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

Abstract

To acquire an appropriate data amount of relevant information from external databases, or the like, for input into a generative AI.SOLUTION: An information processing system manages semantic rule information including semantic rules for acquiring information within a web page whose semantic connections are identified based on the grammar of source codes of the web page as linked information, and navigation rule information including navigation rules that determine the order in which a plurality of web pages are visited when acquiring related information. Then, the system navigates the plurality of web pages based on the navigation rules, and acquires the linked information as the related information from each of the navigated web pages based on the semantic rules.SELECTED DRAWING: Figure 6

Description

[Technical Field]

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

[0002] Generative AI (Artificial Intelligence) is becoming increasingly popular. It extracts features corresponding to information input by users from large amounts of pre-trained data, and derives and outputs appropriate answers.

[0003] Generative AI needs to be trained in advance, but sufficient pre-training is not possible if it is to keep up with information that is not accessible on the internet and has limited visibility, or information that is updated over time, such as the latest updates. As a result, generative AI cannot generate appropriate answers to questions about this information.

[0004] One way to solve this problem is through Retrieval-Augmented Generation (RAG) technology. When generating an answer to a question for a generation AI, RAG retrieves relevant information related to the question from an external database, etc., adds the relevant information to the question, and inputs it into the generation AI, allowing the generation AI to generate and obtain an answer based on the relevant information.

[0005] As a method for obtaining related information from external databases (websites, etc.), a technology has been disclosed that analyzes the tree structure of a website and extracts related data from branch pages by keyword matching (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Patent No. 2020-98596 candidate Summary of the Invention [Problem to be solved by the invention]

[0007] Related information can be obtained from websites using the conventional technology disclosed in Patent Document 1. However, this method is intended to obtain large amounts of data for use in data analysis and machine learning, and is not suitable for RAG.

[0008] That is, in the case of a generation AI that handles text, the number of characters that can be input is limited, and therefore large amounts of data such as those obtained by conventional technologies such as Patent Document 1 cannot be provided as related information to be used in RAG. For this reason, it is desirable to narrow down the related information to the bare minimum necessary so that it fits within the number of characters that can be input to the generation AI and extract it.

[0009] The present invention has been made in consideration of the above-mentioned problems, and aims to obtain an appropriate amount of related information from external databases, etc. and input it into the generation AI. [Means for solving the problem]

[0010] In order to solve the above-mentioned problems, the present invention provides an information processing method executed by an information processing system that acquires related information related to a question from a web page to be input into a generation AI (Artificial Intelligence) that generates an answer to the question, wherein the information processing system has a processor and a memory, and the processor manages semantic rule information including semantic rules for acquiring information within the web page, in which semantic connections are identified based on the grammar of the source code of the web page, as associated information, and crawling rule information including crawling rules that determine the order in which multiple web pages are visited when acquiring the related information, and is characterized in that the information processing system has the following processes: acquire the crawling rules from the crawling rule information; acquire the semantic rules from the semantic rule information; crawl the multiple web pages based on the acquired crawling rules; and acquire the associated information from each of the crawled web pages based on the semantic rules as the related information. [Effects of the Invention]

[0011] According to a representative embodiment of the present invention, it is possible to obtain information related to questions about content that is only available on websites with limited disclosure, or about the latest update information that has just been published on a website, from the website and extract an appropriate amount of data that can be input to the generation AI, thereby realizing a RAG system with high answer accuracy. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram showing the configuration of a computer system according to an embodiment. [Figure 2] FIG. 10 is a diagram showing a chat screen for inputting a question according to the embodiment. [Figure 3] FIG. 10 is a diagram showing a chat screen for outputting an answer according to the embodiment. [Figure 4] 10 is a flowchart showing a question and answer process according to the embodiment. [Figure 5] FIG. 10 is a diagram showing an outline of a keyword extraction process in the question and answer process according to the embodiment. [Figure 6] 10 is a flowchart showing a related data extraction process of the question and answer process according to the embodiment. [Figure 7A] FIG. 10 is a diagram showing an outline of data extraction processing according to the embodiment. [Figure 7B] FIG. 10 is a diagram showing an outline of data extraction processing according to the embodiment. [Figure 8] FIG. 10 is a diagram showing an outline of pruning processing in the related data extraction processing according to the embodiment. [Figure 9] FIG. 1 is a diagram (part 1) showing an outline of a process for additionally acquiring related information according to an embodiment. [Figure 10] FIG. 2 is a diagram (part 2) showing an outline of a process for additionally acquiring related information according to the embodiment. [Figure 11] FIG. 10 is a diagram (part 3) showing an outline of a process for additionally acquiring related information according to the embodiment. [Figure 12] FIG. 10 is a diagram showing an outline of a process for recursively acquiring related information according to the embodiment. [Figure 13]FIG. 10 is a diagram showing an outline of a process for sorting the execution order of recursively acquired related information according to the embodiment. [Figure 14] FIG. 10 is a diagram showing an outline of a process for sorting an execution order based on semantic rules according to an embodiment. [Figure 15] FIG. 10 is a diagram showing an outline of a process for sorting an execution order based on semantic rules according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0013] In the following description, "memory" refers to one or more memory devices, typically a primary storage device. At least one memory device in the memory may be a volatile memory device or a non-volatile memory device.

[0014] In the following description, a "persistent storage device" refers to one or more persistent storage devices. A persistent storage device is typically a non-volatile storage device (e.g., an auxiliary storage device), and specifically, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

[0015] In the following description, the term "storage device" may refer to either the above-mentioned "memory" or the above-mentioned "persistent storage device."

[0016] In the following description, a "processor" refers to one or more processor devices. The at least one processor device is typically a microprocessor device such as a CPU (Central Processing Unit), but may also be another type of processor device such as a GPU (Graphics Processing Unit). The at least one processor device may be a single-core or multi-core. The at least one processor device may also be a processor core. The at least one processor device may also be a processor device in a broader sense, such as a hardware circuit (e.g., an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit)) that performs part or all of the processing.

[0017] In the following description, information that provides an output in response to an input may be described using expressions such as "xxx table," but this information may be data of any structure, or may be a learning model such as a neural network that generates an output in response to an input. Therefore, "xxx table" can be rephrased as "xxx information." In the following description, the configuration of each table is an example, and a table may be divided into two or more tables, or all or part of two or more tables may be one table, or may include several data fields not shown.

[0018] In the following description, processing may be described using a "program" as the subject. However, since a program is executed by a processor to perform a predetermined process using a storage device and / or an interface device, etc., as appropriate, the subject of the process may also be the processor (or a device such as a controller having the processor). A program may be installed in a device such as a computer from a program source. The program source may be, for example, a program distribution server or a computer-readable (e.g., non-transitory) recording medium. In the following description, two or more programs may be realized as one program, and one program may be realized as two or more programs.

[0019] In the following description, functions may be described using expressions such as "xxx unit," but the functions may be realized by one or more computer programs being executed by a processor, or may be realized by one or more hardware circuits (for example, FPGA or ASIC). When a function is realized by a program being executed by a processor, the specified processing is performed using a storage device and / or an interface device, etc., as appropriate, and therefore the function may be at least a part of the processor.

[0020] Furthermore, a process described using a function as the subject may be a process performed by a processor or a device having such a processor. A program may also be installed from a program source. The program source may be, for example, a program distribution computer or a computer-readable recording medium (e.g., a non-transitory recording medium). The description of each function is an example, and multiple functions may be combined into one function, or one function may be divided into multiple functions.

[0021] In the following description, a "computer system" is a system including one or more physical computers. The physical computers may be general-purpose computers or special-purpose computers.

[0022] In addition, the control lines and information lines shown are those that are considered necessary for explanation, and do not necessarily represent all the control lines and information lines that are necessary for implementation. In reality, it can be assumed that almost all components are interconnected.

[0023] Hereinafter, a collection of one or more computers that manage an information processing system and display the display information of this embodiment may be referred to as a management system. When a management computer (hereinafter referred to as a management computer) displays the display information, the management computer is the management system. Furthermore, the combination of a management computer and a display computer is also a management system. Furthermore, in order to speed up and increase the reliability of management processing, multiple computers may be used to achieve processing equivalent to that of a management computer, in which case the multiple computers (including the display computer if the display is performed by the display computer) are the management system. The management computer is an example of an information processing system that executes an information processing method based on an information processing program.

[0024] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

[0025] (Configuration of Overall System S According to the Embodiment) 1 is a diagram showing the configuration of an overall system S according to an embodiment. The overall system S is a system in which a management computer 101, an operation terminal 102, a website 104, and a generation AI 105 are interconnected via a network 103.

[0026] The management computer 101 receives a question input by a user from the input / output device 141 of the operation terminal 102. The management computer 101 executes a keyword extraction process 121, a related data extraction process 122, and a generation AI query process 123, which are recorded in the storage device 112, using the processor 111. The management computer 101 extracts information related to the question (hereinafter referred to as "related information") from the information acquired from the website 104 by executing these processes. The management computer 101 provides the related information and the question to the generation AI 105, and displays the output result of the generation AI 105 on the input / output device 141.

[0027] The management computer 101 has a processor 111 and a storage device 112. The management computer 101 may have an input / output device (not shown). Here, the input / output device is, for example, a touch panel, a display, a keyboard, a mouse, etc. The processor 111 executes a keyword extraction process 121, a related data extraction process 122, and a generation AI query process 123.

[0028] The storage device 112 stores a keyword extraction process 121, a related data extraction process 122, a generation AI query process 123, a generation AI question template 131, semantic rule information 132, and navigation rule information 133. The processes and information stored in the storage device 112 may be stored in different storage devices, or may be stored in a storage device (not shown) connected via the network 103.

[0029] The generation AI question template 131 is template information for a question text that is used when the keyword extraction process 121, the related data extraction process 122, and the generation AI query process 123 generate a question to be sent to the generation AI 105 via the network 103. The question text is generated by inputting the necessary information according to the generation AI question template 131.

[0030] The semantic rule information 132 manages semantic rules that record patterns for extracting related information related to a question from a web page when the related data extraction process 122 obtains information from the website 104 via the network 103. The semantic rules are rules that indicate the semantic connections of information based on the correspondence between elements on a web page, for example. Each semantic rule included in the semantic rule information 132 may be held, for example, as a function for scraping a web page.

[0031] The navigation rule information 133 manages the navigation rule information that records the acquisition order pattern for acquiring HTML information as new page information, such as page forwarding within a website or acquiring a pop-up screen, when the related data extraction process 122 acquires information from the website 104 via the network 103. The navigation rule can also be said to be a rule that determines the navigation order for browsing web pages based on page forwarding and transition rules to subordinate pages. For example, it may be stored as program information that controls a web driver that accesses web pages.

[0032] The semantic rules and traversal rules are dependent on the source code of the web page.

[0033] The generated AI question template 131, semantic rule information 132, and navigation rule information 133 may be created manually, created by some kind of program, or stored in some alternative format.

[0034] The network 103 is a communication path connected by wire or wirelessly, such as, but not limited to, a wired or wireless LAN (Local Area Network).

[0035] Website 104 is a server that runs on multiple computer systems (not shown) on the Internet connected via network 103, and stores and provides information provided by companies, individuals, public institutions, or other unspecified number of people.

[0036] The generation AI 105 is a type of artificial intelligence that runs on multiple computer systems (not shown) on the Internet connected via the network 103, and outputs voice, images, and text corresponding to input based on what it has learned in advance. In the present invention, this refers specifically to artificial intelligence that handles text.

[0037] (Chat screen 200 for entering a question according to an embodiment) 2 is a diagram showing a chat screen 200 for entering a question according to an embodiment of the present invention. The chat screen 200 is displayed on the input / output device 141 of the operation terminal 102. The chat screen 200 includes an input field 201 for a user to enter a question, and a send button 202 for sending the entered question.

[0038] (Chat screen 200 for answer output according to the embodiment) Fig. 3 is a diagram showing a chat screen 200 for outputting answers according to an embodiment. Fig. 3 is a screen similar to the chat screen 200 exemplified in Fig. 2, and includes an input field 201 for a user to input a question and a send button 202 for sending the input question. The chat screen 200 in Fig. 3 newly displays a display field 301 for a question history received as input from a user, and a display field 302 for a computer system's answer to the question.

[0039] (Question and answer process for answering user's questions according to the embodiment) 4 is a flowchart showing an example of a question and answer processing procedure in which the management computer 101 according to the embodiment answers a question from a user. The question and answer processing flow 400 shown in this flowchart may be executed by receiving a question entered via the chat screen 200 displayed on the input / output device 141 of the operation terminal 102. Alternatively, it may be executed by an instruction from some kind of program.

[0040] 4, the processor 111 of the management computer 101 executes a keyword extraction process (S401), a related data extraction process (S402), and a generation AI query process (S403). The question and answer process flow 400 may include other processing steps not shown, and some processes may be omitted, replaced with alternative processes, their execution order may be changed, or they may be executed in parallel, as long as no discrepancies in input and output occur.

[0041] The keyword extraction process (S401) will be described later using Fig. 5, and the related data extraction process (S402) will be described later using Fig. 6. The related information extracted by the related data extraction process can be used for manual analysis by a user or for other purposes.

[0042] In the generation AI query process (S403), the processor 111 combines the question received via the chat screen 200 with the related information of the question acquired from the website by the related data extraction process (S402), and creates a question for the generation AI according to the generation AI question template 131. Then, the processor 111 inputs the question to the generation AI to query it, and acquires the output of the generation AI.

[0043] (Processing Overview of Keyword Extraction Processing in Question and Answer Processing According to an Embodiment) 5 is a diagram showing an overview of the keyword extraction process (S401) of the question and answer processing flow 400 according to the embodiment. In the keyword extraction process 121, the processor 111 executes the generation AI query processing module 501, combines a template 502 stored in the generation AI question template 131 with a question 503 entered in the input field 201 to create a question for the generation AI 105, and inputs the question to the generation AI 105. The processor 111 then receives an output 521 from the generation AI 105 to extract keywords necessary for collecting related information needed to answer the question.

[0044] In Figure 5, the keyword extraction process is exemplified as a process using generation AI 105, but instead of using generation AI 105, it may be replaced with other means such as rule-based data extraction or matching with a keyword list, or it may be performed by combining several means.

[0045] (Related Data Extraction Process According to the Embodiment) 6 is a flowchart illustrating an example of the processing procedure of the related data extraction process (S402) of the question and answer process flow 400 according to the embodiment. In the related data extraction process flow 600 illustrated in this flowchart, the processor 111 executes a navigation rule acquisition process (S601), a semantic rule acquisition process (S602), a data extraction process (S603), a pruning process of acquired information (S604), a determination of whether additional Web information should be acquired (S605), and a process of additional Web information acquisition (S606). The related data extraction process flow 600 may include other processing steps (not shown), and some processes may be omitted, replaced with alternative processes, their execution order may be changed, or executed in parallel, as long as no discrepancies in input and output occur.

[0046] In the navigation rule acquisition process (S601), the processor 111 acquires an appropriate navigation rule according to the keyword of the question acquired in the keyword extraction process (S401) and the target website required to collect related information from the navigation rule information 133. For example, an arbitrary semantic rule may be defined for each website or keyword, or the rule may be acquired by a method such as a keyword search.

[0047] In the semantic rule acquisition process (S602), the processor 111 acquires an appropriate semantic rule from the semantic rule information 132 according to the keywords of the question acquired in the keyword extraction process (S401) and the websites required to collect related information. For example, an arbitrary semantic rule may be defined for each website or keyword, or it may be acquired by a method such as a keyword search.

[0048] In the data extraction process (S603), the processor 111 acquires semantically related information from the source code of the Web page in accordance with the semantic rules acquired in the semantic rule acquisition process (S602).

[0049] 7A and 7B are schematic diagrams illustrating the data extraction process (S603). For example, when there is a website screen that introduces various products by product group in tab display, as illustrated in web screen 701, information is extracted according to semantic rule 702, and extracted information 703 is extracted.

[0050] The semantic rules 702 are rules for identifying semantic connections between pieces of information based on the grammar of the source code of a Web page, for example, and acquiring semantically connected information within the Web page as connected information.

[0051] In the example of Figure 7A, semantic rule 702 is "obtain the Nth (N = 1, 2, ...) element of the group of elements having the X class of HTML, the Nth heading of the group of elements having the Y class, and information on the Z class." Semantic rule 702 indicates that "product group," "product name," "product description," ... are obtained based on correspondence 721a, 721b, 721c, ... between web screen 701 and source code 721 for each Nth element of tab 711 and information 712. As a result of extraction based on semantic rule 702, extracted information 703 including "product group," "product name," and "product description" is extracted as shown in the example of Figure 7B.

[0052] HTML, the source code of modern websites, has a large number of characters per page, exceeding hundreds of thousands to millions of characters. Furthermore, because display is controlled by a combination of CSS and Javascript (registered trademark), it is difficult to obtain semantic connections using HTML alone. Segmenting source code by the number of characters, a common method, rarely results in clustering of related information. Furthermore, semantic connections are often expressed by display switching using CSS or Javascript (registered trademark), making it difficult to gather related information using only the proximity of HTML source code or general keyword matching. In this embodiment, semantically related information written far apart in HTML can be collected in accordance with semantic rules.

[0053] 8 is a schematic diagram illustrating the pruning process (S604) of the related data extraction process according to the embodiment. For example, in the generation AI query process 801, a question for the generation AI 105 is created by combining template information 802 recorded in the generation AI question template 131, question keywords 803 acquired in the keyword extraction process (S401), and related information 804 acquired in the data extraction process (S603). The question is then posed to the generation AI 105 to acquire related information 821. Furthermore, the related information acquired from the website 104 is pruned, leaving only the related information 821 that is directly related to the answer to the question.

[0054] Note that Figure 8 illustrates a pruning process (S604) using generation AI 105, but instead of using generation AI 105, it may be replaced with another means such as rule-based or string matching data pruning, or several means may be combined.

[0055] In the process for determining whether additional Web information should be acquired (S605), it is determined whether there is additional information that should be acquired from the website, based on the navigation rules acquired in the process for acquiring navigation rules (S602). For example, it is determined whether additional information should be acquired by paging forward, or whether more detailed information should be acquired by drilling down on the website. If the process for determining whether additional information should be acquired (S605) finds that there is detailed information that can be acquired by drilling down on the website, the process proceeds to S606; if not, the related data extraction process is terminated.

[0056] 9, 10, and 11 are diagrams showing an outline (parts 1 to 3) of the process of additionally acquiring related information through the process of additionally acquiring Web information (S606).

[0057] Fig. 9 shows an overview of the process for additionally acquiring the first piece of related information. Fig. 9 is an explanatory diagram of an example of the process for additionally acquiring Web information (S606), in which new HTML is acquired from a Web site in accordance with the navigation rule, and a row of related information is additionally acquired. In Fig. 9, when a page forward button 901a is included in an originally displayed Web page 901, the page forward button 901a is pressed to acquire Web page 902. At that time, additional related information 911a (the hatched portion shown in related information 912) acquired from the newly acquired Web page 902 is added to the related information 911 acquired and pruned from the originally displayed Web page 901.

[0058] Fig. 10 shows an overview of the process for additionally acquiring the second piece of related information. Fig. 10 is an explanatory diagram of an example in which, in the process for additionally acquiring Web information (S606), new HTML is acquired from a Web site in accordance with the navigation rule, and a row of related information is additionally acquired. In Fig. 10, a link button 1001a to another page containing more detailed information is pressed in the originally displayed Web page 1001, and Web page 1002 is acquired. At that time, additional related information 1011a (the hatched portion shown in related information 1012) acquired from the newly acquired Web page 1002 is added to the related information 1011 acquired and pruned from the originally displayed Web page 1001.

[0059] Figure 11 shows an overview of the third process, the additional acquisition of related information. Figure 11 illustrates an example of the process for selecting an appropriate link according to a navigation rule in the additional Web information acquisition process (S606) when the structure of the Web site is in a drill-down format with links for obtaining more detailed information, and when conditional branching of the links to be followed occurs depending on the conditions. The generation AI query processing module 1101 combines the generation AI question template information 1102, the condition information 1103, and the HTML information 1104 containing the links obtained from the Web page according to the semantic rules to create a question for the generation AI 105. Then, the module queries the generation AI 105 and obtains from the generation AI 105 answer information 1121 containing the link information to be obtained.

[0060] This allows link information unrelated to the system to be excluded, appropriate links to be selected, and unnecessary information that may cause hallucination to be excluded from the related information provided by RAG. The condition information 1103 handled here is information necessary for selecting a conditional branch, such as information about the operating system and software version, and may be entered manually, may be predefined in a configuration file, or may be obtained by some program such as a system call.

[0061] 11 also illustrates the process of selecting a destination link in the additional Web information acquisition process (S604) using the generation AI 105. However, instead of using the generation AI 105, it may be replaced with another means, such as rule-based data pruning, or several means may be combined.

[0062] As described with reference to Figures 6, 7A, 7B, 8, 9, 10, and 11, the related data extraction process 122 can acquire information related to a question from a website. For example, when considering software update patch support, there may be prerequisite patches for applying the update patch and post-patches for stable operation. It is also conceivable that prerequisite patches and post-patches for the prerequisite and post-patches may also have prerequisite and post-patches. To handle such cases, the aforementioned related data extraction process 122 may be executed recursively to acquire more comprehensive related information.

[0063] 12 is a diagram showing an overview of the recursive related information acquisition process according to the embodiment. As shown in FIG. 12, by recursively executing the related data extraction process 122, it is possible to comprehensively acquire prerequisite patches and post-update patches. In the recursive related information acquisition, a table shown in table 1200 is provided as an excerpt of information acquired from a website using semantic rules. Table 1200 includes an update patch 1201, a prerequisite patch 1202, and a post-update patch 1203.

[0064] Table 1200 stores data on related information (prerequisite patch 1202 and post-update patch 1203) acquired using update patch 1201 as input during the first execution of related data extraction process 122. Prerequisite patch 1202 is a patch that is assumed to have already been executed when update patch 1201 is executed. Post-update patch 1203 is a patch that is executed immediately after update patch 1201 is executed.

[0065] Table 1210 stores data on related information (prerequisite patches and post-patches) that is further obtained by executing the related data extraction process 122 for the second time, with the prerequisite patches and post-patches obtained in the first execution being the target update patches.

[0066] Similarly, table 1220 stores data on related information (prerequisite patches and post-update patches) obtained by the third execution of related data extraction process 122, with the prerequisite patches and post-update patches obtained in the second execution being the target update patches. Note that if both the prerequisite patch and post-update patch do not exist, recursive related information acquisition is not performed, and the patch may be deleted from tables 1200, 1210, 1220, etc.

[0067] As illustrated in Figure 12, by recursively executing the related data extraction process, it is possible to comprehensively obtain information necessary for software updates. Furthermore, as illustrated in Figure 12, when executing the recursive process, it is desirable to set some kind of termination condition to prevent the process from falling into an infinite loop. For example, the termination condition for the execution of the recursive process can be one of the following: setting a maximum number of recursive executions in advance; storing a list of patches that have already been applied so that prerequisite patches and post-patches for applied patches are not obtained by the recursive process; or continuing until no prerequisite patches or post-patches at the end of the list appear, as in table 1220 in Figure 12.

[0068] FIG. 13 is a diagram showing an outline of the process for rearranging the execution order of recursively acquired related information according to an embodiment. Table 1301 in FIG. 13 shows multiple combinations of target patches, prerequisite patches, and post-patches acquired by recursive execution of the related data extraction process 122. Because the acquisition order of patches differs from the execution order of patches that will not cause problems in actual operation, the order may be rearranged based on semantic rules to ensure consistency in the execution order before providing the related information to the generation AI 105, as exemplified in table 1302. An example of rearranging the execution order will be described using FIGS. 14 and 15.

[0069] 14 is a diagram illustrating an outline of the process of sorting the execution order based on semantic rules according to an embodiment. When sorting the execution order of update patches, prerequisite patches, and post-update patches based on semantic rules, a table is first prepared, as shown in table 1400, which includes update patch 1401, prerequisite patch 1402, and post-update patch 1403. Then, the data stored in the first row is extracted from table 1301, and an execution order table 1410 is created in which the prerequisite patch 1402, the target update patch 1401, and the post-update patch 1403 are sorted in this order. As shown in table 1500 in FIG. 15, the first row of table 1400 has been extracted, and the data in the first row has been removed.

[0070] In the process of rearranging the patches into execution order, the first row is again extracted from table 1500, and a temporary table 1520 is created in which the patches are rearranged in the order of prerequisite patch 1502, update patch 1501, and post-patches 1503, just as when execution order table 1410 was created. The update patch sequence used to create temporary table 1520 is then compared with the elements in execution order table 1510 (which is the same as execution order table 1410), and the elements that match are deleted. Temporary table 1520 is inserted in the deleted positions, and a new execution order table 1530 is created. The process illustrated in FIG. 15 is repeated until table 1500 is empty, allowing the patches to be rearranged into an order that does not cause any problems as an execution order.

[0071] In the generation AI query process 123, the relevant information related to the question extracted from the website 104 by the above-mentioned process and the question information received from the input / output device 141 of the operation terminal 102 are shaped according to the information described in the generation AI question template 131 to create a question to be input to the generation AI 105. Then, the generation AI 105 is queried, an output is received, and the received result is displayed on the input / output device 141 of the operation terminal 102.

[0072] (Effects of the embodiment) In the above-described embodiment, one or more web pages are crawled based on the crawling rules, and related information is obtained from each crawled web page based on the semantic rules. Therefore, according to the embodiment, it is possible to obtain chunks of useful information that are semantically connected from a website in accordance with the semantic rules and the crawling rules. Furthermore, it is possible to extract only the appropriate related information in an amount that fits within the number of tokens of the generation AI, enabling highly accurate answers to be obtained using RAG.

[0073] In the above-described embodiment, an instruction to delete information unrelated to the keyword from related information acquired from one or more web pages is input to the generation AI, and the generation AI acquires related information from which information unrelated to the keyword has been deleted in accordance with the instruction. Thus, according to the embodiment, the relevance of each web page to the question is determined, narrowing down the information, and collecting only the information necessary to answer the question as related information.

[0074] In the above-described embodiment, new related information is additionally acquired based on semantic rules from the destination web page transitioned to based on the navigation rules, thereby making it possible to collect related information more widely and comprehensively.

[0075] In the above-described embodiment, detailed information about related information obtained from a transition source web page based on a navigation rule is additionally obtained from a transition destination web page transitioned from the transition source web page based on a semantic rule, thereby making it possible to collect related information more widely, without omissions, and with high comprehensiveness.

[0076] In the above-described embodiment, an instruction to extract related information that matches predetermined condition information from related information obtained from one or more web pages is input to the generation AI, and related information that matches the predetermined condition information is output from the generation AI in response to the instruction. Therefore, according to the embodiment, related information is narrowed down to highly useful information, and highly accurate answers can be obtained using RAG.

[0077] In the above-described embodiment, the second related information is acquired based on the first related information, the navigation rule, and the semantic rule. Therefore, according to the embodiment, by recursively acquiring further related information based on the acquired result of the related information, it is possible to comprehensively acquire related information.

[0078] In the above-described embodiment, a plurality of pieces of information are acquired as the first related information and the second related information, and the processing order of the plurality of pieces of information is changed based on semantic rules. Therefore, according to the embodiment, a plurality of processes (such as software update patches) with a predetermined execution order can be arranged in an executable manner from the acquired related information.

[0079] In the above-described embodiment, the acquired related information and the question are combined to generate a question, the question is input to the generation AI, and an answer is output by the generation AI according to the question. Therefore, according to the embodiment, a highly accurate answer can be obtained using RAG.

[0080] As described above, according to the embodiment of the present invention, the management computer 101 extracts information related to a question based on semantic rules for acquiring data from websites, navigation rules, and query template information for the generation AI, and queries the generation AI by combining the question and related information to obtain an answer, thereby obtaining a correct answer that the generation AI could not have given without the related knowledge. Furthermore, by deleting information that is not directly related to the question from the related information, the risk of the generation AI giving an incorrect answer can be avoided.

[0081] The present invention is not limited to the above-described embodiments, and includes various modifications and equivalent configurations within the spirit of the appended claims. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to configurations including all of the described configurations. Furthermore, part of the configuration of one embodiment may be replaced with the configuration of another embodiment. Furthermore, the configuration of another embodiment may be added to the configuration of one embodiment. Furthermore, part of the configuration of each embodiment may be added, deleted, or replaced with other configurations. [Explanation of symbols]

[0082] S: Overall system, 101: Management computer, 104: Website, 111: Processor, 132: Semantic rule information, 133: Navigation rule information, 105: Generative AI.

Claims

1. An information processing method executed by an information processing system that acquires, from a web page, related information related to a question to be input to a generation AI (Artificial Intelligence) that generates an answer to the question, comprising: the information processing system includes a processor and a memory; the processor: semantic rule information including semantic rules for acquiring information in the web page, in which semantic connections are identified based on the grammar of the source code of the web page, as linked information; managing browsing rule information including browsing rules that define the order in which the plurality of web pages should be browsed when acquiring the related information; acquiring the circulation rule from the circulation rule information; acquiring the semantic rule from the semantic rule information; The plurality of web pages are crawled based on the acquired crawling rule, and the linked information is acquired as the related information from each of the crawled web pages based on the semantic rule. An information processing method characterized by comprising each process.

2. 2. The information processing method according to claim 1, the processor: extracting keywords from a question sentence expressing the question; inputting an instruction to the generation AI to delete information not related to the keyword from the related information acquired from the plurality of web pages; The generating AI acquires the related information from which information unrelated to the keyword has been deleted in response to the instruction. An information processing method characterized by comprising each process.

3. 2. The information processing method according to claim 1, the processor: The related information is newly and additionally acquired based on the semantic rule from the web page to which the transition is made based on the navigation rule.

2. An information processing method comprising:

4. 2. The information processing method according to claim 1, the processor: Detailed information of the related information acquired from the transition source web page based on the navigation rule is additionally acquired from the transition destination web page transitioned from the transition source web page based on the navigation rule, based on the semantic rule.

2. An information processing method comprising:

5. 2. The information processing method according to claim 1, the processor: Accept input of predetermined condition information, inputting an instruction to the generation AI to extract the related information that matches the predetermined condition information from the related information acquired from the plurality of web pages; Acquire the related information that matches the predetermined condition information output from the generation AI in response to the instruction. An information processing method characterized by comprising each process.

6. 2. The information processing method according to claim 1, the processor: Obtaining a first piece of related information; Acquiring second related information based on the first related information, the navigation rule, and the semantic rule. An information processing method characterized by comprising each process.

7. 7. The information processing method according to claim 6, the processor: acquiring a plurality of pieces of information as the first related information and the second related information; The processing order of the plurality of pieces of information is changed based on the semantic rules. An information processing method characterized by comprising each process.

8. 2. The information processing method according to claim 1, the processor: generating a question sentence by combining the acquired related information with the question; Input the question sentence into the generation AI, Obtain the answer output by the generation AI in response to the question. An information processing method characterized by comprising each process.

9. An information processing program for causing a computer to execute the information processing method according to any one of claims 1 to 8.

10. An information processing system that executes the information processing method according to any one of claims 1 to 8.

Images