Information processing device, information processing method, and information processing program

Abstract

[Problem] To address the problem that time and cost are required since an LLM is called every time information collection processing is performed in a conventional technology. [Solution] In order to a

Description

Title of Invention: INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND INFORMATION PROCESSING PROGRAM Technical Field

[0001] The present invention relates to an information processing apparatus that acquires desired information from a file data group containing various information. Background Art

[0002] In recent years, it has been widely spreading to use a large language model (LLM: this abbreviation may hereinafter be used) to acquire predetermined information from a file data group. For example, Patent Literature 1 discloses a technique using an LLM to collect character information related to a predetermined item from a variety of printed materials that do not have a fixed format, such as advertising flyers. Citation List Patent Literature

[0003] Patent Literature 1: Japanese Patent No. 7430437 Summary of Invention Technical Problem

[0004] The technique of Patent Literature 1 gives the LLM a prompt containing various instructions related to collection such as an instruction to specify an item to be collected, thereby collecting one or more pieces of character information about one or more predetermined items from a group of character strings printed on a flyer read by OCR. In the case of such a technique, the LLM is called every time the information collection process is performed, which causes a problem of taking time and being costly. In addition, since the LLM is called every time the process is performed, there is a problem that the extraction accuracy is not constant. Solution to Problem

[0005] Therefore, in order to solve the above problems, the present invention provides the following information processing apparatus and the like. That is, there is provided an information processing apparatus comprising: a prompt setting unit that sets a prompt for specifying desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items; a sample setting unit that sets sample file data that is input to a trained language model together with the set prompt; a script generation unit that inputs the set prompt and the set sample file data to the trained language model to generate a script for acquiring the specified desired information; and a desired information acquisition unit that acquires the specified desired information from the target file data using the generated script.

[0006] Further, there is provided the information processing apparatus further comprising a script storage unit that stores the generated script, in addition to the above features .

[0007] Further, there is provided an information processing apparatus, wherein the script generation unit further inputs correct answer information that is information on a correct answer of the specified desired information for the sample file data to generate the script, in addition to the above features.

[0008] Further, there is provided an information processing apparatus further comprising a script evaluation unit that evaluates a result of acquiring the specified desired information for the sample file data using the generated script in comparison with the correct answer information, in addition to the above features.

[0009] Further, there is provided an information processing apparatus, wherein the script generation unit further inputs a result of evaluation by the script evaluation unit to generate the script, in addition to the above features.

[0010] Further, there is provided an information processing apparatus further comprising a script modification unit that inputs the generated script and part or all of the target file data to the trained language model to modify the generated script to adapt to the target file data, in addition to the above features.

[0011] Further, there is provided an information processing apparatus, wherein the script modification unit further inputs a prompt from which the generated script is created to the trained language model to modify the generated script, in addition to the above features.

[0012] Further, there is provided an information processing apparatus, wherein the script modification unit also modifies a script stored in the script storage unit, in addition to the above features.

[0013] Further, there is provided an information processing apparatus, further comprising a detailed prompt setting unit that sets a detailed prompt that is a detailed prompt for acquiring desired information specified by the set prompt, in addition to the above features .

[0014] Further, there is provided an information processing apparatus further comprising a verification unit that verifies an acquisition result acquired by the desired information acquisition unit, in addition to the above features .

[0015] Further, there is provided an information processing apparatus, wherein the script modification unit modifies a script by reflecting a result of verification by the verification unit, in addition to the above features .

[0016] Further, there is provided an information processing method executed by an information processing apparatus, the information processing method comprising: a prompt setting step of setting a prompt for specifying desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items; a sample setting step of setting sample file data that is input to a trained language model together with the set prompt; a script generation step of inputting the set prompt and sample file data to the trained language model to generate a script for acquiring the desired information; and a desired information acquisition step of acquiring the desired information from the plurality of target file data using the generated script.

[0017] Further, there is provided an information processing program that causes an information processing apparatus to execute: a prompt setting step of setting a prompt for specifying desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items; a sample setting step of setting sample file data that is input to a trained language model together with the set prompt; a script generation step of inputting the set prompt and sample file data to the trained language model to generate a script for acquiring the desired information; and a desired information acquisition step of acquiring the desired information from the plurality of target file data using the generated script.

[0018] Further, there is provided an information processing apparatus comprising: a script holding unit that holds a script for acquiring desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items; a held script modification unit that inputs the held script and part or all of the target file data to the trained language model to modify the held script to adapt to the target file data; and a second desired information acquisition unit that acquires the desired information from the plurality of target file data using the modified script.

[0019] Further, there is provided the information processing apparatus, wherein the held script modification unit further inputs a prompt for modifying the held script to the previously trained language model to modify the held script, in addition to the above features .

[0020] Further, there is provided an information processing method executed by an information processing apparatus, the information processing method comprising: a script holding step of holding a script for acquiring desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items; a held script modification step of inputting the held script and part or all of the target file data to the trained language model to modify the held script to adapt to the target file data; and a second desired information acquisition step of acquiring the desired information from the plurality of target file data using the modified script.

[0021] Further, there is provided an information processing program that causes an information processing apparatus to execute: a script holding step of holding a script for acquiring desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items; a held script modification step of inputting the held script and part or all of the target file data to the trained language model to modify the held script to adapt to the target file data; and a second desired information acquisition step of acquiring the desired information from the plurality of target file data using the modified script. Advantageous Effects of Invention

[0022] The present invention makes it possible to acquire desired information from a file data group containing various information while reducing the time and costs required, and enables reproducible information extraction with a constant extraction accuracy. Brief Description of Drawings

[0023] [Figure 1] Figure 1 is a block diagram showing an example of a functional configuration of an information processing apparatus of a first embodiment. [Figure 2] Figure 2 is a conceptual diagram describing the function of a script generation unit. [Figure 3] Figure 3 is a conceptual diagram showing an example configuration of hardware that implements the information processing apparatus of the first embodiment. [Figure 4] Figure 4 is a flow diagram schematically showing an example of a processing flow of the information processing apparatus of the first embodiment. [Figure 5] Figure 5 is a block diagram showing an example of a functional configuration of an information processing apparatus of a second embodiment. [Figure 6] Figure 6 is a diagram showing setting items and setting examples of a detailed prompt. [Figure 7] Figure 7 is a diagram showing methods of applying each setting item in the detailed prompt. [Figure 8] Figure 8 is a conceptual diagram showing an example of an input screen when a detailed prompt is set. [Figure 9] Figure 9 is a diagram showing an example of various settings when performing verification. [Figure 10] Figure 10 is a conceptual diagram showing an example of converting file data in the PDF format into a text file given position information. [Figure 11] Figure 11 is a conceptual diagram showing another example of converting the format of file data. [Figure 12] Figure 12 is a conceptual diagram showing another way of data reduction. [Figure 13] Figure 13 is a conceptual diagram showing an example of displaying the position on the image as evidence . [Figure 14] Figure 14 is a block diagram showing an example of a functional configuration of an information processing apparatus of a third embodiment. [Figure 15] Figure 15 is a flow diagram showing an example of a processing flow of the information processing apparatus of the third embodiment. Description of Embodiments

[0024] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. Note that the present invention should not be limited to these embodiments in any way, but may be embodied in various aspects without departing from the gist thereof .

[0025] <First Embodiment> <0verview> In acquiring desired information by executing a script (a simple program) on various file data groups, the present invention generates this script by inputting file data, which are sample, and a prompt to a trained language model typified by a large language model (LLM: this abbreviation may hereinafter be used). Here, it is possible to generate a static script that guarantees robustness (the degree of how accurately extraction can be performed from a variety of formats) within the range of the input samples. In addition, since the LLM is called only once when the script is generated, it is possible to reduce the time and costs required by the conventional technique. Further, since the process is performed by a static script, the extraction result is uniquely determined.

[0026] Hereinafter, the functions and the processing flow of the information processing apparatus as well as the content of hardware will be described. Note that the functional blocks of this system described below may be implemented as a combination of hardware and software. Specifically, those that use a computer include a CPU (central processing unit), a main memory, a bus, a secondary storage (e.g., a hard disk drive, a nonvolatile memory, storage media such as CDs and DVDs, and reading drives for those media), an input device used to input information, a printing device, a display device, and other hardware components such as an external peripheral device, an interfaces for the external peripheral device, a communication interface, driver programs for controlling those pieces of hardware, other application programs, and user interface applications. Further, by the arithmetic processing by the CPU according to a program deployed on the main memory, data or the like that is input from input devices, other interfaces, or the like and held on the memory or the hard disk is processed and stored, or an instruction to control each piece of the above hardware and software is generated. Alternatively, the functional blocks of the present system may be implemented by dedicated hardware.

[0027] In addition, the present invention can be implemented not only as a system, but also as a method. Further, such a part of the invention can be configured as software. Furthermore, a program used to cause a computer to execute such software and a recording medium on which the program is fixed are naturally included within the technical scope of this invention (the same applies throughout the present specification).

[0028] <Functional Configuration> Figure 1 is a block diagram showing an example of a functional configuration of an information processing apparatus of the present embodiment. As shown in Figure 1, an information processing apparatus 100 has a prompt setting unit 101, a sample setting unit 102, a script generation unit 103, a script storage unit 104, and a desired information acquisition unit 105.

[0029] <Prompt Setting Unit> The prompt setting unit 101 has a function of setting a prompt for specifying desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items. The target file data may be stored in a storage provided in the information processing apparatus, or may be stored in a storage connected to the information processing apparatus via a communication line or the like. In addition, there are many file formats for the file data, and they include, but not limited to, Word files, PDF (Portable Document Format) files, Excel files, image files, CSV (Comma Separated Values) files, text files, BIM (Building Information Modeling) files, 2D CAD (Computer Aided Design) files, and 3D model files.

[0030] Further, "desired information" means information on a desired item, and information obtained by performing some operation / processing on the information on the item (e.g., the total number, average information, maximum information, spatial relationship such as distance and collision, three-dimensional model information, three-dimensional graphic information, images, and character information). For example, when CSV files related to equipment of a plurality of buildings are targeted and "the total number of elevators" is desired from the file data group, information obtained by extracting the number of elevators of the buildings contained in each file and summing the numbers of them is the desired information. Further, not only the calculation of numerical information but also a processing result of three-dimensional space processing, three-dimensional graphic processing, or the like may be set as the desired information. For example, when three-dimensional model files of a plurality of buildings are targeted and "collision points between pillars and beams" are desired, three-dimensional space information obtained by extracting all models of pillars and beams contained in each file, calculating the presence or absence of collisions between them, and extracting the collision points is the desired information.

[0031] Further, for example, when three-dimensional drawing files of a plurality of buildings are targeted and "inconsistencies between an elevation view and a plan view" are desired, text obtained by extracting all architectural design information on an elevation and a plan contained in each file, checking the presence or absence of consistency when they are reconstructed as a three-dimensional solid, and making the inconsistencies into a sentence is the desired information. In such a three-dimensional space or processing in the three-dimensional space, the present invention that statically performs a fixed process can more often accurately acquire the desired information. A prompt specifying such desired information is set. This prompt may be set by a user, or may be set in advance by a person who provides the present apparatus to the user.

[0032] Note that although a script for acquiring the desired information specified by the set prompt is generated in the script generation unit described later, a prompt that also makes an instruction as to the generation of the script may be set in the prompt setting unit. For example, if "the total number of elevators" is the desired information as in the above example, a prompt may be set which has the instruction content specifying the desired information, such as "Find the total number of elevators from a plurality of given CSV files", or a prompt may be set which has the instruction content to create a script for acquiring the specified desired information, such as "Create a script for finding the total number of elevators from a plurality of given CSV files".

[0033] Further, when the processing of the script is complicated, prompts for generating scripts in a plurality of steps may be set. For example, in the above example of extracting "collision points between pillars and beams" from a three-dimensional model, it is possible to make a division into, for example, (1) a prompt for generating a script for extracting models of pillars and beams contained in each file, (2) a prompt for generating a script for checking for the presence or absence of collisions between them, and (3) a prompt for generating a script for acquiring information in the three-dimensional space obtained by extracting the collision points, and the prompts for generating the respective scripts may be specified.

[0034] <Sample Setting Unit> The sample setting unit 102 has a function of setting sample file data that are input to the trained language model together with the set prompt. The sample file data are file data that are samples when acquiring desired information from a plurality of target file data, for example, as in the above examples, when "the total number of elevators" is set as the desired information, then a CSV file for a plurality of pieces of "building equipment" containing information on the number of elevators is set as the sample file data, when "collision points between pillars and beams" are set as the desired information, then a three-dimensional model file for a plurality of architectural structures containing information on pillars and beams is set as the sample file data, or when "the presence or absence of consistency between an elevation view and a plan view" is set as the desired information, then a 2D drawing file for a plurality of "architectural designs" containing information on the elevation view and the plan view is set as the sample file data.

[0035] Further, a plurality of file data containing the desired information and each having a different format are preferably set as the sample file data. The reason is that this contributes to the generation of a script that can generally acquire desired information even for file data in various formats.

[0036] Note that in many cases, these sample file data are naturally fewer than a plurality of file data targeted for acquisition, but the opposite is also possible. Further, the sample file data may be a part of a plurality of file data targeted for acquisition, or may be other file data. The sample file data may be set by the user, or may be set in advance by a person who provides the present apparatus to the user.

[0037] Note that although the setting of the prompt and the setting of the sample file data have been described as separate configurations for convenience, a configuration is possible such that the prompt and the sample file data are set integrally. That is, by setting a prompt specifying the desired information as well as containing link information to the file data that are samples, the functions of the prompt setting unit and the sample setting unit as described above are implemented to be the same .

[0038] <Script Generation Unit> The script generation unit 103 has a function of generating a script for acquiring the specified desired information by inputting the set prompt and the set sample file data to the trained language model.

[0039] Figure 2 is a conceptual diagram describing the function of the script generation unit. Sample file data 201 and 202 are shown as an example. The sample file data are lists indicating pieces of equipment (elevators, fire extinguishers, and sanitaries) for each building and the numbers of them, one sample file data 201 is written in Japanese, and the other sample file data 202 is mainly written in English.

[0040] These sample file data are input to the LLM together with the prompt. The following description will be given using the LLM, which is typical of trained language models. Note that the trained language model may not only be the LLM but also be a different one as long as it can perform natural language processing tasks, such as a small language model or a multimodal language model that can handle images and other formats.

[0041] Further, the illustrated prompt 203 is: "A plurality of CSV files of building equipment lists will now be given as samples. The lines of them show the types and numbers of pieces of equipment in a plurality of buildings. From these CSV files, create a script that extracts the numbers of elevators and finds the total number of them", and instructs to generate a script that calculates the total number of elevators, which is the desired information.

[0042] The LLM 204 generates a script in response to the input of the prompt 203 and the sample file data 201, 202 like this. By using not only the list written in Japanese but also the list written in English as the sample file data, the generated script 205 causes execution of a process of summing the total number of information in the cells in the rows where the "gxiO" column is and in the column and the total number of information in the cells in the rows where the "Equipment" column is "Elevator" and in the "Number" column. By inputting samples in various notations, it is possible to create a script that can respond to so-called "variation in notation".

[0043] Further, when the processing of the script is complicated as described above, respective scripts may be generated for prompts that are set in a plurality of steps. It is also conceivable that the user modifies the generated script on their own.

[0044] In addition, by inputting sample file data in various formats, it is possible to generate a script that can also appropriately acquire information from file data in each of those formats. In the above example of extracting "collision points between pillars and beams" from a three-dimensional model, a general-purpose script that responds to a variety of model formats can be generated by inputting three-dimensional models in a plurality of different formats as sample data.

[0045] In addition, the script generation unit can be configured to generate a script by further inputting correct answer information that is information on the correct answer of the specified desired information for the sample file data. For example, in the above example if the total number of elevators extracted from the set sample file data is "241", this "241" is the correct answer information. Further inputting this correct answer information to the LLM contributes to the generation of a script that can acquire the desired information more accurately.

[0046] The correct answer information specified here may not only be numerical information, but also be in any format, such as three-dimensional space models, two-dimensional figures, images, or sentences, which are the processing result of the script. Further, not only information on the final result of the script but also correct answer information in the process of processing it may be given as the correct answer information. For example, in the above example of extracting "collision points between pillars and beams" from a three-dimensional model, (1) models of pillars and beams contained in each file, (2) the presence or absence of collisions between them, (3) information in the three dimensional space obtained by extracting the collision points, etc. may be given as correct answer information in the process of processing. Furthermore, since it requires expertise in operation for a user to specify a three-dimensional model that is the correct answer information, it is also desirable to install a three-dimensional viewer or a display unit that makes it possible to easily specify a three-dimensional model or a three-dimensional space that is the correct answer information.

[0047] <Script Storage Unit> The script storage unit 104 has a function of storing the generated script. By having a script storage unit, it is possible to reuse a script that has been set once, and it is possible to acquire a value with reproducibility from a large amount of target file data. Further, it may be stored in association with the acquisition result acquired by the desired information acquisition unit described later using the generated script and the original prompt and the sample file data that are input together.

[0048] <Desired Information Acquisition Unit> The desired information acquisition unit 105 has a function of acquiring the specified desired information from the target file data using the generated script. A process of executing the generated script on the target file data described above to acquire the specified desired information is performed. It is preferable to save the acquired result. This is to supply the acquisition result to post-processing such as verifying or reformatting the acquisition result, as will be described later.

[0049] Further, the acquired information acquired by the desired information acquisition unit may be passed to another script as new target file data. That is, since the generation of a script and the acquisition of desired information from a target file in the present invention can be concatenated, it is possible to build a more complex process.

[0050] As described above, since the process of acquiring desired information is performed by executing a script, the time and costs required can be reduced as compared to the case of calling the LLM each time to acquire the desired information, and reproducible information extraction can be performed with a certain extraction accuracy.

[0051] Furthermore, there has been a problem that "even if you try to prepare a static script in order to solve the problem, only engineers with specialized knowledge can create a script on their own", but by generating a script using the LLM, it is possible to obtain the effect that anyone can create a script on their own.

[0052] <Hardware Configuration> Figure 3 is a conceptual diagram showing an example configuration of hardware that implements the information processing apparatus of the first embodiment. As shown in the figure, an information processing apparatus 300 has a CPU 301 that performs various arithmetic processes, a RAM 302 that is a volatile recording medium, a storage 303 such as a flash memory or an HDD that is a nonvolatile storage medium, a communication interface 304, and an input / output interface 305. The RAM 302 reads a program that performs various arithmetic processes in order to cause the CPU 301 to execute it, and provides a work area (a working area) for the program. Further, a plurality of addresses are assigned to the RAM 302, and a program executed by the CPU 301 can perform processing by determining and accessing the address to exchange data between them (the same applies throughout the present specification).

[0053] Here, the functions of the prompt setting unit 101, the sample setting unit 102, the script generation unit 103, and the desired information acquisition unit 105 of the information processing apparatus 100 of Figure 1 are mainly implemented by the CPU 301 and the RAM 302 in Figure 3. In addition, the function of the script storage unit 104 is mainly implemented by the storage 303 in Figure 3. Further, when a database for storing file data is present externally or a trained language model such as an LLM whose extraction unit is present externally is used, the functions are implemented by exchanging signals and information with each other via the communication interface 304 and the input / output interface 305.

[0054] <Processing Flow> Figure 4 is a flow diagram schematically showing an example of a processing flow of the information processing apparatus of the first embodiment. First, a prompt for specifying desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items is set (S401: a prompt setting step). Then, sample file data that is input to a trained language model together with the set prompt is set (S402: a sample setting step). Then, the set prompt and the set sample file data are input to the trained language model to generate a script for acquiring the specified desired information (S403: a script generation step). Then, the generated script is stored (S404: a script storage step). Then, the specified desired information is acquired from the target file data using the generated script (S405: a desired information acquisition step). Note that the sample setting step may be performed before the prompt setting step, or both steps may be performed integrally. Further, the script storage step may be performed after the desired information acquisition step.

[0055] <Effects> According to the information processing apparatus of the present embodiment, it is possible to acquire desired information from a file data group containing various information while reducing the time and costs required and to perform reproducible information extraction with a certain extraction accuracy.

[0056] <Second Embodiment> <0verview> The present embodiment is an information processing apparatus further including a configuration for evaluating or modifying the generated script, a configuration for setting a more detailed prompt, and a configuration for verifying the acquisition result, on the basis of the first embodiment.

[0057] <Functional Configuration> Figure 5 is a block diagram showing an example of a functional configuration of the information processing apparatus of the present embodiment. As shown in Figure 5, an information processing apparatus 500 has a prompt setting unit 501, a sample setting unit 502, a script generation unit 503, a script storage unit 504, and a desired information acquisition unit 505 as in the first embodiment, and further has a script evaluation unit 506, a script modification unit 507, a detailed prompt setting unit 508, and a verification unit 509. The description of the same components as those in the first embodiment is omitted, and the script evaluation unit 506, the script modification unit 507, the detailed prompt setting unit 508, and the verification unit 509 will be described below.

[0058] <Script Evaluation Unit> The script evaluation unit 506 has a function of evaluating the result of acquiring the specified desired information for the sample file data using the generated script in comparison with the correct answer information. This correct answer information is as described regarding the script generation unit of the first embodiment. By actually acquiring the desired information from the sample file data using the generated script and comparing the result with the correct answer information, it is possible to evaluate how accurate the script is. That is, if the acquired result matches the correct answer information, it can be evaluated as accurate.

[0059] In addition, by configuring the script generation unit so that this evaluation result is further input to the LLM, a script based on the evaluation result can be generated, which can help generate a more accurate script. For example, it can also be configured to continue regenerating a script until the acquired information matches the correct answer information. Note that the evaluation of a script may be performed on a script stored in the script storage unit.

[0060] <Script Modification Unit> The script modification unit 507 has a function of inputting the generated script and part or all of the target file data to the trained language model to modify the generated script to adapt to the target file data. Note that the trained language model used for the modification may be the same as the trained language model used in the script generation unit described above, or may be another.

[0061] For example, if the target file data includes file data in a file format or an item name to be acquired that is not in the sample file data that is input at the time of creating the script, the desired information may not be accurately acquired. In order to prevent such a situation, the generated script is input to the trained language model together with the target file data.

[0062] Further, the script is modified by giving a prompt such as the following example: "This script is a script created with the purpose of 'Select rows indicating elevators in the building equipment list from the CSV file, extract information in the column indicating the number of them, and find the total number1. However, it may not work properly with this file data to be processed. Find out where it doesn't work properly and modify the script". For example, if a script created to acquire "the total number of elevators", causes execution of a process of summing the total number of information in the cells in the rows where the "Hxffi" column is ulPX—] / —" and in the "^" column and the total number of information in the cells in the rows where the "Equipment" column is "Elevator" and in the "Number" column, when the target file data includes file data having cells in the rows where the "ix W" column is and in the column, the information on this file data is omitted from the target of acquisition, and accurate desired information cannot be obtained. Therefore, by inputting the created script and the target file data to the trained language model together with the prompt as described above, a script adapted to the target file data can be generated.

[0063] Modification for adapting it to the file data may be performed by inputting file data one by one from the target file data to the LLM, or may be performed by inputting all file data to the LLM. The prompt from which the script to be modified is created may also further be input.

[0064] Further, a configuration is also possible such that a script stored in the script storage unit described above is also modified in the same way as the modification to the generated script. Since it may be necessary to modify the stored script due to an increase, decrease, or change in the target file data, a configuration is preferable such that this can be modified.

[0065] <Detailed Prompt Setting Unit> The detailed prompt setting unit 508 has a function of setting a detailed prompt that is a detailed prompt for acquiring desired information specified by the set prompt. The setting items of the detailed prompt described below may be items to be applied in the generation of a script, items to be applied to the target file data, items to be applied in a scene where desired information is acquired from the target file data by executing the generated script, or even items applied in the post-processing using the LLM after the execution of the script. The way in which the detailed prompt is applied can be set variously.

[0066] Figure 6 is a diagram illustrating setting items (NAME, DESCRIPTION, and NECESSITY) and setting examples (EXAMPLE) of a detailed prompt, and in this example, a prompt that is set for specifying desired information is "Output collision points between pillars and beams".

[0067] For example, it is shown that a setting item 1 is a setting item that has the name of "GIVING PLURALITY OF DESIRED INFORMATION NAMES" and is for "CORRECT UNDERSTANDING OF DEFINITION OF CHARACTERS". Further, as its description, it is shown that "PLURALITY OF PIECES OF DESIRED INFORMATION TO BE EXTRACTED ARE EXPLICITLY GIVEN", and ' ["$£", "Pillar"] ' is shown as its example. By setting this detailed prompt, an instruction to target not only "ti" but also "Pillar" for extraction is appended to the instruction to "Output collision points between pillars and beams".

[0068] Other setting items of a detailed prompt include "GIVING DESCRIPTION OF DESIRED INFORMATION" in which a description of desired information to be extracted is given for correct understanding of the definition of the characters, "FORMAT SPECIFICATION" in which a user can specify a format to be extracted for suppression of erroneous extraction and unification of the format of the output result, "CONDITION SETTING" in which the content to be extracted is changed according to a specific condition for exception handling or improvement of accuracy, and "PRESENTATION OF SAMPLE SCRIPT" in which a script to be used as a reference when the LLM generates a script is set in advance for improvement of accuracy of generating a script in the LLM. Of course, without being limited to them, a user can set a detailed prompt by providing various setting items on their own.

[0069] Here, methods of applying each setting item described above can be roughly classified into three methods: a method 1 of "NARROWING DOWN FILE CONTENT", a method 2 of "APPENDING PROMPT", and a method 3 of "CHECKING AND REFORMATTING DETECTION RESULT". The method 1 means narrowing down file data to be input before inputting it to the LLM, and the method 2 means appending the set detailed prompt to the main prompt. The method 3 means reformatting the extraction result according to a specified format or the like.

[0070] Figure 7 is a diagram showing methods of applying each setting item in the detailed prompt. As shown in the figure, the method 2 of "APPENDING PROMPT" is applied to each of the items of "GIVING PLURALITY OF DESIRED INFORMATION NAMES", "GIVING DESCRIPTION OF DESIRED INFORMATION", "CONDITION SETTING", and "PRESENTATION OF SAMPLE SCRIPT". Further, all of the methods 1-3 are applied to "FORMAT SPECIFICATION". Application examples will be described later.

[0071] Figure 8 is a conceptual diagram showing an example of an input screen when a detailed prompt is set. As shown in the figure, a prompt of "Output collision points between pillars and beams" that has already been set is displayed on the upper left side of a screen 801. Further, an input field 802 for each of the setting items of "DESIRED INFORMATION NAME", "FORMAT SPECIFICATION", "CONDITION SETTING", and "SAMPLE SCRIPT" is displayed on the left side of the screen. Further, an input field for another setting item is further displayed through the user's operation. By accepting the user's entry operation to each of these input fields, a detailed prompt is set. Further, the script generation unit applies the set detailed prompt to generate a script, or the desired information acquisition unit applies the set detailed prompt to acquire the desired information. By setting a detailed prompt in this way, the accuracy of acquiring the desired information can be improved.

[0072] An example of applying a detailed prompt as described above is shown. For example, if "'Pillar1 also refers to 1 ft' "is set in the item of "DESIRED INFORMATION NAME" and "Output a blank if no pillar can be extracted" is set in the item of "CONDITION SETTING", this detailed prompt is appended to the main prompt and is input. Then, after three-dimensional models written as "Pillar" are also extracted as pillars by applying "DESIRED INFORMATION NAME", collision points between pillars and beams are output. Further, if no pillar can be extracted from the target file data, a blank is output.

[0073] In addition, when setting a detailed prompt, the detailed prompt setting unit can be configured to automatically set each setting item according to the set prompt. Alternatively, it may be configured to accept settings made by the user and recommend some settings for the items that are not set. By accepting settings for the detailed prompt from the user, it is possible to flexibly and appropriately perform extraction according to the user's intention, but it may be difficult for the user to set an optimal prompt. Therefore, in this way, the functions of automatic filling of a detailed prompt and suggestion are effective. These functions enable simplification and improvement of operability.

[0074] Further, in order to achieve simplification and improvement of operability, it is also preferable to configure the desired information acquisition unit to perform a test in which the set prompt and the detailed prompt are applied before the execution process of the script on all of the target file data. This test is to execute the script on a part of the target file data. Based on the result of this test, the user can judge whether the set prompt and detailed prompt bring their desired result or not, and can reset the prompt and the like to obtain a more desirable acquisition result.

[0075] As shown in Figure 8, a "TEST" button 804 is displayed on the left side of the screen 801 under the input fields 802 for the detailed prompt and the display field 803 for the prompt, and a test is performed using this button. Then, the test result is displayed in the dotted frame 805 drawn in the figure for description. Based on the extraction result of this test, the user can reset the prompt or perform the actual acquisition process using the script from which the test result is obtained.

[0076] Note that although this example uses a configuration in which the detailed prompt is set separately from the setting of the prompt specifying the desired information, a configuration is possible such that the prompt and the detailed prompt are set successively or integrally.

[0077] <Verification Unit> The verification unit 509 has a function of verifying the acquisition result acquired by the desired information acquisition unit. There are various specific ways of verification by the verification unit. The verification described below may be performed by inputting a prompt to the LLM or by executing a coded script.

[0078] Figure 9 is a diagram showing an example of various settings when performing verification. As shown in the figure, it is possible to make four types of settings: "USE OF DETAILED PROMPT", "VERIFICATION PROMPT", "VERIFICATION SCRIPT", and "FILE-TO-FILE COMPARISON".

[0079] In "USE OF DETAILED PROMPT", the extraction settings set in the detailed prompt are used for verification. For example, when a format is specified in "FORMAT SPECIFICATION", which is a setting item for the detailed prompt, the extraction result is compared to the specified format to perform a verification of "Is the extraction result within a predetermined similarity range of the specified format?". In addition, using the settings in "CONDITION SETTING", a verification of "If the extraction has failed, has a blank been output?" is performed.

[0080] Further, "VERIFICATION PROMPT" is for independently setting a prompt for verification, for example, a prompt such as "Check if this information indicates the material of elevator doors" is set and input. Further, "VERIFICATION SCRIPT" is for independently setting a script for verification, for example, as shown in the figure, an instruction to return "Error" when the detection result is less than "1" and return "OK" otherwise is input for verification.

[0081] In addition, in "FILE-TO-FILE COMPARISON", at the time of extraction from a plurality of files, the validity of the extraction result is checked by comparing it with those of the other files. As shown as an example, when the extracted files A_D are compared to each other, the file A, the file C, and the file D are all written only with numbers and symbols, while the file B is written mainly in kanji. In such a case, a verification result that "FILE B IS NOT VALID COMPARED TO OTHERS" is obtained.

[0082] Here, when the acquisition result is verified by the verification unit, the script modification unit described above can be configured to modify the script by reflecting the verification result. For example, if the information on the extraction result is an error as described above, an instruction of "The information on XXX was extracted in the previous extraction, but the verification result showed an error. Modify the script considering this error result." is input to the LLM to modify the script.

[0083] <Hardware Configuration> The information processing apparatus of the present embodiment is similar to that of the first embodiment and can be implemented by the hardware configuration shown in Figure 3. The functions of the script evaluation unit 506, the script modification unit 507, the detailed prompt setting unit 508, and the verification unit 509, which are components specific to the information processing apparatus of the second embodiment shown in Figure 5, are mainly implemented by the CPU 301 and the RAM 302 in Figure 3.

[0084] <Processing Flow> The processing flow of the information processing apparatus of the present embodiment is basically the same as the processing flow of the information processing apparatus of the first embodiment. In addition, the information processing apparatus of the present embodiment further includes: a script evaluation step of evaluating the result of acquiring the desired information for the sample file data using the generated script in comparison with the correct answer information; a script modification step of inputting the generated script and part or all of the target file data to the trained language model to modify the generated script to adapt to the target file data; a detailed prompt setting step of setting a detailed prompt that is a detailed prompt for acquiring desired information specified by the set prompt; and a verification step of verifying the acquisition result acquired in the desired information acquisition step.

[0085] <Effects> According to the information processing apparatus of the present embodiment, it is possible to provide an information processing apparatus capable of improving the accuracy and robustness of acquiring desired information.

[0086] <Additional Functions> On the basis of the first embodiment or the second embodiment, it is also possible to build an information processing apparatus to which the following functional components are further added.

[0087] <Filtering Unit> A filtering unit has a function of excluding predetermined file data of the target file data from the target of acquisition by the desired information acquisition unit. There are various approaches of filtering. For example, filtering is performed based on file attributes (e.g., the extension, the size, the file name, and data items associated with the file) . In addition, an individual keyword is set and filtering is performed based on this keyword. Specifically, a script for filtering by performing a semantic search, clustering, a character search, or the like is generated, and filtering is performed by executing the script. Further, filtering may be performed by giving an instruction to the LLM. For example, by inputting an instruction of "Does this file contain content about elevators?", filtering can be performed based on whether content about elevators is contained or not.

[0088] With the function of such a filtering unit, for example, it is possible to respond to the request of a user who wants to acquire area from building outlines out of 500 target file data but wants to skip elevation views as targets .

[0089] <Conversion Unit> A conversion unit has a function of converting the format of the target file data according to the characteristics of the file data. Note that when this information processing apparatus has the conversion unit, the desired information acquisition unit performs acquisition from the file data converted by the conversion unit. For example, when the file data is in the PDF format, it is converted to the text file format. Further, in the case of image data, it is converted into the text file format by performing character recognition using OCR. Further, in the case of three-dimensional data, it may be converted into two-dimensional data or an image. Performing such conversions contributes to improvement of robustness.

[0090] Furthermore, at the time of conversion into a text file, it is also preferable to perform conversion so that position information in text is given. Figure 10 is a conceptual diagram showing an example of converting file data in the PDF format into a text file to which position information is given. Figure 10(a) shows a portion of a file in the PDF format showing a building area table of a house. Figure 10(b) shows an example of conversion into the JSON format in which the texts of "SHAPE", "CALCULATION FORMULA", and "AREA" are written together with their position information. Figure 10(c) shows an example of conversion into an Array format in which the text is written together with the position information in a similar manner. In this way, by giving position information, extraction is performed in consideration of the relationship between texts, which contributes to improvement of robustness. Further, Figure 10(d) shows an example in which the text showing a table showing the "AREA" and "NUMBER OF TATAMI MATS" of each of "WESTERNSTYLE ROOM 1", "WESTERN-STYLE ROOM 2", and "WESTERN-STYLE ROOM 3" in the original file is converted into text not having position information but having table information (in the dotted frame in the figure). By converting it in this way, it can be targeted for acquisition while the table structure is retained.

[0091] Figure 11 is a conceptual diagram showing another example of converting the format of file data. As shown in Figure 11, in the case of a file including figures (a kitchen plan view), it is converted into the SVG format, thereby being converted into text with figure information, so that it becomes easier to perform acquisition by a script. Conversely, two-dimensional vector information may be converted into a raster image or the like, or data conversion from three dimensions to two dimensions may be performed.

[0092] <Data Reduction Unit> A data reduction unit has a function of reducing the data of the target file data. Data reduction can be performed in various ways. For example, the data is reduced by narrowing down the target file data to specific spaces, properties, areas, specific pages, sections, paragraphs, tables, etc. It is also possible to perform filtering by file attributes (e.g., page numbers or positions), perform filtering by individual keywords (semantic searches, clustering, character searches), or perform filtering by spatial positions (x, y, and z coordinates). It is also possible to perform filtering by inputting an instruction to the LLM (e.g., "Does this file contain content about elevators?").

[0093] Figure 12 is a conceptual diagram showing another way of data reduction. As shown in Figure 12(a), a specific portion 1202 is cut out from one file data 1201 and it is checked whether intended data is contained therein. As a result, if the intended data is contained, the portion is supplied to the acquisition process for acquiring the desired information. If it is not contained, a portion 1203 obtained by shifting the previous portion is cut out and checked in the same way. In this way, in the check before acquisition, portions that do not contain the intended data are excluded from the target of acquisition, thereby enabling improvement of processing efficiency. Note that this check may be performed by executing a script, may be performed using the LLM, or may be performed using other approaches. Further, such processing may be performed in the three dimensional space.

[0094] Further, as shown in Figure 12 (b) , the above described check is performed while shifting the specific portion, and the portions containing the intended data are temporarily stored in a memory 1204. Then, it is checked whether the processing memory 1204 contains all the elements necessary for acquiring the intended data, and if it contains all, the portions held in the memory supplied to the acquisition processing in the desired information acquisition unit. On the other hand, when it does not contain all, the specific portion is further shifted and the series of processes from the above check is repeated.

[0095] <Evidence Display Unit> An evidence display unit has a function of, for an acquisition result acquired by the desired information acquisition unit, displaying the evidence of acquisition of the acquisition result.

[0096] In this information processing method, desired information can be acquired by setting a prompt for specifying the desired information and executing a script generated by inputting it to the LLM together with the sample file data. On the other hand, it may be difficult to set a prompt for generating an optimal script to obtain the desired information. Therefore, by displaying the evidence that has led to the acquisition result (e.g., an intermediate process), it is possible to show the user "what to do and how to do it in order to improve the acquisition result".

[0097] Figure 13 is a conceptual diagram showing an example of displaying the position on the image as evidence. The figure shows a file in the PDF format showing a building area table of a house, and the position of the information (characters) acquired from this file is highlighted by surrounding it in a bold dotted frame 1301 on the image. This enables the user to see where the acquisition result is acquired from. In addition, in a script for processing a three-dimensional model, it is also useful to display the process of processing the script, to highlight the information used at that time, etc., for confirmation of the evidence.

[0098] Further, an indication explaining the reason for the acquisition may be displayed. For example, as preprocessing before acquisition, an instruction of "Output the reason for outputting this information as an acquisition result for each detailed process" is input to the LLM. Then, after acquisition, the acquisition result is shown, and the evidence is displayed, for example, "Since Z-01301 is located 140 px to the right of the text of 'DRAWING NUMBER1 and matches the specified regular expression, Z-01301 is extracted as a drawing number". In this way, since the effect of the prompt set on their own and the script generated thereby on the acquisition result is shown, the user can examine various settings, which can help optimize the acquisition of the desired information.

[0099] cThird Embodiment> <0verview> The present embodiment is an information processing apparatus similar to the information processing apparatus of the first embodiment or the second embodiment.

[0100] <Functional Configuration> Figure 14 is a block diagram showing an example of a functional configuration of the information processing apparatus of the present embodiment. As shown in Figure 14, an information processing apparatus 1400 has a script holding unit 1401, a held script modification unit 1402, and a second desired information acquisition unit 1403.

[0101] <Script Holding Unit> The script holding unit 1401 has a function of holding a script for acquiring desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items. In the first embodiment, the script generated by the script generation unit is targeted for modification by the script modification unit, but in the present embodiment, a script created and set in advance is held in the script holding unit. Since the content of the held script is the same as that of the scripts in the first embodiment and the second embodiment, duplicated description will be omitted.

[0102] cHeld Script Modification Unit> The held script modification unit 1402 has a function of inputting the held script and part or all of the target file data to the trained language model to modify the held script to adapt to the target file data.

[0103] The function of the held script modification unit is the same as the function of the script modification unit in the first embodiment and the second embodiment except that the modification target is a script held by the script holding unit. Therefore, duplicated description is omitted here.

[0104] In addition, a configuration is also possible such that a prompt for modifying the held script is further input to the trained language model to modify the script. This prompt is for providing an instruction on how to make modifications, for example, for specifying a file type or format to adapt to as a target of acquisition, or for specifying data to be removed.

[0105] <Second Desired Information Acquisition Unit> The second desired information acquisition unit 1403 has a function of acquiring the desired information from the plurality of target file data using the modified script. The function of the second desired information acquisition unit is the same as the function of the desired information acquisition unit in the first embodiment and the second embodiment except that the acquisition of the desired information is performed by a script modified by the held script modification unit. Therefore, duplicated description is omitted here.

[0106] <Hardware Configuration> The information processing apparatus of the present embodiment is similar to the first embodiment or the second embodiment and can be implemented by the hardware configuration shown in Figure 3. The functions of the held script modification unit 1402 and the second desired information acquisition unit 1403 of the information processing apparatus of the third embodiment shown in Figure 14 are mainly implemented by the CPU 301 and the RAM 302 in Figure 3. Further, the function of the script holding unit 1401 is mainly implemented by the storage 303 in Figure 3.

[0107] <Processing Flow> Figure 15 is a flow diagram showing an example of a processing flow of the information processing apparatus of the present embodiment. First, a script for acquiring desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items is held (S1501: a script holding step) . Then, the held script and part or all of the target file data are input to the trained language model to modify the held script to adapt to the target file data (S1502: a held script modification step). Then, the desired information is acquired from the plurality of target file data using the modified script (S1503: a second desired information acquisition step).

[0108] <Effects> According to the information processing apparatus of the present embodiment, by making it possible to modify a script set and held in advance, desired information can be acquired by a script adapted to target file data. Reference Signs List

[0109] 100, 500: information processing apparatus 101, 501: prompt setting unit 102, 502: sample setting unit 103, 503: script generation unit 104, 504: script storage unit 105, 505: desired information acquisition unit 506: script evaluation unit 507: script modification unit 508: detailed prompt setting unit 509: verification unit 301: CPU 302: RAM 3 03 : storage 5 304: communication interface 305: input / output interface 1400: information processing apparatus 1401: script holding unit 1402: held script modification unit 10 1403: second desired information acquisition unit

Claims

1. An information processing apparatus comprising: a prompt setting unit that sets a prompt for specifying desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items;a sample setting unit that sets sample file data that is input to a trained language model together with the set prompt;a script generation unit that inputs the set prompt and the set sample file data to the trained language model to generate a script for acquiring the specified desired information; anda desired information acquisition unit that acquires the specified desired information from the target file data using the generated script.2 . The information processing apparatus according to claim 1, further comprising a script storage unit that stores the generated script.3 . The information processing apparatus according to claim 1 or 2, wherein the script generation unit further inputs correct answer information that is information on a correct answer of the specified desired information for the sample file data to generate the script.4 . The information processing apparatus according to claim 3, further comprising a script evaluation unit that evaluates a result of acquiring the specified desired information for the sample file data using the generated script in comparison with the correct answer information.5 . The information processing apparatus according to claim 4, wherein the script generation unit further inputs a result of evaluation by the script evaluation unit to generate the script.6 . The information processing apparatus according to claim 1 or 2, further comprising a script modification unit that inputs the generated script and part or all of the target file data to the trained language model to modify the generated script to adapt to the target file data.7 . The information processing apparatus according to claim 6, wherein the script modification unit further inputs a prompt from which the generated script is created to the trained language model to modify the generated script.8 . The information processing apparatus according to claim 6 or 7, wherein the script modification unit also modifies a script stored in the script storage unit.9 . The information processing apparatus according to claim 1 or 2, further comprising a detailed prompt setting unit that sets a detailed prompt that is a detailed prompt for acquiring desired information specified by the set prompt.

10. The information processing apparatus according to claim 1 or 2, further comprising a verification unit that verifies an acquisition result acquired by the desired information acquisition unit.

11. The information processing apparatus according to claim 10, wherein the script modification unit modifies a script by reflecting a result of verification by the verification unit.12 . An information processing method executed by an information processing apparatus, the information processing method comprising:a prompt setting step of setting a prompt for specifying desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items;a sample setting step of setting sample file data that is input to a trained language model together with the set prompt;a script generation step of inputting the set prompt and the set sample file data to the trained language model to generate a script for acquiring the specified desired information; anda desired information acquisition step of acquiring the specified desired information from the target file data using the generated script.

13. An information processing program that causes an information processing apparatus to execute:a prompt setting step of setting a prompt for specifying desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items;a sample setting step of setting sample file data that is input to a trained language model together with the set prompt;a script generation step of inputting the set prompt and the set sample file data to the trained language model to generate a script for acquiring the specified desired information; anda desired information acquisition step of acquiring the specified desired information from the target file data using the generated script.

14. An information processing apparatus comprising: a script holding unit that holds a script for acquiring desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items;a held script modification unit that inputs the held script and part or all of the target file data to a trained language model to modify the held script to adapt to the target file data; anda second desired information acquisition unit that acquires the desired information from the target file data using the modified script.

15. The information processing apparatus according to claim 14, wherein the held script modification unit further inputs a prompt for modifying the held script to the trained language model to modify the held script.16 . An information processing method executed by an information processing apparatus, the information processing method comprising:a script holding step of holding a script for acquiring desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items;a held script modification step of inputting the held script and part or all of the target file data to a trained language model to modify the held script to adapt to the target file data; anda second desired information acquisition step of acquiring the desired information from the target file data using the modified script.

17. An information processing program that causes an information processing apparatus to execute:a script holding step of holding a script for acquiring desired information, which is information about a desired item, from a plurality of target file data containing a plurality of items;a held script modification step of inputting the held script and part or all of the target file data to a trained language model to modify the held script to adapt to the target file data; anda second desired information acquisition step of acquiring the desired information from the target file data using the modified script.INTERNATIONAL SEARCH REPORT International application No. PCT / JP2024 / 0146I9A. CLASSIFICATION OF SUBJECT MATTER GWF9 / 455(2O18.Ol)i;GW^ / 30(2O18.O^ FI: G06F9 / 455 100; G06F8 / 30 According to International Patent Classification (IPC) or to both national classification and IPC B. FIELDS SEARCHED Minimum documentation searched (classification system followed by classification symbols) G06F9 / 455: G06F8 / 30 Documentation searched other than minimum documentation to the extent that such documents are included in the fields searched Published examined utility model applications of Japan 1922-1996 Published unexamined utility model applications of Japan 1971-2024 Registered utility model specifications of Japan 1996-2024 Published registered utility model applications of Japan 1994-2024 Electronic data base consulted during the international search (name of data base and, where practicable, search terms used) C. DOCUMENTS CONSIDERED TO BE RELEVANT Category* Citation of document, with indication, where appropriate, of the relevant passages Relevant to claim No. Y A jgffl ft®, C h a t GPTIfty i A I H ® 09 October 2023, no. 923, pp. 30-41, (NIKKEI PERSONAL COMPUTING), non-official translation (FUKUDA, Kazuhiro, Leave Office Document Creation to ChatGPT: Delegation Techniques for the AI Era) p. 30, right column, lines 2-16, fig. 1, etc. 1-2, 6-17 3-5 Y A GRAHAM, J. Advanced Data Analysis (ChatGPT Enterprise version) [online], 09 December 2023, [retrieved on 13 June 2024], retrieved from the Internet: <URL: https: / / web.archive.org / web / 2023120911275 3 / https : / / help.openai .com / en / articles / 8437071 -advanced-data-analy sis-chatgpt-enterprise- version> particularly, line 31, etc. 1-2, 6-17 3-5 | | Further documents are listed in the continuation of Box C. | | See patent family annex. * Special categories of cited documents: “A” document defining the general state of the art which is not considered to be of particular relevance “D” document cited by the applicant in the international application ,4E” earlier application or patent but published on or after the international filing date •4L” document which may throw doubts on priority claim(s) or which is cited to establish the publication date of another citation or other special reason (as specified) “O” document referring to an oral disclosure, use, exhibition or other means “P” document published prior to the international filing date but later than the priority date claimed “T” later document published after the international filing date or priority date and not in conflict with the application but cited to understand the principle or theory underlying the invention “X” document of particular' relevance; the claimed invention cannot be considered novel or cannot be considered to involve an inventive step when the document is taken alone “Y” document of particular relevance; the claimed invention cannot be considered to involve an inventive step when the document is combined with one or more other such documents, such combination being obvious to a person skilled in the art document member of the same patent family Date of the actual completion of the international search Date of mailing of the international search report 13 June 2024 25 June 2024 Name and mailing address of the ISA / JP Authorized officer Japan Patent Office (ISA / JP) 3-4-3 Kasumigaseki, Chiyoda-ku, Tokyo 100-8915 Japan Telephone No.

Images