Analysis program, information processing device, and analysis method

Abstract

This analysis program causes a computer to execute: a step for acquiring input information indicating a technical idea; a step for acquiring comparison information with which the technical idea is compared; a step for dividing the technical idea per structural element; a step for generating a first instruction sentence that includes an instruction to compare each of a plurality of structural elements obtained by the division with the comparison information; and a step for outputting an evaluation result indicating the relevance between each of the plurality of structural elements and the comparison information, on the basis of results obtained from a large-scale language model by inputting the first instruction sentence into the large-scale language model.

Description

Analysis program, information processing device, and analysis method

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

[0002] In recent years, various large-scale language models have been developed. Large-scale language models are language models trained using a huge amount of text data, and are trained to be able to process a variety of natural languages.

[0003] Non-Patent Document 1 discloses a method for improving the output accuracy of ChatGPT, an example of a large-scale language model. Specifically, Non-Patent Document 1 discloses that the accuracy of the output is improved by inputting the instruction "Let's think step by step" into ChatGPT.

[0004] "The Spell to Make ChatGPT Smarter," Nikkei Newspaper, [online], March 24, 2023, [Retrieved September 6, 2023], Internet <URL: https: / / www.nikkei.com / article / DGXZQOUC22BVO0S3A320C2000000 / >

[0005] While large-scale language models can be applied to a variety of services, there are still few examples of them being implemented as concrete services. In this regard, new techniques that utilize large-scale language models to compare input information that represents technical ideas with other information are desired.

[0006] In one example of the present disclosure, an analysis program is provided, which causes a computer to execute the steps of acquiring input information indicating a technical idea, acquiring comparison information of an object to be compared with the technical idea, dividing the technical idea into constituent elements, generating a first instruction statement including an instruction to compare each of the divided constituent elements with the comparison information, and outputting an evaluation result indicating a relevance between each of the plurality of constituent elements and the comparison information based on a result obtained from a large-scale language model by inputting the first instruction statement into the large-scale language model.

[0007] In one example of the present disclosure, the dividing step includes a step of generating a second instruction sentence for dividing the technical idea into constituent elements, and a step of inputting the second instruction sentence into the large-scale language model.

[0008] In one example of the present disclosure, the dividing step includes dividing the technical idea into constituent elements according to a predetermined rule.

[0009] In one example of the present disclosure, the first instruction sentence includes an instruction to cause the large-scale language model to output the positions of the plurality of constituent elements in the comparison information.

[0010] In one example of the present disclosure, the comparison information includes a patent document, and the description location includes at least one of a paragraph number in the patent document, a figure number in the patent document, and a claim number in the patent document.

[0011] In one example of the present disclosure, the first instruction sentence includes an instruction to cause the large-scale language model to output a reason for extracting the description position.

[0012] In one example of the present disclosure, the first instruction sentence includes an instruction to cause the large-scale language model to output a degree of association between each of the plurality of constituent elements and the comparison information.

[0013] In one example of the present disclosure, the analysis program further causes the computer to perform the steps of generating a third instruction sentence that summarizes the comparison information from a predetermined perspective, and outputting a summary of the comparison information based on a result obtained from the large-scale language model by inputting the third instruction sentence into the large-scale language model.

[0014] In another example of the present disclosure, an information processing device is provided. The information processing device includes a control unit for controlling the information processing device. The control unit executes the following processes: acquiring input information indicating a technical idea; acquiring comparison information of an object to be compared with the technical idea; dividing the technical idea into constituent elements; generating a first instruction statement including an instruction to compare each of the divided constituent elements with the comparison information; and outputting an evaluation result indicating a relevance between each of the plurality of constituent elements and the comparison information based on a result obtained from a large-scale language model by inputting the first instruction statement into the large-scale language model.

[0015] Another example of the present disclosure provides a computer-executable analysis method, the analysis method including the steps of: acquiring input information indicating a technical idea; acquiring comparison information of an object to be compared with the technical idea; dividing the technical idea into constituent elements; generating a first instruction statement including an instruction to compare each of the divided constituent elements with the comparison information; and outputting an evaluation result indicating a relevance between each of the plurality of constituent elements and the comparison information based on a result obtained from a large-scale language model by inputting the first instruction statement into the large-scale language model.

[0016] The above and other objects, features, aspects and advantages of the present invention will become apparent from the following detailed description of the invention taken in conjunction with the accompanying drawings.

[0017] 5. A diagram showing an example of a device configuration of an information processing system. A diagram for explaining a relevance assessment function. A schematic diagram showing an example of a hardware configuration of an information processing device. A schematic diagram showing an example of a hardware configuration of a user terminal. A diagram showing an example of a data flow between an information processing device, a user terminal, and a server. A diagram showing an example of a setting screen displayed in step S110 shown in FIG. 5. A diagram showing an example of a setting screen displayed in step S112 shown in FIG. 5. A diagram showing an example of a setting screen displayed in steps S114 and S118 shown in FIG. 5. A diagram showing an example of an instruction sentence. A diagram showing an example of an instruction sentence. A diagram showing an example of answer information generated by a large-scale language model. A diagram showing an example of an instruction sentence. A diagram showing an example of an evaluation result screen displayed in step S150 shown in FIG. A diagram for explaining an example of generating input information from an image. A diagram schematically showing a relevance assessment process according to another example. A diagram showing an example of an instruction sentence.

[0018] Hereinafter, each embodiment according to the present invention will be described with reference to the drawings. In the following description, the same parts and components are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed description thereof will not be repeated. Note that each embodiment and each modified example described below may be selectively combined as appropriate.

[0019] <A. Information Processing System 10> First, the device configuration of the information processing system 10 will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of the device configuration of the information processing system 10.

[0020] 1, the information processing system 10 includes an information processing device 100, a user terminal 200, and a server 300. The information processing device 100, the user terminal 200, and the server 300 are configured to be able to communicate with each other via a network NW (for example, the Internet).

[0021] The information processing device 100 is a notebook or desktop personal computer (PC), a tablet terminal, a smartphone, or any other computer equipped with a communication function. The number of information processing devices 100 constituting the information processing system 10 may be one, or may be two or more. The information processing device 100 is operated, for example, by a company "A."

[0022] The user terminal 200 is, for example, a notebook or desktop PC, a tablet terminal, a smartphone, or any other computer with a communication function. The number of user terminals 200 constituting the information processing system 10 may be one, or may be two or more. The information processing device 100 is owned, for example, by user "A," who is a general user.

[0023] The server 300 is a laptop or desktop personal computer (PC), a tablet terminal, a smartphone, or any other computer with a communication function. The number of servers 300 constituting the information processing system 10 may be one, or two or more. The server 300 is operated by, for example, company "B."

[0024] The server 300 stores a large-scale language model 324. The large-scale language model 324 is a language model trained on a huge amount of text data, more than several billion pieces of text, and is trained to be able to process various natural languages. The large-scale language model 324 is also called an LLM (Large Language Model). The large-scale language model 324 is trained so that when a directive is input, it generates an output corresponding to the directive.

[0025] Examples of the large-scale language model 324 include the GPT series such as GPT-3 (Generative Pretrained Transformer) and GPT-4, PaLM (Scaling Language Modeling with Pathways), LLaMA (Large Language Model Meta AI), and the well-known LLM. In addition to the GPT series, various large-scale language models may be used, such as a transformer-based large-scale language model such as BERT (Bidirectional Encoder Representations from Transformers), a CNN (Convolutional Neural Network), an RNN (Recurrent Neural Network), and an LSTM (Long Short Term Memory).

[0026] For example, company "B" publishes an API (Application Programming Interface) for using the functions of the large-scale language model 324. This allows designers and general users of company "A" to use the functions of the large-scale language model 324 through the API.

[0027] The various processes described in this specification may be implemented in the information processing device 100, the user terminal 200, the server 300, or other computers.

[0028] Furthermore, although the above description has been given of an example in which the information processing system 10 includes the server 300, the information processing system 10 does not have to include the server 300. In this case, the information processing system 10 is configured with one or more information processing devices 100 and one or more user terminals 200.

[0029] <B. Processing Overview> The information processing device 100 provides the user "A" with a function for evaluating the relevance between input information indicating a technical idea and comparison information.

[0030] A "technical idea" is a technical means for solving a technical problem. A technical idea is defined by a combination of elements (hereinafter referred to as "constituent elements") that constitute an invention. Examples of input information that represents a technical idea include claims described in a patent, constituent elements obtained by summarizing technical documents using a large-scale language model, and constituent elements obtained by summarizing images using a large-scale language model.

[0031] The "comparison information" is information to be compared with the input information. The comparison information may be publicly known information that is publicly available to the general public, or may be private information that is kept secret within the company.

[0032] Public information includes published patent documents and published non-patent documents, while private information includes, for example, information that is kept secret within a company (for example, patent documents and technical documents before publication).

[0033] Examples of patent documents include unexamined patent publications, patent gazettes, published patent publications, republished patents, and utility model publications. For example, patent documents include bibliographic information, a description, claims, drawings, and an abstract. Examples of bibliographic information include an application number, a publication number, a patent registration number, an application date, a publication date, a registration date, an applicant, a patent owner, a title of the invention, an agent, and an application country. Examples of non-patent documents include academic papers, newspaper articles, books, and web pages.

[0034] In the following, "public information" will be used as an example of "comparison information," but the comparison information is not limited to public information.

[0035] An overview of the function of evaluating the relevance between input information 123 and public information 125 will be described with reference to Fig. 2. Fig. 2 is a diagram for explaining the relevance evaluation function.

[0036] The information processing device 100 acquires input information 123 indicating a technical idea. The input information 123 may be acquired from any source. For example, the input information 123 may be acquired from the user terminal 200 described above.

[0037] The information processing device 100 also acquires public information 125 to be compared with the input information 123. The source from which the public information 125 is acquired is arbitrary. As one example, the public information 125 is acquired from a patent database 124 (see FIG. 3 ) described below that is stored in a storage device within the information processing device 100. As another example, the public information 125 is acquired from a patent database in a server managed by the Patent Office. As yet another example, the public information 125 is acquired from a patent database managed by another server.

[0038] The relevance evaluation between the input information 123 and the public information 125 is realized by using the large-scale language model 324. More specifically, the information processing device 100 first divides the technical idea shown in the input information 123 into constituent elements. Then, the information processing device 100 generates a directive 126 (first directive) including an instruction to compare each of the divided constituent elements with the public information 125.

[0039] The directive 126 is, for example, registered in advance as a template in the information processing device 100. The information processing device 100 generates the directive 126 by specifying various pieces of information in argument sections 127 to 129 within the directive 126.

[0040] More specifically, the acquired input information 123 is specified in the argument part 127. The acquired public information 125 is specified in the argument part 129.

[0041] Each divided component is sequentially specified in the argument section 128. For example, if the technical idea indicated in the input information 123 is divided into N components (N is an integer equal to or greater than 2), the information processing device 100 specifies each of the N components in the argument section 128. As a result, the information processing device 100 generates directives 126 corresponding to the number of divided components. In the example of FIG. 2, N directives 126 are generated.

[0042] The generated instruction sentence 126 is input to the large-scale language model 324. When the large-scale language model 324 receives the instruction sentence 126 as input, it generates a response according to the instruction sentence 126.

[0043] The information processing device 100 inputs the directive 126 into the large-scale language model 324, and based on the results obtained from the large-scale language model 324, outputs an evaluation result 130 indicating the relevance between each of the divided constituent elements and the public information 125. The relevance may be expressed as a numerical value indicating the degree of relevance, or may be expressed as an explanatory sentence. In the example of FIG. 2, the relevance is expressed as a numerical value.

[0044] The evaluation result 130 from the information processing device 100 may be output to any destination. As an example, the output destination is the user terminal 200. The evaluation result 130 output to the user terminal 200 is displayed on the display of the user terminal 200, for example.

[0045] The evaluation function is effective, for example, when conducting invalidation document research. More specifically, the user specifies the patent invention to be invalidated as input information 123. The information processing device 100 then breaks down the patent invention into its constituent elements and compares each element with the public information 125. The information processing device 100 then outputs evaluation results 130 indicating the extent to which each element is disclosed in the public information 125. This allows the user to significantly reduce the time required for literature screening and literature peer review.

[0046] The use of the evaluation function is not limited to invalidation document searches. As another example, the evaluation function may be used in prior art searches. In this case, the user specifies the invention to be the subject of the prior art search in the input information 123. The information processing device 100 then breaks down the invention into its constituent elements and compares each element with the public information 125. The information processing device 100 then outputs an evaluation result 130 indicating the degree to which each element is disclosed in the public information 125. This allows the user to easily determine whether the invention being searched possesses novelty or inventive step. As a result, the user can significantly reduce the time required for literature screening and literature peer review.

[0047] As described above, the information processing device 100 uses the large-scale language model 324 to compare the input information 123 indicating a technical idea with the public information 125, and provides new value to the user.

[0048] <C. Hardware Configuration> Next, the hardware configurations of the information processing device 100 and the user terminal 200 shown in FIG. 1 will be described in order with reference to FIGS. 3 and 4. FIG.

[0049] The hardware configuration of server 300 shown in FIG. 1 is similar to that of information processing device 100, and therefore description thereof will not be repeated.

[0050] (C1. Information Processing Apparatus 100) First, the hardware configuration of the information processing apparatus 100 shown in Fig. 1 will be described with reference to Fig. 3. Fig. 3 is a schematic diagram showing an example of the hardware configuration of the information processing apparatus 100.

[0051] The information processing device 100 includes a control device 101, a read only memory (ROM) 102, a random access memory (RAM) 103, a communication interface 104, a display interface 105, an input interface 107, and an auxiliary storage device 120. These components are connected to a bus 110.

[0052] The control device 101 is configured, for example, by at least one integrated circuit. The integrated circuit may be configured, for example, by at least one central processing unit (CPU), at least one graphics processing unit (GPU), at least one application specific integrated circuit (ASIC), at least one field programmable gate array (FPGA), or a combination thereof.

[0053] The control device 101 controls the operation of the information processing device 100 by executing various programs such as an analysis program 122 and an operating system. Upon receiving an execution command for one of the various programs, the control device 101 reads the program from the auxiliary storage device 120 or the ROM 102 into the RAM 103. The RAM 103 functions as a working memory and temporarily stores various data required for executing the various programs.

[0054] A LAN (Local Area Network), an antenna, etc. are connected to the communication interface 104. The information processing device 100 exchanges data with external devices via the communication interface 104. The external devices include, for example, a user terminal 200, a server 300, and other communication devices.

[0055] A display 106 is connected to the display interface 105. The display interface 105 sends an image signal for displaying an image to the display 106 in accordance with a command from the control device 101 or the like. The display 106 is, for example, a liquid crystal display, an organic EL (Electro Luminescence) display, or other display device. The display 106 may be configured integrally with the information processing device 100 or may be configured separately from the information processing device 100.

[0056] An input device 108 is connected to the input interface 107. The input device 108 is, for example, a mouse, a keyboard, a touch panel, or any other device capable of receiving user operations. The input device 108 may be configured integrally with the information processing device 100 or may be configured separately from the information processing device 100.

[0057] The auxiliary storage device 120 is, for example, a hard disk, a flash memory, an SSD (Solid State Drive), or other storage medium. The auxiliary storage device 120 stores the analysis program 122, the patent database 124, the instruction statement 126, and the like. The patent database 124 includes a plurality of pieces of public information 125. The analysis program 122, the patent database 124, and the instruction statement 126 may not be stored in the auxiliary storage device 120, but may also be stored in a memory area (e.g., cache memory, etc.) of the control device 101, the ROM 102, the RAM 103, an external device, or the like.

[0058] The analysis program 122 may be provided not as a standalone program but as part of an arbitrary program. In this case, various processes defined in the analysis program 122 are realized in cooperation with an arbitrary program such as the analysis program 222 described below. Even a program that does not include some of these modules does not deviate from the spirit of the analysis program 122 according to this embodiment. Furthermore, some or all of the functions provided by the analysis program 122 may be realized by dedicated hardware. Furthermore, the information processing device 100 may be configured in the form of a so-called cloud service in which at least one server executes part of the processing of the analysis program 122.

[0059] (C2. User Terminal 200) Next, the hardware configuration of the user terminal 200 shown in Fig. 1 will be described with reference to Fig. 4. Fig. 4 is a schematic diagram showing an example of the hardware configuration of the user terminal 200.

[0060] The user terminal 200 includes a control device 201, a ROM 202, a RAM 203, a communication interface 204, a display interface 205, an input interface 207, and an auxiliary storage device 220. These components are connected to a bus 210.

[0061] The control device 201 is configured, for example, by at least one integrated circuit. The integrated circuit may be configured, for example, by at least one CPU, at least one GPU, at least one ASIC, at least one FPGA, or a combination thereof.

[0062] The control device 201 controls the operation of the user terminal 200 by executing various programs such as an analysis program 222 and an operating system. Upon receiving an execution command for one of the programs, the control device 201 reads the program from the auxiliary storage device 220 or the ROM 202 into the RAM 203. The RAM 203 functions as a working memory and temporarily stores various data required for executing the program.

[0063] A LAN, an antenna, and the like are connected to the communication interface 204. The user terminal 200 exchanges data with external devices via the communication interface 204. The external devices include, for example, the information processing device 100, the server 300, and other communication devices. The user terminal 200 may be configured to be able to download the analysis program 222 from the information processing device 100.

[0064] A display 206 is connected to the display interface 205. The display interface 205 sends an image signal for displaying an image to the display 206 in accordance with a command from the control device 201 or the like. The display 206 is, for example, a liquid crystal display, an organic EL display, or other display device. The display 206 may be configured integrally with the user terminal 200 or may be configured separately from the user terminal 200.

[0065] An input device 208 is connected to the input interface 207. The input device 208 is, for example, a mouse, a keyboard, a touch panel, or any other device capable of receiving user operations. The input device 208 may be configured integrally with the user terminal 200 or may be configured separately from the user terminal 200.

[0066] The auxiliary storage device 220 is, for example, a hard disk, a flash memory, an SSD (Solid State Drive), or other storage medium. The auxiliary storage device 220 stores the analysis program 222 and the like. The analysis program 222 may be stored in a storage area of ​​the control device 201 (for example, a cache memory), the ROM 202, the RAM 203, an external device (for example, a server), or the like, without being limited to the auxiliary storage device 220.

[0067] The analysis program 222 may be provided not as a standalone program but as part of an arbitrary program. In this case, various processes defined in the analysis program 222 are realized in cooperation with an arbitrary program such as the above-described analysis program 122. Even a program that does not include some of these modules does not deviate from the spirit of the analysis program 222 according to this embodiment. Furthermore, some or all of the functions provided by the analysis program 222 may be realized by dedicated hardware. Furthermore, the user terminal 200 may be configured in the form of a so-called cloud service in which at least one server executes part of the processing of the analysis program 222.

[0068] <D. Data Flow> Next, the operation of the information processing system 10 related to the evaluation of the relevance between the input information 123 and the public information 125 will be described with reference to Figures 5 to 13. Figure 5 is a diagram showing an example of data flow between the information processing device 100, the user terminal 200, and the server 300.

[0069] In the following, the explanation will be given on the assumption that the above-mentioned relevance assessment function provided by the information processing device 100 is applied to the review of public information 125, but the application of the relevance assessment function is not limited to the review of public information 125.

[0070] (D1. Step S110) First, the process of step S110 shown in Fig. 5 will be described with reference to Fig. 6. Fig. 6 is a diagram showing an example of a setting screen 400A displayed in step S110.

[0071] The setting screen 400A accepts the setting of search conditions for searching the public information 125 from the above-mentioned patent database 124 (see FIG. 3 ). By setting the search conditions, the user can specify one or more pieces of public information 125 to be compared with the input information 123. The setting screen 400A includes, for example, a selection field 410 and a setting field 412.

[0072] The selection field 410 accepts a selection input of a search type. For example, search types that can be specified in the selection field 410 include "full-text search," "field search," "command search," "semantic search," and "number search."

[0073] The setting field 412 accepts input of various search conditions. Examples of search conditions that can be input into the setting field 412 include the patent document number, the type of the number (for example, application number or publication number), and the country in which the application was published.

[0074] Although the above description has been given of an example in which the search conditions are input by the user, the search conditions may be generated automatically based on the input information 123 .

[0075] (D2. Step S112) Next, the process of step S112 shown in Fig. 5 will be described with reference to Fig. 7. Fig. 7 is a diagram showing an example of a setting screen 400B displayed in step S112.

[0076] The setting screen 400B is displayed, for example, by scrolling the setting screen 400A. The setting screen 400B accepts settings for the review conditions. As an example, the setting screen 400B includes a setting field 420, a selection field 422, and a setting field 424. The setting field 424 includes an editing field 425.

[0077] The setting field 420 accepts settings for review targets. For example, the setting field 420 accepts input of the maximum number of public information 125 to be reviewed and the review portion of the public information 125. Examples of review portions that can be specified include the abstract, the title of the invention, the top claim in the claims (i.e., claim 1), the technical field, the background art, the problem to be solved by the invention, the effects of the invention, the means for solving the problem, and the specification.

[0078] The selection field 422 accepts the selection of the peer review type. Selectable peer review types include, for example, "search perspective / comparative invention," "claim relevance evaluation," "user instructions such as summarization / information extraction," and "user tag / evaluation." The user can use the relevance evaluation function by selecting "claim relevance evaluation."

[0079] The setting field 424 receives the setting of the input information 123 indicating the technical idea. The input information 123 can be set in various ways.

[0080] In one aspect, the user selects the "Claim Text" button in the setting field 424. In this case, the user can input the claim himself / herself in the edit field 425. For example, the user can copy and paste a claim from a document or a claim proposal that the user has created into the edit field 425. Alternatively, the user may input the claim directly into the edit field 425.

[0081] In another aspect, the user selects the "Publication Number" button in the setting field 424. In this case, the user inputs a publication number such as a patent number or publication number into the setting field 424 and presses the "Get Claim Text" button. Based on this, the information processing device 100 refers to the above-mentioned patent database 124 and searches for a patent document corresponding to the input publication number. Then, the information processing device 100 reflects the claims described in the searched patent document in the editing field 425.

[0082] In yet another aspect, the user selects the "Technical Text" button in the setting field 424. In this case, the user specifies a document that expresses a technical idea. Examples of such a document include technical papers, newspaper articles, and invention documents. The user then presses the "Get Claim Text" button. Based on this, the information processing device 100 extracts constituent features from the specified document by having the large-scale language model 324 summarize the specified document.

[0083] More specifically, the information processing device 100 generates an instruction for extracting technical features from the input document and inputs the instruction to the large-scale language model 324. As an example, the instruction includes an instruction to "identify technical features based on the [target document] and generate a Japanese patent claim." The [target document] is an argument that specifies the document to be summarized. This causes the large-scale language model 324 to extract constituent elements from the input document. The extracted constituent elements are reflected in the edit field 425.

[0084] (D3. Steps S114, S116, S118) Next, the processes of steps S114, S116, and S118 shown in Fig. 5 will be described with reference to Fig. 8 and Fig. 9. Fig. 8 is a diagram showing an example of a setting screen 400C displayed in steps S114 and S118.

[0085] Settings screen 400C is displayed, for example, by scrolling the above-described settings screen 400B. As an example, settings screen 400C includes a selection field 430, an edit field 432, a selection field 436, a selection field 438, and a start button 440. Edit field 432 includes a display field 434 for configuration requirements.

[0086] The selection field 430 accepts the selection of a method for dividing the input information 123 into constituent elements, which is set in the above-described edit field 425 (see FIG. 7 ). As an example, the selection field 430 includes a “line break” button, an “automatic division by AI” button, and a “manual input” button.

[0087] In a certain situation, it is assumed that the user selects the "Line Break" button in the selection field 430. In this case, the information processing device 100 divides the technical idea contained in the input information 123 into constituent elements in accordance with a predetermined division rule (step S116). In the example of FIG. 8, a claim, which is an example of the input information 123, is divided. The division rule includes dividing into predetermined keywords. Examples of the predetermined keywords include line break codes, "is," "and," and "consisting of."

[0088] In another aspect, it is assumed that the user selects the “Automatic division by AI” button in the selection field 430. In this case, the information processing device 100 divides the claim into constituent features using the large-scale language model 324 (step S116).

[0089] More specifically, the information processing device 100 generates a directive (second directive) for dividing the technical idea as the input information 123 into constituent elements. Fig. 9 is a diagram showing an example of the generated directive 156. The directive 156 is stored in advance in the auxiliary storage device 120 of the information processing device 100 as a template, for example.

[0090] The directive 156 is defined to divide the input information 123 specified in the argument section 157 into constituent elements. In the argument section 157, for example, a claim, which is an example of the input information 123, is specified.

[0091] Thereafter, the information processing device 100 inputs the instruction statement 156 that specifies the input information 123 to the large-scale language model 324. Upon receiving the instruction statement 156, the large-scale language model 324 generates an answer corresponding to the instruction statement 156. The generated answer is output to the information processing device 100. In the generated answer, the input information 123 is indicated for each constituent element.

[0092] In yet another aspect, it is assumed that the user selects the “Manual Input” button in selection field 430. In this case, information processing device 100 allows the user to freely edit the constituent features of the claim in edit field 432.

[0093] The input information 123 divided by the method selected in the selection field 430 is displayed for each constituent element in a display field 434 of the editing field 432. The display fields 434 are displayed in the editing field 432 according to the number of divided constituent elements.

[0094] As an example, when a claim input as input information 123 is divided into constituent features #1 to #4, four display columns 434 are displayed in edit column 432. In the example of Fig. 8, constituent feature #1 is displayed in display column 434A, constituent feature #2 is displayed in display column 434B, constituent feature #3 is displayed in display column 434C, and constituent feature #4 is displayed in display column 434D.

[0095] The user can freely edit each of the constituent elements displayed in the display field 434 .

[0096] The user can also increase or decrease the number of display columns 434 displayed in the edit column 432. As an example, the user can increase the number of blank display columns 434 by pressing an "Add component requirement" button. The user can delete the display column 434 corresponding to the "Delete component requirement" button. Furthermore, the user can delete all display columns 434 displayed in the edit column 432 by pressing a "Delete all component requirements" button.

[0097] The selection field 436 accepts the selection of an item to be summarized regarding the public information 125. The information processing device 100 generates a summary of the public information 125 according to the item to be summarized selected in the selection field 436. The generation method will be described later.

[0098] The selection field 438 accepts the selection of the large-scale language model 324 to be used. The large-scale language model 324 selected in the selection field 438 is used for evaluating the relevance between the input information 123 and the public information 125, and the like.

[0099] When the user presses the start review button 440, the user terminal 200 transmits the information entered on the setting screens 400A to 400C (see FIGS. 6 to 8) to the information processing device 100.

[0100] In the above, an example has been described in which the processes of steps S112, S114, S116, and S118 are executed after the process of step S110, but the processes of steps S112, S114, S116, and S118 may be executed before the process of step S110. In this case, the search conditions in step S110 may be automatically set using the divided constituent elements #1 to #4 as conditions for a similar search. When the search conditions are automatically set, the process of step S110 does not necessarily have to be executed.

[0101] (D3. Step S120) Next, in step S120, the information processing device 100 searches for public information 125 that matches the search conditions set on the setting screen 400A (see FIG. 6 ) from among the public information 125 registered in the patent database 124. The information processing device 100 uses the public information 125 that matches the search conditions as a population to be compared with the input information 123.

[0102] (D4. Step S122) Next, in step S122, the information processing apparatus 100 generates the above-mentioned directive statement 126 (see FIG. 2) to be input to the large-scale language model 324.

[0103] 10 is a diagram illustrating an example directive 126. Preferably, directive 126 includes a plurality of instructions 131A-131D.

[0104] The information processing device 100 specifies input information 123 for the argument part 127 of each of the instructions 131A to 131D. The information processing device 100 also specifies public information 125 for the argument part 129 of each of the instructions 131A to 131D.

[0105] Furthermore, the information processing apparatus 100 sequentially specifies each of the constituent elements divided on the setting screen 400C for the argument section 128 of each of the instructions 131A to 131D.

[0106] As an example, assume that the technical idea contained in input information 123 is divided into constituent elements #1 to #4. In this case, information processing device 100 specifies constituent element #1 in argument section 128 for each of instructions 131A to 131D, thereby generating a first directive 126. Next, information processing device 100 specifies constituent element #2 in argument section 128 for each of instructions 131A to 131D, thereby generating a second directive 126. Next, information processing device 100 specifies constituent element #3 in argument section 128 for each of instructions 131A to 131D, thereby generating a third directive 126. Next, information processing device 100 specifies constituent element #4 in argument section 128 for each of instructions 131A to 131D, thereby generating a fourth directive 126. In this way, the information processing device 100 repeats the generation of the directive 126 according to the number of divided constituent elements #1 to #4.

[0107] Instruction 131A specifies instructions for extracting the description content related to each of constituent elements #1 to #4 from public information 125. This allows the user to easily understand whether each of constituent elements #1 to #4 is disclosed in public information 125.

[0108] Instruction 131B specifies instructions for causing large-scale language model 324 to extract the description positions of constituent features #1 to #4 in public information 125. The description positions may be indicated by paragraph numbers in the patent document, by claim numbers in the patent document, or by the number of characters or lines from a predetermined reference position in the patent document. This allows the user to easily grasp the description positions of constituent features #1 to #4 in public information 125.

[0109] Instruction 131C specifies an instruction for causing large-scale language model 324 to output the reason for extracting the description positions of constituent features #1 to #4 in public information 125. In other words, instruction 131C specifies an instruction for outputting the reason for determining that each of constituent features #1 to #4 is described in each description position. This provides the user with information for determining whether each of constituent features #1 to #4 is disclosed in public information 125.

[0110] Instruction 131D specifies an instruction for causing large-scale language model 324 to output the degree of association between each of constituent features #1 to #4 and public information 125. This allows the user to easily understand the degree to which each of constituent features #1 to #4 is associated with public information 125.

[0111] The information processing device 100 transmits the instruction statement 126 generated in step S122 to the server 300.

[0112] In the above description, an example has been given in which four instructions 131A to 131D are defined in instruction statement 126, but it is not necessary that all of instructions 131A to 131D be defined in instruction statement 126. As an example, at least one of instructions 131A to 131D may be defined in instruction statement 126.

[0113] In addition, although the above description has been given on the assumption that the entire text of the public information 125 is specified in the argument section 129, if the public information 125 contains a large amount of text, the information processing device 100 may divide the public information 125 and then compare the public information 125 with the constituent elements #1 to #4. This allows the user to use the API without any restrictions, even if the amount of data that can be input is limited in the terms of use for the API for using the large-scale language model 324. This also prevents the comparison result from being overlooked.

[0114] (D5. Step S124) Next, in step S124, the server 300, upon receiving the instruction statement 126 from the information processing device 100, inputs the instruction statement 126 into the large-scale language model 324. As a result, the large-scale language model 324 generates an answer corresponding to the instruction statement 126.

[0115] 11 is a diagram showing answer information 326, which is an example of an answer generated by the large-scale language model 324. In the example of FIG. 11, the answer information 326 is shown in a table format, but the format of the answer information 326 is arbitrary. The output format of the large-scale language model 324 is, for example, specified in advance in the instruction statement 126, and the large-scale language model 324 outputs the answer information 326 in accordance with the output format specified in the instruction statement 126.

[0116] The response information 326 includes an evaluation result 327 for each piece of public information 125 searched in step S120. In the example of Fig. 11, each piece of evaluation result 327 is associated with an identifier of the public information 125. The identifier of the public information 125 is specified by, for example, an application number, a publication number, or a registered publication number.

[0117] Furthermore, the evaluation result 327 includes a "description", a "description location", a "reason for extraction", and a "relevance score" for each of the constituent elements #1 to #4 (see FIG. 8) set on the setting screen 400C. The "description" is the result output from the large-scale language model 324 in response to the above-mentioned instruction 131A (see FIG. 10). The "description location" is the result output from the large-scale language model 324 in response to the above-mentioned instruction 131B (see FIG. 10). The "reason for extraction" is the result output from the large-scale language model 324 in response to the above-mentioned instruction 131C (see FIG. 10). The "relevance score" is the result output from the large-scale language model 324 in response to the above-mentioned instruction 131D (see FIG. 10).

[0118] The server 300 transmits the response information 326 generated in step S124 to the information processing device 100.

[0119] (D6. Step S126) Next, in step S126, information processing device 100 generates a directive 166 shown in Fig. 12. Fig. 12 is a diagram showing an example of directive 166.

[0120] The directive 166 is, for example, stored in advance as a template in the auxiliary storage device 120 of the information processing device 100. The directive 166 (third directive) defines an instruction for summarizing the public information 125 from a predetermined perspective.

[0121] As an example, instruction 166 includes at least one of instructions 167A summarizing public information 125 with respect to the technical field, instructions 167B summarizing public information 125 with respect to the problem that the invention is intended to solve, instructions 167C summarizing public information 125 with respect to the effects of the invention, and instructions 167D summarizing public information 125 with respect to the action / function.

[0122] More specifically, each of the instructions 167A to 167D includes an argument portion 168. For each of the argument portions 168, the information processing device 100 specifies the public information 125 to be summarized.

[0123] The information processing device 100 transmits the instruction statement 166 generated in step S126 to the server 300.

[0124] (D7. Step S128) Next, in step S128, the server 300, upon receiving the instruction statement 166 from the information processing device 100, inputs the instruction statement 166 to the large-scale language model 324. As a result, the large-scale language model 324 outputs a summary of the public information 125 in accordance with the instruction statement 166.

[0125] The summaries output from the large-scale language model 324 include a summary of public information 125 related to the technical field, a summary of public information 125 related to the problems that the invention aims to solve, a summary of public information 125 related to the effects of the invention, and a summary of public information 125 related to the actions and functions.

[0126] The server 300 transmits the response information generated in step S128 to the information processing device 100. In this way, the information processing device 100 outputs a summary of the public information 125 based on the result obtained from the large-scale language model 324 by inputting the directive statement 166 into the large-scale language model 324.

[0127] (D8. Steps S140, S150) Next, in step S140, information processing device 100 generates an evaluation result screen based on response information 326 received from server 300. The evaluation result screen is written in a language such as HTML (HyperText Markup Language), for example.

[0128] Next, in step S150, the user terminal 200 displays the evaluation result screen generated by the information processing device 100. Fig. 13 is a diagram showing an example of the evaluation result screen 400D. The evaluation result screen 400D is displayed on the display 206 of the user terminal 200, for example.

[0129] The evaluation result screen 400D includes a display field 450 that displays information regarding the input information 123, a display field 452 that displays information regarding the public information 125, a display field 454 that displays the evaluation results regarding the relevance of the input information 123 and the public information 125, and a display field 456 that displays information regarding a summary of the public information 125.

[0130] The display field 450 displays information entered in the setting screens 400A and 400B (see FIGS. 6 and 7) described above. As an example, the display field 450 displays an "ID" (Identification), a "Review Type", a "Model", the "Number of Acquired Items", a "Status", an "Acquisition Time", a "Review Subject", and a "Constituent Requirement".

[0131] "ID" is an identifier for uniquely identifying the peer review evaluation results. "Review type" corresponds to the item set in the selection field 422 (see FIG. 7) described above. "Model" indicates the type of large-scale language model 324 used during the relevance evaluation, and corresponds to the information set in the selection field 438 (see FIG. 8) described above. "Number of acquired items" corresponds to the number of public information 125 compared with the input information 123. "Status" indicates whether the relevance evaluation process was completed successfully. "Acquisition time" indicates the time required from the start to the end of the relevance evaluation process. "Review target" corresponds to the information selected in the setting field 420 (see FIG. 7) described above. "Configuration requirements" displays the configuration requirements set in the editing field 432 (see FIG. 8) described above.

[0132] The display field 452 displays information related to the public information 125 that was the subject of the relevance evaluation. For example, the display field 452 displays the "Title of the Invention," "Solution," "Selected Drawing," "Application Number," "Publication Number," "Legal Status," and "Applicant."

[0133] Display field 454 displays the evaluation results generated from the response information 326 (see FIG. 11 ) of the large-scale language model 324. As an example, the evaluation results display the relevance scores of each of constituent elements #1 to #4 with the public information 125, the positions of constituent elements #1 to #4 in the public information 125, the contents of constituent elements #1 to #4 described in the public information 125, and the reasons for determining the relevance scores. The evaluation results also include a total score of the relevance between constituent elements #1 to #4 and the public information 125.

[0134] A summary of the public information 125 is displayed in the display field 456. The summary is the result obtained by inputting the directive 166 (see FIG. 12 ) described above into the large-scale language model 324. The summary includes a summary of the public information 125 related to the technical field, a summary of the public information 125 related to the problem to be solved by the invention, a summary of the public information 125 related to the effects of the invention, and a summary of the public information 125 related to the action / function.

[0135] As described above, by extracting the comparison results between the input information 123 and the public information 125 separately for components #1 to #4, the user can significantly reduce the time it takes to read the lengthy public information 125. As a result, the user can significantly reduce the time it takes to conduct invalidity searches and prior art searches.

[0136] 13, the comparison result of one piece of public information 125 is shown in display fields 452, 454, and 456. In reality, however, display fields 452, 454, and 456 are displayed on evaluation result screen 400D for the number of pieces of public information 125 whose relevance has been evaluated. The evaluation results of the public information 125 are sorted, for example, in order of the total score of the relevance evaluation.

[0137] <E1. Others> Next, other examples of the above-described embodiment will be described.

[0138] In the above-described setting screens 400B and 400C, an example has been described in which the input information 123 is generated by dividing a claim entered as text into constituent features, but the input information 123 indicating the technical idea does not necessarily have to be generated from text. For example, the input information 123 indicating the technical idea may be generated from an image.

[0139] 14 is a diagram for explaining an example of generating input information 123 from an image IM. In the example of FIG. 14, an image IM showing a pet bottle is shown.

[0140] The large-scale language model 324 according to this example is trained to be able to interpret the content of images. Examples of large-scale language models that can interpret the content of images include CLIP (Contrastive Language-Image Pretraining), BLIP (Bootstrapping Language Image Pre-training for unified vision-language understanding and generation), and ViLBERT (Vision-and-Language BERT).

[0141] The information processing device 100 inputs an image IM and an instruction 176 to the large-scale language model 324. The instruction 176 is defined to interpret the objects shown in the image IM by their constituent elements. As a result, the large-scale language model 324 divides the objects shown in the image IM into constituent elements and outputs input information 123 indicating a technical idea.

[0142] 14 is an actual output result of ChatGPT. In this output result, the pet bottle shown in the image IM is divided into the constituent elements of the "bottle cap," "neck portion," "body portion," and "bottom portion."

[0143] By extracting the technical idea from the image in this way, even a designer who is not familiar with creating claims can easily generate the input information 123.

[0144] <E2. Others> Next, still another example of the above embodiment will be described.

[0145] 10, the large-scale language model 324, upon receiving the directive 126, outputs the location where each constituent element is described in the public information 125 as an example of the evaluation result. In the above example, the paragraph number in the public information 125 was output as the description location. In contrast, in this example, the drawing number in the public information 125 is output as the description location.

[0146] 15 is a diagram illustrating the relevance evaluation process in this example. As shown in FIG. 15, the public information 125 includes one or more drawings 125F. The drawings 125F are, for example, image data. The format of the images is not particularly limited.

[0147] The information processing apparatus 100 according to this example uses the large-scale language model 324 to compare the input information 123 including the drawing 125F in the public information 125.

[0148] More specifically, the large-scale language model 324 according to this example is trained to be able to interpret the content of an image. The information processing device 100 inputs a drawing 125F and a predetermined instruction sentence to the large-scale language model 324. The instruction sentence is defined to describe an object shown in the drawing 125F. Upon receiving the instruction sentence, the large-scale language model 324 outputs an explanation of the object shown in the drawing 125F.

[0149] Thereafter, information processing device 100 instructs large-scale language model 324 to compare input information 123 with the explanatory text of drawing 125F. As a result, information processing device 100 evaluates the relevance between input information 123 and drawing 125F included in public information 125. Information processing device 100 also evaluates the relevance between input information 123 and a document (such as a specification) included in public information 125. This evaluation is as described above, and therefore the description thereof will not be repeated.

[0150] This outputs an evaluation result 130. The evaluation result 130 includes not only the relevance score between each constituent feature and the public information 125, but also the description position of each constituent feature in the public information 125. The description position may be indicated by a paragraph number of the specification in the public information 125, or by a drawing number in the public information 125.

[0151] <E3. Others> Next, still another example of the above embodiment will be described.

[0152] In the above description, there is no particular evaluation as to whether or not it is possible to rationalize the combination of each piece of public information 125. In contrast, the information processing device 100 according to the present example evaluates whether or not it is possible to rationalize the combination of each piece of public information 125.

[0153] 16 is a diagram showing an example of a directive 186 used in the rationale evaluation. The directive 186 is stored in advance as a template in the auxiliary storage device 120 of the information processing device 100. The directive 186 includes argument portions 187, 188, 189A, and 189B.

[0154] The directive 186 is prescribed to determine whether the inventive step can be denied for the combination of patent documents specified in the argument sections 189A and 189B with respect to the input information 123 specified in the argument section 187. The input information 123 set in the argument section 187 is, for example, a claim.

[0155] A candidate for the main citation is specified in the argument section 189A. As an example, the information processing device 100 specifies one of the public information 125 specified in the above-mentioned answer information 326 (see FIG. 11 ) in the argument section 189A. Preferably, the information processing device 100 specifies, among the public information 125 specified in the answer information 326, the public information 125 having the highest relevance score with the input information 123 in the argument section 189A. In this case, the information processing device 100 may specify the entire text of the public information 125 in the argument section 189A, or may specify a part of the public information 125 (for example, an abstract or a specification) in the argument section 189A.

[0156] A secondary citation candidate is specified in argument section 189A. As an example, information processing device 100 specifies one of public information 125 specified in the above-described answer information 326 (see FIG. 11 ) in argument section 189B. The document specified in argument section 189B is different from the document specified in argument section 189A. Preferably, information processing device 100 specifies, as a secondary citation, other public information 125 having a relevance score equal to or greater than a predetermined value (e.g., 4 or greater) for a constituent element having a relevance score with respect to the primary citation equal to or less than a predetermined value (e.g., 2 or less) in argument section 189B. In this case, information processing device 100 may specify the entire text of public information 125 in argument section 189B, or may specify a portion of public information 125 (e.g., a summary) in argument section 189B.

[0157] Preferably, the information processing device 100 specifies the content of the examination criterion "Part III, Chapter 2, Section 2, Inventive Step" in the argument section 188. This allows the large-scale language model 324 to understand the content of the examination criterion and determine whether or not it is possible to rationalize the combination of the public information 125 specified in the argument sections 189A and 189B.

[0158] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the above description, and is intended to include all modifications within the meaning and scope of the claims.

[0159] 10 Information processing system, 100 Information processing device, 101 Control device, 102 ROM, 103 RAM, 104 Communication interface, 105 Display interface, 106 Display, 107 Input interface, 108 Input device, 110 Bus, 120 Auxiliary storage device, 122 Analysis program, 123 Input information, 124 Patent database, 125 Public information, 125F Drawing, 126 Instruction, 127 Argument part, 128 Argument part, 129 Argument part, 130 Evaluation result, 131A Instruction, 131B Instruction, 131C Instruction, 131D Instruction, 156 Instruction, 157 Argument part, 166 Instruction, 167A Instruction, 167B Instruction, 167C Instruction, 167D Instruction, 168 Argument part, 176 Instruction, 186 Instruction statement, 187 argument section, 188 argument section, 189A argument section, 189B argument section, 200 user terminal, 201 control device, 202 ROM, 203 RAM, 204 communication interface, 205 display interface, 206 display, 207 input interface, 208 input device, 210 bus, 220 auxiliary storage device, 222 analysis program, 300 server, 324 large-scale language model, 326 response information, 327 evaluation result, 400A setting screen, 400B setting screen, 400C setting screen, 400D evaluation result screen, 410 selection field, 412 setting field, 420 setting field, 422 selection field, 424 setting field, 425 editing field, 430 selection field, 432 editing field, 434 display field, 434A display field, 434B Display column, 434D display column, 436 selection column, 438 selection column, 440 start button, 450 display column, 452 display column, 454 display column, 456 display column, IM image, NW network.

Claims

1. An analysis program, wherein the analysis program causes a computer to execute: a step of acquiring input information indicating a technical idea; a step of acquiring comparison information to be compared with the technical idea; a step of dividing the technical idea into constituent elements; a step of generating a first instruction sentence including an instruction to compare each of the plurality of divided constituent elements with the comparison information; and a step of outputting an evaluation result indicating the relevance between each of the plurality of constituent elements and the comparison information based on a result obtained from the large language model by inputting the first instruction sentence into the large language model.

2. The analysis program according to claim 1, wherein the dividing step includes: a step of generating a second instruction sentence for dividing the technical idea into constituent elements; and a step of inputting the second instruction sentence into the large language model.

3. The analysis program according to claim 1, wherein the dividing step includes a step of dividing the technical idea into constituent elements according to a predetermined rule.

4. The analysis program according to any one of claims 1 to 3, wherein the first instruction sentence includes an instruction for causing the large language model to output the description positions of the plurality of constituent elements in the comparison information.

5. The analysis program according to claim 4, wherein the comparison information includes patent documents, and the description position includes at least one of a paragraph number in the patent document, a figure number in the patent document, and a claim number in the patent document.

6. The analysis program according to claim 4, wherein the first instruction sentence includes an instruction for causing the large language model to output the reason for extracting the description position.

7. The analysis program according to any one of claims 1 to 3, wherein the first instruction sentence includes an instruction for causing the large language model to output the degree of relevance between each of the plurality of constituent elements and the comparison information.

8. The analysis program according to any one of claims 1 to 3, wherein the analysis program further causes the computer to execute: a step of generating a third instruction sentence for summarizing the comparison information from a predetermined perspective; and a step of outputting a summary of the comparison information based on a result obtained from the large language model by inputting the third instruction sentence into the large language model.

9. An information processing apparatus, comprising a control unit for controlling the information processing apparatus, wherein the control unit performs: a process of acquiring input information indicating a technical idea; a process of acquiring comparison information to be compared with the technical idea; a process of dividing the technical idea into constituent elements; a process of generating a first instruction sentence including an instruction to compare each of the plurality of divided constituent elements with the comparison information; and a process of outputting an evaluation result indicating the relevance between each of the plurality of constituent elements and the comparison information based on a result obtained from the large language model by inputting the first instruction sentence into the large language model.

10. An analysis method executed by a computer, comprising: a step of acquiring input information indicating a technical idea; a step of acquiring comparison information to be compared with the technical idea; a step of dividing the technical idea into constituent elements; a step of generating a first instruction sentence including an instruction to compare each of the plurality of divided constituent elements with the comparison information; and a step of outputting an evaluation result indicating the relevance between each of the plurality of constituent elements and the comparison information based on a result obtained from the large language model by inputting the first instruction sentence into the large language model.

Images