Verification Support Device, Verification Support Method, and Program

Abstract

Provided is a verification support device or the like capable of assisting in verifying the applicability of description requirements in patent application documents. 【Solution means】A verification support device according to an aspect of the present disclosure includes a reception means for receiving a claim and a specification in which an invention according to the claim is described as an input sentence, an instruction for extracting terms used in the claim from the input sentence, and an instruction for generating an explanatory text regarding the presence or absence of a definition of the extracted terms in the specification, an input means for inputting the instructions to a large language model, an acquisition means for acquiring a generated sentence generated based on the input instructions, and an output means for outputting, based on the generated sentence, the extracted terms and explanatory texts related to the presence or absence of definitions of the respective extracted terms.

Description

Technical Field

[0001] The present disclosure relates to a verification support device, a verification support method, and a program.

Background Art

[0002] Techniques for assisting in creating specifications and the like related to patent applications using large language models (LLMs) are disclosed. For example, Patent Document 1 discloses a system that automatically checks the description contents of a patent specification and claims before a patent application. Specifically, in Patent Document 1, the sentences of the patent specification and the claims are input, the input sentences are analyzed, and based on the analysis results, the consistency and appropriateness of the specification and the claims are evaluated.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The specification and the claims related to a patent application need to be created in accordance with the description requirements. For example, the invention according to the claim in the claims must not exceed the scope described in the detailed description of the invention in the specification. If the invention according to the claim exceeds the scope described in the detailed description of the invention, the description of the claims violates the support requirement. Also, for example, the claims must be described so that the invention for which a patent is sought can be clearly understood. If the invention for which a patent is sought is not clearly described, the description of the claims violates the clarity requirement.

[0005] Before filing a patent application, it is important to verify whether the descriptions in the specification and claims meet such description requirements.

[0006] Patent Document 1 describes a system for automatically checking the descriptions in the patent specification and claims, but there is no disclosure regarding verification of whether the description requirements are met. Also, there is no description of specific processing related to the checking.

[0007] One of the objectives of the present disclosure is made in view of the above problems, and it is to provide a verification support device and the like that can assist in verifying the applicability of the description requirements in patent application documents.

Means for Solving the Problems

[0008] A verification support device according to an aspect of the present disclosure includes a reception means for receiving a claim and a specification in which an invention related to the claim is described as an input sentence, an instruction for extracting terms used in the claim from the input sentence, and an instruction for generating an explanatory text regarding the presence or absence of definitions of the extracted terms in the specification, an input means for inputting them to a large language model, an acquisition means for acquiring a generated sentence generated based on the input instruction, and an output means for outputting, based on the generated sentence, the extracted terms and explanatory texts related to the presence or absence of definitions of each of the extracted terms.

[0009] A verification support method according to an aspect of the present disclosure receives a claim and a specification in which an invention related to the claim is described as an input sentence, inputs an instruction for extracting terms used in the claim from the input sentence and an instruction for generating an explanatory text regarding the presence or absence of definitions of the extracted terms in the specification to a large language model, acquires a generated sentence generated based on the input instruction, and outputs, based on the generated sentence, the extracted terms and explanatory texts related to the presence or absence of definitions of each of the extracted terms.

[0010] A program according to an aspect of the present disclosure causes a computer to execute a process of receiving a claim and a specification in which the invention related to the claim is described as an input sentence, an instruction to extract terms used in the claim from the input sentence, and an instruction to generate a description regarding the presence or absence of a definition of the extracted terms in the specification, inputting the instructions into a large language model, a process of obtaining a generated sentence generated based on the input instructions, and a process of outputting, based on the generated sentence, the extracted terms and descriptions related to the presence or absence of definitions of the respective extracted terms.

Advantages of the Invention

[0011] According to the present disclosure, it is possible to assist in verifying the adequacy of the description requirements in patent application documents.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Mode for Carrying Out the Invention

[0013] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0014] <First Embodiment> The outline of the verification support device according to the first embodiment will be described.

[0015] The verification support device of the present disclosure uses a large language model to support verification regarding the descriptions in the specification and claims. A large language model is a language model constructed using a large amount of dataset and technologies such as deep learning. For example, a large language model generates a sentence according to the input sentence. The large language model may be possessed by a server or the like that can communicate with the verification support device, or may be possessed by the verification support device. The large language model may have the ability to execute complex language tasks such as text generation, question answering, and summarization.

[0016] The verification support device inputs a predetermined instruction into the large language model. The verification support device also acquires the text generated by the large language model according to the predetermined instruction. Then, the verification support device outputs the information including the generated text to the terminal device. The terminal device is a device including an input / output device that is used by the user. The user is, for example, a person involved in creating a specification or the like. The verification support device may be communicably connected to the terminal device.

[0017] FIG. 1 is a first block diagram showing an example of the functional configuration of the verification support device. As shown in FIG. 1, the verification support device 100 includes a reception unit 110, an input unit 120, an acquisition unit 130, and an output unit 140.

[0018] The reception unit 110 receives an input text. The input text is information to be input into the large language model. The input text may be text data including the content of the claims and the specification in which the invention related to the claims is described. For example, the reception unit 110 receives the description of the claims and the description of the specification input in the terminal device as the input text. At this time, the specification is the specification in which the invention related to the claims is described. That is, the specification may be the specification corresponding to the claims.

[0019] In the present disclosure, the "claims" refer to the individual claims described in the claims of the patent application. The claims define the technical scope of the invention and are texts indicating the content of the invention for which protection is sought. In the present disclosure, the "specification" is a part of the patent application documents and refers to a document including a detailed description of the invention. The specification describes in detail the configuration, operation, effects, etc. of the invention so that those skilled in the art can implement the invention.

[0020] Thus, the reception unit 110 receives the claims and the specification in which the invention related to the claims is described as the input text. The reception unit 110 is an example of reception means.

[0021] The input unit 120 inputs a predetermined instruction into the large language model. Specifically, the input unit 120 inputs an instruction to extract terms used in the claims from the input sentence. Further, the input unit 120 inputs a further instruction for the terms extracted according to the instruction to extract terms. For example, the input unit 120 inputs an instruction to generate an explanatory text regarding the presence or absence of the definition of the extracted terms in the specification. Specifically, the input unit 120 may input an instruction to extract sentences in the specification related to the definition of the extracted terms, an instruction to determine whether the definition of the extracted terms is described, and the like. Note that the specific instructions are not limited to this example. Note that the "terms" may refer to specific words or phrases used in the claims and the specification. The terms may be used to express the technical features and components of the invention.

[0022] In this way, the input unit 120 inputs an instruction to extract terms used in the claims from the input sentence and an instruction to generate an explanatory text regarding the presence or absence of the definition of the extracted terms in the specification into the large language model. The input unit 120 is an example of an input means.

[0023] The acquisition unit 130 acquires information generated by the large language model according to the instruction. For example, when the instruction is input into the large language model, a new generated sentence corresponding to the instruction is generated by the large language model. The acquisition unit 130 acquires the generated sentence.

[0024] In this way, the acquisition unit 130 acquires the generated sentence generated based on the input instruction. The acquisition unit 130 is an example of an acquisition means.

[0025] The output unit 140 outputs various types of information. For example, the output unit 140 outputs a list of the extracted terms included in the generated sentence. Further, the output unit 140 may output an explanatory text related to the presence or absence of a definition for each of the extracted terms included in the generated sentence. Further, the output unit 140 may rearrange the extracted terms and output an explanatory text related to the presence or absence of a definition for each of the rearranged terms. Alternatively, the output unit 140 may output the acquired generated sentence as it is. For example, when information is output to the terminal device by the output unit 140, the information may be displayed on a display included in the terminal device.

[0026] As described above, the output unit 140 outputs the extracted terms and an explanatory text related to the presence or absence of a definition for each of the extracted terms, based on the generated sentence. The output unit 140 is an example of an output means.

[0027] Next, an example of the operation of the verification support device 100 will be described with reference to FIG. 2. In the present disclosure, each step of the flowchart is represented using a number assigned to each step, such as "S1".

[0028] FIG. 2 is a flowchart for explaining an example of the operation of the verification support device.

[0029] The reception unit 110 receives the claims and the specification in which the invention related to the claims is described as an input sentence (S1).

[0030] The input unit 120 inputs an instruction to extract the terms used in the claims from the input sentence and an instruction to generate an explanatory text regarding the presence or absence of a definition of the extracted terms in the specification, into the large language model (S2).

[0031] The acquisition unit 130 acquires the generated sentence generated based on the input instruction (S3).

[0032] The output unit 140 outputs the extracted terms and an explanatory text related to the presence or absence of a definition for each of the extracted terms, based on the generated sentence (S4).

[0033] In this way, the verification support device 100 receives the claims and the specification in which the invention related to the claims is described as input sentences. Further, the verification support device 100 inputs an instruction to extract the terms used in the claims from the input sentences and an instruction to generate an explanatory text regarding the presence or absence of definitions of the extracted terms in the specification into the large language model. Furthermore, the verification support device 100 acquires the generated sentences generated based on the input instructions. Then, the verification support device 100 outputs, based on the generated sentences, the extracted terms and the explanatory texts related to the presence or absence of definitions of each of the extracted terms.

[0034] If the definition of the term used in the claims is not described in the specification, the description of the claims in the claims may not meet the description requirements. Therefore, before filing a patent application, verification regarding the description requirements may be performed on the created specification and the like. At this time, users such as the specification creator and the inventor can conduct considerations regarding the definition of terms, such as confirming the presence or absence of the definition of the term used in the claims by using the information output by the verification support device 100. That is, the verification support device 100 can support the verification of the applicability of the description requirements in the patent application documents. Furthermore, by supporting the consideration regarding the definition of terms, it is also possible to prevent the scope of rights that is not intended by the applicant when the claims are interpreted later.

[0035] <Second Embodiment> Next, the verification support device according to the second embodiment will be described. In the second embodiment, a further example regarding the verification support device described in the first embodiment will be described. Note that the description of the content overlapping with the first embodiment will be partially omitted.

[0036] FIG. 3 is a diagram schematically showing an example of a configuration including a verification support device. As shown in FIG. 3, the verification support device 100 is communicably connected to a server device 200 and a terminal device 300 via a wired or wireless network. The server device 200 is a device having a large language model. The server device 200 manages the large language model. The terminal device 300 is a device used by a user. The user is, for example, a person involved in specification creation. The user may be an inventor, an attorney, a patent engineer, a clerical worker related to specification creation, or the like.

[0037] In the example of FIG. 3, the verification support device 100 may be a server device. The verification support device 100 may be implemented in a cloud server. On the other hand, the configuration including the verification support device 100 is not limited to this example. For example, the verification support device 100 may be included in the terminal device 300. That is, the verification support device 100 may be constructed in the terminal device 300. Also, the verification support device 100 may manage a large language model. That is, the verification support device 100 may have a large language model.

[0038] FIG. 4 is a first block diagram including an example of the functional configuration of the verification support device. The server device 200 includes a model management unit 210. The model management unit 210 stores a large language model and related programs and the like. When a prompt is input, the model management unit 210 generates a sentence indicating an answer corresponding to the prompt. Then, the model management unit 210 outputs the generated sentence. At this time, the sentence generated in the model management unit 210 is referred to as a generated sentence. Note that the large language model may be, for example, GPT-2, GPT-3, GPT-3.5, GPT-4, or GPT-4o. Also, the large language model may be Claude3, Claude3.5, T5, BERT, RoBERTa, ELECTRA, or the like. The large language model is not limited to this example.

[0039] The terminal device 300 has an input device 310 such as a keyboard and a touch panel. Further, the terminal device 300 has an output device 320 such as a display and a speaker. The terminal device 300 is, for example, a personal computer, a smartphone, a tablet terminal, etc., but is not limited to this example. The terminal device 300 only needs to have an input device 310 capable of inputting the claims and the specification, and an output device 320 capable of outputting the information output by the verification support device 100.

[0040] The verification support device 100 includes a reception unit 110, an input unit 120, an acquisition unit 130, an output unit 140, and a prompt storage unit 150.

[0041] The reception unit 110 receives an input sentence. For example, a user inputs the claims and the specification using the input device 310 of the terminal device 300. The information input in the terminal device 300 is received by the reception unit 110. At this time, the reception unit 110 receives the claims and the specification input in the terminal device 300 as an input sentence. The claims received as the input sentence may be one claim or two or more claims.

[0042] Note that the input of the claims and the specification may be performed by the user inputting text data. Also, the input of the claims and the specification may be performed by the user specifying a target data file. In this case, the user specifies a data file including the description of the claims and a data file including the description of the specification. The reception unit 110 reads the specified file. Then, the reception unit 110 may receive the claims and the specification included in the read file as an input sentence.

[0043] The input unit 120 inputs an instruction to the large language model by inputting an instruction to the model management unit 210 of the server device 200. Specifically, the input unit 120 inputs a prompt including the instruction to the model management unit 210.

[0044] At this time, the input unit 120 may input a prompt including an instruction based on a predetermined prompt. Specifically, the prompt storage unit 150 stores the prompt. The stored prompt includes a predetermined instruction in advance. Further, the stored prompt may also include an instruction regarding information to be referred to. The input unit 120 inputs a prompt based on the prompt stored in the prompt storage unit 150 to the model management unit 210. For example, assume that a first prompt including an instruction to extract terms used in the claims from the input sentence is stored. At this time, the first prompt also includes an instruction to refer to the input sentence. When the input unit 120 inputs the first prompt to the model management unit 210, the input unit 120 may input the first prompt to the model management unit 210 together with the input sentence which is the information to be referred to.

[0045] In addition, the input unit 120 may input a further prompt to the model management unit 210 by using the generated sentence generated in the model management unit 210. For example, when the first prompt is input, the model management unit 210 outputs a generated sentence generated based on the first prompt. In this case, the generated sentence includes the terms used in the claims extracted from the input sentence. Here, assume that a second prompt including an instruction to generate a description regarding the presence or absence of the definition of the extracted terms in the specification is stored in the prompt storage unit 150. The second prompt also includes an instruction to refer to the extracted terms. Therefore, when the input unit 120 inputs the second prompt to the model management unit 210, the input unit 120 may input the second prompt by designating the extracted terms which are the information to be referred to.

[0046] Note that the order of the prompts input by the input unit 120 may be determined in advance. For example, after the input sentence is received, the input unit 120 inputs the first prompt, and when the generated sentence based on the first prompt is output, the input unit 120 inputs the second prompt, and such an order may be determined. The information indicating the order of inputting the prompts may be stored in the prompt storage unit 150.

[0047] The acquisition unit 130 acquires the generated sentence output from the model management unit 210. That is, the acquisition unit 130 acquires the generated sentence generated based on the input prompt.

[0048] The output unit 140 outputs information based on the acquired generated sentence to the terminal device 300. At this time, in the output device 320 of the terminal device 300, the information based on the generated sentence is output.

[0049] Also, the output unit 140 may output an interface for prompting an input to the terminal device 300. At this time, in the output device 320 of the terminal device 300, the interface for prompting an input is output.

[0050] Next, a specific example of the operation of the verification support device 100 will be described using the sequence diagram of FIG. 5. FIG. 5 is a first sequence diagram for explaining an example of the operation of the verification support device.

[0051] First, in the terminal device 300, a claim and a specification describing the invention related to the claim are input by the user (S101). At this time, for example, an interface as shown in FIG. 6 may be output in the output device 320 of the terminal device 300. FIG. 6 is a first diagram showing an example of an input interface. In the example of FIG. 6, an input form capable of inputting a specification and a claim is shown. For example, the user inputs the text of the claim into the claim input form. Also, the user inputs the text of the specification into the specification input form. The input information is transmitted to the verification support device 100. At this time, when the "start" button in the example of FIG. 6 is selected, the information input to the verification support device 100 may be transmitted. Note that an interface such as the example of FIG. 6 may be displayed on the output device 320, for example, by the output unit 140 outputting an interface to the terminal device 300.

[0052] The reception unit 110 of the verification support device 100 receives the claim and the specification input in the terminal device 300 as input sentences (S102).

[0053] Upon receiving the input sentence, the input unit 120 inputs the first prompt to the model management unit 210 of the server device 200 (S103). Specifically, when the input sentence is received, the input unit 120 reads out the first prompt that is set as the prompt to be read from among the prompts stored in the prompt storage unit 150. The first prompt includes an instruction to extract the terms used in the claims from the input sentence. For example, the first prompt includes a sentence such as "Please extract all the terms described in the claims of the 'input sentence'." Here, the "input sentence" indicates an instruction to refer to the input sentence. Therefore, the input unit 120 inputs the input sentence, which is the information to be referred to, and the first prompt to the model management unit 210. At this time, the first prompt may include additional sentences. For example, the first prompt may further include an explanation of the terms to be extracted, such as "The terms include phrases for each part of speech, phrases grouped together including modifiers, and coined words, etc." Also, the first prompt may further include a sentence specifying an exception, such as "Please ignore the words 'the foregoing','means', 'part', and'step'."

[0054] The model management unit 210 of the server device 200 generates a first generated sentence based on the input first prompt (S104). In this case, in the model management unit 210, a first generated sentence corresponding to the instruction included in the first prompt is generated. Specifically, in the model management unit 210, the terms used in the claims are extracted from the input sentence. Then, in the model management unit 210, a first generated sentence including the extracted terms is generated. The model management unit 210 transmits the first generated sentence to the verification support device 100.

[0055] The acquisition unit 130 acquires the first generated sentence (S105). Then, in response to the acquisition of the first generated sentence, the input unit 120 inputs the second prompt to the model management unit 210 (S106). Specifically, when the first generated sentence is acquired, the input unit 120 reads out the second prompt that is set as the prompt to be read out among the prompts stored in the prompt storage unit 150. The second prompt includes an instruction to generate an explanatory text related to the presence or absence of a definition, including the text in the specification related to the definition for each extracted term included in the first generated sentence. That is to say, it can also be said that the second prompt is obtained based on the first generated sentence and a predetermined prompt. For example, the second prompt includes a sentence such as "Please determine whether the term 'extracted term' is defined in the specification of the 'input sentence'. If there is a definition, please quote the corresponding sentence." At this time, the "extracted term" indicates an instruction to refer to the extracted term included in the first generated sentence. Therefore, the input unit 120 inputs the second prompt including an instruction to refer to the first generated sentence to the model management unit 210. At this time, the second prompt may include additional text. For example, the second prompt may include a sentence specifying the details of the information to be output, such as "Please also output the paragraph number of the paragraph including the corresponding sentence." Also, the second prompt may include a sentence specifying the output form, such as "Indicate the target term at the beginning and output a dividing line at the end of the answer for each term."

[0056] The model management unit 210 generates a second generated sentence based on the input second prompt (S107). In this case, in the model management unit 210, a second generated sentence corresponding to the instruction included in the second prompt is generated. Specifically, in the model management unit 210, for each extracted term, it is determined whether the term is defined in the specification. Also, when the term is defined, the corresponding sentence is extracted. Then, in the model management unit 210, a second generated sentence including an explanatory text related to the presence or absence of a definition, including the text in the specification related to the definition of the extracted term, is generated. The model management unit 210 transmits the second generated sentence to the verification support device 100.

[0057] The acquisition unit 130 acquires the second generated sentence (S108). The output unit 140 outputs an explanatory text regarding the verification result (S109). Specifically, the output unit 140 outputs a sentence based on the second generated sentence to the terminal device 300. At this time, the output unit 140 may output the second generated sentence as it is, or may output a sentence generated using the second generated sentence. For example, assume that the second generated sentence includes an explanatory text related to the presence or absence of a definition for each term. At this time, the output unit 140 may store the extracted term and the corresponding explanatory text in an array, rearrange the order, and output them. Further, the output unit 140 may output, for each claim, the extracted term that appears in the claim and the corresponding explanatory text. In this case, the output unit 140 may also output, for example, the paragraph number of the paragraph including the sentence in the specification related to the definition of the term included in the claim for each claim. Further, the output unit 140 may emphasize and display the explanatory text regarding the term determined to have no definition among the extracted terms.

[0058] Then, the output device 320 of the terminal device 300 displays the explanatory text regarding the verification result (S110). FIG. 7 is a first diagram showing an example of the displayed explanatory text. The example of FIG. 7 is an example in which the second generated sentence is output. In the example of FIG. 7, explanatory texts of verification results related to the presence or absence of definitions are shown for terms "passerby", "face authentication", and "gate terminal". For example, the term "passerby" is determined to have a "definition" in the specification. And the sentence in the specification related to the definition is shown. In the case of this example, the paragraph number is also shown. Thus, the second generated sentence may include the paragraph number of the paragraph including the sentence in the specification related to the definition of the extracted term for each extracted term.

[0059] Also, the term "gate terminal" is determined to have "no definition" in the specification. And as an explanatory text related to the presence or absence of a definition, the sentence "The term 'gate terminal' appears multiple times, but no clear definition is found." is shown. Thus, the explanatory text related to the presence or absence of a definition may include not only the presence or absence of a definition and the text in the specification, but also supplementary explanatory text for the case where there is a definition or the case where there is no definition.

[0060] The operation example of the verification support device 100 is not limited to this example. For example, in the example of FIG. 5, information is output in the process of S110, but the timing at which the information is output is not limited to this example. For example, the output unit 140 may output a sentence based on the acquired generated sentence to the terminal device 300 each time the generated sentence is acquired.

[0061] Also, the processes of S107 and S108 may be performed for each extracted term. For example, assume that there are two extracted terms. At this time, in the process of S107, for the first term among the extracted terms, a second generated sentence including an explanatory text related to the presence or absence of a definition, which includes the text in the specification related to the definition of the extracted term, is generated. In the process of S108, the second generated sentence generated for the first term is acquired. Next, for the second term among the extracted terms, the processes of S107 and S108 are performed again. That is, the processes of S107 and S108 may be repeatedly performed as many times as the number of extracted terms. In this case, as many second generated sentences as the number of extracted terms are acquired. And in the process of S109, an explanatory text integrating the acquired second generated sentences may be output.

[0062] Also, in the example of FIG. 5, the next prompt is input in response to the acquisition of the generated sentence. This is not limited to this example. For example, when an input sentence is received, the input unit 120 may input a prompt including both the instruction in the first prompt and the instruction in the second prompt.

[0063] Next, a further example of the operation of the verification support apparatus 100 will be described using the sequence diagram of FIG. 8. This operation example is an operation example of the verification support apparatus 100 when there is no definition description for the extracted terms. This operation example is performed, for example, after the operation of FIG. 5.

[0064] FIG. 8 is a second sequence diagram for explaining an example of the operation of the verification support apparatus. The input unit 120 inputs the third prompt to the model management unit 210 of the server apparatus 200 (S201). Specifically, when it is determined whether there is a definition for the extracted terms, the input unit 120 reads out the third prompt that is set as the prompt to be read out among the prompts stored in the prompt storage unit 150. The third prompt includes an instruction for specifying terms that are not defined in the specification among the extracted terms. That is, it can be said that the third prompt is obtained based on the second generated sentence and a predetermined prompt. For example, the third prompt includes a sentence such as "Please specify the terms determined in the 'second generated sentence' as not being defined in the specification among the 'extracted terms'." At this time, "the'second generated sentence'" indicates an instruction to refer to the second generated sentence. Therefore, the input unit 120 inputs the third prompt including an instruction to refer to the second generated sentence to the model management unit 210. At this time, the third prompt may include additional sentences. For example, the third prompt may include a sentence specifying details of an instruction such as "Regarding the terms, ignore the differences between 'part' and'means' and consider them the same terms."

[0065] Note that in the process of S108, when the second generated sentences are acquired for the number of extracted terms, in the process of S201, an instruction to refer to the explanatory text obtained by integrating the acquired second generated sentences, or an array in which each of the second generated sentences is stored, may be indicated.

[0066] The model management unit 210 of the server device 200 generates a third generated sentence based on the input third prompt (S202). In this case, in the model management unit 210, a third generated sentence corresponding to the instruction included in the third prompt is generated. Specifically, in the model management unit 210, a term determined not to be defined in the specification is identified. Then, in the model management unit 210, a third generated sentence including the term determined not to be defined in the specification is generated. The model management unit 210 transmits the third generated sentence to the verification support device 100.

[0067] The acquisition unit 130 acquires the third generated sentence (S203). Further, in response to the acquisition of the third generated sentence, the input unit 120 inputs a fourth prompt to the model management unit 210 (S204). Specifically, when the third generated sentence is acquired, the input unit 120 reads out the fourth prompt set as the prompt to be read out among the prompts stored in the prompt storage unit 150. The fourth prompt includes an instruction to generate a sentence explaining the term for each identified term. That is, it can also be said that the fourth prompt is obtained based on the third generated sentence and a predetermined prompt. The sentence explaining the term can be a sentence that can serve to explain the definition of the term. That is, it can also be said that the fourth prompt includes an instruction to generate candidates for sentences that can serve to explain the definition of the term to the large language model. Hereinafter, the sentence explaining the term is also referred to as a candidate sentence.

[0068] For example, the fourth prompt includes a sentence such as "In the specification of the 'input sentence', please generate a paragraph explaining the definition of the'specified term'." Here, the "input sentence" indicates an instruction to refer to the input sentence, and the'specified term' indicates an instruction to refer to the specified term. Therefore, the input unit 120 inputs a fourth prompt including an instruction to refer to the specification of the input sentence and an instruction to refer to the specified term into the model management unit 210. Here, the input unit 120 may input an instruction to repeat the process. For example, the input unit 120 may issue an instruction to repeat the process until paragraphs explaining the definitions are generated for all of the specified terms. Also, the fourth prompt may include additional sentences. For example, the fourth prompt may include an example sentence indicating a predetermined term and a paragraph explaining the definition of the predetermined term, and may include an instruction to generate a paragraph with reference to the example sentence.

[0069] The model management unit 210 generates a fourth generated sentence based on the input fourth prompt (S205). In this case, in the model management unit 210, a fourth generated sentence corresponding to the instruction included in the fourth prompt is generated. Specifically, in the model management unit 210, for each of the specified terms, a paragraph including candidate sentences is generated, and then a fourth generated sentence including the candidate sentences is generated in the model management unit 210. The model management unit 210 transmits the fourth generated sentence to the verification support device 100.

[0070] The acquisition unit 130 acquires the fourth generated sentence (S206). Then, in response to the acquisition of the fourth generated sentence, the input unit 120 inputs the fifth prompt to the model management unit 210 (S207). Specifically, when the fourth generated sentence is acquired, the input unit 120 reads out the fifth prompt that is set as the prompt to be read out among the prompts stored in the prompt storage unit 150. The fifth prompt includes an instruction to identify a sentence similar to the candidate sentence from the specification, and an instruction to indicate the location in the specification where the candidate sentence should be inserted based on the identified sentence. That is to say, it can also be said that the fifth prompt is obtained based on the fourth generated sentence and a predetermined prompt. For example, the fifth prompt includes a sentence such as "Please consider where to insert the sentence explaining the term in the specification of the 'input sentence'. First, identify a sentence similar to the sentence explaining the term. Then, based on the identified sentence, indicate the location in the specification where it should be inserted." At this time, the "sentence explaining the term" indicates an instruction to refer to the candidate sentence included in the fourth generated sentence. Therefore, the input unit 120 inputs the fifth prompt including an instruction to refer to the specification of the input sentence and an instruction to refer to the sentence explaining the term to the model management unit 210. Here, the input unit 120 may input an instruction to repeat the process. For example, the input unit 120 may give an instruction to repeat the process until the location where the sentence explaining the term is to be inserted is considered for all the identified terms. Also, the fifth prompt may include additional sentences. For example, the fifth prompt may include an instruction indicating the paragraph number of the insertion location. Also, the fifth prompt may include a sentence specifying the output form.

[0071] The model management unit 210 generates a fifth generated sentence based on the input fifth prompt (S208). In this case, in the model management unit 210, a fifth generated sentence corresponding to the instruction included in the fifth prompt is generated. Specifically, in the model management unit 210, for each specified term, a sentence similar to the candidate sentence is specified from the specification, and based on the specified sentence, the location in the specification where the candidate sentence should be inserted is considered. Then, in the model management unit 210, a fifth generated sentence including the location in the specification where the candidate sentence should be inserted is generated. The model management unit 210 transmits the fifth generated sentence to the verification support device 100. Note that the location in the specification where the candidate sentence should be inserted may be before or after the specified sentence, or may be the end of the chapter including the specified sentence, and is not limited thereto.

[0072] The acquisition unit 130 acquires the fifth generated sentence (S209). The output unit 140 outputs an explanatory text regarding the candidate sentence (S210). Specifically, the output unit 140 outputs a sentence based on the fifth generated sentence to the terminal device 300. At this time, the output unit 140 may output the fifth generated sentence as it is, or may output a sentence generated using the fifth generated sentence.

[0073] Then, the output device 320 of the terminal device 300 displays the explanatory text regarding the candidate sentence (S211). FIG. 9 is a second diagram showing an example of the displayed explanatory text. The example of FIG. 9 is an example in which the fifth generated sentence is output. In the example of FIG. 9, for the term "gate terminal" determined not to be defined in the specification, the candidate sentence and the location where the candidate sentence should be inserted are shown. Also, the location where the candidate sentence should be inserted is explicitly indicated using a paragraph number. Furthermore, a sentence explaining the reason for the location where the candidate sentence should be inserted is also shown.

[0074] The operation example of the verification support device 100 is not limited to this example. For example, in the example of FIG. 8, information is output in the process of S110, but the timing at which the information is output is not limited to this example. For example, the output unit 140 may output a sentence based on the acquired generated sentence to the terminal device 300 each time a generated sentence is acquired.

[0075] Also, the processes of S205 and S206 may be performed for each specified term. For example, assume that there are two specified terms. At this time, in the process of S205, for the first term among the specified terms, a fourth generated sentence including a candidate sentence is generated. In the process of S206, the fourth generated sentence generated for the first term is acquired. Next, for the second term among the specified terms, the processes of S205 and S206 are performed again. That is, the processes of S205 and S206 may be repeatedly performed the number of times equal to the number of specified terms. In this case, the fourth generated sentences are acquired the number of times equal to the number of specified terms. Then, in the process of S207, an explanatory text obtained by integrating the acquired fourth generated sentences, or a fifth prompt including an instruction to refer to an array in which each of the fourth generated sentences is stored may be input to the model management unit 210.

[0076] Similarly, the processes of S208 and S209 may be performed for each candidate sentence. For example, assume that there are two candidate sentences. At this time, in the process of S208, for the first candidate sentence among the candidate sentences, a fifth generated sentence including a location in the specification where the candidate sentence should be inserted is generated. In the process of S209, the fifth generated sentence generated for the first candidate sentence is acquired. Next, for the second candidate sentence among the candidate sentences, the processes of S208 and S209 are performed again. That is, the processes of S208 and S209 may be repeatedly performed the number of times equal to the number of candidate sentences. In this case, the fifth generated sentences are acquired the number of times equal to the number of candidate sentences. Then, in the process of S210, an explanatory text obtained by integrating the acquired fifth generated sentences may be output.

[0077] Also, in the example of FIG. 8, according to the acquisition of the generated sentence, the next prompt is input. Not limited to this example, for example, in the process of S201, the input unit 120 may input a prompt including the instruction in the third prompt, the instruction in the fourth prompt, and the instruction in the fifth prompt.

[0078] As described above, the verification support device 100 of the second embodiment receives the claims and the specification in which the invention related to the claims is described as input sentences. Further, the verification support device 100 inputs an instruction to extract terms used in the claims from the input sentences and an instruction to generate an explanatory text regarding the presence or absence of definitions of the extracted terms in the specification into the large language model. Furthermore, the verification support device 100 acquires the generated sentences generated based on the input instructions. Then, the verification support device 100 outputs, based on the generated sentences, the extracted terms and the explanatory texts related to the presence or absence of definitions of the respective extracted terms.

[0079] If the definition of the term used in the claims is not described in the specification, the description of the claims including the claims may not satisfy the description requirements. Therefore, before filing a patent application, verification regarding the description requirements may be performed on the created specification and the like. At this time, a user such as the specification creator can perform considerations regarding the definition of the term, such as confirming the presence or absence of the definition of the term used in the claims by using the information output by the verification support device 100. Therefore, since the verification work by hand is reduced in the verification support device 100, efficient verification of the description requirements becomes possible. That is, the verification support device 100 can support the verification of the applicability of the description requirements in the patent application documents.

[0080] In addition, the verification of the applicability of the description requirements is highly specialized. Therefore, it may be difficult to verify unless the operator is skilled in patent applications. On the other hand, since the verification support device 100 can output information related to the applicability of the description requirements, it can encourage even inexperienced operators to efficiently verify the applicability of the description requirements.

[0081] Furthermore, the verification support device 100 may perform the following processing. That is, the verification support device 100 inputs an input sentence and a first prompt including an instruction to extract terms used in the claims from the input sentence. Further, the verification support device 100 obtains a first generated sentence generated based on the first prompt, and inputs a second prompt including an instruction to generate an explanatory text related to the presence or absence of a definition, including the sentences in the specification related to the definition of each of the extracted terms included in the first generated sentence. The verification support device 100 obtains a second generated sentence generated based on the second prompt. Then, based on the first generated sentence and the second generated sentence, the verification support device 100 outputs, for each extracted term, the sentence in the specification related to the definition of the term and the explanatory text related to the verification result of the presence or absence of the definition.

[0082] Thereby, the verification support device 100 can enable the user to grasp the terms and information regarding the presence or absence of their definitions, so that it can effectively verify the description requirements such as the support requirement and the clarity requirement. In addition, since the verification support device 100 can specify the sentences in the relevant specification for each term used in the claims, it becomes possible to enable the user to perform a comprehensive verification of the description requirements. Therefore, the verification support device 100 can contribute to reducing the risk of oversight and improving the accuracy of verification in the verification of the description requirements. Furthermore, the verification support device 100 can enable the user to quickly grasp the contradictions and deficiencies in the claims and the specification related to the patent application, and can improve the quality of the documents.

[0083] In addition, the verification support device 100 may input an instruction to identify terms not defined in the specification among the extracted terms, and an instruction to generate a candidate sentence, which is a sentence explaining the term, for each identified term, and obtain a generated sentence including the candidate sentences for each identified term. Thereby, since the verification support device 100 can appropriately complement the explanation of the terms, it can contribute to the creation of a specification that satisfies requirements such as the support requirement and the clarity requirement. That is, the verification support device 100 can contribute to improving the quality of the specification.

[0084] Further, the verification support device 100 may input an instruction to specify a sentence similar to the candidate sentence from the specification and an instruction to specify a location in the specification where the candidate sentence should be inserted based on the specified sentence, and obtain a generated sentence including the location where the candidate sentence should be inserted. Thereby, since the verification support device 100 can indicate an appropriate insertion location of the candidate sentence, it is possible to support the complement of the description of the term while maintaining the structure of the specification.

[0085] As described above, the verification support device 100 can structure and output a list of terms, the presence or absence of term definitions, candidate sentences, etc. Therefore, the verification support device 100 can efficiently perform the verification of the description requirements. The user can easily refer to the necessary information, and the productivity of the verification work is improved.

[0086] Note that in a large language model, there may be an upper limit on the amount of text that can be processed at one time in units such as the number of tokens. When a plurality of prompts such as the first prompt to the fifth prompt are used step by step, it is expected to execute the process up to the upper limit of the amount of text for each pair of input and output based on each prompt. Thereby, even when the amount of processing is large as in the present disclosure, an effect of easily preventing an error due to the upper limit of the amount of text can be obtained. Further, even when accuracy is required for the process as in the present disclosure, a decrease in accuracy can be prevented.

[0087] [Modification Example 1] The verification support device 100 may input an instruction to explain the interpretation of a claim including the term to a large language model using a sentence related to the definition of the term.

[0088] Specifically, it is assumed that the second prompt instructs that the second generated sentence includes the extracted terms, the sentences related to the definitions of the extracted terms, and the paragraph numbers of the paragraphs containing such sentences. In this case, the input unit 120 inputs a sixth prompt including an instruction for explaining the interpretation of the components of the claims to the model management unit 210 based on the second generated sentence. That is to say, it can also be said that the sixth prompt is obtained based on the second generated sentence and a predetermined prompt. For example, the sixth prompt includes a sentence such as "Please explain how each component of the claims should be interpreted using the claims and the specification of the 'input sentence' and the paragraphs included in the'second generated sentence'. In this regard, please consider the definitions of the extracted terms."

[0089] The acquisition unit 130 acquires the sixth generated sentence generated based on the sixth prompt. The sixth generated sentence includes sentences explaining the interpretation of each component of the claims. Then, the output unit 140 outputs a sentence explaining the interpretation of the components of the claims based on the sixth generated sentence.

[0090] Thus, the second generated sentence includes the paragraph numbers of the paragraphs in the specification containing the sentences related to the definitions of the extracted terms for each of the extracted terms. At this time, the verification support device 100 may input a prompt including an instruction for generating a sentence explaining the interpretation of the components of the claims based on the specification and the second generated sentence. Then, the verification support device 100 acquires the generated sentence generated based on the prompt, and may output a sentence explaining the interpretation of the components of the claims based on the generated sentence generated based on the prompt.

[0091] Thereby, the verification support device 100 can show the user the interpretation of the components of the claims that can be considered from the current description of the specification.

[0092] Note that the interpretation of claims is not limited to the interpretation of components as described above. For example, the interpretation of claims may include the interpretation of individual terms and the relationship between components. For the interpretation of claims, reference information is referred to. The scope of the reference information may be the entire specification, a specific section of the specification such as "Detailed Description of the Invention", or a combination of the specification and the drawings.

[0093] Also, the interpretation may be performed in consideration of general knowledge in the technical field, based only on the description in the specification, or in consideration of relevant prior art documents. As a form of interpretation, it may be performed in natural language, as structured data such as JSON (JavaScript Object Notation), or as a diagrammatic representation such as a UML (Unified Modeling Language) diagram.

[0094] [Modification Example 2] In the above example, an example in which the verification support device 100 performs verification related to the presence or absence of a term definition has been described. Not limited to this example, the verification support device 100 can perform various verifications.

[0095] For example, the verification support device 100 may perform verification as to whether the terms used in the claims are used in another expression in the specification.

[0096] In this case, first, in the terminal device 300, the claims and the specification in which the invention related to the claims is described are input by the user. The reception unit 110 receives the claims and the specification input in the terminal device 300 as input sentences. The input unit 120 inputs a first prompt in response to the input sentence being received. The first prompt includes an instruction to extract the terms used in the claims from the input sentence. The acquisition unit 130 acquires a first generated sentence based on the first prompt.

[0097] Then, in response to the acquisition of the first generated sentence, the input unit 120 inputs the seventh prompt to the model management unit 210. The seventh prompt includes an instruction to verify the presence or absence of similar terms, which are terms with similar meanings in the specification, for the extracted terms. For example, the seventh prompt includes a sentence such as "Please check whether there are similar terms, which are terms with similar meanings, in the specification of the 'input sentence' for the 'extracted term'." At this time, the '"extracted term"' indicates an instruction to refer to the extracted terms included in the first generated sentence. The '"input sentence"' indicates an instruction to refer to the input sentence. Therefore, the input unit 120 inputs the seventh prompt including an instruction to refer to the first generated sentence and an instruction to refer to the specification of the input sentence to the model management unit 210.

[0098] At this time, the seventh prompt may include additional sentences. For example, the seventh prompt may include an instruction to check the usage of similar terms in the specification and verify whether they should be unified into the terms of the claims. Also, the seventh prompt may include an instruction to specify the output format. An example of an instruction to specify the output format is an instruction such as "Please output from the perspective of whether each extracted term is a term that appears only in the claims, whether there are similar terms, and if there are similar terms, what they are, where they appear, and whether they should be unified into the terms of the claims."

[0099] The model management unit 210 generates a seventh generated sentence based on the input seventh prompt. In this case, in the model management unit 210, the presence or absence of similar terms in the specification is verified for the extracted terms. Then, in the model management unit 210, a seventh generated sentence including an explanatory text regarding the presence or absence of similar terms in the specification for the extracted terms is generated. The model management unit 210 transmits the seventh generated sentence to the verification support device 100.

[0100] The acquisition unit 130 acquires the seventh generated sentence. Then, the output unit 140 outputs the sentence based on the seventh generated sentence to the terminal device 300. The output device 320 of the terminal device 300 displays an explanatory text regarding the presence or absence of similar terms of the extracted terms based on the seventh generated sentence. FIG. 10 is a third diagram showing an example of the displayed explanatory text. The example of FIG. 10 is an example in which the seventh generated sentence is output. In the example of FIG. 10, explanatory texts regarding the presence or absence of similar terms are shown for the terms "passerby" and "face data".

[0101] For example, the term "passerby" is not a term that appears only in the claims, but is a term for which there are similar terms. And as similar terms of "passerby", the terms "visitor" and "member" are shown. Further, it is also shown that the similar terms should not be unified with "passerby" used in the claims. In this way, the output unit 140 may output an explanatory text regarding the presence or absence of similar terms for each of the extracted terms.

[0102] Note that the above-mentioned similar terms refer to other expressions used in the specification that have meanings and functions similar or equivalent to the terms described in the claims. These similar terms may be used to assist in understanding the invention, but since they may cause confusion in the interpretation of the claims, it is preferable to use them appropriately. For example, "transmission unit" and "transmission means", "control device" and "controller", etc. are cited as examples of similar terms.

[0103] Note that when receiving the input of the input sentence, the verification support device 100 may output an interface capable of function selection. FIG. 11 is a second diagram showing an example of the input interface. In the example of FIG. 11, an input form capable of inputting the specification and the claims is shown. Further, in this example, an input form capable of function selection is also shown. For example, by a pull-down, a function of verifying the presence or absence of the definition of a term and a function of verifying an alternative expression of a term may be selectable. The user makes a selection of the function displayed by the pull-down, for example, together with the input of the input sentence. The verification support device 100 may change the processing to be performed according to the selected function.

[0104] This modification example has been described by taking as an example the verification of whether the terms used in the claims are used in a different expression in the specification. However, the examples of verification are not limited to this. For example, the context of the specification may be analyzed, and verification of contradictions in the sentences in the specification may be performed.

[0105] [Modification Example 3] In the above example, an example in which the prompts are stored in advance in the prompt storage unit 150 of the verification support device 100 has been described. Not limited to this example, the prompts may be stored in the terminal device 300. For example, the verification support device 100 may perform an input to the model management unit 210 of the prompt in response to the transmission of the prompt from the terminal device 300.

[0106] [Modification Example 4] The verification support device 100 may input an instruction to output the probability of satisfying the description requirements to the model management unit 210 for each claim. That is, the verification support device 100 may input a prompt including an instruction to calculate the probability of satisfying the description requirements for each claim to the model management unit 210 based on the verification result of the presence or absence of the definition of the extracted terms. Then, the verification support device 100 outputs a sentence based on the generated sentence generated in response to the prompt to the terminal device 300. At this time, the generated sentence includes the probability of satisfying the description requirements of each claim. Note that the prompt may include, for example, an instruction to calculate the probability of satisfying the description requirements based on the second generated sentence for each claim. At this time, the prompt may be input to the model management unit 210 together with the second generated sentence. Alternatively, the prompt may include, for example, an instruction to calculate the probability of satisfying the description requirements based on the seventh generated sentence for each claim. At this time, the prompt may be input to the model management unit 210 together with the seventh generated sentence. The examples of the prompt are not limited to this. Note that the probability may be displayed in multiple levels such as "○", "△", "×", etc., or may be indicated by numerical values such as "0%" to "100%". Other quantitative numerical values may also be used.

[0107] As a result, the verification support device 100 can provide the user with materials for determining whether the current claims meet the description requirements.

[0108] [Modification Example 5] The verification support device 100 may have a large language model. FIG. 12 is a second block diagram including an example of the functional configuration of the verification support device. As in this example, the verification support device 100 may have a model management unit 210. In this case, the verification support device 100 may perform operations similar to the above-described operation examples. [Modification Example 6] In the above-described embodiment, an example of extracting terms by inputting a prompt including an instruction to extract terms used in the claims into a large language model has been described. The term extraction method is not limited to this example. For example, term extraction may be performed by a statistics-based, graph-based, or machine learning-based method. Also, term extraction may be performed by a method of extracting sentences included in a dataset using a dataset including specific sentences. [Modification Example 7]

[0109] The input of the prompt from the input unit 120 to the model management unit 210 is realized, for example, by data transmission through an API (Application Programming Interface).

[0110] Not limited to this example, the input of the prompt may be realized by input through a database in which the prompt is recorded. Also, when the large language model operates in a local environment (when the model management unit 210 is provided in the verification support device 100), the input of the prompt may be performed as an input to the large language model constructed in the local environment. As a result, even if the network is offline, the execution of the process becomes possible. Also, the processing speed can be improved.

[0111] [Modification Example 8] In the above example, the output of the sentence based on the fifth generated sentence to the terminal device 300 was described. At this time, a specification in which the candidate sentence is reflected may be output. That is, the verification support device 100 may output a specification in which the candidate sentence is inserted at the location in the specification where the candidate sentence should be inserted.

[0112] For example, after the fifth generated sentence is acquired in the process of S209, the input unit 120 inputs the eighth prompt to the model management unit 210 in response to the acquisition of the fifth generated sentence. Specifically, when the fifth generated sentence is acquired, the input unit 120 reads out the eighth prompt that is set as the prompt to be read out among the prompts stored in the prompt storage unit 150. The eighth prompt includes an instruction to generate a specification in which the candidate sentence is inserted. For example, the eighth prompt includes a sentence such as "Please insert the 'candidate sentence' into the 'location in the specification where the candidate sentence should be inserted'. Then, please output the inserted specification." At this time, the "candidate sentence" indicates an instruction to refer to the candidate sentence included in the fourth generated sentence or the fifth generated sentence. Also, the "location in the specification where the candidate sentence should be inserted" indicates an instruction to refer to the location in the specification where the candidate sentence should be inserted, which is included in the fifth generated sentence. Therefore, the input unit 120 inputs the eighth prompt including the instruction to refer to the specification of the input sentence, the instruction to refer to the candidate sentence, and the instruction to refer to the location in the specification where the candidate sentence should be inserted to the model management unit 210.

[0113] The model management unit 210 generates an eighth generated sentence based on the input eighth prompt. In this case, in the model management unit 210, an eighth generated sentence corresponding to the instruction included in the eighth prompt is generated. Specifically, in the model management unit 210, a specification in which the candidate sentence is inserted is generated at the location in the specification where the candidate sentence should be inserted. Then, the model management unit 210 transmits the eighth generated sentence to the verification support device 100.

[0114] The acquisition unit 130 acquires the eighth generated sentence. The output unit 140 outputs a description including the generated specification. Specifically, the output unit 140 outputs a sentence based on the eighth generated sentence to the terminal device 300. At this time, the output unit 140 may output the eighth generated sentence as it is, or may output a sentence generated using the eighth generated sentence.

[0115] In this way, the verification support device 100 may output a specification in which the candidate sentence is reflected. Thereby, the verification support device 100 can provide the user with a specification in which the candidate sentence is inserted, so that the manual verification work of the description requirements is reduced, and efficient specification creation and verification of the description requirements are possible.

[0116] [Modification Example 9] In the present disclosure, the claims and the specification that can be received as the input sentence may be created manually, or may be generated by a large language model.

[0117] For example, as an example of supporting the verification of the applicability of the description requirements in the patent application documents in the verification support device, the verification of the quality of the patent application documents generated by the large language model may be supported. In other words, for the purpose of supporting the verification of the quality of the patent application documents generated by the large language model, the verification support device in the present disclosure may be used.

[0118] First, claims and a specification are generated in the large language model. The claims and the specification generated at this time are natural sentences. For example, the terminal device 300 or another device communicable with the terminal device 300 inputs a prompt for instructing the generation of the claims and the specification to the model management unit 210. In response to the input, the large language model of the model management unit 210 generates the claims and the specification. The model management unit 210 outputs the generated claims and specification to the terminal device 300 or another device communicable with the terminal device 300. The above is an example, and the example in which the claims and the specification are generated in the large language model is not limited to this.

[0119] The claims and specifications generated by the large language model are stored, for example, in the terminal device 300 or another device capable of communicating with the terminal device 300. Then, the user uses the input device 310 of the terminal device 300 to input the claims and specifications generated by the large language model. At this time, the input of the claims and specifications may be performed by the user inputting text data. Also, the input of the claims and specifications may be performed by the user designating the target data file. The reception unit 110 receives the claims and specifications input in the terminal device 300 as input sentences.

[0120] Note that the claims and specifications may be generated by the model management unit 210, or may be generated by a large language model different from the model management unit 210.

[0121] When the claims and specifications are generated by the model management unit 210, the generated claims and specifications may be stored in the server device 200 or the verification support device 100. In this case, the reception unit 110 may read out the claims and specifications stored in the server device 200 or the verification support device 100, and receive the read claims and specifications as input sentences.

[0122] In this way, the verification support device 100 may receive the claims and specifications generated by the large language model as input sentences. Thereby, the verification support device 100 can contribute to improving the quality of the patent application documents created by the large language model. Also, the verification support device 100 can assist users and the like involved from the stage of creating patent application documents by the large language model to easily confirm the quality of the patent application documents, or give suggestions for quality improvement.

[0123] Note that, among the claims and specifications received as input sentences, either one may be a natural sentence generated by the large language model. Even in this case, the same effects as described above are achieved.

[0124] Note that instead of the user's input, the terminal device 300 or another device capable of communicating with the terminal device 300 may execute processing at a predetermined or arbitrary timing, and the claims and specification generated by the large language model may be input to the reception unit 110 of the verification support device 101. The predetermined timing may be the time when the claims and specification generated by the large language model are stored in the terminal device 300 or another device capable of communicating with the terminal device 300, the time when the claims and specification are generated by the large language model, the time when the terminal device 300 or another device capable of communicating with the terminal device 300 receives the generated claims and specification from the large language model, or a time determined based on these times. In this case, the same effects as described above are achieved. Furthermore, the generation and verification of patent application documents can be integrated, and the user can be assisted more efficiently.

[0125] [Modification Example 10] In the present disclosure, the sentences included in a plurality of prompts may be described in one prompt. For example, assume that the third prompt includes a sentence such as "Please identify the terms among the 'extracted terms' that are determined not to be defined in the specification in the 'second generated sentence'.", and the fourth prompt includes a sentence such as "Please generate a paragraph explaining the definition of the'specified terms' in the specification of the 'input sentence'." In this case, the third prompt may be described as "Please identify the terms among the 'extracted terms' that are determined not to be defined in the specification in the'second generated sentence'. Then, please generate a paragraph explaining the definition of the specified terms in the specification." and the third prompt may include the sentence of the fourth prompt. In this case, the third prompt serves the role of the fourth prompt. As a result, since a generated sentence including a plurality of processing results can be obtained by inputting a single prompt, the processing flow can be simplified. Further, when there is an upper limit on the amount of text that can be processed at one time in a large language model, the time required to obtain a generated sentence can be shortened by inputting a prompt including a plurality of instructions at one time. Note that although examples of the third prompt and the fourth prompt are described above, the example of describing the sentences included in a plurality of prompts in one prompt is not limited to this. For example, it may be the first prompt and the second prompt, or the second prompt and the third prompt.

[0126] <The Third Embodiment> Next, the verification support device according to the third embodiment will be described. In the third embodiment, an example of performing verification regarding description requirements using a model that has learned predetermined information will be described. Note that the description of the content overlapping with the first and second embodiments will be partially omitted.

[0127] Figure 13 is a second block diagram including an example of the functional configuration of the verification support device. As shown in Figure 13, the verification support device 101 is communicably connected to the server device 201 and the terminal device 300 via a wired or wireless network. The server device 201 is a device having a learned model. The server device 201 manages the learned model. The learned model may be a large language model or a model generated by various other methods.

[0128] The server device 201 includes a model management unit 211. A learned model such as a large language model and related programs are stored therein. The model management unit 211 returns a predetermined output for the input information.

[0129] The verification support device 101 includes a reception unit 110, an input unit 121, an acquisition unit 131, an output unit 141, a prompt storage unit 150, and a model generation unit 160.

[0130] The model generation unit 160 generates a learned model. Specifically, the model generation unit 160 generates a learned model by training using the information of filed patent application documents and the examination data corresponding to the patent application documents as training data. The filed patent application documents include the filed predetermined claims and the predetermined specification in which the invention related to the predetermined claims is described. Further, the examination data indicates the examination result including the points regarding the description requirements for the predetermined claims. More specifically, the examination data indicates the explanation that the predetermined claims do not satisfy the description requirements based on the description of the predetermined specification. For example, the examination data includes the explanation that the predetermined claims do not satisfy the support requirement and the explanation that the support requirement is not satisfied based on the description of the predetermined specification.

[0131] For example, when the examination in each patent office of each country is carried out multiple times, for each examination in each country, the examination data includes the examination result including the points regarding the description requirements for the claims to be examined.

[0132] That is, the model generation unit 160 generates a learned model by learning the relationship between the examination data indicating the examination result including the points pointed out regarding the description requirements for a given claimed claim, the given claimed claim, and the given specification in which the invention related to the given claimed claim is described. The learned model generated by such learning may be a model that outputs information regarding the presence or absence of points for modification for the claimed claim and the specification to satisfy the description requirements by inputting the claimed claim and the corresponding specification. Hereinafter, the information regarding the presence or absence of points for modification is also referred to as modification point information.

[0133] For example, the model generation unit 160 may generate a learned model by learning the relationship between the examination data indicating the examination result including the points pointed out regarding the description requirements for a given examined claimed claim, the given claimed claim, and the given specification in which the invention related to the given claimed claim is described.

[0134] The modification point information may be information indicating whether modification is necessary to satisfy the description requirements, or may be information indicating an amendment for satisfying the description requirements. For example, the learned model may be a model that outputs an amendment to the description in the specification so that the input claimed claim and specification satisfy the support requirements. The learning method may be various known methods such as, for example, neural networks, support vector machines, decision trees, and random forests. Alternatively, the learning may be performed as fine-tuning of a large language model.

[0135] Note that the examination data may be data of examination documents proposed by the Patent Office such as a notice of reasons for refusal and a refusal decision. Further, the examination data may be opinions of a patent attorney, a patent engineer, etc. regarding the filed patent application documents. Further, the filed patent application documents and the examination data may be acquired from the terminal device 300 or may be acquired from a database (not shown). Further, the examination data may be information regarding a product considered to be practicing the invention described in the claims of the patent application documents. Specifically, it includes the name of the product, model number, specification, instruction manual, product catalog, images showing the appearance and internal structure of the product, operation explanation videos, etc. Further, it also includes materials explaining how the product implements the patented invention and materials showing the correspondence between the patent claims and the product. Further, considering standard essential patents, the examination data may be information regarding the standard specifications of the relevant technical field.

[0136] The generated learned model exists in the model management unit 211. Note that, as described in the first and second embodiments, the model management unit 211 may be included in the verification support device 101. Further, the process of generating the model may be performed by another device communicably connected to the verification support device 101.

[0137] The input unit 121 inputs the input sentence received by the reception unit 110 to the model management unit 211. That is, the input unit 121 inputs the input sentence to a learned model that has learned the relationship between examination data indicating an examination result including a pointing-out regarding the description requirements for a given claim filed, the given claim, and a given specification in which the invention related to the given claim is described.

[0138] When the learned model is a large language model, the input unit 121 inputs, for example, a prompt including the input sentence, an instruction to verify whether the claim and the specification in the input sentence satisfy the description requirements, and an instruction to output an amendment to the specification when the description requirements are not satisfied.

[0139] The acquisition unit 131 acquires the correction point information. Specifically, due to the input by the input unit 121, the correction point information is output by the model management unit 211. The acquisition unit 131 acquires the correction point information from the model management unit 211. That is, the acquisition unit 131 acquires the correction point information including the presence or absence of correction points for meeting the description requirements for the specification in the input sentence output by the learned model.

[0140] The output unit 141 outputs the correction point information. Specifically, the output unit 141 outputs the correction point information to the terminal device 300. The output device 320 of the terminal device 300 displays the correction point information. For example, the output unit 141 outputs the correction point information including whether the claims in the current input sentence meet the support requirements, and if the support requirements are not met, in which paragraphs of the specification what corrections should be made. The correction point information may refer to information specifically indicating the correction points for the specification and claims in the patent application. For example, the correction point information may be information including the presence or absence of correction points for meeting the description requirements for the specification received as the input sentence output by the learned model. The correction point information may include, in addition to the judgment result of whether correction is necessary, proposals for specific correction locations and correction contents if correction is necessary.

[0141] Note that the input unit 121 may further input the specification in which corrections based on the output correction point information have been made to the model management unit 211. That is, the input unit 121 may input the corrected specification in which the corrections based on the correction point information are reflected in the specification in the input sentence and the claims in the input sentence to the learned model. Note that the corrected specification may refer to the specification in which the corrections based on the correction point information output by the learned model are reflected with respect to the original specification received as the input sentence. The corrected specification may be the specification in which the changes or additions necessary to meet the description requirements and the like are reflected. The corrected specification may be input to the learned model again for further verification.

[0142] At this time, the corrected specification may be generated by the input unit 121 based on the correction point information. Alternatively, the corrected specification may be input by the user. In this case, the corrected specification is transmitted from the terminal device 300 to the verification support device 101.

[0143] The model management unit 211 outputs correction point information for the corrected specification. The acquisition unit 131 acquires the correction point information for the corrected specification. Then, the output unit 141 may output the correction point information for the corrected specification. At this time, the output unit 141 may output the correction point information for the specification before correction and the correction point information for the corrected specification.

[0144] Next, the operation of the verification support device 101 will be described using the sequence diagram of FIG. 14. FIG. 14 is a third sequence diagram for explaining an example of the operation of the verification support device. In this operation example, an example of verifying the description requirements when a pre-generated learned model exists will be described.

[0145] First, at the terminal device 300, the user inputs a claim and the specification in which the invention related to the claim is described (S301). The reception unit 110 receives the claim and the specification input at the terminal device 300 as input sentences (S302).

[0146] In response to the input sentence being received, the input unit 121 inputs the input sentence to the model management unit 211 of the server device 201 (S303). At this time, when the learned model is a large language model, the input unit 121 inputs, for example, a prompt including the input sentence, an instruction to verify whether the claim and the specification in the input sentence meet the description requirements, and an instruction to output an amendment to the specification when the description requirements are not met.

[0147] The model management unit 210 outputs correction point information in response to the input from the input unit 121 (S304). The acquisition unit 131 acquires the correction point information (S305).

[0148] The output unit 141 outputs the correction point information to the terminal device 300 (S306). At this time, the output unit 141 may output the correction point information as it is, or may output a sentence generated using the correction point information. Then, the output device 320 of the terminal device 300 displays a sentence regarding the correction point information (S307).

[0149] The operation example of the verification support device 101 is not limited to this example. For example, after the process of S307, the input unit 121 may input the corrected specification in which the correction based on the correction point information is reflected, and the claims in the input sentence, to the model management unit 211.

[0150] Also, for example, after the process of S305, the input unit 121 may generate a corrected specification in which the correction based on the correction point information is reflected, and input the corrected specification and the claims in the input sentence to the model management unit 211. The acquisition unit 131 acquires the correction point information for the corrected specification. Then, the output unit 141 outputs the correction point information for the corrected specification. At this time, the output unit 141 may output the correction point information for the corrected specification and the correction point information for the specification before correction.

[0151] As described above, the verification support device 101 of the third embodiment inputs an input sentence to a learned model that has learned the relationship between the examination data indicating the examination result including the points pointed out regarding the description requirements for a predetermined claim filed, the predetermined claim, and the predetermined specification in which the invention related to the predetermined claim is described. Also, the verification support device 101 acquires the correction point information including the presence or absence of correction points for satisfying the description requirements for the specification in the input sentence, output by the learned model. Then, the verification support device 101 outputs the correction point information.

[0152] Thereby, the verification support device 101 can quickly output the correction content for satisfying the description requirements. Therefore, the verification support device 101 can reduce the burden on the person in charge of creating the patent application documents and improve the quality of the patent application documents.

[0153] Further, the verification support device 101 may input the amended specification in which the amendments based on the amendment point information in the input text are reflected, and the claims in the input text, into the learned model. Then, the verification support device 101 may acquire the amendment point information for the amended specification.

[0154] Thereby, the verification support device 101 can evaluate the effectiveness of the amendment based on the amendment point information.

[0155] As a specific example, the learned model can add the following descriptions to the amendment point information. For example, when the specification describes that "when the measured value A of the sensor exceeds the threshold value T, the control operation C is executed", the learned model may propose to add the description that "even when the measured value A of the sensor is greater than or equal to the threshold value T, the control operation C is executed". This is the result of learning cases in which it has been pointed out in past examination data that the expression "exceeds" does not include "greater than or equal to".

[0156] As another example, when the specification describes that "the operation mode of the device is switched from mode X to mode Y", the learned model may propose to add as amendment point information "the specific procedures and conditions when the operation mode of the device is switched from mode X to mode Y". This is the result of learning cases in which it has been pointed out in past examination data that the description of the specific implementation means for mode switching is insufficient.

[0157] Furthermore, as an example regarding numerical ranges, when the specification describes that "the thickness of member P is 1 mm to 5 mm", the learned model may propose to add "the influence when the thickness of member P is less than 1 mm or exceeds 5 mm, and the reason why the range of 1 mm to 5 mm is optimal". This is the result of learning cases in which an explanation of the critical significance of numerical limitations has been required in past examination data.

[0158] In this way, the trained model can, based on learning, establish the relationship between the examination data indicating the examination results including the points regarding the description requirements for a given claimed item that has been examined, the given claimed item, and the given specification in which the invention related to the given claimed item is described, and make a specific proposal for enriching the description in the specification. Thereby, it becomes possible not only to meet the description requirements but also to assist in creating a specification for securing a stronger scope of rights.

[0159] <Example of the hardware configuration of the verification support device> The hardware constituting the verification support device of the first, second, and third embodiments described above will be described. FIG. 15 is a block diagram showing an example of the hardware configuration of a computer device constituting the verification support device in each embodiment. In the computer device 90, the verification support device and the verification support method described in each embodiment and each modification are realized. For example, the verification support device and the like described in each embodiment and each modification may have the hardware configuration shown in FIG. 15.

[0160] As shown in FIG. 15, the computer device 90 includes a processor 91, a RAM (Random Access Memory) 92, a ROM (Read Only Memory) 93, a storage device 94, an input / output interface 95, a bus 96, and a drive device 97. Note that the verification support device and the like may be realized by a plurality of electric circuits.

[0161] The memory device 94 stores a program (computer program) 98. The processor 91 executes the program 98 of this verification support device using the RAM 92. Specifically, for example, the program 98 includes programs that cause a computer to execute the processes shown in FIGS. 2, 5, 8, 14, etc. When the processor 91 executes the program 98, the functions of each component of this verification support device are realized. The program 98 may be stored in the ROM 93. Also, the program 98 may be recorded on the recording medium 80 and read using the drive device 97, or may be transmitted from an external device (not shown) to the computer device 90 via a network (not shown).

[0162] The input / output interface 95 exchanges data with peripheral devices (such as a keyboard, a mouse, a display device, etc.) 99. The input / output interface 95 functions as a means for acquiring or outputting data. The bus 96 connects each component.

[0163] Note that there are various modification examples for the method of realizing the verification support device. For example, each component included in the verification support device can be realized as a dedicated device. Also, the verification support device can be realized based on a combination of a plurality of devices.

[0164] A processing method of recording a program for realizing each component in the functions of each embodiment on a recording medium, reading the program recorded on the recording medium as code, and executing it on a computer is also included in the scope of each embodiment. That is, a computer-readable recording medium is also included in the scope of each embodiment. Also, the recording medium on which the above-described program is recorded, and the program itself are included in each embodiment.

[0165] The recording medium is, for example, a floppy (registered trademark) disk, hard disk, optical disk, magneto-optical disk, CD (Compact Disc)-ROM, magnetic tape, non-volatile memory card, or ROM, but is not limited to this example. Also, the program recorded on the recording medium is not limited to a program that executes processing alone, but also includes programs that operate on an OS (Operating System) in cooperation with other software and the functions of expansion boards to execute processing within the scope of each embodiment.

[0166] As described above, the present invention has been described with reference to the embodiments, but the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

[0167] Also, the above embodiments and modified examples can be combined as appropriate. Further, the functions of Embodiments 1 to 3 and the functions of each modified example in the present disclosure can all be performed simultaneously within a non-contradictory range. Alternatively, among the functions of Embodiments 1 to 3 and the functions of each modified example, any function can be performed simultaneously within a non-contradictory range. In that case, in Embodiments 1 to 3 and each modified example, processing may be performed on common inputs. For example, for a predetermined claim and specification received as an input sentence, the processing described in Embodiments 1 to 3 and each modified example may be sequentially performed. Alternatively, any function among each embodiment and each modified example may be processed in order. In that case, the processing result such as the generated sentence obtained by the previous processing may be input to the large language model together with the prompt in the subsequent processing, or may be used as part of the prompt in the subsequent processing.

[0168] <Supplementary Note> Some or all of the above embodiments may also be described as follows in the supplementary note, but are not limited thereto.

[0169] [Supplementary Note 1] Receiving means for receiving a claim and a specification in which the invention according to the claim are described as an input sentence, Input means for inputting into a large language model an instruction to extract terms used in the claim from the input text and an instruction to generate an explanatory text regarding the presence or absence of definitions of the extracted terms in the specification; Acquisition means for acquiring a generated text generated based on the input instruction; Output means for outputting, based on the generated text, the extracted terms and explanatory texts related to the presence or absence of definitions for each of the extracted terms, comprising: A verification support device.

[0170] [Appendix 2] The input means inputs the input text and a first prompt including an instruction to extract terms used in the claim from the input text; The acquisition means acquires a first generated text generated based on the first prompt; The input means inputs, for each of the extracted terms included in the first generated text, a second prompt including an instruction to generate an explanatory text related to the presence or absence of a definition, including a passage in the specification related to the definition of the term; The acquisition means acquires a second generated text generated based on the second prompt; The output means outputs, based on the first generated text and the second generated text, for each extracted term, a passage in the specification related to the definition of the term and an explanatory text related to the verification result of the presence or absence of a definition; The verification support device according to Appendix 1.

[0171] [Appendix 3] The input means inputs an instruction to identify terms not defined in the specification among the extracted terms and an instruction to generate a candidate text, which is a text explaining the term, for each identified term; The acquisition means acquires a generated text including the candidate text for each identified term; The verification support device according to Appendix 1 or 2.

[0172] [Appendix 4] The input means inputs an instruction to specify, from the specification, a sentence similar to the candidate sentence, and an instruction to indicate a location in the specification where the candidate sentence should be inserted based on the specified sentence. The acquisition means acquires a generated sentence including the location where the candidate sentence should be inserted. The verification support device according to Supplementary Note 3.

[0173] [Supplementary Note 5] The second generated sentence includes the paragraph number of the paragraph in the specification that contains the sentences related to the definition of the term for each extracted term. The input means inputs a prompt including an instruction to generate a sentence explaining the interpretation of the components of the claim based on the specification and the second generated sentence. The acquisition means acquires the generated sentence generated based on the prompt. The output means outputs a sentence explaining the interpretation of the components of the claim based on the generated sentence generated based on the prompt. The verification support device according to Supplementary Note 2.

[0174] [Supplementary Note 6] The input means inputs to the large language model an instruction to extract terms used in the claim from the input sentence, and an instruction to verify the presence or absence of similar terms, which are terms with similar meanings in the specification, for the extracted terms. The acquisition means acquires a generated sentence including an explanatory text regarding the presence or absence of similar terms for each of the extracted terms, generated based on the input instruction. The output means outputs, for each extracted term, an explanatory text regarding the presence or absence of similar terms based on the generated sentence regarding the presence or absence of similar terms. The verification support device according to Supplementary Note 1.

[0175] [Supplementary Note 7] The input means inputs the input sentence into a trained model that has learned the relationship between the examination data indicating the examination results including the points pointed out regarding the description requirements for a given claimed claim, the given claimed claim, and the given specification in which the invention related to the given claimed claim is described, The acquisition means acquires correction point information including whether there are correction points for satisfying the description requirements for the specification in the input sentence, output by the trained model, The output means outputs the correction point information. The verification support device according to Supplementary Note 1.

[0176] [Supplementary Note 8] The input means inputs the corrected specification in which the correction based on the correction point information is reflected in the specification in the input sentence and the claims in the input sentence into the trained model, The acquisition means acquires correction point information for the corrected specification. The verification support device according to Supplementary Note 7.

[0177] [Supplementary Note 9] Receives the claims and the specification in which the invention related to the claims is described as an input sentence, Inputs an instruction to extract the terms used in the claims from the input sentence and an instruction to generate a description regarding the presence or absence of definitions of the extracted terms in the specification into a large language model, Acquires the generated sentence generated based on the input instruction, Outputs the extracted terms and the descriptions related to the presence or absence of definitions of each of the extracted terms based on the generated sentence. Verification support method.

[0178] [Supplementary Note 10] A process of receiving the claims and the specification in which the invention related to the claims is described as an input sentence, A process of inputting an instruction to extract the terms used in the claims from the input sentence and an instruction to generate a description regarding the presence or absence of definitions of the extracted terms in the specification into a large language model, A process of obtaining a generated sentence generated based on the input instruction, and Based on the generated sentence, a process of outputting, for each of the extracted terms and the extracted terms, an explanatory text related to the presence or absence of a definition, and causing a computer to execute the process. Program.

[0179] [Appendix 11] Receiving means for receiving, as an input sentence, a claim generated by a large language model and a specification describing the invention related to the claim, Input means for inputting, to the large language model, an instruction to extract terms used in the claim from the input sentence and an instruction to generate an explanatory text related to the presence or absence of a definition of the extracted terms in the specification, Obtaining means for obtaining a generated sentence generated based on the input instruction, Output means for outputting, based on the generated sentence, the extracted terms and, for each of the extracted terms, an explanatory text related to the presence or absence of a definition, and comprising: Verification support device.

[0180] [Appendix 12] Receiving, as an input sentence, a claim generated by a large language model and a specification describing the invention related to the claim, Inputting, to the large language model, an instruction to extract terms used in the claim from the input sentence and an instruction to generate an explanatory text related to the presence or absence of a definition of the extracted terms in the specification, Obtaining a generated sentence generated based on the input instruction, Outputting, based on the generated sentence, the extracted terms and, for each of the extracted terms, an explanatory text related to the presence or absence of a definition. Verification support method.

[0181] [Appendix 13] A process of receiving, as an input sentence, a claim generated by a large language model and a specification describing the invention related to the claim, and A process of inputting an instruction to extract terms used in the claim from the input text and an instruction to generate an explanatory text regarding the presence or absence of the definition of the extracted terms in the specification into a large language model, A process of obtaining a generated text generated based on the input instruction, Based on the generated text, a process of causing a computer to execute an output process of the extracted terms and the explanatory text related to the presence or absence of the definition of each of the extracted terms, Program.

[0182] Also, part or all of the configurations described in Appendices 2 to 8 that are subordinate to Appendix 1 described above may be subordinate to Appendices 9 to 13 in the same subordinate relationship as Appendices 2 to 8. Furthermore, within the scope not departing from the above-described embodiments, part or all of the configurations described as appendices can similarly be made subordinate to various hardware, software, various recording means for recording software, or systems.

Explanation of Reference Numerals

[0183] 100, 101 Verification support device 110 Reception unit 120, 121 Input unit 130, 131 Acquisition unit 140, 141 Output unit 150 Prompt storage unit 200, 201 Server device 210, 211 Model management unit 300 Terminal device 310 Input device 320 Output device

Claims

1. a receiving means for receiving, as input text, a claim and a specification in which the invention according to the claim is described; an input means for inputting a first instruction for extracting terms contained in the claims from the input sentence and a second instruction for generating an explanatory sentence regarding whether or not the extracted terms are defined in the specification into a large-scale language model; an acquisition means for acquiring a generated sentence generated based on the first instruction and the second instruction; an output means for outputting an explanatory sentence regarding the presence or absence of a definition for each of the terms extracted based on the generated sentence, the input means inputs, in response to acquisition of the generated sentence, a third instruction for identifying a term that is not defined in the specification among the extracted terms, and a fourth instruction for generating a candidate sentence that explains the definition of the identified term; The acquiring means acquires a generated sentence including the candidate sentence, the output means outputs the candidate sentences for each of the identified terms based on a generated sentence including the candidate sentences. Verification support device.

2. the input means inputs the input sentence and inputs a first prompt including the first instruction; the acquiring means acquires a first generated sentence generated based on the first prompt; the input means inputs a second prompt including the second instruction, the second prompt including an instruction to generate, for each of the extracted terms included in the first generated sentence, an explanatory sentence regarding the presence or absence of a definition, the explanatory sentence including a sentence in the specification related to the definition of the term; the acquiring means acquires a second generated sentence generated based on the second prompt; the output means outputs, for each extracted term, a sentence in the specification related to a definition of the term and an explanatory sentence regarding a verification result of the presence or absence of a definition, based on the first generated sentence and the second generated sentence. The verification support device according to claim 1 .

3. the input means inputs an instruction to specify a sentence similar to the candidate sentence from the specification, and an instruction to indicate a location in the specification where the candidate sentence should be inserted based on the specified sentence; the acquiring means acquires a generated sentence including a position where the candidate sentence should be inserted; 3. The verification support device according to claim 1.

4. the second generated sentence includes, for each extracted term, a paragraph number of a paragraph in the specification that includes a sentence related to a definition of the term; the input means inputs a prompt including an instruction to generate a sentence explaining an interpretation of a component of the claim based on the specification and the second generated sentence; The acquiring means acquires a generated sentence based on the prompt, the output means outputs a sentence explaining an interpretation of the element of the claim based on a generated sentence generated based on the prompt. The verification support device according to claim 2.

5. the input means inputs to the large-scale language model an instruction to extract terms used in the claims from the input sentence and an instruction to verify, for the extracted terms, the presence or absence of similar terms in the specification that are terms similar in meaning to the terms; The acquiring means acquires generated sentences generated based on an input instruction, the generated sentences including explanatory sentences regarding the presence or absence of the similar terms for each of the extracted terms; the output means outputs, for each extracted term, an explanatory sentence regarding the presence or absence of the similar term, based on the generated sentence regarding the presence or absence of the similar term. The verification support device according to claim 1 .

6. The input means inputs the input sentence into a trained model that has learned the relationship between examination data indicating examination results including indications regarding description requirements for a specific claim filed, and a specific specification in which the specific claim and the invention related to the specific claim are described; The acquisition means acquires correction information including the presence or absence of corrections to satisfy description requirements for the specification in the input sentence, the corrections being output by the trained model; The output means outputs the correction point information. The verification support device according to claim 1 .

7. The input means inputs a corrected specification in which the specification in the input sentence is modified based on the modification point information and a claim in the input sentence into the trained model; The acquiring means acquires amendment information for the amended specification. The verification support apparatus according to claim 6.

8. The input means inputs an instruction to insert the candidate sentence based on the location, the acquiring means acquires a generated sentence including a specification into which the candidate sentence is inserted based on the portion; The verification support device according to claim 3.

9. A claim and a specification in which the invention according to the claim is described are accepted as an input sentence; inputting a first instruction for extracting terms contained in the claims from the input sentence and a second instruction for generating an explanatory sentence regarding whether or not the extracted terms are defined in the specification into a large-scale language model; acquiring a generated sentence generated based on the first instruction and the second instruction; outputting an explanatory sentence regarding the presence or absence of a definition for each of the terms extracted based on the generated sentence; In response to obtaining the generated sentence, inputting a third instruction for identifying a term that is not defined in the specification among the extracted terms, and a fourth instruction for generating a candidate sentence that explains a definition of the identified term; A generated sentence including the candidate sentence is obtained, outputting the candidate sentence for each identified term based on a generated sentence including the candidate sentence; Verification support methods.

10. A process of accepting a claim and a specification in which the invention related to the claim is described as an input sentence; A process of inputting a first instruction for extracting terms included in the claims from the input sentence and a second instruction for generating an explanatory sentence regarding whether or not the extracted terms are defined in the specification into a large-scale language model; A process of acquiring a generated sentence generated based on the first instruction and the second instruction; A process of outputting an explanatory sentence regarding the presence or absence of a definition for each of the terms extracted based on the generated sentence; a process of inputting a third instruction for identifying a term that is not defined in the specification among the extracted terms in response to acquisition of the generated sentence, and a fourth instruction for generating a candidate sentence that explains the definition of the identified term; A process of obtaining a generated sentence including the candidate sentence; outputting the candidate sentence for each identified term based on a generated sentence including the candidate sentence; program.

Images