Information processing device, information processing method, and program

Abstract

The present invention enables easy and efficient synchronization between a document and a database.　A server 1 comprises: an item definition acquisition unit 51; information extraction means such as an information extraction control unit 52 and an extraction model 70; a DB generation unit 53; and an update control unit 54. The item definition acquisition unit 51 acquires a list L that defines items corresponding to information to be extracted from a web document W. The information extraction control unit 52 and the extraction model 70 extract information corresponding to each of the items of the list L from the web document W. The DB generation unit 53 generates an extraction item DB 72 in which the items defined by the list L are associated with the extracted information. When the update of the web document W or the list L has been detected, the update control unit 54 causes the information extraction means to re-execute extraction processing, and updates the extraction item DB 72 on the basis of the result of the re-execution.

Description

Information processing device, information processing method, and program

[0001] The present invention relates to an information processing device, an information processing method, and a program.

[0002] Conventionally, there are technologies relating to databases and document systems, such as Notion and Filemaker (see, for example, Patent Document 1).

[0003] Japanese Patent Application Laid-Open No. 2021-149560

[0004] However, in the conventional technology described above, data is first collected and a table is created, and then content such as documents is generated. This creates a need for a solution to the problem of complicated data input UIs and document output UIs, but conventional technology, including the technology of Patent Document 1, is unable to fully meet this need.

[0005] The present invention has been made in view of the above circumstances, and has as its object to enable synchronization of content and a database to be performed simply and efficiently.

[0006] In order to achieve the above object, an information processing device of one aspect of the present invention comprises: item definition acquisition means for acquiring item definition information that defines items corresponding to information to be extracted from content; information extraction means for extracting the information related to the items of the item definition information from the content; structured data generation means for generating structured data that associates the items defined by the item definition information with the extracted information; and update control means for, when an update of the content or the item definition information is detected, causing the information extraction means to function and re-execute the extraction process, and updating the structured data based on the result of the re-execution. Each of the information processing devices of one aspect of the present invention is a method and a program corresponding to the information processing method and program of one aspect of the present invention, respectively.

[0007] According to the present invention, it is possible to easily and efficiently synchronize content with a database.

[0008] 2A to 2C are diagrams showing an overview of the present service that can be realized by an information processing system to which a server according to an embodiment of the information processing device of the present invention is applied. FIG. 2B is a diagram showing an example of the configuration of an information processing system to which a server according to an embodiment of the information processing device of the present invention is applied. FIG. 3 is a diagram showing the hardware configuration of a server according to an embodiment of the present invention. FIG. 4 is a functional block diagram of a server according to an embodiment of the present invention. FIG. 5 is a flowchart showing the processing operation of the server of FIGS. 2 to 4. FIG. 6 is a flowchart showing detailed processing operation of the server of FIGS. 2 to 4. FIG. 7 is a diagram showing an example of a list of items to be extracted that the server of FIGS. 2 to 4 acquires. FIG. 8 is a diagram showing the processing by the server of FIGS. 2 to 4 to extract item information from the document body. FIG. 9 is a diagram showing automatic reflection processing (when a web document is updated) by the server of FIGS. 2 to 4. FIG. 10 is a diagram showing automatic reflection processing (when a list is updated) by the server of FIGS. 2 to 4. FIG. 11 is a diagram showing technical details of the extraction processing by the server of FIGS. 2 to 4. FIG. 12 is a diagram showing an example of display of structured data by the server of FIGS. 2 to 4. FIG. 13 is a diagram showing display updates when an item is added to the server of FIGS. 2 to 4. FIG. 14 is a diagram showing the sorting function of the server of FIGS. 2 to 4. FIG. 15 is a diagram showing a first example of the filtering function of the server of FIGS. 2 to 4. FIG. 16 is a diagram showing a second example of the filtering function of the server of FIGS. 2 to 4. Figure 17 shows an example of an annotated list of the servers of Figures 2 to 4. Figure 18 shows the annotated list of Figure 17 and prompts generated from that list. Figure 19 shows an example of implicit information extraction of the servers of Figures 2 to 4.

[0009] An embodiment of the present invention will be described below with reference to the drawings. First, with reference to Fig. 1, an overview of a service (hereinafter referred to as "this service") that can be realized by an information processing system (see Fig. 2 described later) to which a server according to an embodiment of the information processing device of the present invention is applied will be described.

[0010] FIG. 1 is a diagram showing an outline of the service that can be realized by an information processing system to which a server according to an embodiment of the information processing device of the present invention is applied.

[0011] This service allows users to build databases that fit their natural workflow, regardless of the order in which documents are created and items are defined.

[0012] 1 shows an application running on server 1. User A creates web document W and creates list L, which is item definition information. List L defines items such as "Name," "Account Name," "Department," "Email Address," and "Telephone Number." In step S1, user A sends list L to server 1, along with an instruction to extract information corresponding to (related to) each of the items in list L. When server 1 acquires list L, server 1 executes a process to extract information corresponding to the items in list L from web document W.

[0013] In this case, the server 1 first analyzes the web document W, extracts information corresponding to each item in the list L (step S2), and executes a process (step S3) to display the extracted results as a panel P. For example, in response to an extraction instruction such as "person," information such as "Yamada Taro" in the "Name" field and "Development Department" in the "Department" field is extracted and displayed on the panel P.

[0014] When User A or the web document server updates Web document W, or when User A updates list L (step S4), Server 1 detects the update (step S5), automatically executes a re-extraction process (step S6), and updates the contents of panel P (step S7). Information extracted from multiple Web documents is integrated and displayed in a tabular form as a database. User A can sort and filter the database displayed in tabular form to find the information he or she needs.

[0015] In this way, the server 1 automatically extracts related information from the list L of items defined by the user A from the web document W created in free format, and generates a structured database (see Table 102 in Figure 12). This database is automatically updated as the document or list is updated, so the latest information is always maintained.

[0016] In this way, it is possible to build a database that suits the user's natural work flow, regardless of the order in which documents are created and item definitions are performed.In addition, because structured data is automatically updated as documents and item definitions are updated, the database can always be kept up to date, preventing data inconsistencies caused by manual updates.

[0017] Although not shown in Figure 1, this service may use an extraction model 70 (see Figure 4) as a machine learning model to infer and extract information that is not explicitly stated in the content from the context.

[0018] That is, the server 1 can use the machine learning model to infer and extract information that is not explicitly stated in the content from the context, thereby enabling automatic information complementation to extract information that is not explicitly stated by the document creator.

[0019] In this service, although not shown in Fig. 1, annotation information for each item may be included in the item definition information to adjust the extraction results. That is, the item definition information includes annotation information for each item, and the server 1 can adjust the extraction results based on the annotation information. This allows the user to flexibly adjust the format and content of the extraction results.

[0020] Although not shown in Fig. 1, this service may display structured data in a tabular format, allowing sorting and filtering operations. That is, the server 1 may be equipped with a display control unit that displays structured data in a tabular format and an operation reception unit that receives sorting or filtering operations on the structured data. This allows the generated database to be efficiently operated and analyzed.

[0021] Although not shown in Figure 1, this service may integrate information extracted from multiple pieces of content into a single database. That is, the server 1 may be equipped with an integration unit that integrates the relevant information extracted from multiple pieces of content into a single structured data. This makes it possible to build a unified database from multiple documents.

[0022] Although not shown in Fig. 1, this service may allow content creation and item definition information creation to be performed in any order. In other words, content creation and item definition information creation can be performed in any order on the server 1. This allows for flexible support of the user's workflow and enables database construction in a more natural format.

[0023] Although not shown in Fig. 1, this service may re-execute the extraction process for all existing content in response to updates to the item definition information. In other words, the server 1 can re-execute the extraction process for all existing content in response to updates to the item definition information. This maintains the consistency of the entire database and makes it easy to add or change items.

[0024] Although not shown in Fig. 1, this service may provide alternative information when the relevant information is not included in the content. That is, when the relevant information corresponding to an item is not included in the content, the server 1 sets predetermined alternative information as the relevant information. This prevents data loss and makes it possible to build a complete database.

[0025] Next, the configuration of an information processing system that realizes the provision of the above-described service, i.e., an information processing system to which a server according to an embodiment of the information processing device of the present invention is applied, will be described with reference to Fig. 2. Fig. 2 is a diagram showing an example of the configuration of an information processing system to which a server according to an embodiment of the information processing device of the present invention is applied.

[0026] 2 is configured to include a server 1, a user terminal 2, an external system 3, and a web document server 4. The server 1, the user terminal 2, the external system 3, and the web document server 4 are connected to each other via a network N such as the Internet.

[0027] Server 1 is an information processing device managed by the service provider of this service (Figure 1). Server 1 executes various processes to realize this service while appropriately communicating with user terminal 2, external system 3, and web document server 4. User terminal 2 is an information processing device operated by user A, and is composed of any of a smartphone, tablet, personal computer, etc.

[0028] The external system 3 is an external information processing system that operates in cooperation with the server 1. The web document server 4 is a server that publishes and provides various content (for example, pages introducing recipes and books) on the Internet.

[0029] FIG. 3 is a block diagram showing an example of a hardware configuration of a server in the information processing system shown in FIG.

[0030] The server 1 includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a bus 14, an input / output interface 15, an input unit 16, an output unit 17, a memory unit 18, a communication unit 19, and a drive 20.

[0031] The CPU 11 executes various processes according to programs recorded in the ROM 12 or programs loaded from the storage unit 18 into the RAM 13. The RAM 13 also stores data and the like necessary for the CPU 11 to execute various processes.

[0032] The CPU 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output interface 15 is also connected to this bus 14. An input unit 16, an output unit 17, a storage unit 18, a communication unit 19, and a drive 20 are connected to the input / output interface 15.

[0033] The input unit 16 is configured with, for example, a keyboard and is used to input various types of information. The output unit 17 is configured with, for example, a display such as an LCD, a speaker, and the like and outputs various types of information as images and sounds. The storage unit 18 is configured with, for example, a DRAM (Dynamic Random Access Memory) and is used to store various types of data. The communication unit 19 communicates with other devices (for example, the user terminal 2, external system 3, and web document server 4 in FIG. 2 ) via a network N including the Internet.

[0034] Removable media 30, such as a magnetic disk, optical disk, magneto-optical disk, or semiconductor memory, is appropriately attached to the drive 20. Programs read from the removable media 30 by the drive 20 are installed in the storage unit 18 as needed. The removable media 30 can also store various data stored in the storage unit 18 in the same way as the storage unit 18.

[0035] Although not shown, the user terminal 2, external system 3, and web document server 4 in Fig. 2 can also have basically the same hardware configuration as that shown in Fig. 3. Therefore, a description of the hardware configuration of the user terminal 2, external system 3, and web document server 4 will be omitted.

[0036] The various hardware and software components constituting the information processing system of FIG. 2, including the server 1 of FIG. 3, work together to execute various processes for providing the present service of FIG.

[0037] FIG. 4 is a functional block diagram showing an example of the functional configuration of the server of FIG. 3 in the information processing system of FIG.

[0038] 4, the CPU 11 of the server 1 functions as an item definition acquisition unit 51, an information extraction control unit 52, a DB generation unit 53, an update control unit 54, a display control unit 55, an operation reception unit 56, an integration unit 57, and an external system output unit 58. In addition, an area of ​​the storage unit 18 of the server 1 is provided with an extraction model 70, a document DB 71, an extracted item DB 72, and a user information DB 73.

[0039] The extraction model 70 is, for example, a large-scale language model or generative AI, and is controlled by the information extraction control unit 52 based on past learning results, executes processing of the instructed content, and outputs the processing results to the source of the instruction via the information extraction control unit 52.

[0040] The document DB 71 stores web documents such as personal self-introduction pages that are made public to users, for example, members of this service, and members can update the contents of their own pages by instructing them themselves.

[0041] The extracted item DB 72 is a database that stores the items of the list L acquired from the user and the information corresponding to each item in association with each other. The items and information are updated each time an extraction process is performed.

[0042] The user information DB 73 stores user information. Specifically, the user's name, address, telephone number, email address, membership information, etc. The information shown here is an example and is not limited to this.

[0043] The item definition acquisition unit 51 acquires item definition information (list L) that defines items corresponding to information to be extracted from content, and stores the acquired item definition information in the document DB 71.

[0044] The information extraction control unit 52 and the extraction model 70 (machine learning model) constitute the information extraction means (the portion enclosed by the dashed line in FIG. 4 ). The information extraction control unit 52 controls the extraction model 70 to execute an extraction process. The information extraction means analyzes web documents W, such as web pages on the network N (e.g., web pages about cooking instructions) or self-introduction pages made available to members by the server 1, and executes an extraction process to extract relevant information related to the items in the list L from the web documents W. Note that the content and content providers shown here are merely examples and are not limited to these. Specifically, the information extraction control unit 52 controls the extraction model 70 to extract relevant information related to the items in the item definition information from the web documents W. In addition to the web documents W, the content also includes, for example, text data including text data and image data.

[0045] The DB generation unit 53 generates structured data in which the items defined by the item definition information are associated with the extracted information, and stores the generated structured data in the extracted item DB 72.

[0046] When the update control unit 54 detects an update to the content or item definition information, it causes the information extraction control unit 52 to re-execute the extraction process and updates the data in the extracted item DB 72. As a result, the changes to the content or item definition information are automatically reflected in the database.

[0047] The display control unit 55 displays in a table format the data of the extracted item DB 72. Specifically, the display control unit 55 organizes the extracted items and information and presents them on the user terminal 2 in a format (such as a table) that is easy for the user to view.

[0048] The operation reception unit 56 receives sorting and filtering operations for the displayed table (structured data). The operation reception unit 56 provides functions such as sorting data according to specific conditions and filtering by specific keywords based on instructions from the user.

[0049] The integration unit 57 integrates information extracted from multiple contents into a single structured data set. The integration unit 57 compiles information extracted from different documents into a single database, enabling cross-sectional analysis.

[0050] The external system output unit 58 converts the data (structured data) in the extracted item DB 72 into various file formats such as CSV files, spreadsheets, or other database formats, and transmits the converted data directly to the external system 3 or outputs it via an API. By linking with other systems using the external system output unit 58, various files and databases can be utilized in a wider range of ways. In other words, the external system output unit 58 outputs the data in the extracted item DB 72 to the external system 3 in any of the formats of a CSV file, a spreadsheet, or via an API.

[0051] 5 is a flowchart showing the basic processing operations of the server 1 according to an embodiment of the present invention. In step S1, the item definition acquisition unit 51 acquires a list L that defines items corresponding to information to be extracted from the web document W. In step S2, the information extraction control unit 52 extracts information corresponding to the items in the list L from the web document W. In step S3, the DB generation unit 53 generates an extracted item DB 72 that associates the items defined in the list L with the extracted information. In step S4, if the update control unit 54 detects an update to the web document W, it causes the information extraction means, which includes the information extraction control unit 52 and the extraction model 70, to re-execute the extraction process and updates the extracted item DB 72 based on the results of the re-execution.

[0052] FIG. 6 is a flowchart showing a more detailed processing operation of the server 1 according to an embodiment of the present invention. In step S11, the server 1 acquires the web document body. In step S12, the server 1 acquires a list defining items to be extracted. In step S13, the server 1 extracts information corresponding to each item in the list from the web document body. In step S14, the server 1 displays the information of the extracted items on the screen along with the web document body. In step S15, the server 1 determines whether the body or the list has been updated. If the body has been updated, the process proceeds to step S16; if the list has been updated, the process proceeds to step S18. In step S16, the server 1 re-extracts the item information from the updated body. In step S17, the server 1 updates the display of the extracted item information. In step S18, the server 1 re-extracts the information from the updated list. In step S19, the server 1 updates the display of the items and information in the updated list. If it is determined in step S15 that there is no update, the server 1 proceeds to step S20, where it displays the information in a table format.

[0053] Figure 8 shows the process of extracting item information from the body of a web document on server 1. As shown in Figure 8, server 1 extracts "Name," "Account Name," "Department," and "Telephone Number" from the body of the web document according to list L acquired from the user. A web document W and a panel P are displayed on screen 81.

[0054] 9 shows the automatic reflection process when a document is updated on server 1. As shown in Fig. 9, when an "email address" is written in the body of the document, it is immediately reflected in the extracted items. In response to the email address item 91 added to list L, an email address 92 is automatically added and displayed on panel P.

[0055] Fig. 10 is a diagram showing the automatic reflection process when updating a list on the server 1. As shown in Fig. 10, when "phone number" is deleted and "address" is added in list L, this is automatically reflected in panel P in the form that "phone number" disappears and "address" is added.

[0056] 11 shows the technical details of the extraction process on the server 1. A prompt 101 is generated from a list L of extraction items declared by the user, and this prompt 101 is provided to a large-scale language model and a generation AI to execute the process of extracting items from the body of a web document.

[0057] 12 is a diagram showing an example of the display of structured data on server 1. Information on items extracted from the text of multiple web documents is output as a list in the form of a table 102. By displaying this table 102, it is possible to view the information on each extracted item all at once.

[0058] 13 is a diagram showing how the display is updated when an item is added to server 1. When a "phone number" item 103 is added to list L, a "phone number" column 104 is added to table 102. This function allows changes to item definitions to be immediately reflected in the display.

[0059] 14 is a diagram showing the sorting function of the server 1. In the table 102, rows can be rearranged by specifying the "Department" column 105. This makes it possible to organize (narrow) and display information based on specific items.

[0060] FIG. 15 is a diagram showing a first example of the filtering function of server 1. FIG. 16 is a diagram showing a second example of the filtering function of server 1. A search box 106 is provided in table 102 shown in FIGS. 15 and 16. By entering a keyword in this search box 106, the user can perform filtering to display only rows containing that keyword. For example, by narrowing down the search using keywords such as "development" in FIG. 15 and "Tokyo" in FIG. 16, only related information can be displayed in table 102. In FIG. 15, only people who belong to the "development" department are displayed in table 102, and in FIG. 16, only people who live in "Tokyo" are displayed in table 102.

[0061] 17 is a diagram showing an example of an annotated list (item definition information) of the server 1. Annotations 107 can be added to each item in the list L. For example, by adding an annotation to the "Address" item such as "Please remove any "resident" at the end and keep the address concise," the extraction results are adjusted so that the "resident" at the end of the "Address" is removed. Also, by adding an annotation to the "Phone Number" item such as "If not found, please mark as "Unknown," this is reflected in the prompt 101 (see FIG. 18) that lists the extraction conditions, and information about people who do not have a "Phone Number" written in the text is displayed as "Unknown."

[0062] Fig. 18 shows the annotated list of Fig. 17 and a prompt generated from the list. In Fig. 17, as a method of adjusting the extraction results by adding annotations to each item in list L, a prompt 101 including annotations is generated from list L of extraction items declared by the user, and is given to extraction model 70 (large-scale language model or generative AI).

[0063] Figure 19 is a diagram showing an example of implicit information extraction by server 1. Figure 19 shows an example of extracting information by inferring information that is not explicitly stated (e.g., "Does water need to be boiled?") from the context of a sentence such as "boil spaghetti" that is published in a document such as a cooking recipe, for example, a web document that introduces how to make a dish. Processing that includes information extraction and inference from such context interpretation is something that can only be done by AI or similar.

[0064] The gist of the embodiment will be described below. The server 1 that generates structured data from content according to the embodiment solves the problem of efficiently utilizing the vast amount of unstructured information that is generated and shared inside and outside an organization.

[0065] In modern society, information that exists as free-form natural language text is difficult to reuse or analyze, and conventional database systems require strict schema definition and manual data formatting. Furthermore, because document creation and database construction are separate processes, there are problems such as the hassle of dual management and the likelihood of information inconsistencies.

[0066] Furthermore, deciphering the information and context implicit in documents and extracting them as structured data required advanced manual interpretation, which was costly and time-consuming. This meant that the latent knowledge existing within an organization was not fully utilized, leading to delayed decision-making, a decline in quality, and stagnation in business efficiency.

[0067] The function of the information extraction means in the server 1 is called an infobox, and the function of detecting updates to the web document W or list L and executing a re-extraction process to update the extracted item DB 72 is called a literary database.

[0068] Infobox is a function that automatically extracts related information based on predefined items from content (e.g., web document W) written by a user in natural language using an extraction model 70 (machine learning model).

[0069] For a specific document type (e.g., minutes, reports, specifications), the user writes the "item name" of the information they want to extract and the extraction rules (e.g., parts containing specific keywords, date format, numeric range, etc.) in the "item definition information (list L)."

[0070] By simply linking the target web document W to this definition information, the infobox analyzes the web document W, extracts information corresponding to the defined items, and presents it.

[0071] When using infobox, it is important to extract specific items such as "current job," "recent status," "address," and "research topic," rather than giving general instructions to the extraction model 70 to generate a summary.

[0072] The literary database is a system that aggregates structured information extracted by infobox and functions as a database that can be displayed in a list, sorted, filtered, etc. It has the characteristic that a high-quality database is formed semi-automatically through the everyday act of document creation, without any special database construction work.

[0073] The name "Literary" was inspired by the idea of ​​"Literary Programming" proposed by Donald Knuth, and aims to combine natural and easy-to-understand document writing for humans with efficient data processing by machines.

[0074] In conventional database + document systems such as Notion and Filemaker, data is collected, tables are created, and then documents are generated, but a major feature of literary databases is that they operate in the reverse order.

[0075] It should be noted that methods that do not use the extraction model 70 can also be considered as methods for achieving the same functionality. For example, in a method of extraction using pattern matching without using AI, a date such as "2025 / 04 / 17" can be found using a pattern of four-digit number / one, two-digit number / two-digit number. Other possible methods include using morphological analysis to determine nouns or verbs, or using a dictionary of place names, personal names, store names, etc.

[0076] Furthermore, there are methods that use markup specifically for extraction, such as "Place name: Tokyo Tower," or that extract items one by one and then list them at the end, rather than creating a list and extracting them all at once.However, these methods make it difficult to extract information that requires contextual understanding, such as "need to boil water" from "boil spaghetti."

[0077] By using the extraction model 70 (machine learning model), it becomes possible to extract information that is not directly stated in the web document W and implicit information inferred from the context as structured data.

[0078] This complements information that humans tend to omit as implicit assumptions, increasing the completeness of the information and enabling deeper analysis and utilization.

[0079] It is also possible to adjust the extraction results by adding annotations to the item definition information (list L). For example, adding an annotation to the "Address" item such as "Please remove the word 'resident' at the end and make the address concise," will result in more formatted information being extracted.

[0080] The server 1 of the embodiment can improve the efficiency of information management in a variety of businesses, not limited to a specific industry or occupation.

[0081] Specific application examples include the following: 1. Management work: Automatically extracts progress and issues from project reports, supporting rapid situation understanding and decision-making. 2. Planning and development work: Clarifies decisions and responsible parties from specifications and meeting minutes, reducing rework and confirmation work. 3. Sales activities: Structures customer needs, proposals, and next actions from sales negotiation records, improving the accuracy of sales strategies. 4. Supports recording response history in CRM: Infobox and generation AI work together to automatically generate response history drafts from customer response notes to enter into the CRM system, significantly reducing the time required to create records. 5. Inside sales call log analysis: Infobox extracts customer interests, issues, key person information, etc. from voice-recognized call logs, contributing to the formulation of sales strategies. 6. Contract creation support: Infobox is used to efficiently extract specific clauses and conditions from past contract documents and related materials, supporting the creation of drafts for new contracts. 7. Structuring meeting minutes and negotiation logs: Infobox automatically extracts information such as company name, person in charge, decisions, issues, and next actions from free-form minutes and negotiation logs, streamlining post-meeting action management and information sharing.

[0082] Additionally, making infobox output machine-readable reduces the need for exception handling when inputting data into other systems. For example, it becomes possible to define status as an enumeration such as "completed," "in progress," or "unknown," and to define fees as a numeric value. This will facilitate the opening of APIs for external use of databases created from documents, further expanding the concept of literary databases.

[0083] The server 1 of this embodiment achieves seamless integration into existing workflows and automatic creation of information assets. Users do not need to make major changes to their existing document creation processes; the various documents created in daily work are automatically converted into structured information assets. This allows information scattered throughout an organization to be systematically organized and made usable, without any special awareness or additional workload.

[0084] Another important benefit is that it promotes the sharing of perspectives between the "readers (information users, managers)" and the "writers (information creators)" and the creation of high-quality information. Through the item definition information (List L), readers can clarify their "perspective" on which information items are important, and can set standards for information sharing within the organization. By being aware of the items that readers consider important, writers can prevent information omissions and create higher-quality documents. This reduces misunderstandings during collaborative work and promotes smoother communication.

[0085] The server 1 of the embodiment streamlines information sharing in business projects and helps team members share their knowledge and experiences. For corporate use, it is necessary to record experiences (thought processes / trial processes) to synchronize the team, and infoboxes and literary databases fulfill this role. In communities, it also promotes divergent information sharing and facilitates joint use.

[0086] Literary databases are a great fit for history (work logs) that have a set format, such as troubleshooting and meeting minutes. This makes it easy to filter and sort large amounts of information and access the information you need.

[0087] As described above, according to the server 1 of the embodiment, a high-quality structured database is semi-automatically constructed through daily document creation activities, and its value is further enhanced by the advanced information extraction capabilities of the extraction model 70.

[0088] This will not only improve individual work efficiency, but also promote knowledge sharing and utilization throughout the organization, supporting more productive and creative activities. In addition, machine-readable data output will facilitate integration with existing systems and is expected to serve as an information infrastructure for the digital transformation of organizations.

[0089] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and modifications, improvements, etc. within the scope of achieving the object of the present invention are considered to be included in the present invention.

[0090] In the above embodiment, examples of application to management tasks, planning and development tasks, sales activities, etc. have been described, but the server 1 of the embodiment can be used for a wider variety of purposes. For example, as a management system for estimates, invoices, and receipts, information such as amounts, dates, and business partners within documents can be automatically extracted and stored in a database. In this case, even if there are variations in notation even for the same business partner, or if prices are written in a mixed format, including and excluding tax, the information can be extracted in an appropriate, unified form.

[0091] It can also be applied to building databases of papers and books, streamlining library management by automatically extracting author names, publication years, keywords, abstracts, etc. Furthermore, it can extract information such as nutrients and ingredients from food composition tables and create a database, or scan numerous posters and flyers and organize their contents in a table format.

[0092] Furthermore, the server 1 of the embodiment can not only directly display the extracted items to the user, but also function as an intermediate processor to pass the extracted items to a downstream AI system for further processing. This makes it possible to build an advanced automated system in which only necessary information is selected from a large amount of unstructured data and AI generates the next action or proposal based on that information.

[0093] Furthermore, the system configuration shown in FIG. 2 and the hardware configuration of the server 1 shown in FIG. 3 are merely examples for achieving the object of the present invention, and are not particularly limited.

[0094] Furthermore, the functional block diagram shown in Fig. 4 is merely an example and is not particularly limited. That is, it is sufficient if the information processing system in Fig. 2 is provided with a function that can execute the various processes described above as a whole, and the functional blocks and databases used to realize this function are not particularly limited to the example in Fig. 4.

[0095] Furthermore, the locations of the functional blocks and databases are not limited to those shown in Fig. 4 and may be arbitrary. For example, at least some of the functional blocks and databases located on the server 1 side may be provided on the user terminal 2 side, the external system 3 side, the web document server 4 side, or another information processing device not shown.

[0096] The above-described series of processes can be executed by hardware or software, and each functional block can be configured by hardware alone, software alone, or a combination of both.

[0097] When a series of processes is executed by software, the programs constituting the software are installed onto a computer or the like from a network or a recording medium. The computer may be a computer incorporated into dedicated hardware. The computer may also be a computer capable of executing various functions by installing various programs, such as a server, a general-purpose smartphone, or a personal computer.

[0098] The recording medium containing such a program may be composed not only of a removable medium (not shown) that is distributed separately from the device main body in order to provide the program to the user, but also of a recording medium that is provided to the user in a state that is pre-installed in the device main body.

[0099] In this specification, the steps describing the program to be recorded on the recording medium include not only processes that are performed in chronological order, but also processes that are not necessarily performed in chronological order but are performed in parallel or individually.

[0100] In summary, an information processing device to which the present invention is applied is sufficient as long as it has the following configuration, and can take on a variety of different embodiments. That is, an information processing device to which the present invention is applied (for example, the server 1 in FIGS. 2 to 4) comprises: (1) item definition acquisition means (for example, the item definition acquisition unit 51 in FIG. 4) that acquires item definition information (for example, list L in FIG. 1) that defines items corresponding to information to be extracted from content (for example, web document W in FIG. 1); information extraction means (for example, the information extraction control unit 52 and extraction model 70 in FIG. 4) that extracts the information related to the items of the item definition information (for example, list L in FIG. 1) from the content; structured data generation means (for example, the DB generation unit 53 in FIG. 4) that generates structured data (for example, the extracted item DB 72 in FIG. 4) that associates the items defined by the item definition information (for example, list L in FIG. 1) with the extracted information; and an update control means (e.g., update control unit 54 in FIG. 4) that, when an update of the content (e.g., web document W in FIG. 1) or the item definition information (e.g., list L in FIG. 1) is detected, causes the information extraction means (e.g., information extraction control unit 52 and extraction model 70 in FIG. 4) to function and re-execute the extraction process, and updates the structured data (e.g., extracted item DB 72 in FIG. 4) based on the result of the re-execution.

[0101] In this way, it is possible to build a database that matches the user's natural workflow, regardless of the order of document creation and item definition. Furthermore, because structured data is automatically updated as documents and item definitions are updated, the database can always be kept up to date, preventing data inconsistencies caused by manual updates. As a result, content (e.g., web documents W) and databases (e.g., extracted item DB 72) can be synchronized easily and efficiently.

[0102] (2) Furthermore, the information extraction means (e.g., the information extraction control unit 52 and the extraction model 70 in FIG. 4) can use a machine learning model (e.g., the extraction model 70 in FIG. 4) to infer and extract information that is not explicitly stated in the content (e.g., the web document W in FIG. 1) from the context.

[0103] This makes it possible to extract information that is not explicitly stated by the document author through automatic information completion. For example, from a recipe like the one shown in Figure 19, it is possible to infer and extract information that is not explicitly stated, such as "you need to boil water," from the context.

[0104] (3) Furthermore, the item definition information may include annotation information for each item, and the information extraction means (e.g., the information extraction control unit 52 and extraction model 70 in FIG. 4) may adjust the extraction results based on the annotation information.

[0105] This allows users to flexibly adjust the format and content of the extracted results. For example, as shown in Figures 17 and 18, users can adjust the extracted results by adding an annotation to "Address" such as "Please remove "resident" at the end and make the address concise."

[0106] (4) The system may further include a display control means (e.g., the display control unit 55 in FIG. 4) for displaying the structured data in a tabular format, and an operation receiving means (e.g., the operation receiving unit 56 in FIG. 4) for receiving a sorting or filtering operation on the structured data.

[0107] This allows for efficient manipulation and analysis of the generated database. For example, as shown in Figure 14, you can rearrange the rows by specifying the "Department" column, or you can narrow down the table by keywords such as "Development" or "Tokyo" as shown in Figures 15 and 16.

[0108] (5) Furthermore, the system may further include an integration means (e.g., integration unit 57 in FIG. 4) for integrating relevant information extracted from multiple contents (e.g., web document W in FIG. 1 and an internal company self-introduction page) into a single structured data.

[0109] This makes it possible to build a unified database from multiple documents. As shown in Figure 12, multiple pieces of personal information can be displayed together in one table.

[0110] (6) Furthermore, the creation of the content (for example, the web document W in FIG. 1) and the creation of the item definition information (for example, the list L in FIG. 1) can be performed in any order.

[0111] This allows for flexible support for the user's workflow, enabling database construction in a more natural format: users can create documents first and then define fields, or define fields first and then create documents, depending on their workflow.

[0112] (7) Furthermore, the information extraction means (e.g., the information extraction control unit 52 and extraction model 70 in Figure 4) can re-execute the extraction process for all existing content (e.g., web document W in Figure 1) in response to updates to the item definition information (e.g., list L in Figure 1).

[0113] This maintains consistency across the database and makes it easy to add or change fields: if a field definition changes, the new field information is automatically extracted from all documents and the database is updated.

[0114] (8) Furthermore, if the relevant information corresponding to the item is not described in the content (e.g., web document W in Figure 1), predetermined alternative information can be set as the relevant information.

[0115] This prevents data loss and allows a complete database to be constructed. As shown in Figure 17, if the "telephone number" is not entered, the value "unknown" can be set.

[0116] (9) Furthermore, the system may further include an output means (e.g., external system output unit 58 in FIG. 4) that outputs the structured data to an external system (e.g., external system 3 in FIG. 4) in any of the following forms: a CSV file, a spreadsheet, or via an API.

[0117] This makes it possible to link the generated database (structured data) with other systems and utilize the structured data in those systems, thereby enabling the extracted structured data to be utilized in a wider range of applications.

[0118] DESCRIPTION OF SYMBOLS 1: Server, 2: User terminal, 3: External system, 4: Web document server, 11: CPU, 12: ROM, 13: RAM, 14: Bus, 15: Input / output interface, 16: Input unit, 17: Output unit, 18: Storage unit, 19: Communication unit, 20: Drive, 30: Removable media, 51: Item definition acquisition unit, 52: Information extraction control unit, 53: DB generation unit, 54: Update control unit, 55: Display control unit, 56: Operation acceptance unit, 57: Integration unit, 58: External system output unit, 70: Extraction model 70: Document DB, 72: Extracted item DB, 73: User information DB, A: User, L: List, N: Network, P: Panel, W: Web document

Claims

1. An information processing device comprising: an item definition acquisition means for acquiring item definition information that defines items corresponding to information to be extracted from content; an information extraction means for extracting the information corresponding to the items of the item definition information from the content; a structured data generation means for generating structured data that associates the items defined by the item definition information with the extracted information; and an update control means for, when an update to the content or the item definition information is detected, causing the information extraction means to function and re-execute the extraction process, and updating the structured data based on the result of the re-execution.

2. The information processing device according to claim 1, wherein the information extraction means extracts information that is not explicitly stated in the content by inferring it from the context.

3. The information processing device according to claim 1 or 2, wherein the item definition information includes annotation information for each item, and the information extraction means adjusts the extraction results based on the annotation information.

4. The information processing device according to claim 1, further comprising: a display control means for displaying the structured data in a table format; and an operation receiving means for receiving at least one of sorting and filtering operations on the structured data.

5. The information processing device according to claim 1, further comprising: integration means for integrating relevant information extracted from a plurality of contents into a single structured data.

6. The information processing device according to claim 1, wherein the creation of the content and the creation of the item definition information can be performed in any order.

7. The information processing device according to claim 1, wherein the extraction process is re-executed for all of the existing content in response to an update of the item definition information.

8. The information processing device according to claim 1, wherein, when the extracted relevant information is not included in the content, predetermined alternative information is set as the relevant information.

9. The information processing device according to claim 1, further comprising: output means for outputting the structured data to an external system in any one of the following formats: a CSV file, a spreadsheet, or via an API.

10. An information processing method executed by an information processing device, comprising: an item definition acquisition step of acquiring item definition information that defines items corresponding to information to be extracted from content; an information extraction step of extracting information related to items of the item definition information from the content; a structured data generation step of generating structured data that associates the items defined by the item definition information with the extracted information; and an update control step of, when an update to the content or the item definition information is detected, re-executing the extraction process by the information extraction step and updating the structured data based on the result of the re-execution.

11. A program that causes a computer to execute control processing including: an item definition acquisition step of acquiring item definition information that defines items corresponding to information to be extracted from content; an information extraction step of extracting information related to items in the item definition information from the content; a structured data generation step of generating structured data that associates the items defined by the item definition information with the extracted information; and an update control step of, when an update to the content or the item definition information is detected, re-executing the extraction processing by the information extraction step and updating the structured data based on the result of the re-execution.

Images