Information processing system, information processing method, and program

Abstract

To provide a system configured to properly acquire information included in text information from a generative AI device.SOLUTION: An information processing system includes: a text information acquisition unit which acquires text information including characters; an item acquisition unit which acquires an extraction item regarding target extraction information to be extracted from the text information, based on operation input of a user; a query information specifying unit which specifies extraction query information to be associated with a setting candidate item corresponding to the extraction item, based on a query database that stores the setting candidate item to be set as an extraction target and query information to be input to a generative AI device that has been trained to extract the extraction information from the text information, in association with each other; a transmission unit which transmits the extraction query information and the text information to the generative AI device; and an extraction information acquisition unit which acquires the extraction information included in the text information, from the generative AI device.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] An accounting processing system that uses a generation AI device to automatically journalize from text data has been disclosed (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6895632 Summary of the Invention [Problem to be solved by the invention]

[0004] The accounting processing system described in Patent Document 1 extracts transaction information from supporting data using image recognition processing, and generates text information including transaction details from the extracted transaction information using character recognition processing.The accounting processing system also generates journal entry data from the text information using a journal entry generation AI that has learned to select a combination of account items corresponding to the transaction details.This enables the accounting processing system to perform appropriate automatic journal entries.

[0005] However, in the accounting processing system described in Patent Document 1, although a journal entry generation AI is used, if the query information input into the journal entry generation AI to generate journal entry data is inappropriate, a problem arises in that journal entry data cannot be generated appropriately from text information.

[0006] Therefore, in order to solve the above problem, the present invention aims to provide a system that can appropriately acquire information contained in text information from a generation AI device. [Means for solving the problem]

[0007] An information processing system according to one embodiment of the present invention comprises a text information acquisition unit that acquires text information including characters; an item acquisition unit that acquires extraction items related to extraction information to be extracted from the text information based on user input; a query information identification unit that identifies extraction query information associated with the candidate setting items corresponding to the extraction items based on a query database that associates and stores candidate setting items that may be set as extraction targets with query information to be input to a generation AI device that has been trained to extract the extraction information from the text information; a transmission unit that transmits the extraction query information and the text information to the generation AI device; and an extraction information acquisition unit that acquires the extraction information contained in the text information from the generation AI device.

[0008] An information processing method according to one embodiment of the present invention includes a computer that executes the following steps: a text information acquisition unit that acquires text information including characters; acquiring extraction items related to extraction information to be extracted from the text information based on user input; identifying extraction query information associated with the candidate setting items corresponding to the extraction items based on a query database that stores and associates candidate setting items that can be set as extraction targets with query information to be input to a generation AI device that has been trained to extract the extraction information from the text information; transmitting the extraction query information and the text information to the generation AI device; and acquiring the extraction information contained in the text information from the generation AI device.

[0009] A program according to one embodiment of the present invention causes a computer to perform the following operations: a text information acquisition unit that acquires text information including characters; acquires extraction items related to extraction information to be extracted from the text information based on user input; identifies extraction query information associated with the candidate setting items corresponding to the extraction items based on a query database that stores candidate setting items that can be set as extraction targets and query information to be input to a generation AI device that has been trained to extract the extraction information from the text information; transmits the extraction query information and the text information to the generation AI device; and acquires the extraction information contained in the text information from the generation AI device. [Effects of the Invention]

[0010] According to the present invention, a system can be provided that can appropriately acquire information contained in text information from a generation AI device. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram showing an overview of an extraction system. [Figure 2] 1 is a database showing an example of a query database 101a. [Figure 3] 1 is a database showing an example of a text database 101b. [Figure 4] 10 is a database showing an example of a label database 101c. [Figure 5] 10 is a flowchart showing a processing procedure of the extraction system. [Figure 6] FIG. 10 is a diagram illustrating an example of target text information. [Figure 7] FIG. 10 is a diagram showing an example of extracted query information and extracted block text information input to a generation AI device. [Figure 8] FIG. 10 is a diagram showing an example of extracted information output from a generation AI device. [Figure 9] FIG. 2 illustrates an example of a hardware configuration of a computer. DETAILED DESCRIPTION OF THE INVENTION

[0012] An extraction system 10 according to one embodiment of the present invention will be described in detail below with reference to the drawings. However, the embodiment described below is merely an example, and is not intended to exclude various modifications or applications of techniques not explicitly described below. In other words, the present invention can be implemented by various modifications or combinations of the embodiments without departing from the spirit of the invention. Furthermore, in the following description of the drawings, identical or similar parts are denoted by identical or similar reference numerals.

[0013] In this embodiment, the terms "unit," "device," and "system" do not simply refer to physical means, but also include cases where the functions of the "unit," "device," or "system" are realized by software. Furthermore, the functions of one "unit," "device," or "system" may be realized by two or more physical means or devices, and the functions of two or more "units," "devices," or "systems" may be realized by one physical means or device. Furthermore, the various functions described below of each of the multiple devices constituting the extraction system 10 may be configured to be executed by other devices in the multiple devices.

[0014] === Overview of Extraction System 10 === <<Configuration Overview>> An outline of the extraction system 10 will be described with reference to Fig. 1. Fig. 1 is a diagram showing an outline of the extraction system 10.

[0015] The extraction system 10 is a system that extracts text desired by a user from text contained in a digital document generated based on a target image using an optical character recognition device.

[0016] The target image is, for example, an image of various contracts, papers, etc. The target image is, for example, an image specified in a graphic format such as JPEG (Joint Photographic Experts Group), TIFF (Tagged Image File Format), or PNG (Portable Network Graphics), or an image in PDF (Portable Document Format) data.

[0017] The extraction system 10 includes, for example, an extraction device 100, an optical character recognition device 200, and a user terminal 300.

[0018] The extraction device 100 is a device that outputs the result of character recognition of a target image based on the result of character recognition of the target image by an optical character recognition device.

[0019] The optical character recognition device 200 is a device that performs character recognition on a target image.

[0020] The extraction device 100 and the optical character recognition device 200 may be, for example, a cloud computer, a server computer, a personal computer (e.g., a desktop, laptop, tablet, etc.), a media computer platform (e.g., a cable or satellite set-top box, a digital video recorder), a handheld computer device (e.g., a PDA, an email client, etc.), or any other type of computer or communication platform. At least a portion of the processing in the extraction device 100 and the optical character recognition device 200 may be implemented by one or more computers (for example, but not limited to, cloud computing configured by one or more computers).

[0021] The user terminal 300 is a device that receives user operation input and displays various types of information.

[0022] The user terminal 300 may be, for example, a smartphone, a mobile phone (feature phone), a personal computer (e.g., a desktop, laptop, tablet, etc.), a media computing platform (e.g., a cable or satellite set-top box, a digital video recorder), a handheld computing device (e.g., a personal digital assistant (PDA), an email client, etc.), a wearable device (e.g., a glasses-type device, a watch-type device, etc.), or another type of computer or communication platform.

[0023] <<Processing Overview>> An overview of the processing of the extraction system 10 will be described with reference to FIG.

[0024] First, in step S10, the extraction device 100 acquires, from the optical character recognition device 200, text information generated based on the target image (hereinafter referred to as "target text information").

[0025] In step S11, the extraction device 100 identifies text information of a predetermined group of characters included in the target text information (hereinafter referred to as "block text information"). The extraction device 100 assigns a label (hereinafter referred to as "extraction label") indicating the classification of the block text information to each block text information. The extraction device 100 associates the block text information with the extraction label and stores them in the storage unit 101 (label database D101c described below).

[0026] The block text information may be, for example, a group of multiple sentences, or a single sentence of multiple words. Specifically, in a contract or a paper, the block text information may be, for example, at least one sentence, at least one article, clause, or paragraph, at least one chapter, or a collection of information in a table format.

[0027] In step S12, the user terminal 300 receives an operation input from the user and acquires items (hereinafter referred to as "extracted items") that are categories related to the information to be extracted from the text information (hereinafter referred to as "extracted information"). The user terminal 300 transmits the extracted items to the extraction device 100.

[0028] In step S13, the extraction device 100 refers to the label database D111 to identify an extraction label corresponding to the extraction item. Then, the extraction device 100 refers to the text database D101b to identify block text information (hereinafter referred to as "extracted block text information") related to the identified extraction label.

[0029] In step S14, the extraction device 100 refers to the query database D112 to identify candidate setting items corresponding to the identified extraction label, and based on the candidate setting items, identifies query information (hereinafter referred to as "extraction query information") to be input to the generation AI (Artificial Intelligence) device.

[0030] The generative AI device is a device that has the functionality of a large-scale language model (hereinafter referred to as "generative AI") with a huge data set based on a deep learning model. The generative AI is, for example, a text generation AI trained to extract extracted information based on query information, such as ChatGPT or Google Bard.

[0031] The query information is a sentence to be input to the generation AI device, or a prompt, which is an instruction sentence given to the generation AI. In other words, the extraction query information is an instruction sentence to be input to the generation AI device in order to extract specific text information from the block text information.

[0032] In step S15, the extraction device 100 transmits the extracted block text information and the extracted query information to the generation AI device.

[0033] In step S16, the extraction device 100 obtains, from the generation AI device, extraction information that is the result of extracting text information from the block text information.

[0034] This allows the extraction device 100 to appropriately extract information desired by the user from the target text information using the generation AI device.

[0035] ===Extraction device 100=== As shown in FIG. 1, the extraction device 100 includes a memory unit 101, a text information acquisition unit 102, a label setting determination unit 103, an item acquisition unit 104, a block text information identification unit 105, a query information identification unit 106, a transmission unit 107, an extraction information acquisition unit 108, and a display processing unit 109.

[0036] The storage unit 101 includes, for example, a query database D101a, a text database D101b, and a label database D101c.

[0037] The query database D101a will be described with reference to Fig. 2. Fig. 2 shows an example of the query database D101a. The query database D101a is a database that stores candidate setting items, which are items that can be set as extraction targets, in association with query information. The query database D101a is a database for identifying queries to be input to the generation AI device.

[0038] As shown in Fig. 2, the query database D101a includes items such as [Candidate ID], [Candidate Setting Item], and [Query Information]. [Candidate ID] stores identification information that can uniquely identify a candidate setting item. [Candidate Setting Item] stores text information that indicates a candidate setting item. [Query Information] stores query information to be input to the generation AI device.

[0039] The query information is a prompt optimized to allow the generation AI device to extract extraction information related to candidate setting items from text information. The query information includes items such as extraction requirements and extraction examples. The extraction requirements are information for extracting appropriate information, such as information on the characteristics of the information to be extracted and the extraction method (e.g., extraction requirements T21 in Figure 7, which will be described later). The extraction examples are information for improving the accuracy of extraction, such as information on examples of extraction results in response to input (e.g., extraction examples T22 in Figure 7, which will be described later).

[0040] By using the query database D101a, the extraction device 100 enables a user to input query information to the generation AI device in order to obtain desired extraction information from the generation AI.

[0041] The text database D101b will be described with reference to Fig. 3. Fig. 3 shows an example of the text database D101b. The text database 101b is a database that stores block text information and extraction labels that indicate classifications of the block text information in association with each other.

[0042] As shown in FIG. 3, the text database D101b includes items such as [Text ID], [Text Information], [Block ID], [Block Text Information], and [Extraction Label]. [Text ID] stores identification information that can uniquely identify the target text information. [Text Information] stores the target text information. [Block ID] stores identification information that can uniquely identify the block text information. [Block Text Information] stores block text information of a predetermined group of the target text information. [Extraction Label] stores a label that indicates the classification of the block text information.

[0043] By using the text database D101b, the extraction device 100 can appropriately extract, from the query database D101a, query information for obtaining extraction information desired by the user from the generation AI.

[0044] The label database D101c will be described with reference to Fig. 4. Fig. 4 shows an example of the label database D101c. The label database 101c is a database that stores labels indicating classifications and text information in association with each other.

[0045] As shown in FIG. 4, the label database D101c includes items such as [Label ID], [Extraction Label], and [Text Features]. [Label ID] stores identification information that can uniquely identify a label. [Label] stores an extraction label to be assigned to block text information. [Text Features] stores text features of text included in block text information to which an extraction label is associated. Text features are information that indicate, for example, a word, a combination of multiple words, or a word or combination of words included in a tabular area. Specifically, for example, the text feature corresponding to the extraction label "Company Name" is the condition "text that includes all of 'Party A', 'Party B', and 'Contract'."

[0046] By using the label database D101c, the extraction device 100 can appropriately assign extraction labels to block text information included in the target text information, thereby generating the text database D101b. That is, the extraction device 100 (the label setting determination unit 103, which will be described later) can assign extraction labels to each of the block text information including a word (e.g., "Article 2"), a combination of multiple words (e.g., the combination of "A," "B," and "Contract"), and a tabular region (e.g., a combination of a ruled line and words surrounded by the ruled line), which are included in the block text information.

[0047] The text information acquisition unit 102 acquires, for example, from the optical character recognition device 200, text information generated in the optical character recognition device 200 based on a target image.

[0048] The label setting determination unit 103 identifies block text information included in the target text information. Then, the label setting determination unit 103 associates the extraction label with the block text information based on the correspondence between the extraction label and the text features of the text included in the text information. The label setting determination unit 103 associates the extraction label with the block text information and stores it in the text database D101b.

[0049] Specifically, the label setting determination unit 103 first analyzes the text included in the target text information, for example, by morphological analysis, and classifies the target text information into block text information. For example, if the target image is a contract and the label setting determination unit 103 identifies the text "Article 2" in the target text information, the label setting determination unit 103 classifies the sentence defined in "Article 2" of the contract as block text information.

[0050] The label setting determination unit 103 may then associate extraction labels with the block text information based on, for example, the label database D101c. Specifically, the label setting determination unit 103 refers to the label database D101c and associates, with block text information that satisfies any of the text features, a label corresponding to the text feature as an extraction label. The label setting determination unit 103 associates the extraction labels with the block text information and stores them in the text database D101b. Note that the method for analyzing the text included in the target text information is, for example, morphological analysis.

[0051] The label setting determination unit 103 may associate the extraction label with the block text information, for example, by inputting the block text information into a trained model that has been trained using the extraction label and text features included in the text information as training data. Specifically, the label setting determination unit 103, for example, inputs the block text information into the trained model, and acquires from the trained model at least one extraction label that indicates a probability that it is associated with the block text information equal to or greater than a predetermined threshold. The label setting determination unit 103 associates, for example, the extraction label that indicates the highest probability among the acquired extraction labels with the block text information, and stores the associated extraction label in the text database D101b.

[0052] The item acquisition unit 104 acquires extracted items based on operation input by the user to the user terminal 300. Specifically, for example, when the target image is a contract, the item acquisition unit 104 acquires the "company name" input by the user to the user terminal 300 from the user terminal 300 as an extracted item.

[0053] The block text information identification unit 105 refers to the text database D101b to identify extracted block text information, which is block text information corresponding to the extraction item. Specifically, the block text information identification unit 105 identifies, for example, "Company Name," which is an extraction label corresponding to the extraction item "Company Name," in the text database D101b. Then, the block text information identification unit 105 refers to the text database D101b to identify "Article 2...," which is block text information related to the extraction label "Company Name."

[0054] The query information identification unit 106 identifies extracted query information by referring to the query database D101a. Specifically, the query information identification unit 106 identifies, for example, "company name" as a setting candidate item corresponding to "company name" as an extracted item in the query database D101a. The query information identification unit 106 identifies, in the query database D101a, extracted query information related to the setting candidate item "company name."

[0055] The transmission unit 107 transmits extraction query information for acquiring extraction information from the generation AI device. Specifically, the transmission unit 107 transmits the extraction query information and the block text information to the generation AI device.

[0056] The extraction information acquisition unit 108 acquires the extraction information included in the text information from the generation AI device. Specifically, when the target image is a contract, the extraction information acquisition unit 108 acquires extraction information indicating, for example, the extracted item "company name" from the contract, such as "A Co., Ltd."

[0057] The display processing unit 109 displays various types of information on the display unit. For example, the display processing unit 109 displays information shown in Fig. 7 and Fig. 8, which will be described later, on the display unit.

[0058] <<First Modification>> When the target text information includes data in a table format, the label setting determination unit 103 may determine whether or not there is text information corresponding to the extracted item in the text information included in the table format area.

[0059] Specifically, the label setting determination unit 103 identifies the tabular text information of the block B30 shown in FIG. 6. The label setting determination unit 103 associates an extraction label with the block B30 based on the text features of the label database D101c. At this time, the label setting determination unit 103 may associate the extraction label with the title of the block B30 ("Rental Requirements" in FIG. 6). The label setting determination unit 103 determines whether the text information contained in the block B30, which is related to the extraction label corresponding to the extraction item, includes text information corresponding to the extraction information. At this time, the label setting determination unit 103 may identify the ruled lines of the block B30 and determine whether the text information written in each area surrounded by the ruled lines corresponds to the extraction item.

[0060] Then, when the label setting determination unit 103 determines that there is text information (item) corresponding to the extraction item, the extraction device 100 acquires, as extraction information, information related to the item corresponding to the extraction item in the tabular data.

[0061] On the other hand, if the label setting determination unit 103 determines that there is no text information (item) corresponding to the extracted item, the extraction device 100 transmits the extracted query information and the extracted block text information to the generation AI device, as described above, and obtains the extracted information from the generation AI device.

[0062] In this way, the extraction device 100 may be configured to acquire extracted information rule-based when the target text information contains information with a clear correspondence between items and text content, and to acquire extracted information using the generation AI device when the rule-based acquisition of extracted information is not possible. This enables the extraction device 100 to reduce the frequency of using the generation AI device and improve the accuracy of acquiring appropriate extracted information.

[0063] <<Second Modification>> The extraction device 100 may identify extraction information based on a comparison result between first extraction information acquired when the label setting determination unit 103 determines that there is text information (item) corresponding to the extraction item and second extraction information acquired from the generation AI device when the label setting determination unit 103 determines that there is no text information (item) corresponding to the extraction item. Specifically, the extraction device 100 may transmit the comparison result to the user terminal 300 and identify extraction information based on a selection result of the first extraction information or the second extraction information based on a user's operation input. Furthermore, the extraction device 100 may identify extraction information that indicates a higher index of extraction accuracy between the index of extraction accuracy indicated by the first extraction information and the index of extraction accuracy indicated by the second extraction information. This enables the extraction device 100 to improve the accuracy of acquiring appropriate extraction information.

[0064] <<Third Modification>> In the above description, the label setting determination unit 103 generates the text database D101b, but this is not limiting. If the text database D101b is not generated, the label setting determination unit 103 identifies "text features" associated with the labels in the label database D101c corresponding to the extraction items. Then, based on the identified text features, the block text information identification unit 105 associates extraction labels corresponding to the text features with block text information included in the target text information that satisfies the text features, and identifies the block text information as extracted block text information. In other words, the extraction device 100 does not need to include the text database D101b and may be configured to identify extracted block text information based on the identified text features. This enables identification of extracted block text information with a simple system configuration.

[0065] ===Optical character recognition device 200=== Returning to Fig. 1, we will now explain the configuration of optical character recognition device 200. Optical character recognition device 200 is a device that, when a target image is input, recognizes characters included in the target image and generates, for example, text information, an accuracy index, and coordinates (hereinafter simply referred to as "generated information") for each recognized character.

[0066] The optical character recognition device 200 may be configured to include a plurality of optical character recognition devices, and may be configured to transmit to the extraction device 100, for example, generated information that is the character recognition result with the highest accuracy among the character recognition results obtained by each of the plurality of optical character recognition devices.

[0067] As shown in FIG. 1 , the optical character recognition device 200 includes, for example, a storage unit 210, a transmission / reception unit 220, and a processing unit 230. The storage unit 210 stores various information. The processing unit 230 performs processing for character recognition. The transmission / reception unit 220 transmits and receives various information to and from the extraction device 100. The processing unit 230 analyzes images using, for example, a neural network trained to distinguish between characters. The neural network includes, for example, multiple convolutional network layers and recurrent network layers. The processing unit 230 segments, for example, a target image into pages, blocks, lines, or characters. Character recognition is performed on characters included in the segmented images to generate, for example, first generated information for each character. The processing unit 230 may generate generated information for, for example, a group of segmented images (for example, a target image, a block image, or a line image).

[0068] ===User terminal 300=== The configuration of the user terminal 300 will be described with reference to Fig. 1. As shown in Fig. 1, the user terminal 300 includes functional units such as a storage unit 310, a transmission / reception unit 320, and a display processing unit 330. Each functional unit is a function realized by, for example, a processor 1001 reading out a program stored in a memory 1002.

[0069] The storage unit 410 stores various types of information. The user information 311 stores user names, attribute information, etc. in association with a user ID as a primary key. The user name is an arbitrary name registered by the user. The attribute information is information such as the user's name and address. The transmission / reception unit 320 transmits and receives various types of information to and from the extraction device 100. The various types of information acquired by the transmission / reception unit 320 are stored in the storage unit 310. The display processing unit 330 displays the various types of information acquired from the extraction device 100 on the display unit.

[0070] ===Processing Procedure=== The processing procedure of the extraction system 10 will be described with reference to Figures 5, 6, 7, and 8. Figure 5 is a flowchart showing the processing procedure of the extraction system 10. Figure 6 is a diagram showing an example of target text information. Figure 7 is a diagram showing an example of extraction query information and extraction block text information input to the generation AI device. Figure 8 is a diagram showing an example of extraction information output from the generation AI device.

[0071] In step S100, the extraction device 100 acquires a target image from a predetermined device and transmits the target image to the optical character recognition device 200.

[0072] In step S101, the optical character recognition device 200 segments the target image to generate target text information of the target image, and transmits the target text information to the extraction device 100.

[0073] In step S102, the extraction device 100 stores the target text information in the text database D101b. At this time, the extraction device 100 divides the target text information into block text information and stores the block text information.

[0074] Specifically, as shown in Figure 6, the block text information is information that indicates, for example, block B10, which is block text information contained in "Article 2," block B20, which is block text information contained in the bullet points of "Property Description," and block B30, which is the tabular area of ​​"Rental Requirements" and the block text information contained in that tabular area.

[0075] In step S103, the extraction device 100 associates an extraction label with each block of text information that satisfies the text features in the label database D101c, and stores the extraction label in the text database D101b in association with the block of text information.

[0076] Specifically, the extraction device 100 identifies block text information with block ID "1001002" in the text database D101b shown in Fig. 3, which satisfies the text feature "text including all of 'A', 'B', and 'contract'" of label ID "2002" in the label database D101c shown in Fig. 4. When the extraction device 100 can identify the block text information, it associates the extraction label "company name" corresponding to the text feature in the label database D101c with the block text information, and stores the associated information in the text database D101b.

[0077] In step S104, the extraction device 100 acquires extraction items, which are categories related to the extraction information, from the user terminal 300. Specifically, the extraction device 100 acquires extraction items indicating "company name," "address," etc. That is, the extraction device 100 acquires information indicating the item contents that the user desires to extract.

[0078] In step S105, the extraction device 100 identifies extracted block text information associated with the extraction label in the text database D101b corresponding to the extraction item.

[0079] Specifically, the extraction device 100 identifies, for example, the extraction label "company name" in the text database D101b that corresponds to the extraction item "company name." The extraction device 100 identifies, as the extracted block text information, the block text information "Article 2..." in the text database D101b that is associated with the extraction label "company name."

[0080] In step S106, the extraction device 100 identifies candidate setting items in the query database D101a that correspond to the extraction items, and identifies query information that corresponds to the identified candidate setting items as extracted query information.

[0081] Specifically, the extraction device 100 identifies, for example, a setting candidate item "company name" in the query database D101a corresponding to the extraction item "company name." The extraction device 100 identifies query information in the query database D101a corresponding to the setting candidate item "company name" as extracted query information (information shown in "extraction requirement" and "extraction example" in FIG. 2).

[0082] In step S107, the extraction device 100 transmits the extracted block text information and the extracted query information to the generation AI device. The extracted block text information and the extracted query information transmitted to the generation AI device will be described below with reference to FIG.

[0083] As shown in FIG. 7, block text information T10 is a predetermined range of text information included in the target text information, and in this case, the block text information is "Article 2...". Extraction query information T20 also includes extraction requirements T21 and extraction examples T22. As shown in FIG. 7, extraction requirements T21 are information for extracting appropriate information, and here, they specify specific descriptions of the extraction target ("landlord" and "borrower") and the output format (JSON array format). Extraction examples T22 are information for improving the convenience of users who use the extracted information, and here, they provide examples of input sentences and specify the extraction format of the extraction results for those examples.

[0084] In step S108, the extraction device 100 acquires the extraction information from the generation AI device. For example, the extraction device 100 displays the extraction information shown in Figure 8 on the display unit of the user terminal 300. As shown in Figure 8, the extraction device 100 displays the extraction target (here, the company name of the "lender" and the company name of the "borrower") on the display unit in the output format (here, JSON format) specified in the extraction requirement T21.

[0085] If the user desires to obtain a plurality of pieces of extraction information, the extraction device 100 repeats the processes from step S104 to step S108.

[0086] In this way, the extraction system 10 can input prompts into the generation AI device to enable the user to properly obtain the information they desire from the generation AI device, thereby enabling the user to properly obtain the information.

[0087] ===Hardware Configuration=== An example of a hardware configuration for implementing the extraction device 100, the optical character recognition device 200, and the user terminal 300 on a computer will be described with reference to Fig. 9. Fig. 9 is a diagram showing an example of the hardware configuration of a computer.

[0088] As shown in FIG. 9, the computer 1000 includes a processor 1001, a memory 1002, a storage device 1003, an input I / F unit 1004, a data I / F unit 1005, a communication I / F unit 1006, and a display unit 1007.

[0089] The processor 1001 is a control unit that controls various processes in the computer 1000 by executing programs stored in the memory 1002 .

[0090] The memory 1002 is a storage medium such as a RAM (Random Access Memory), etc. The memory 1002 temporarily stores the program code of the program executed by the processor 1001 and data required when the program is executed.

[0091] The storage device 1003 is a non-volatile storage medium such as a hard disk drive (HDD), flash memory, etc. The storage device 1003 stores an operating system and various programs for realizing the above-mentioned components.

[0092] The input I / F unit 1004 is a device for receiving input from a user. Specific examples of the input I / F unit 1004 include a keyboard, a mouse, a touch panel, various sensors, and a wearable device. The input I / F unit 1004 may be connected to the computer 1000 via an interface such as a USB (Universal Serial Bus).

[0093] The data I / F unit 1005 is a device for inputting data from outside the computer 1000. A specific example of the data I / F unit 1005 is a drive device for reading data stored in various storage media. The data I / F unit 1005 may be provided outside the computer 1000. In this case, the data I / F unit 1005 is connected to the computer 1000 via an interface such as a USB.

[0094] The communication I / F unit 1006 is a device for performing data communication via the Internet N, either wired or wirelessly, with devices external to the computer 1000. The communication I / F unit 1006 may be provided external to the computer 1000. In this case, the communication I / F unit 1006 is connected to the computer 1000 via an interface such as a USB.

[0095] The display unit 1007 is a device for displaying various types of information. Specific examples of the display unit 1007 include a liquid crystal display, an organic EL (Electro-Luminescence) display, and a display of a wearable device. The display unit 1007 may be provided outside the computer 1000. In this case, the display unit 1007 is connected to the computer 1000 via, for example, a display cable. Furthermore, when a touch panel is used as the input I / F unit 1004, the display unit 1007 can be configured as an integrated unit with the input I / F unit 1004.

[0096] ===Summary=== <1> The extraction system 10 of this embodiment includes a text information acquisition unit 102 that acquires text information including characters, an item acquisition unit that acquires extraction items related to the extraction information to be extracted from the text information based on a user's operation input, a query information identification unit 106 that identifies extraction query information associated with the setting candidate items corresponding to the extraction items based on a query database D101a that associates and stores candidate setting items that can be set as the extraction target with query information to be input to a generation AI device trained to extract extraction information from text information, a transmission unit 107 that transmits the extraction query information and text information to the generation AI device, and an extraction information acquisition unit 108 that acquires the extraction information included in the text information from the generation AI device. This enables the extraction system 10 to appropriately extract information desired by the user from the target text information from the generation AI device.

[0097] <2> The extraction system 10 of this embodiment further includes a block text information identification unit 105 that identifies extracted block text information associated with extraction labels corresponding to extraction items, each block text information being a predetermined group of characters included in text information. The transmission unit 107 transmits the extraction query information and the extracted block text information to the generation AI device, and the extraction information acquisition unit 108 acquires the extraction information included in the extracted block text information from the generation AI device. This allows the extraction system 10 to appropriately extract information desired by the user from the target text information with a smaller amount of processing.

[0098] <3> The extraction system 10 of this embodiment further includes a block text information identification unit 105 that identifies extracted block text information associated with extraction labels corresponding to extraction items based on a text database D101b that stores block text information, which is a predetermined group of characters included in text information, and extraction labels, which are classifications of the predetermined group of characters, in association with each other. The transmission unit 107 transmits the extraction query information and the extracted block text information to the generation AI device, and the extraction information acquisition unit 108 acquires the extraction information included in the extracted block text information from the generation AI device. This allows the extraction system 10 to appropriately extract information desired by the user from the target text information from the generation AI device with less processing load.

[0099] <4> The extraction system 10 in this embodiment further includes a label setting determination unit 103 (label setting unit) that associates extraction labels with block text information based on the correspondence between the extraction labels and the text features of the text included in the text information (e.g., estimation results using the label database D101c and a trained model). This allows the extraction system 10 to appropriately extract information desired by the user from the target text information using a generation AI device with a smaller amount of processing.

[0100] <5> Furthermore, the extraction system 10 of this embodiment further includes a label setting determination unit 103 (determination unit) that, when the text information includes tabular data, determines whether or not the tabular data contains items corresponding to the extraction items, and the extraction information acquisition unit 108 acquires, as extraction information, information related to the items corresponding to the extraction items in the tabular data when the label setting determination unit 103 (determination unit) determines that there are items corresponding to the extraction items. This enables the extraction system 10 to appropriately extract information desired by the user from the target text information using a less processing load from the generation AI device.

[0101] <6> Furthermore, in the present embodiment, the transmitting unit 107 in the extraction system 10 transmits the extracted query information and the extracted block text information to the generating AI device when the label setting determining unit 103 (determination unit) determines that there is no item corresponding to the extracted item. This allows the extraction system 10 to reduce the frequency of using the generating AI device and improve the accuracy of obtaining appropriate extracted information. [Explanation of symbols]

[0102] 10...Extraction system, 100...Extraction device, 101...Memory unit, 102...Text information acquisition unit, 103...Label setting determination unit, 104...Item acquisition unit, 105...Block text information identification unit, 106...Query information identification unit, 107...Transmission unit, 108...Extraction information acquisition unit, 109...Display processing unit, 200...Optical character recognition device, 300...User terminal.

Claims

1. a text information acquisition unit that acquires text information including characters; an item acquisition unit that acquires extraction items related to extraction information to be extracted from the text information based on an operation input by a user; a query information identification unit that identifies extraction query information associated with the candidate setting item corresponding to the extraction item based on a query database that associates and stores query information to be input to a generation AI device that has been trained to extract the extraction information from the text information; and a transmission unit that transmits the extracted query information and the text information to the generation AI device; an extraction information acquisition unit that acquires the extraction information included in the text information from the generation AI device; An information processing system comprising:

2. a block text information specifying unit that specifies extraction labels associated with each piece of block text information, which is a predetermined group of characters included in the text information, and that specifies extracted block text information associated with the extraction labels corresponding to the extraction items; The transmission unit transmits the extracted query information and the extracted block text information to the generation AI device; The extraction information acquisition unit acquires the extraction information included in the extraction block text information from the generation AI device. The information processing system according to claim 1 .

3. a block text information specifying unit that specifies extracted block text information associated with the extraction label corresponding to the extraction item based on a text database that stores block text information, which is a predetermined group of characters included in the text information, and extraction labels, which are classifications of the predetermined group of characters, in association with each other; The transmission unit transmits the extracted query information and the extracted block text information to the generation AI device; The extraction information acquisition unit acquires the extraction information included in the extraction block text information from the generation AI device. The information processing system according to claim 1 .

4. a label setting unit that associates the extraction label with the block text information based on a correspondence between the extraction label and a text feature that is a feature of text included in the text information; 4. The information processing system according to claim 2 or 3.

5. a determination unit that, when the text information includes table-format data, determines whether or not there is an item corresponding to the extracted item in the table-format data; when the determination unit determines that there is an item corresponding to the extraction item, the extraction information acquisition unit acquires, as the extraction information, information related to the item corresponding to the extraction item in the tabular data. The information processing system according to any one of claims 2 to 4.

6. The transmission unit transmits the extracted query information and the extracted block text information to the generation AI device when the determination unit determines that there is no item corresponding to the extracted item. The information processing system according to claim 5 .

7. The computer a text information acquisition unit that acquires text information including characters; acquiring extraction items related to extraction information to be extracted from the text information based on a user's operation input; Identifying extraction query information associated with the candidate setting item corresponding to the extraction item based on a query database that associates and stores candidate setting items that can be set as extraction targets with query information to be input to a generation AI device that has been trained to extract the extraction information from the text information; Sending the extracted query information and the text information to the generating AI device; acquiring the extracted information included in the text information from the generating AI device; An information processing method that performs the above.

8. On the computer, a text information acquisition unit that acquires text information including characters; acquiring extraction items related to extraction information to be extracted from the text information based on a user's operation input; Identifying extraction query information associated with the candidate setting item corresponding to the extraction item based on a query database that associates and stores candidate setting items that can be set as extraction targets with query information to be input to a generation AI device that has been trained to extract the extraction information from the text information; Sending the extracted query information and the text information to the generating AI device; acquiring the extracted information included in the text information from the generating AI device; A program that executes the following.

Images