Information processing system, information processing program, and information processing method

Abstract

To provide an information processing system for performing knowledge management suitable to share or utilize information stored in an organization.SOLUTION: This information processing system includes a knowledge database for storing knowledge data groups about a plurality of knowledge stored in an organization and a plurality of tags attached to them, a generative AI interface part for communicating with an information server for providing a chat function of generative AI, and a retrieval processing unit for generating a narrow-down menu on the basis of the plurality of tags according to transmission of a question sentence or the like from a user terminal, displaying a retrieval screen including the narrow-down menu in the user terminal, performing a drill-down retrieval based on a narrow-down condition transmitted from the user terminal to narrow down the plurality of knowledge, transferring the narrowed down knowledge in association with the question sentence or the like to the information server with reference of them as a retrieval condition, and displaying an obtained answer sentence as a retrieval result in the user terminal.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an information processing system that performs knowledge management suitable for sharing or utilizing information accumulated within an organization such as a company or an association. The present invention also relates to an information processing system that performs knowledge management suitable for sharing or utilizing information available from a public website. The present invention also relates to an information processing program and an information processing method used in such an information processing system. [Background technology]

[0002] Companies, organizations, and other such organizations have traditionally used generative AI (artificial intelligence) to improve the efficiency of various tasks, such as information search, document creation, and customer service. However, the following issues have arisen when sharing or utilizing information accumulated within an organization.

[0003] The first challenge is that while generative AI generates answers and documents based on learned information, there is a desire for organizations to reflect their own unique knowledge and culture in knowledge management. Here, "knowledge" generally refers to knowledge, but in the business world, "knowledge" refers to the knowledge, experience, and added-value systematic information such as case studies accumulated by organizations such as companies and groups. For example, to reflect a company's unique knowledge and culture in knowledge management, a system is needed to incorporate internal information and properly pass it on to generative AI.

[0004] The second challenge is that when using generative AI to reflect an organization's unique knowledge in knowledge management, it is necessary to avoid passing knowledge from different contexts or different organizations to the generative AI in a mixed manner, thereby preventing the generative AI from providing false information (lies).

[0005] The third challenge is that when using generative AI to reflect an organization's unique knowledge in knowledge management, knowledge management is not possible without the knowledge in the first place, but since outdated knowledge can lead to errors, it is necessary to keep the organization's unique knowledge fresh.

[0006] The fourth challenge is that the use of generative AI tends to be limited because there are employees with varying levels of IT literacy within companies and organizations. To expand the use of generative AI regardless of the level of IT literacy, a system with easy operability and learning functions is required.

[0007] In order to build an information processing system for knowledge management, it is possible to consider using cloud computing, which utilizes infrastructure generally provided as a service. However, there are the following issues when building an information processing system using cloud computing.

[0008] The first challenge is that many companies and organizations have an internal policy that requires that important business documents and other data be stored on-premises. On-premises refers to a system usage model in which the hardware and software required for system operation and infrastructure construction, such as servers and network equipment, are owned and operated within the organization. Furthermore, even if companies want to migrate to cloud computing, they have a large amount of important business documents stored on existing on-premises equipment, making it difficult to migrate to cloud computing.

[0009] The second challenge is that when using cloud computing to reflect an organization's unique knowledge in generative AI, various business documents, etc., which constitute important explicit knowledge, must be registered in a knowledge database on the cloud. Selecting and registering these documents from the large amount of business documents stored on on-premise devices within the organization requires a lot of time and cost, making it impossible to immediately apply this technology to business operations.

[0010] The third challenge is that even if important business documents are selected and registered in a knowledge database on the cloud, if the originals are still stored on on-premise devices, they must be synchronized so that the copies on the knowledge database are updated every time the originals are updated. Otherwise, there is a risk that the generation AI will generate incorrect answers based on outdated knowledge.

[0011] Meanwhile, chatbots for customer support purposes have become widespread on corporate homepages. The term "chatbot" is a combination of "chat," which means real-time communication using data communication lines on a computer network, and "bot," which is derived from robot. However, in addition to the third challenge mentioned above, the following challenges existed when chatbots were made public.

[0012] The fourth challenge is that the following preparations are required before releasing a chatbot. In the case of a script-based chatbot, it is necessary to create in advance response scenarios that anticipate customer questions, while in the case of a learning-based chatbot, it is necessary to have the system learn from pre-prepared combinations of questions and answers. In either case, the premise is that a collection of Q&As is prepared in advance that anticipate the types of inquiries customers will have, and it is then necessary to create a script and prepare learning data based on that.

[0013] The fifth challenge is that preparing the Q&A collection mentioned above takes time and money, and it is impossible to know whether the prepared Q&A collection is useful for resolving customer problems until the chatbot is actually published on the website and put into operation.

[0014] As a related technology, Patent Document 1 discloses a portable electronic encyclopedia that has information from an encyclopedia database and communication information via a communication network, and is equipped with an AI learning function, which enables various searches based on observed plants and animals, region, season, etc., and has functions such as inputting extended data and recording extracted data.

[0015] This portable electronic encyclopedia has an audio input means for capturing sounds such as cries, an image input means for capturing images of shapes and contours, an odor input means for searching for and identifying odors, a tactile input means for sensing the feel of an object, a calculation means for measuring and recognizing air pressure, temperature, etc., an extended data input means for inputting various auxiliary data such as text, a communication network function, as well as a built-in database, an AI function for comparing the characteristics and properties of newly input auxiliary data to search and self-learn, such as selecting and comparing related information and similar objects, and an extracted data storage means for storing and accumulating the information searched for by this means. [Prior art documents] [Patent documents]

[0016] [Patent Document 1] JP 2004-70889 (paragraphs 0005-0009, Figure 3) Summary of the Invention [Problem to be solved by the invention]

[0017] According to Patent Document 1, when only vague information is obtained, such as when the input information is duplicated or there is an extreme lack of information, the AI ​​function with learning capabilities extracts the missing information from the built-in database based on the information input as auxiliary data, extracts related data and similar objects, etc., and compares these to narrow down the information, thereby enabling a more detailed electronic search, and the data extracted by these AI functions can be stored.

[0018] However, Patent Document 1 does not take into consideration the sharing or utilization of information accumulated within organizations such as companies and groups, nor does it disclose a configuration suitable for this. Furthermore, the number of species of plants and animals observed around the world is enormous, and it is likely that the pre-installed database will contain data on a wide variety of plants and animals. Therefore, when the AI ​​function extracts missing information from the database, extracts related data and similar objects, and compares them to narrow down the information, there is a risk that information on different plants and animals or outdated information may be mixed in, resulting in the provision of false information.

[0019] In view of the above, a first object of the present invention is to provide an information processing system that performs knowledge management suitable for sharing and utilizing information accumulated within an organization. A second object of the present invention is to prevent knowledge from different contexts or different organizations from being mixed and passed to the generation AI when using a generation AI to reflect the organization's unique knowledge in knowledge management. A third object of the present invention is to maintain the freshness of the organization's unique knowledge. A fourth object of the present invention is to provide an information processing system that is easy to operate and has a learning function, in order to expand the use of generation AI regardless of the level of IT literacy.

[0020] Furthermore, a fifth object of the present invention is to realize an information processing system suitable for sharing or utilizing information stored within an organization, without the need to select necessary information from the large amount of business documents, etc. stored in on-premise devices within the organization and register it in a knowledge database, or to update copies in the knowledge database in the same way as updating the originals on the on-premise devices.

[0021] In addition, a sixth object of the present invention is to make it possible to publish and operate a chatbot for customer support on a website without creating an operating script or preparing learning data based on a Q&A collection that predicts what kind of inquiries customers will have in advance. Also, a seventh object of the present invention is to provide an information processing program and an information processing method for use in the information processing system described above. [Means for solving the problem]

[0022] In order to solve at least some of the above problems, an information processing system according to a first aspect of the present invention is an information processing system that provides information desired by a user to a user terminal used by the user by communicating with the user terminal via a network, and is equipped with: a knowledge database that stores a group of knowledge data relating to multiple pieces of knowledge accumulated in an organization and multiple tags assigned to that knowledge; a generation AI interface unit that communicates via the network with an information server that provides a chat function for the generation AI; and a search processing unit that, in response to a question or request sent from the user terminal, generates a narrowing-down menu based on the multiple tags, displays a search screen including the narrowing-down menu on the user terminal, performs a drill-down search based on the narrowing conditions sent from the user terminal to narrow down the multiple pieces of knowledge, links the question or request to referencing the narrowed-down knowledge as a search condition, and passes it to the information server via the generation AI interface unit, and displays the answer obtained from the information server as a search result on the user terminal.

[0023] Furthermore, an information processing program according to a first aspect of the present invention is an information processing program used in an information processing system that provides information desired by a user to a user terminal by communicating with the user terminal via a network, and causes a CPU to execute the following steps in response to transmission of a question or request from the user terminal: using a knowledge database that stores a group of knowledge data relating to a plurality of knowledge items accumulated in an organization and a plurality of tags assigned to the knowledge items, generate a narrowing-down menu based on the plurality of tags, and display a search screen including the narrowing-down menu on the user terminal; narrowing down the plurality of knowledge items by performing a drill-down search based on the narrowing-down conditions transmitted from the user terminal; communicating with an information server that provides a chat function of the generation AI via a network, and transferring the narrowed-down knowledge item to the information server in association with the question or request as a search condition; and displaying an answer obtained from the information server as a search result on the user terminal.

[0024] Furthermore, an information processing method according to a first aspect of the present invention is an information processing method for providing information desired by a user to a user terminal used by the user by communicating with the user terminal via a network, and includes the steps of: in response to transmission of a question or request from the user terminal, using a knowledge database that stores a group of knowledge data relating to a plurality of knowledge items accumulated in an organization and a plurality of tags assigned to the knowledge items, generating a narrowing-down menu based on the plurality of tags, and displaying a search screen including the narrowing-down menu on the user terminal; performing a drill-down search based on the narrowing-down conditions transmitted from the user terminal to narrow down the plurality of knowledge items; communicating with an information server that provides a chat function of the generation AI via a network, linking the question or request to reference the narrowed-down knowledge as a search condition, and transferring it to the information server; and displaying an answer obtained from the information server as a search result on the user terminal.

[0025] According to the first aspect of the present invention, by using the power of generative AI to expand a knowledge database that stores business know-how and technologies accumulated over many years by organizations such as companies and groups, it is possible to realize highly accurate answers using the chat function of the generative AI and to provide an information processing system, etc. that performs knowledge management suitable for sharing or utilizing information accumulated within an organization.

[0026] Furthermore, by linking the search criteria for referencing narrowed-down knowledge in the organization's knowledge database with a question or request statement and passing it to the information server of the generation AI, when using the generation AI to reflect the organization's own knowledge in knowledge management, it is possible to avoid passing a mixture of knowledge from different contexts or different organizations to the generation AI.

[0027] Furthermore, in the information processing system according to the second aspect of the present invention, the search processing unit sets inapplicability information to the tag of knowledge that has been designated in advance as inapplicable, and when narrowing down multiple pieces of knowledge stored in the knowledge database, it excludes knowledge that has a tag with the inapplicability information set, thereby preventing the knowledge from being passed on to the information server.

[0028] According to the second aspect of the present invention, if knowledge that was created a long time ago and needs to be updated is designated as inapplicable, the search processing unit sets inapplicability information in the tag of the knowledge designated in advance as inapplicable, and excludes that knowledge when narrowing down multiple pieces of knowledge, thereby maintaining the freshness of the organization's unique knowledge.

[0029] In addition, in the information processing system according to the third aspect of the present invention, the generation AI interface unit configures the information server to launch a log output program when information regarding the question or request sentence cannot be obtained even when referring to the narrowed-down knowledge, so that when information regarding the question or request sentence cannot be obtained even when referring to the narrowed-down knowledge, the generation AI interface unit receives from the information server, along with historical information regarding the execution of the chat function of the generation AI, information identifying the question or request sentence and information indicating that there is no knowledge to refer to, and stores this information in a log database.

[0030] According to the third aspect of the present invention, knowledge management can clarify knowledge that is necessary to answer questions and requests from users but is not stored in the organization's knowledge database.By replenishing the organization's knowledge database based on the results, it is possible to maintain the freshness of the organization's unique knowledge.

[0031] In addition, in the information processing system according to the fourth aspect of the present invention, the search processing unit stores the filtering conditions used when narrowing down knowledge by performing a drill-down search in the knowledge database, displays a search screen on the user terminal that includes multiple tabs for selecting filtering conditions, and narrows down the multiple pieces of knowledge according to the filtering conditions corresponding to the selected tab.

[0032] Furthermore, in the information processing system according to the fifth aspect of the present invention, the search processing unit stores the narrowing-down conditions used when performing a drill-down search to narrow down knowledge in the knowledge database for each use of the knowledge, and displays a search screen on the user terminal that shows multiple narrowing-down conditions for each use of the knowledge.

[0033] According to the fourth or fifth aspect of the present invention, an information processing system with simple operability and learning functions can be provided by having employees with high literacy in the use of knowledge and generative AI prepare appropriate search narrowing conditions for each application in advance, thereby expanding the use of generative AI regardless of the level of IT literacy.

[0034] As a result of the above, an information processing system that combines a knowledge management system and generative AI can be constructed, and the unique knowledge held within an organization such as a company or organization can be reflected in the generative AI, making it easy for all employees to utilize that knowledge.

[0035] Further, an information processing system according to a sixth aspect of the present invention is an information processing system that provides information desired by a user to a user terminal used by the user by communicating with the user terminal via a network, the information processing system including a content collection server that collects content via the network and stores it in a content database, and when update or deletion of already collected content is detected via the network, updates or deletes the corresponding content stored in the content database; and a knowledge database that stores a group of knowledge data relating to a plurality of knowledge pieces and a plurality of tags assigned to the knowledge pieces, and stores knowledge data including the content collected by the content collection server in the knowledge database in response to update or deletion of the content by the content collection server. a content collection interface unit that updates the knowledge data by updating or deleting content included in the knowledge data stored in the user terminal; a generation AI interface unit that communicates via a network with an information server that provides a chat function for the generation AI; and a search processing unit that, in response to a question or request sent from the user terminal, generates a narrowing-down menu based on the plurality of tags, displays a search screen including the narrowing-down menu on the user terminal, performs a drill-down search based on the narrowing-down conditions sent from the user terminal to narrow down the plurality of knowledge items, links the question or request to reference the narrowed-down knowledge as a search condition, and passes it to the information server via the generation AI interface unit, and displays the answer obtained from the information server on the user terminal as a search result.

[0036] Here, the content collection server may include at least one of a web crawler that extracts content via an intra-organizational network from an intra-organizational portal web server that operates a portal site within the organization and provides content to the user terminals, a file crawler that extracts content via an intra-organizational network from an intra-organizational file server that stores files of content used within the organization, and a groupware crawler that extracts content via an intra-organizational network from a groupware server that operates groupware within the organization and provides content to the user terminals.

[0037] In this case, the content collection interface unit stores knowledge data including content collected from various servers within the organization in a knowledge database, and detects updates or deletions to the content and updates the knowledge data.This makes it possible to realize an information processing system that is suitable for sharing or utilizing information stored within an organization, without the need to select necessary information from the large amount of business documents, etc. stored on on-premises equipment within the organization and register it in the knowledge database, or to update the copy on the knowledge database in the same way as updating the original on the on-premises equipment.

[0038] Alternatively, the content collection server may include a web crawler that extracts content from publicly available websites over a wide area network. In this case, the content collection interface unit stores knowledge data including content such as product manuals collected from publicly available websites in a knowledge database, and detects updates or deletions to the content and updates the knowledge data.This makes it possible to publish and operate a chatbot on a website for customer support purposes without having to create an operating script or prepare learning data based on a Q&A collection that anticipates what kind of inquiries customers will have in advance. [Brief explanation of the drawings]

[0039] [Figure 1] 1 is a diagram showing an example of the configuration of an entire system including an information processing system according to a first embodiment of the present invention. [Figure 2] 2 is a block diagram showing an example of the configuration of an information processing server shown in FIG. 1. FIG. [Figure 3] 4 is a flowchart showing an example of the operation of the information processing system according to the first embodiment of the present invention. [Figure 4] FIG. 10 is a schematic diagram showing an example of a search screen displayed on a user terminal. [Figure 5] FIG. 10 is a schematic diagram showing an example of a search screen on which a prompt is displayed. [Figure 6] FIG. 10 is a diagram showing an example of the configuration of an entire system including an information processing system according to second and third embodiments of the present invention. [Figure 7] FIG. 7 is a block diagram showing an example of the configuration of a content collection server 400 shown in FIG. 6. [Figure 8] FIG. 10 is a schematic diagram illustrating an example of the operation of the information processing system according to the second embodiment of the present invention. [Figure 9] FIG. 7 is a block diagram showing an example of the configuration of the information processing server shown in FIG. 6. [Figure 10] 10 is a flowchart showing an example of the operation of the information processing system according to the second embodiment of the present invention. [Figure 11] FIG. 10 is a schematic diagram showing an example of the structure of knowledge data passed to a generation AI. [Figure 12] FIG. 7 is a block diagram showing an example of the configuration of a content collection server 500 shown in FIG. 6. [Figure 13] FIG. 10 is a schematic diagram illustrating an example of the operation of an information processing system according to a third embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0040] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. Note that the same components are designated by the same reference numerals and redundant description will be omitted. First Embodiment Fig. 1 is a diagram showing an example of the configuration of an overall system including an information processing system according to a first embodiment of the present invention. As shown in Fig. 1, this overall system includes an information processing server 100 constituting the information processing system according to the first embodiment of the present invention, an information server 200 providing a chat function for the generation AI, and at least one user terminal used by a user (Fig. 1 shows, as an example, a plurality of user terminals 301, 302, ...). The information processing server 100 to the user terminal 301, etc. have the function of communicating with each other via a network such as the Internet or a mobile communication network.

[0041] The information processing server 100 provides the information requested by the user to the user terminal 301, etc., by communicating with the user terminal 301, etc., used by the user via a network. Here, the information processing server 100 can integrate and link its own applications with the applications and web services of the information server 200 using an API (Application Programming Interface). An API is a set of rules or specifications that stipulate procedures and data formats for calling and using the functions and managed data of a certain computer program (software) from other external programs.

[0042] The information server 200 includes, for example, a search engine that searches stored data and provides a chat function as an interactive service using AI that generates original data using trained data. Examples of such interactive AI chat services include Open AI ChatGPT, Microsoft Bing AI, Google BERT, and Meta OPT.

[0043] The information processing server 100 or the information server 200 may be configured as a single server device, or may be configured as multiple server devices that are connected to each other via a LAN (local area network) or the like and operate in cooperation with each other.

[0044] For example, the information processing server 100 is installed in data center A of a company or cloud provider that operates this information processing service business. Also, the information server 200 is installed in data center B of a company or cloud provider that operates an interactive AI chat service business.

[0045] The user terminal 301 may be, for example, a mobile phone such as a smartphone or feature phone that can connect to a network such as the Internet or a mobile communication network, a tablet terminal, or a personal computer. In the following, as an example, a case where a personal computer is used as the user terminal 301 will be described.

[0046] <Information processing server configuration example 1> Fig. 2 is a block diagram showing an example of the configuration of the information processing server shown in Fig. 1. As shown in Fig. 2, the information processing server 100 includes an operation unit 101, a display unit 102, an audio input / output unit 103, a communication circuit 104, an interface 105, a CPU (Central Processing Unit) 110, a storage unit 120, and a cache memory 130. The interface 105 to the cache memory 130 are connected to one another via a bus line. Note that some of the components shown in Fig. 1 or 2 may be omitted or modified, or other components may be added to the components shown in Fig. 1 or 2.

[0047] The operation unit 101 is composed of, for example, a keyboard, a mouse, etc., and is used to input various commands and data to the information processing server 100. The display unit 102 includes, for example, an LCD (liquid crystal display), etc., and displays an operation screen, etc. The audio input / output unit 103 includes, for example, a microphone, an amplifier, a speaker, etc., and converts audio signals into electrical signals and electrical signals into audio signals.

[0048] The communication circuit 104 has a function of performing data communication with the information server 200, the user terminal 301, etc. via a network such as a wired LAN, a wireless LAN, the Internet, or a mobile communication network. The interface 105 is connected to the operation unit 101 to the communication circuit 104, and transmits various commands and data between them and the CPU 110.

[0049] The CPU 110 performs various calculations and data processing in accordance with various software (including information processing programs) stored in the storage unit 120. The recording medium (storage medium) in the storage unit 120 may be a hard disk, flexible disk, magneto-optical disk, solid state drive, magnetic tape, ROM (read only memory), CD-ROM, DVD-ROM, or the like. The cache memory 130 is configured with RAM, etc.

[0050] In addition to the above software, the storage unit 120 stores at least one knowledge database (in FIG. 2, as an example, multiple knowledge databases DB1, DB2, DB3, ... are shown) that stores a group of knowledge data related to multiple pieces of knowledge accumulated in an organization such as a company or an organization and multiple tags assigned to that knowledge, and may further store a log database DB0 that stores information related to logs.

[0051] Furthermore, the knowledge database DB1 etc. may further store prompt information relating to a plurality of prompts representing commands to the information server 200. Such data and information is read out as needed and written to the cache memory 130. Note that a plurality of knowledge databases DB1, DB2, DB3, etc. may be provided for each narrowing-down condition in a drill-down search and each use of knowledge.

[0052] Here, the CPU 110 and the information processing program stored in the storage unit 120 constitute a generation AI interface unit 111 and a search processing unit 112 as functional blocks.

[0053] The generation AI interface unit 111 uses an API to communicate via a network with the information server 200, which provides the chat function of the generation AI, and can integrate or link the applications of the information processing server 100 with the applications and web services of the information server 200.

[0054] For example, in response to a question or request sent from the user terminal 301, the search processing unit 112 generates a narrowing-down menu based on multiple tags stored in the knowledge database DB1 or the like, and displays a search screen including the narrowing-down menu on the user terminal 301. When the user of the user terminal 301 inputs narrowing-down conditions for a drill-down search into the user terminal 301, the user terminal 301 transmits the narrowing-down conditions to the information processing server 100.

[0055] The search processing unit 112 performs a drill-down search based on the narrowing-down conditions sent from the user terminal 301, narrows down the multiple pieces of knowledge stored in the knowledge database DB1 or the like, links the reference to the narrowed-down knowledge as a search condition with the question or request statement, and passes it to the information server 200 via the generation AI interface unit 111, and displays the answer statement obtained from the information server 200 as the search result on the user terminal 301.

[0056] <Example of operation of the first embodiment> Next, an example of the operation of the information processing system according to the first embodiment of the present invention will be described below. Fig. 3 is a flowchart showing an example of the operation of the information processing system according to the first embodiment of the present invention.

[0057] 3, in response to a question or request sent from a user terminal 301 or the like, the search processing unit 112 of the information processing server 100 uses a knowledge database DB1 or the like to generate a narrowing-down menu based on multiple tags, and causes a search screen including the narrowing-down menu to be displayed on the user terminal 301 or the like. The knowledge database DB1 or the like stores a group of knowledge data relating to multiple pieces of knowledge accumulated in an organization and multiple tags assigned to that knowledge.

[0058] Fig. 4 is a schematic diagram showing an example of a search screen displayed on a user terminal. The search screen shown in Fig. 4 includes a drill-down navigation screen 301a for performing a drill-down search using a narrowing menu, a list of knowledge narrowed down by the drill-down search (knowledge list) 301b, and an AI chat screen 301c containing a question or request and a response to that question in the chat function of the generation AI.

[0059] 4, a plurality of format names, a plurality of posting user names, etc. are displayed as search options used in drill-down searches. Note that various operation buttons may be arranged in a portion of the search screen (the left end of FIG. 4).

[0060] The search processing unit 112 may store the filtering conditions used when narrowing down knowledge by performing a drill-down search in a knowledge database DB1 or the like, display a search screen including multiple tabs for selecting filtering conditions on the user terminal 301 or the like, and narrow down multiple pieces of knowledge according to the filtering conditions corresponding to the selected tab.

[0061] 4, the knowledge list 301b includes, as examples of multiple tabs, a "New Knowledge" tab, a "Draft" tab, and a "Trash" tab. When a user selects one of these tabs, a search screen is displayed that includes at least one piece of knowledge narrowed down according to the narrowing conditions corresponding to the selected tab.

[0062] In addition, the search processing unit 112 may store the narrowing down conditions used when performing a drill-down search to narrow down knowledge in a knowledge database DB1 or the like for each use of the knowledge, and display a search screen on a user terminal 301 or the like that shows multiple narrowing down conditions for each use of the knowledge.

[0063] By utilizing these functions, anyone can easily narrow down knowledge and pass it to generative AI. The use of generative AI in business is clearly limited to the target product, project, customer, or application. By taking advantage of these characteristics, employees with high literacy in knowledge and generative AI can prepare appropriate search and refinement conditions for each application in advance, providing an information processing system with simple operability and learning functions, thereby expanding the use of generative AI regardless of the level of IT literacy.

[0064] For example, you can save the search refinement criteria as a tab, giving it a name such as "For creating replies to customer inquiries about product A" or "Complete set of design standards for project X." By clicking on that tab, you can always narrow down and search the knowledge database using that condition, and by launching AI chat from there, anyone can utilize AI that appropriately applies their company's knowledge.

[0065] Next, in step S2 of FIG. 3, the search processing unit 112 of the information processing server 100 performs a drill-down search based on the narrowing-down conditions transmitted from the user terminal 301 or the like to narrow down a plurality of knowledge items.

[0066] As an example, when a knowledge database DB1 or the like stores a group of document information including multiple document information accumulated in an organization and an index including multiple index records, the search processing unit 112 performs a drill-down search of the document information using the index.

[0067] Here, each piece of document information included in the document information group corresponds to each index record included in the index. The document information has multiple document items defined by a document data structure for each document type, and each of these document items stores document item data of tag type or character string type.

[0068] The index record has fields corresponding to the refinement menu of the drill-down search. Each field stores tag-type data that matches the tag type associated with the refinement menu in the menu tag correspondence table, and that is document item data stored in the document item of the document information corresponding to the index record or data associated with the document item data in the tag table.

[0069] In such a case, the search processing unit 112 searches the fields corresponding to each index record included in the index in a search using the menu items of the refinement menu in the drill-down search as conditions.

[0070] Next, in step S3 of Figure 3, when a prompt call is requested from a user terminal 301, etc., the search processing unit 112 of the information processing server 100 reads out a set of prompts to be used for the target business from a knowledge database DB1, etc., displays them on the user terminal 301, etc., and allows the user to select a prompt.

[0071] Next, in step S4 of Figure 3, the generation AI interface unit 111 of the information processing server 100 communicates with the information server 200, which provides the chat function of the generation AI, via a network, and the search processing unit 112 of the information processing server 100 transfers to the information server 200 via the generation AI interface unit 111 the search conditions for referencing narrowed-down knowledge and the prompt read from the knowledge database DB1, etc., when a prompt is selected by the user terminal 301, etc., in conjunction with the question or request statement.

[0072] When linking a question sentence or a request sentence with reference to the narrowed-down knowledge as a search condition, the search processing unit 112 may, for example, control the information server 200 via the generation AI interface unit 111 to search for similar sentences for the question sentence or the request sentence using the narrowed-down knowledge as the target.

[0073] Furthermore, the search processing unit 112 may display a search screen including a selection means for selecting whether or not to use a group of knowledge data stored in the knowledge database DB1, etc., on the user terminal 301, etc. This allows the user to select whether or not to use knowledge accumulated in the organization.

[0074] When the user selects to use knowledge stored in a knowledge database DB1 or the like, the search processing unit 112 links the reference to the narrowed-down knowledge as a search condition with the question or request statement and passes it to the information server 200 via the generation AI interface unit 111.

[0075] 4, the AI ​​chat screen 301c has a "Reset" button and a "Save" button as examples of selection methods. When the user clicks the "Reset" button, referring to the narrowed-down knowledge is not considered a search condition, and the information server 200 generates a response sentence based on general learned information.

[0076] On the other hand, if the user clicks the "Save" button, referring to the narrowed-down knowledge becomes a search condition, and the information server 200 generates a response sentence by referring only to the narrowed-down knowledge. Since the knowledge list 301b shown in Fig. 4 displays the name of the higher-level knowledge "current project" as the reference knowledge, the search processing unit 112 may set referring to this knowledge as a search condition in the chat function of the generation AI of the information server 200.

[0077] However, the "Search Results" in the knowledge list 301b shows that the number of lower-level knowledge items narrowed down by the drill-down search is 25, raising the question of whether all of this large number of knowledge items should be set as search conditions in the chat function of the generation AI.

[0078] Therefore, the search processing unit 112 of the information processing server 100 may display a search screen including the knowledge reference mode switch 301sw shown in Figure 4 on the user terminal 301, etc., as a selection means for selecting whether the search condition is to refer to some of the multiple knowledge items narrowed down by performing a drill-down search and displayed in the knowledge list 301b, or to refer to all of that knowledge item.

[0079] When obtaining an answer to a specific question, in order to use several pieces of knowledge that are closely related to the question, the knowledge reference mode selector switch 301sw is set to the left side, "Related only" mode. This mode is suitable for applications such as answering Q&A-style questions about operation methods and solutions in customer support, or creating documents by referring to knowledge.

[0080] To realize this mode, the search processing unit 112 further narrows down the knowledge by searching for similar sentences for at least a portion of the question or request sentence entered by the user into the user terminal 301, etc., for multiple knowledge items narrowed down in the drill-down search, and sets the search condition for the chat function of the generation AI to refer to knowledge with a high degree of similarity.

[0081] On the other hand, if you need to check whether a question strictly meets all knowledge conditions, investigate where the problem is if there is one, or compare a question comprehensively with all knowledge, set the knowledge reference mode selector switch 301sw to the "Comprehensive" mode on the right. This mode is suitable for applications such as checking industrial product specifications against past failure cases, or checking all clauses that must be observed when creating legal documents such as contracts.

[0082] To achieve this mode, the search processing unit 112 sets a search condition in the chat function of the generation AI to refer to all knowledge narrowed down by the drill-down search without performing a similar sentence search. Note that if all knowledge cannot be passed to the generation AI at once due to reasons such as an upper limit on the number of characters of knowledge that can be passed to the generation AI, the search processing unit 112 performs an operation of passing the knowledge to the generation AI in multiple batches.

[0083] Furthermore, the search processing unit 112 may set inapplicability information to the tag of knowledge that has been designated in advance as inapplicable based on the freshness of the knowledge, and when narrowing down a plurality of pieces of knowledge stored in the knowledge database DB1 or the like, exclude the knowledge having the tag with the inapplicability information set, thereby preventing the knowledge from being handed over to the information server 200. Note that it is desirable that, for example, an "ON" button and an "OFF" button be provided on the search screen as a setting means for setting whether or not to enable this function.

[0084] If knowledge that was created a long time ago and needs to be updated is designated as inapplicable, the search processing unit 112 sets inapplicability information to the tag of the knowledge designated in advance as inapplicable, and excludes the knowledge when narrowing down a plurality of knowledge, thereby keeping the organization's unique knowledge fresh.

[0085] Important internal knowledge is documented explicit knowledge (knowledge that can be explained or expressed using text, diagrams, or mathematical formulas) such as various internal manuals, business manuals, operation manuals, maintenance manuals, or customer service manuals. Furthermore, these are revised annually or monthly and managed differently for each business division and product. Therefore, keeping them updated appropriately and up-to-date is the most effective way to ensure that generative AI can properly utilize them.

[0086] Therefore, if this explicit knowledge is stored in a knowledge database such as DB1 used in the drill-down navigation, it will be possible to list related knowledge based on events that trigger updates, such as applications, projects, products, or parts used, etc. This will make it easier to keep knowledge up to date.

[0087] Furthermore, by adding a flag to each piece of knowledge as a tag indicating whether it is applicable to the generating AI, if the knowledge becomes outdated and needs to be updated but has not yet been updated, it can be removed from the filtering criteria simply by adding a flag indicating that it is inapplicable as a tag, making it easy to avoid the risk of passing incorrect or outdated information to the generating AI.

[0088] Furthermore, when using prompts stored in a knowledge database DB1 or the like, the search processing unit 112 displays a search screen on the user terminal 301 or the like, including an AI chat screen with options for selecting at least one of those prompts, and passes the prompt selected by the user terminal 301 or the like to the information server 200 via the generation AI interface unit 111.

[0089] Figure 5 is a schematic diagram showing an example of a search screen on which a prompt is displayed. The input window at the bottom of the AI ​​chat screen 301c shown in Figure 4 displays the letter "P" in a circle, which is an icon for launching a dialog box for selecting a prompt. Clicking this icon displays the "Prompt Selection" dialog box D1 shown in Figure 5.

[0090] The "Prompt Selection" dialog box D1 shown in Figure 5 displays the first prompt, "Have AI write a blog (simplified version)," and the second prompt, "Create an FAQ about Company A." Note that "FAQ" stands for Frequently Asked Questions. Each of these prompts has switches (1) and (2) for setting whether or not to use the prompt.

[0091] To use the first prompt, switch (1) is turned on, and a dialog box D2 for creating the first command is displayed. In the dialog box D2, the user enters the desired theme in the (Enter theme here) field and clicks the "Apply" button, and the chat function of the generation AI provided in the information server 200 executes the first command.

[0092] Furthermore, when using the second prompt, setting switch (2) to ON displays a dialog box D3 for creating a second command. In the dialog box D3, when the user enters the desired information in the "XX" field and clicks the "Apply" button, the chat function of the generation AI provided in the information server 200 executes the second command.

[0093] In this way, by managing prompts in a knowledge database and making them applicable from the AI ​​chat screen, even inexperienced users can select appropriate prompts and make use of the generative AI.

[0094] The multiple prompts stored in the knowledge database DB1 or the like may include at least one generally applicable basic prompt and multiple task prompts that can be selected for each task or work. For example, one basic prompt may be set as a prompt that is always applied to a task, and multiple task prompts that can be selected for each of various tasks or works in the organization may be set as options.

[0095] Basic prompts may be commands that are generally applicable, such as "Write concise and polite business documents" or "Reply in both Japanese and English." On the other hand, business prompts may be commands such as "Create reply emails for customer support," "Check design documents against company regulations," or "Create FAQs from inquiry logs." It is desirable to define multiple prompts with titles that immediately indicate their usefulness, set them up so that they can be shared, and then call them up at any time from the AI ​​chat screen.

[0096] The information server 200 may also be configured so that the generation AI interface unit 111 launches a log output program when it cannot obtain information about the question or request sentence even after referring to the narrowed-down knowledge. This is a function of the API installed in the information server 200, and utilizes the function of launching other programs when the generation AI responds.

[0097] If information regarding the question or request sentence cannot be obtained even by referring to the narrowed-down knowledge, the generation AI interface unit 111 receives from the information server 200, along with historical information regarding the execution of the chat function of the generation AI, information identifying the question or request sentence and information indicating that no knowledge to refer to exists, and stores this information in the log database DB0.

[0098] This makes it possible to clarify knowledge that is necessary for answering questions and requests from users but that is not stored in the organization's knowledge database within knowledge management, and by replenishing the organization's knowledge database based on the results, it is possible to keep the organization's unique knowledge fresh.

[0099] For example, the log stored in the log database DB0 includes all the execution conditions of the AI ​​chat (occurrence date and time (e.g., year, month, day, minute, second), chat panel ID, user ID, project ID, saved filtering conditions, etc.), as well as information identifying the question or request passed to the information server 200, and information indicating that there is no knowledge to refer to.

[0100] In addition, or instead, when the generation AI interface unit 111 obtains information regarding a question or request sentence by referring to the narrowed down knowledge, it may store in the log database DB0 information identifying the referenced knowledge and information identifying the answer sentence obtained from the information server 200, along with historical information regarding the execution of the generation AI's chat function.

[0101] By recording both appropriate answers and cases where an answer could not be given due to lack of knowledge in the log, those in charge of promoting the use of knowledge can periodically check the log and see how useful this information processing system is (it is also possible to calculate the return on investment, etc.). At the same time, by identifying and properly organizing knowledge that is needed on-site but not prepared as knowledge (tacit knowledge), it can contribute to improving work efficiency, etc.

[0102] Next, in step S5 of Fig. 3, the search processing unit 112 of the information processing server 100 displays the answer obtained from the information server 200 as a search result on the user terminal 301. On the AI ​​chat screen 301c shown in Fig. 4, in response to the question "What should I do when I go on a business trip?", the answer "When you go on a business trip, you need to follow the following procedures..." is displayed.

[0103] In response to the request, "You are a FAQ creation master. Please create an FAQ about Company A's telework regulations...," a response showing several questions and answers about Company A's telework regulations is displayed. In this way, the generation AI can convert difficult internal documents and other information into easy-to-understand response text, which is also useful for promoting user understanding.

[0104] According to the first embodiment of the present invention, by using the power of a generating AI to expand a knowledge database DB1 or the like that stores business know-how and technologies that have been accumulated over many years by organizations such as companies and groups, it is possible to realize highly accurate answers using the chat function of the generating AI and to provide an information processing system or the like that performs knowledge management suitable for sharing or utilizing information accumulated within an organization.

[0105] <Second embodiment> 6 is a diagram showing an example of the configuration of an entire system including information processing systems according to second and third embodiments of the present invention. The information processing system according to the second embodiment of the present invention is suitable for sharing or utilizing information stored in on-premise devices within an organization, and in addition to the information processing server 100, further includes a content collection server 400 for collecting and updating knowledge data stored in the information processing server 100.

[0106] As shown in Figure 6, a content collection server 400, an in-house portal web server 401, an in-house file server 402, and a groupware server 403, along with user terminals 301, 302, etc., are placed in an office or the like within an organization such as a company or organization, and are placed in an on-premises environment.

[0107] The user terminal 301 to the groupware server 403 communicate with each other via an internal network of the organization, and can also communicate with the information processing server 100, the information server 200, and other servers outside the organization via a wide area network. However, in order to maintain an on-premise environment within the organization, the content collection server 400 to the groupware server 403 cannot be accessed from outside the organization.

[0108] The intra-organization portal web server 401 operates a portal site that is used only within the organization, provides content including text (document data or character data) or image data to the user terminals 301, etc., and manages schedules and reports, thereby facilitating the coordination and sharing of information within the organization. Such intra-organization portal sites cannot be used by anyone, and are limited to those within the organization.

[0109] The internal file server 402 stores and manages files of content including text or image data used within the organization. For example, in the case of a company, the internal file server 402 stores and manages files such as various internal business documents, technical documents, customer data, transaction history, or personnel and financial information.

[0110] The groupware server 403 operates groupware such as Notes within the organization and provides content including text or image data to the user terminals 301. For example, groupware has functions such as document sharing, e-mail, electronic bulletin boards, and schedule management (calendars and schedules), and is used for exchanging messages and sharing information among members within the organization.

[0111] The content collection server 400 extracts and collects content including text, image data, etc. from the in-house portal web server 401, the in-house file server 402, or the groupware server 403 via the in-house network using a crawler that retrieves each content.

[0112] Here, a crawler is a program or function that periodically acquires information, such as documents or images, from other devices or the web and automatically creates a database. Crawlers are primarily used to create databases and indexes for search engines, and are also used for various statistical surveys and other purposes.

[0113] For example, a crawler may request a copy of an HTML document, follow the links in the document, and then collect other documents. If the crawler finds a new document, it registers it as a file in the database. If the crawler finds that a registered file does not have a corresponding document, it deletes the file from the database.

[0114] In this way, the content collection server 400 collects content via the organization's internal network and stores it in the content database, and when it detects an update or deletion of already collected content via the organization's internal network, it updates or deletes the corresponding content stored in the content database.

[0115] <Content collection server configuration example 1> FIG. 7 is a block diagram showing an example of the configuration of the content collection server 400 shown in FIG. 6, and FIG. 8 is a schematic diagram for explaining an example of the operation of the information processing system according to the second embodiment of the present invention.

[0116] The content collection server 400 shown in Fig. 7 includes a real-time clock 106 in addition to the components of the information processing server 100 shown in Fig. 2, and is equipped with a CPU 410 and a storage unit 420 instead of the CPU 110 and the storage unit 120. Note that some of the components shown in Fig. 6 or 7 may be omitted or modified, or other components may be added to the components shown in Fig. 6 or 7.

[0117] The communication circuit 104 has the function of performing data communication between the user terminals 301, 302, etc. and the in-organization portal web server 401 to the groupware server 403 via an in-organization network such as a wired LAN or wireless LAN, and also performing data communication between the information processing server 100, information server 200, etc. outside the organization via a wide area network such as the Internet or a mobile communication network. The interface 105 is connected to the operation unit 101 to the communication circuit 104, and transmits various commands and data between them and the CPU 410.

[0118] The CPU 410 performs various calculations and data processing in accordance with various software (including information processing programs) stored in the storage unit 420. In addition to the above software, the storage unit 420 also stores a content database DB4 that stores content collected by the content collection server 400.

[0119] Here, web crawler 411, file crawler 412, groupware crawler 413, and content collection controller 414 are configured as functional blocks by CPU 410 and information processing programs stored in storage unit 420. Alternatively, at least one of web crawler 411 to groupware crawler 413 may be configured as needed.

[0120] The web crawler 411 extracts the content of the portal site from the content database DB401 (Figure 8) of the internal portal web server 401, which operates a portal site within the organization and provides content including text or image data to user terminals 301, etc., via the internal network.

[0121] In addition, the file crawler 412 extracts content such as business documents from a content database DB402 (Figure 8) of an internal file server 402, which stores content files including text or image data used within the organization, via the internal network.

[0122] Furthermore, the groupware crawler 413 extracts groupware content via the organization's internal network from the content database DB403 (Figure 8) of the groupware server that operates groupware within the organization and provides content including text or image data, etc. to user terminals 301, etc.

[0123] In addition, when extracting content, the web crawler 411 to the groupware crawler 413 also acquire, as additional information about the content, obtainable attributes such as the title, location (e.g., URL or file hierarchy), timestamp (e.g., time information indicating the date the content was created or updated), or author, for each piece of content.

[0124] Location is an important element for identifying content, and is also useful for showing users the underlying knowledge through the chat function of the generation AI. Folder names (e.g., product names or uses) can be extracted from location information and mapped as tags for drill-down searches in the knowledge database.

[0125] The content collection controller 414 collects content extracted by the web crawler 411, file crawler 412, or groupware crawler 413, and stores the collected content in the content database DB4. At that time, the content collection controller 414 also stores additional information such as title, location, and timestamp in the content database DB4.

[0126] For this purpose, the content collection controller 414 has a timer function that uses the timing signal output from the real-time clock 106, which performs timing operations, and activates the extraction operations of the web crawler 411, file crawler 412, or groupware crawler 413 at a preset timing, for example, once a day or once an hour.

[0127] For example, each time content is collected, the content collection controller 414 compares the timestamp attached to the content with the timestamp of the previous collection, and if the time represented by the timestamp is newer, it detects the content at that location as being "updated" and updates the content database DB4.

[0128] In addition, the content collection controller 414 detects content that is included in the list of content stored in the content database DB4 but was not collected during the current collection as being "deleted" and deletes it from the content database DB4.

[0129] <Information processing server configuration example 2> Fig. 9 is a block diagram showing an example of the configuration of the information processing server shown in Fig. 6. In the information processing server 100 shown in Fig. 9, a generation AI interface unit 111, a search processing unit 112, and a content collection interface unit 113 are configured as functional blocks by a CPU 110 and an information processing program stored in a storage unit 120. Note that some of the components shown in Fig. 9 may be omitted or modified, or other components may be added to the components shown in Fig. 9.

[0130] 7 communicates with the information processing server 100 via a wide area network, and the content collection controller 414 of the content collection server 400 transmits extracted content to the content collection interface unit 113 of the information processing server 100 based on the extraction results of the web crawler 411, the file crawler 412, or the groupware crawler 413. The content collection controller 414 also transmits information about "updated" or "deleted" content to the content collection interface unit 113.

[0131] As a result, in the information processing server 100, the content collection interface unit 113 stores knowledge data including content collected by the content collection server 400 in a knowledge database DB1 that stores a group of knowledge data relating to multiple pieces of knowledge and multiple tags assigned to those pieces of knowledge.

[0132] In addition, the content collection interface unit 113 updates or deletes the content included in the knowledge data stored in the knowledge database DB1 in response to the update or deletion of content by the content collection server 400, thereby updating the knowledge data.

[0133] In this way, the content collection interface unit 113 updates the knowledge data stored in the knowledge database DB1 in synchronization with content extraction by the web crawler 411, the file crawler 412, or the groupware crawler 413. In other respects, the second embodiment may be similar to the first embodiment.

[0134] As a result, when utilizing AI to take advantage of the knowledge contained in each organization's unique explicit knowledge (various business documents, etc.), the content on the organization's business server can be left as is, and the content can be utilized without the hassle of selecting important content and registering it in a knowledge database.

[0135] Furthermore, the original can be managed in the same place as before, and when the original is updated, the content in the knowledge data stored in the information processing system is automatically updated. Furthermore, when the original is deleted, the content in the knowledge data stored in the information processing system is also deleted so that it will no longer be referenced.

[0136] <Example of operation of the second embodiment> Next, an example of the operation of the information processing system according to the second embodiment of the present invention will be described with reference to Fig. 10. Fig. 10 is a flowchart showing an example of the operation of the information processing system according to the second embodiment of the present invention.

[0137] As a premise, the content collection controller 414 of the content collection server 400 shown in FIG. 7 collects content via a network and stores it in the content database DB4, and when it detects via the network that already collected content has been updated or deleted, it updates or deletes the corresponding content stored in the content database DB4.

[0138] Furthermore, the content collection interface unit 113 of the information processing server 100 shown in Figure 9 stores knowledge data including content collected by the content collection server 400 in a knowledge database DB1 that stores a group of knowledge data related to multiple pieces of knowledge and multiple tags assigned to those pieces of knowledge.

[0139] In addition, the content collection interface unit 113 updates or deletes the content included in the knowledge data stored in the knowledge database DB1 in response to the update or deletion of content by the content collection server 400, thereby updating the knowledge data.

[0140] Referring to Figure 8, at a business location such as research and development, production, sales, or maintenance, a user uses a user terminal 301 to make an inquiry using a spoken language query (question or request) to the knowledge management system "Solution Desk" operated by the information processing server 100.

[0141] In step S1 of Figure 10, in response to a query sent from the user terminal 301, the search processing unit 112 of the information processing server 100 generates a narrowing down menu based on multiple tags using the knowledge database DB1 updated by the content collection interface unit 113, and displays a search screen including the narrowing down menu on the user terminal 301.

[0142] When a user inputs search criteria using the user terminal 301, in step S2 of Figure 10, the search processing unit 112 of the information processing server 100 performs a drill-down search based on the search criteria sent from the user terminal 301 to narrow down multiple knowledge items.

[0143] Next, in step S3 of Figure 10, when a prompt call is requested from the user terminal 301, the search processing unit 112 of the information processing server 100 reads out a set of prompts to be used for the target business from the knowledge database DB1, displays them on the user terminal 301, and allows the user to select a prompt.

[0144] Next, in step S4 of Figure 10, the generation AI interface unit 111 of the information processing server 100 communicates with the information server 200, which provides the chat function of the generation AI, via a network, and the search processing unit 112 of the information processing server 100 links both the knowledge narrowed down by the filtering conditions in step S2 and the prompt selected by the user terminal 301 in step S3 with the query statement and passes them to the information server 200 via the generation AI interface unit 111 (see Figure 8).

[0145] In this case, the knowledge may be transferred to the information server 200 in a structured data format that includes information on the source of the citation. For example, the knowledge may be transferred to the information server 200 in a nested format (e.g., a standard format such as JSON (JavaScript Object Notation) or XML (Extensible Markup Language)) using parentheses or the like to make each structure clear.

[0146] The structure used in such a data format may include information about the source of the knowledge, such as a title, a location such as a link or page information for reference, a timestamp such as a creation date or update date, a format, or various fields and their values ​​(tags).

[0147] Furthermore, if the original content (e.g., a document) is in PDF (Portable Document Format), it may include page information on which the content was located, or it may include the entire text extracted by matching each page with the "bookmarks" included in the PDF.

[0148] Referring to Figure 4, when a user operates the user terminal 301 to set the knowledge reference mode switch 301sw to the "related only" mode on the left side, the top few pieces of knowledge with the highest similarity in the results of the similar sentence search are passed to the generation AI.

[0149] On the other hand, if the user operates the user terminal 301 to set the knowledge reference mode selector switch 301sw to the "all-inclusive" mode on the right side, all of the knowledge displayed in the knowledge list 301b as a result of the drill-down search is passed to the generation AI. Note that if it is not possible to pass all of the knowledge to the generation AI at once because there is a limit to the number of characters of knowledge that can be passed to the generation AI, the knowledge up to the limit is passed to the generation AI in multiple batches.

[0150] Figure 11 is a schematic diagram showing an example of the structure of knowledge data passed to the generation AI. When multiple pieces of knowledge are consecutive, the structure of the knowledge data passed to the generation AI uses a structure in which each piece of knowledge is nested using parentheses such as {}, < >, or ( ) so that it is clear that each piece of knowledge is separate. Also, even within one piece of knowledge, each item is expressed in a nested structure using parentheses so that the knowledge title, knowledge location, creation date, author, text body, etc. can be identified.

[0151] Furthermore, in the case of large documents where the text body is divided into multiple pages, the knowledge data is expressed in an even deeper nested structure so that each page corresponds to the content of the text body, and is passed to the generation AI.

[0152] In addition, when the information processing server 100 passes knowledge data to the information server 200, if the information server 200 determines that the knowledge required for the answer does not exist, the search processing unit 112 of the information processing server 100 may set a prompt so that it can return this information to the information processing server 100.

[0153] Alternatively, when the information server 200 determines that it can provide an answer, the search processing unit 112 of the information processing server 100 may set a prompt so that the answer can return information about the source of the quote, such as the title, location (e.g., page number of a PDF), or timestamp. This makes it possible to show the user, for example, the location where the source document is located, or, in the case of a PDF, the page on which the content is written.

[0154] Instead of the user instructing such a prompt each time, the search processing unit 112 may manage the prompt itself as part of the knowledge, set it as a "basic prompt," and automatically insert it into the knowledge data handed over to the information server 200 each time.

[0155] The information server 200 provides an interactive AI chat service, and by searching knowledge data using the narrowed-down knowledge as a search condition, generates response data representing a response to a query and transmits it to the information processing server 100 (see Figure 8).

[0156] In step S5 of Fig. 10, the search processing unit 112 of the information processing server 100 displays the answer sentence obtained from the information server 200 as a search result on the user terminal 301. By using the above prompt, it is possible to indicate the source of the quote in the answer sentence.

[0157] According to the second embodiment of the present invention, the content collection interface unit 113 stores knowledge data including content collected from various servers within the organization in a knowledge database, and detects updates or deletions to the content and updates the knowledge data. This makes it possible to realize an information processing system or the like that is suitable for sharing or utilizing information stored within an organization, without the need to select necessary information from a large amount of business documents, etc. stored in on-premise devices within the organization and register it in the knowledge database, or to update copies in the knowledge database in the same way as updating the originals on the on-premise devices.

[0158] <Third embodiment> 6 again, the information processing system according to the third embodiment of the present invention is suitable for operating a chatbot for customer support without creating an operation script or preparing learning data, and in addition to the information processing server 100, further includes a content collection server (for web) 500 for collecting and updating knowledge data stored in the information processing server 100. Furthermore, the entire system is provided with a homepage publishing web server 501.

[0159] The content collection server 500 has a function of communicating with the information processing server 100, the information server 200, the user terminals 301, 302, . . . and the homepage publishing web server 501 via a wide area network.

[0160] The content collection server 500 or the homepage publishing web server 501 may be configured as one server device, or may be configured as a plurality of server devices that are connected to each other via a LAN or the like and operate in cooperation with each other.

[0161] The homepage public web server 501 operates public websites such as corporate homepages. The content collection server 500 collects content from such public websites, and may be located anywhere that allows access to the public websites via a network. For example, the content collection server 500 and homepage public web server 501 may be installed in a data center or the like.

[0162] <Content collection server configuration example 2> FIG. 12 is a block diagram showing an example of the configuration of the content collection server 500 shown in FIG. 6, and FIG. 13 is a schematic diagram for explaining an example of the operation of the information processing system according to the third embodiment of the present invention.

[0163] The content collection server 500 shown in Fig. 12 includes a CPU 510 and a storage unit 520 instead of the CPU 410 and the storage unit 420 of the content collection server 400 shown in Fig. 7. Note that some of the components shown in Fig. 12 may be omitted or modified, or other components may be added to the components shown in Fig. 12.

[0164] The communication circuit 104 has a function of performing data communication between the information processing server 100 and the homepage publishing web server 501 via a network such as a wired LAN, a wireless LAN, the Internet, or a mobile communication network. The interface 105 is connected to the operation unit 101 to the communication circuit 104, and transmits various commands and data between them and the CPU 510.

[0165] The CPU 510 performs various calculations and data processing in accordance with various software (including information processing programs) stored in the storage unit 520. In addition to the above software, the storage unit 520 also stores a content database DB5 that stores content collected by the content collection server 500.

[0166] Here, the CPU 510 and the information processing program stored in the storage unit 520 constitute a web crawler 511 and a content collection controller 512 as functional blocks.

[0167] The web crawler 511 extracts content including text, image data, etc. from public websites such as corporate homepages published by the homepage publishing web server 501 via a wide area network such as the Internet.

[0168] In addition, when extracting content, the web crawler 511 also acquires, as additional information about each piece of content, obtainable attributes such as the title, location (e.g., URL or file hierarchy), timestamp (e.g., time information indicating the date the content was created or updated), or author.

[0169] The content collection controller 512 collects the content extracted by the web crawler 511 and stores the collected content in the content database DB5. At that time, the content collection controller 512 also stores additional information such as the title, location, and timestamp in the content database DB5.

[0170] For this purpose, the content collection controller 512 has a timer function that uses the timing signal output from the real-time clock 106, which performs timing operations, and activates the extraction operation of the web crawler 511 at a preset timing, for example, once a day or once an hour.

[0171] In addition, the content collection controller 512 has the same functions as the content collection controller 414 shown in Figure 7, such as detecting that content has been "updated" or "deleted" on a public website based on the extraction results of the web crawler 511, and updating or deleting the corresponding content in the content database DB5.

[0172] The content collection server 500 communicates with the information processing server 100 via a wide area network, and the content collection controller 512 of the content collection server 500 transmits the extracted content to the content collection interface unit 113 of the information processing server 100 based on the extraction results of the web crawler 511. The content collection controller 512 also transmits information about the "updated" or "deleted" content to the content collection interface unit 113.

[0173] As a result, in the information processing server 100, the content collection interface unit 113 stores knowledge data including content collected by the content collection server 500 in a knowledge database DB1 that stores a group of knowledge data related to multiple pieces of knowledge and multiple tags assigned to those pieces of knowledge.

[0174] In addition, the content collection interface unit 113 updates or deletes the content included in the knowledge data stored in the knowledge database DB1 in response to the update or deletion of content by the content collection server 500, thereby updating the knowledge data.

[0175] In this way, the content collection interface unit 113 updates the knowledge data stored in the knowledge database DB1 in synchronization with the content extraction by the web crawler 511. In other respects, the third embodiment may be similar to the first or second embodiment.

[0176] This makes it possible to publish a chatbot on a company's website for customer support. This eliminates the need for the prior script creation, learning, and preparation of a Q&A collection that were previously required. Instead, the web crawler 511 automatically collects HTML content from the website, as well as publicly available PDF documents such as product manuals, and the AI ​​generator responds to inquiries via chat based on this information.

[0177] Furthermore, if there is knowledge (new FAQs) to be added to customer support, it is possible to add that knowledge data to the knowledge database DB1 as needed. Furthermore, using knowledge that includes both content automatically acquired from websites and content directly registered as knowledge data, the generation AI can provide customer support via chat after confirming its operation.

[0178] Furthermore, when a company's homepage or the like is updated, the web crawler 511 is periodically started to reflect the updated information in the knowledge data, so that the generation AI can always provide customer support with the latest information. Also, deletion of content is automatically reflected in the knowledge data, so the generation AI no longer refers to that knowledge.

[0179] <Example of operation of the third embodiment> 13, a web crawler 511 of the content collection server 500 shown in FIG. 12 extracts content from a public website such as a company homepage, and a content collection controller 512 automatically transfers the content to the information processing server 100. The content collection server 500 can also extract publicly available PDFs and automatically transfer them to the information processing server 100. The content collection server 500 may also assign tags to the content according to the URL.

[0180] As a result, the content collection interface unit 113 of the information processing server 100 shown in Figure 9 stores knowledge data including content collected by the content collection server 500 in a knowledge database DB1 that stores a group of knowledge data related to multiple pieces of knowledge and multiple tags assigned to those pieces of knowledge.

[0181] In addition, the content collection interface unit 113 updates or deletes the content included in the knowledge data stored in the knowledge database DB1 in response to the update or deletion of content by the content collection server 500, thereby updating the knowledge data.

[0182] Furthermore, the content collection interface unit 113 adds new FAQs to the knowledge database DB1 as needed, in addition to content automatically acquired from public websites such as corporate homepages, and after verifying their operation, makes the chatbot public.

[0183] On the other hand, a user uses a user terminal 301 to make an inquiry to the knowledge management system "Solution Desk" operated by the information processing server 100 using a conversational inquiry sentence (question sentence or request sentence).

[0184] In response to a query sent from the user terminal 301, the search processing unit 112 of the information processing server 100 uses the knowledge database DB1 updated by the content collection interface unit 113 to generate a narrowing down menu based on multiple tags, and displays a search screen including the narrowing down menu on the user terminal 301.

[0185] When a user inputs a search condition using the user terminal 301, the search processing unit 112 of the information processing server 100 performs a drill-down search based on the search condition sent from the user terminal 301 to narrow down a plurality of knowledge items.

[0186] Furthermore, the generation AI interface unit 111 of the information processing server 100 communicates with the information server 200, which provides the chat function of the generation AI, via a network, and the search processing unit 112 of the information processing server 100 links both the narrowed-down knowledge and the selected or set prompt with the query statement and passes them to the information server 200 via the generation AI interface unit 111.

[0187] The information server 200 provides an interactive AI chat service, and by searching knowledge data using the narrowed-down knowledge as a search condition, generates response sentence data representing a response sentence to a query sentence and transmits it to the information processing server 100.

[0188] The search processing unit 112 of the information processing server 100 displays the chatbot screen on the user terminal 301 together with the answer text obtained from the information server 200. The answer from the generation AI on the chatbot screen displays, as source information, a link to the content that serves as the basis on a public website.

[0189] According to the third embodiment of the present invention, the content collection interface unit 113 stores knowledge data including content such as product manuals collected from publicly available websites in a knowledge database, and detects updates or deletions to the content to update the knowledge data. This makes it possible to publish and operate a chatbot on a website for customer support purposes without creating an operating script or preparing learning data based on a Q&A collection that anticipates what kind of inquiries customers will have in advance.

[0190] Furthermore, in the above-described embodiment, by passing structured knowledge data to the generation AI, the following features can be realized, which can strengthen knowledge and guarantee the basis for the generation AI's answers.

[0191] (1) Users can use the “basic prompts” registered as knowledge to instruct the generation AI to perform the same functions for each inquiry. (2) If knowledge does not exist, or if knowledge exists but is not useful for answering a question, the system can inform the user or record the information in a log. This allows the system to recognize that it will be important to add knowledge for that question in the future, leading to improvements in the system. (3) Generative AI sometimes mixes content from different contexts due to text correlation (a situation known as "AI lying"). However, since reliability of answers is required in customer support, the location of the underlying content (such as a URL) can be provided along with the answer. In other words, by disclosing the answer as "a company guaranteeing the content of original content," it is possible to balance the practical and useful operation of generative AI with the issue of providing guarantees to users.

[0192] The present invention is not limited to the above-described embodiments, and many modifications within the technical spirit of the present invention are possible by those skilled in the art. For example, although the information server 200 and the homepage publishing web server 501 are provided separately from the information processing system in Figures 1 and 6, the information processing system may include them.

[0193] The configurations of the information server 200, the internal portal web server 401 to the groupware server 403, and the homepage publishing web server 501 may be the same as those shown in Fig. 2. Furthermore, it is also possible to omit some of the operations from the embodiments described above. [Industrial Applicability]

[0194] The present invention can be used in information processing systems that perform knowledge management suitable for sharing or utilizing information accumulated within organizations such as companies and groups, information processing systems that perform knowledge management suitable for sharing or utilizing information available from public websites, and information processing programs and information processing methods used in such information processing systems. [Explanation of symbols]

[0195] 100...information processing system, 101...operation unit, 102...display unit, 103...audio input / output unit, 104...communication circuit, 105...interface, 106...real-time clock, 110...CPU, 111...generating AI interface unit, 112...search processing unit, 113...content collection interface unit, 120...storage unit, 130...cache memory, 200...information server, 301, 302,...user terminal, 301a...drill-down navigation screen, 301b...knowledge list, 301c...AI chat screen, 301sw...knowledge reference mode Switch for switching between the two, 400...content collection server, 401...web server for internal portal, 402...internal file server, 403...groupware server, 410...CPU, 411...web crawler, 412...file crawler, 413...groupware crawler, 414...content collection controller, 420...storage unit, 500...content collection server (for web), 501...web server for publishing homepage, 510...CPU, 511...web crawler, 512...content collection crawler, 520...storage unit

Claims

1. An information processing system that provides information desired by a user to a user terminal by communicating with the user terminal via a network, a knowledge database that stores a group of knowledge data relating to a plurality of pieces of knowledge accumulated in an organization and a plurality of tags assigned to the pieces of knowledge; a generation AI interface unit that communicates with an information server that provides a chat function for the generation AI via a network; a search processing unit that generates a narrowing down menu based on the plurality of tags in response to transmission of a question sentence or a request sentence from the user terminal, displays a search screen including the narrowing down menu on the user terminal, narrows down the plurality of knowledges by performing a drill-down search based on the narrowing down conditions transmitted from the user terminal, links a reference to the narrowed down knowledge as a search condition with the question sentence or the request sentence, transfers the narrowed down knowledge to the information server via the generation AI interface unit, and displays an answer sentence obtained from the information server as a search result on the user terminal; An information processing system comprising:

2. An information processing system that provides information desired by a user to a user terminal by communicating with the user terminal via a network, a content collection server that collects content via a network and stores it in a content database, and that, when detecting an update or deletion of already collected content via the network, updates or deletes the corresponding content stored in the content database; a content collection interface unit that stores knowledge data including content collected by the content collection server in a knowledge database that stores a group of knowledge data related to a plurality of pieces of knowledge and a plurality of tags assigned to the pieces of knowledge, and updates or deletes content included in the knowledge data stored in the knowledge database in response to an update or deletion of content by the content collection server, thereby updating the knowledge data; a generation AI interface unit that communicates with an information server that provides a chat function for the generation AI via a network; a search processing unit that generates a narrowing down menu based on the plurality of tags in response to transmission of a question sentence or a request sentence from the user terminal, displays a search screen including the narrowing down menu on the user terminal, narrows down the plurality of knowledges by performing a drill-down search based on the narrowing down conditions transmitted from the user terminal, links a reference to the narrowed down knowledge as a search condition with the question sentence or the request sentence, transfers the narrowed down knowledge to the information server via the generation AI interface unit, and displays an answer sentence obtained from the information server as a search result on the user terminal; An information processing system comprising:

3. 3. The information processing system according to claim 1, wherein the search processing unit controls the information server via the generation AI interface unit to perform a similar sentence search for the question sentence or request sentence within the narrowed-down knowledge.

4. 3. The information processing system of claim 1, wherein the search processing unit displays on the user terminal a search screen in which a drill-down navigation screen for performing a drill-down search using a narrowing menu, a list of knowledge narrowed down by the drill-down search, and an AI chat screen containing a question or request statement and a response statement for the chat function of the generation AI are arranged side by side.

5. 3. The information processing system according to claim 1, wherein the search processing unit displays a search screen on the user terminal, the search screen including a selection means for selecting whether or not to use the group of knowledge data stored in the knowledge database, and when it is selected to use the group of knowledge data stored in the knowledge database, the search screen is linked to the question sentence or request sentence as a search condition to refer to the narrowed-down knowledge, and is passed to the information server via the generation AI interface unit.

6. 3. The information processing system according to claim 1, wherein the search processing unit displays on the user terminal a search screen including a selection means for selecting whether the search condition is to refer to a portion of the plurality of knowledge items narrowed down by performing a drill-down search, or to refer to all of the plurality of knowledge items, and when the search condition is selected to refer to a portion of the plurality of knowledge items, the search processing unit further narrows down the knowledge items by performing a similar sentence search on at least a portion of the message of the question sentence or request sentence within the plurality of knowledge items, and sets the search condition to refer to knowledge items with a high degree of similarity.

7. 3. The information processing system according to claim 1, wherein the search processing unit stores in the knowledge database the filtering conditions used when performing a drill-down search to narrow down the knowledge, displays on the user terminal a search screen including multiple tabs for selecting filtering conditions, and narrows down the multiple knowledge items according to the filtering conditions corresponding to the selected tab.

8. 3. The information processing system according to claim 1, wherein the search processing unit stores the narrowing down conditions used when performing a drill-down search to narrow down knowledge in the knowledge database for each use of the knowledge, and displays a search screen on the user terminal that shows multiple narrowing down conditions for each use of the knowledge.

9. 3. The information processing system according to claim 1, wherein the search processing unit sets inapplicability information to tags of knowledge that has been designated in advance as inapplicable, and when narrowing down the multiple pieces of knowledge stored in the knowledge database, excludes knowledge that has a tag with the inapplicability information set, thereby preventing the knowledge from being passed on to the information server.

10. the knowledge database further stores prompt information relating to a plurality of prompts representing instructions to the information server; An information processing system as described in claim 1 or 2, wherein the search processing unit displays a search screen on the user terminal, including an AI chat screen with options for selecting at least one of the plurality of prompts, and transfers the selected prompt to the information server via the generation AI interface unit.

11. 11. The information processing system of claim 10, wherein the plurality of prompts includes at least one basic prompt that is generally applicable and a plurality of task prompts that are selectable for each task or work.

12. An information processing system as described in claim 1 or 2, wherein the generation AI interface unit configures the information server to launch a log output program when information regarding the question or request statement cannot be obtained even when referring to the narrowed-down knowledge, so that when information regarding the question or request statement cannot be obtained even when referring to the narrowed-down knowledge, the generation AI interface unit receives from the information server, along with historical information regarding the execution of the chat function of the generation AI, information identifying the question or request statement and information indicating that there is no knowledge to refer to, and stores this information in a log database.

13. An information processing system as described in claim 1 or 2, wherein when the generation AI interface unit obtains information regarding the question or request sentence by referring to the narrowed down knowledge, it stores information identifying the referenced knowledge and information identifying the answer sentence in a log database along with historical information regarding the execution of the chat function of the generation AI.

14. The content collection server: a web crawler that extracts content from an internal portal web server that operates a portal site within the organization and provides content to the user terminal via an internal network of the organization; a file crawler that extracts content from an internal file server that stores files of content used within the organization via an internal network; a groupware crawler that extracts content from a groupware server that operates groupware within the organization and provides content to the user terminal via an intra-organization network; 3. The information processing system according to claim 2, further comprising at least one of:

15. The information processing system of claim 2 , wherein the content collection server includes a web crawler that extracts content from publicly available web sites over a wide area network.

16. 3. The information processing system of claim 2, wherein when the search processing unit transfers knowledge data related to the narrowed-down knowledge to the information server via the generation AI interface unit, the search processing unit uses a structured data format including a title, location, timestamp, format, field and its value, or, in the case of a PDF including bookmarks, the entire text extracted by matching the bookmarks with each page, as information on the source of the knowledge.

17. 17. An information processing system according to claim 2 or 16, wherein the search processing unit automatically inserts into the knowledge data a prompt that instructs the information server to reply to the query if it determines that the knowledge required for the answer does not exist, or that instructs the information server to indicate the source of the query in the answer if it determines that the information server can provide an answer.

18. An information processing program used in an information processing system that provides information desired by a user to a user terminal by communicating with the user terminal via a network, a step of generating a narrowing-down menu based on a plurality of tags using a knowledge database that stores a group of knowledge data relating to a plurality of pieces of knowledge accumulated in an organization and a plurality of tags assigned to the pieces of knowledge in response to transmission of a question or request from the user terminal, and displaying a search screen including the narrowing-down menu on the user terminal; a step of narrowing down the plurality of knowledge items by performing a drill-down search based on the narrowing-down conditions transmitted from the user terminal; a step of communicating with an information server that provides a chat function of the generation AI via a network, linking the search criteria of the narrowed-down knowledge with the question or request sentence, and transferring the search criteria to the information server; a step of displaying the answer sentence obtained from the information server on the user terminal as a search result; An information processing program that causes a CPU to execute the above.

19. An information processing program used in an information processing system that provides information desired by a user to a user terminal by communicating with the user terminal via a network, a step (a) of collecting content via a network and storing it in a content database, and, when updating or deleting already collected content via the network is detected, updating or deleting the corresponding content stored in the content database; a step (b) of storing the knowledge data including the content collected in the step (a) in a knowledge database that stores a group of knowledge data relating to a plurality of pieces of knowledge and a plurality of tags assigned to the pieces of knowledge, and updating or deleting the content included in the knowledge data stored in the knowledge database in response to the update or deletion of the content in the step (a), thereby updating the knowledge data; a step (c) of generating a narrowing-down menu based on the plurality of tags in response to transmission of a question or request from the user terminal, and displaying a search screen including the narrowing-down menu on the user terminal; a step (d) of narrowing down the plurality of knowledge items by performing a drill-down search based on the narrowing-down conditions transmitted from the user terminal; (e) communicating with an information server that provides a chat function for the generation AI via a network, and linking the search criteria for referring to the narrowed-down knowledge with the question or request sentence and transferring it to the information server; a step (f) of displaying the answer sentence obtained from the information server on the user terminal as a search result; An information processing program that causes a CPU to execute the above.

20. An information processing method for providing information desired by a user to a user terminal by communicating with the user terminal via a network, comprising: a step of generating a narrowing-down menu based on a plurality of tags using a knowledge database that stores a group of knowledge data relating to a plurality of pieces of knowledge accumulated in an organization and a plurality of tags assigned to the pieces of knowledge in response to transmission of a question or request from the user terminal, and displaying a search screen including the narrowing-down menu on the user terminal; narrowing down the plurality of knowledge items by performing a drill-down search based on the narrowing-down conditions transmitted from the user terminal; communicating with an information server that provides a chat function for the generation AI via a network, linking the search criteria for referring to the narrowed-down knowledge with the question or request sentence and transferring it to the information server; a step of displaying the answer sentence obtained from the information server on the user terminal as a search result; An information processing method comprising:

21. An information processing method for providing information desired by a user to a user terminal by communicating with the user terminal via a network, comprising: a step (a) of collecting content via a network and storing the collected content in a content database, and, when updating or deleting a content that has already been collected via the network, updating or deleting the corresponding content stored in the content database; a step (b) of storing the knowledge data including the content collected in step (a) in a knowledge database that stores a group of knowledge data relating to a plurality of pieces of knowledge and a plurality of tags assigned to the pieces of knowledge, and updating or deleting the content included in the knowledge data stored in the knowledge database in response to the update or deletion of the content in step (a), thereby updating the knowledge data; (c) generating a refinement menu based on the plurality of tags in response to transmission of a question or request from the user terminal, and displaying a search screen including the refinement menu on the user terminal; (d) performing a drill-down search based on the filtering conditions transmitted from the user terminal to narrow down the plurality of knowledge items; (e) communicating with an information server that provides a chat function for the generation AI via a network, and linking the search criteria for referring to the narrowed-down knowledge with the question or request sentence and transferring it to the information server; a step (f) of displaying the answer sentence obtained from the information server on the user terminal as a search result; An information processing method comprising:

Images