementioned message and sending it to the first component, The third component comprises a first subcomponent, a second subcomponent, and a third subcomponent. The first subcomponent comprises a function for processing using a database and management system, and a function for acquiring public information and sharing it within the third component. The aforementioned database has the function of storing publicly available information, The management system includes a function to retrieve the public information from the public information group according to a first query, The second subcomponent has the function of creating the first instruction, the first query, and the second instruction, The first instruction statement includes the first instruction and the first document, The first instruction includes a procedure for extracting the first statement, which is presented as fact, from the first document. The first query includes a request to collect information containing the first description from the publicly available information set. The second instruction statement includes the second instruction, the first description, and the public information; The second instruction includes a procedure for generating the first evaluation, The first evaluation expresses the consistency between the publicly available information and the first description. The third subcomponent is an information processing system that has the function of creating the message based on the first evaluation.

3. The second component has the function of receiving a third instruction and transmitting a second evaluation to the third component. The large-scale language model has a function to generate the second evaluation according to the third instruction sentence, The third component includes a function to receive the second evaluation and share it internally, and a function to create the third instruction statement and send it to the second component. The aforementioned database has the function of storing business information sets, The management system has a function to retrieve business information from the business information group according to a second query and share it within the third component. The second subcomponent has the function of creating the second query and the third directive, The second query includes a request to collect information containing the first description from the business information group, The third instruction statement includes the third instruction, the first description, and the business information, The third instruction includes a procedure for generating the second evaluation, The second evaluation expresses the consistency between the business information and the first description. The information processing system according to claim 2, wherein the third subcomponent has a function to create the message based on the second evaluation.

4. When the first evaluation is lower than the set evaluation, and the second evaluation is higher than the set evaluation, The information processing system according to claim 3, wherein the message indicates that information included in the business information group is included in the first description, even though it is not included in the public information group.

5. The second component has the function of receiving a fourth instruction and transmitting the type to the third component. The large-scale language model has a function to classify the first description into the type according to the fourth instruction, The third component includes a function for transmitting the fourth instruction to the second component, and a function for receiving the type and sharing it internally within the third component. The second subcomponent has the function of creating the fourth instruction statement, The fourth instruction statement includes the fourth instruction and the first description, The information processing system according to claim 2, wherein the fourth instruction includes a step of classifying the first description into the type.

6. The second component has the function of receiving a fifth instruction and transmitting the first source to the third component. The large-scale language model has a function to identify the first source according to the fifth instruction, The third component includes a function for transmitting the fifth instruction to the second component, and a function for receiving the first source and sharing it internally within the third component. The second subcomponent has the function of creating the fifth instruction statement, The fifth instruction statement includes the fifth instruction, the first description, and the first document, The information processing system according to claim 5, wherein the fifth instruction includes a step of extracting the first source relating to the first description from the first document.

7. The aforementioned database has a function for storing knowledge graphs, The aforementioned knowledge graph includes a first node and a second node, The first subcomponent has the function of creating the first node and the second node, The first node and the second node each have a first label, The first label is assigned to the node that stores the description identified as fact, The first node is, A first attribute value pair that can store a second description, A second attribute value pair capable of storing the aforementioned type, A third attribute value pair that can store information identifying the second document, A fourth attribute value pair capable of storing information identifying a second source, The second description above is a description cited as fact in the second document included in the publicly available information set, The second source mentioned above is the source relating to the second description mentioned above, as stated in the second document. The second node is connected to the first label, The first attribute value pair capable of storing the first description, The second attribute value pair capable of storing the aforementioned type, A third attribute value pair capable of storing information identifying the first document, The information processing system according to claim 6, comprising: a fourth attribute value pair capable of storing information identifying the first source.

8. The first component has the function of receiving and providing a third node, The third component has a function to transmit the third node to the first component, The management system includes a function to retrieve the third node from the knowledge graph according to the third query and share it within the third component, The second subcomponent has the function of creating the third query, The information processing system according to claim 7, wherein the third query includes a request to collect nodes from the knowledge graph that are classified into the type and to create a result set, and a request to collect from the result set the third nodes whose fourth attribute value pairs do not contain information that identifies documents in the result set.

9. The first component has the function of receiving and providing hearsay channels, The aforementioned hearsay path starts at the third node and includes a chain of originating and destination nodes. The third component has a function to transmit the hearsay path to the first component, The management system includes a function to obtain the hearsay path from the knowledge graph according to the fourth query and share it within the third component, The second subcomponent has the function of creating the fourth query, The fourth query includes a request to collect from the result set a chain of the source node and the destination node, starting from the third node. The information processing system according to claim 8, wherein the fourth attribute value pair of the destination node stores information that identifies the document stored in the third attribute value pair of the source node.

10. The large-scale language model has a function to generate a third description according to the first instruction, The first instruction includes a procedure for extracting the third statement expressed as an opinion or commentary from the first document, The aforementioned knowledge graph includes a fourth node, The first subcomponent has the function of creating the fourth node, The fourth node is provided with a second label, The second label is assigned to a node that stores a description expressed as an opinion or commentary. The fourth node is, A fifth attribute value pair capable of storing the third description, The second attribute value pair capable of storing the aforementioned type, The information processing system according to claim 7, comprising: a third attribute value pair capable of storing information that identifies the first document.

11. The first component has the function of receiving and providing a third node, The third component has a function to transmit the third node to the first component, The management system includes a function to retrieve the third node from the knowledge graph according to a fifth query and share it within the third component, The second subcomponent has the function of creating the fifth query, The fifth query is a request to collect nodes classified into the type from the knowledge graph and create a result set, The request includes collecting the third node from the result set, The third node is provided with the first label, The information processing system according to claim 10, wherein the fourth attribute value pair stores information stored in the third attribute value pair of the fourth node having the second label.

12. An information processing method comprising the first to ninth steps, In the first step described above, the first component receives a document and transmits it to the second component. In the second step, the second component receives the document and transmits the first instruction to the third component. The second component comprises a first subcomponent and a second subcomponent, The first subcomponent is equipped with a function to perform processing using a database and a management system, The aforementioned database has the function of storing public information and business information, The second subcomponent has a function for creating the first instruction statement, The first instruction statement includes the first instruction and the document, The first instruction includes a procedure for extracting statements identified as facts from the document, In the third step, the third component receives the first instruction and generates the description using a large-scale language model. In the fourth step, the third component transmits the description to the second component, In the fifth step, the second component receives the description and shares it within the second component, In the sixth step, the second subcomponent creates a first query and shares it within the second component. The first query includes a request to collect information containing the description from the publicly available information set, In the seventh step, the first subcomponent uses the management system to retrieve public information from the database according to the first query and shares it within the second component. In the eighth step, the second subcomponent creates a second instruction and sends it to the third component. The second instruction statement includes the second instruction, the description and the public information, The second instruction includes a procedure for generating the first evaluation, The first evaluation above expresses the consistency between the publicly available information and the description, In the ninth step, the third component receives the second instruction and uses the large-scale language model to generate the first evaluation. In the tenth step, the third component transmits the first evaluation to the second component. In the 11th step, the second component receives the first evaluation and shares it internally within the second component. In the twelfth step, the second subcomponent creates a second query which is shared within the second component. The second query includes a request to collect information containing the description from the business information group, In the 13th step, the first subcomponent uses the management system to retrieve business information from the database according to the second query and shares it within the second component. In the 14th step, the second subcomponent creates a third instruction and sends it to the third component. The third instruction statement includes the third instruction, the description, and the business information, The third instruction includes a procedure for generating a second evaluation, The second evaluation above expresses the consistency between the business information and the description, In the 15th step, the third component receives the third instruction and generates the second evaluation using the large-scale language model. In the sixteenth step, the third component transmits the second evaluation to the second component. In step 17, the second component receives the second evaluation and shares it within the second component. The second component comprises a third subcomponent, The third subcomponent has the function of creating a message based on the first evaluation and the second evaluation, In step 18, the third subcomponent creates the message and sends it to the first component. When the first evaluation is lower than the set evaluation, and the second evaluation is higher than the set evaluation, The aforementioned message conveys that information included in the business information group is included in the description, even though it is not included in the public information group. In the 19th step, the first component is an information processing method that receives and provides the message.

13. An information processing method comprising the 19th step, followed by the 20th to 28th steps, In step 20 above, the second subcomponent creates a fourth instruction and sends it to the third component. The fourth instruction statement includes the fourth instruction and the description, The fourth instruction includes a procedure for classifying the descriptions into types, In the 21st step, the third component receives the fourth instruction and generates the type using the large-scale language model. In step 22 above, the third component transmits the type to the second component, In step 23 above, the second component receives the type and shares it within the second component, In step 24, the second component creates a fifth instruction and sends it to the third component. The fifth instruction includes the fifth instruction, the description and the document, The fifth instruction above includes a procedure for extracting the source relating to the description from the document, In step 25, the third component receives the fifth instruction and uses the large-scale language model to identify the source. In step 26, the third component transmits the source to the second component. In step 27 above, the second component receives the source and shares it within the second component, In step 28 above, the first subcomponent creates a node, The aforementioned database has a function for storing knowledge graphs, The aforementioned knowledge graph includes the aforementioned nodes, The node is equipped with a label, The aforementioned label is assigned to the node that stores the description identified as fact, The node has the label and, A first attribute value pair capable of storing the above description, A second attribute value pair capable of storing the aforementioned type, A third attribute value pair capable of storing information identifying the aforementioned document, The information processing method according to claim 12, comprising a fourth attribute value pair capable of storing information that identifies the aforementioned source.

Images