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

Abstract

A technology is provided that increases a user's sense of satisfaction with the results of a determination of the possibility of acquiring rights related to intellectual property. [Solution] The information processing device includes a related technology acquisition unit that acquires related technology information related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in natural language sentences; a positive opinion generation unit that generates a positive opinion regarding the possibility of acquiring rights to the intellectual property to be analyzed using a first large-scale language model; a negative opinion generation unit that generates a negative opinion regarding the possibility of acquiring rights to the intellectual property to be analyzed using a second large-scale language model; and a conclusion generation unit that generates a conclusion regarding the possibility of acquiring rights to the intellectual property to be analyzed using a third large-scale language model.

Description

[Technical Field]

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

[0002] Patent Document 1 discloses a technology for determining the possibility of acquiring rights for information related to intellectual property entered by a user. In this technology, a rank indicating the possibility of acquiring the rights is presented to the user as a result of the determination. Furthermore, if the intellectual property is an invention, the degree of agreement with similar documents for each constituent element of the invention is presented to the user as a result of the determination. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2019-179493 Summary of the Invention [Problem to be solved by the invention]

[0004] In the technology described in Patent Document 1, the ranking of the possibility of obtaining rights, the degree of agreement with similar documents, etc., as described above, may not be enough to convince users of the judgment results. Therefore, there is a need to increase the user's satisfaction with the judgment results.

[0005] The present disclosure has been made in consideration of the above-mentioned problems, and one exemplary purpose thereof is to provide a technology that increases a user's sense of satisfaction with the results of a determination of the possibility of acquiring rights related to intellectual property. [Means for solving the problem]

[0006] An information processing device according to an exemplary aspect of the present disclosure includes a related technology acquisition means for acquiring related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence; a positive opinion generation means for generating a positive opinion regarding the possibility of obtaining rights for the intellectual property to be analyzed based on the analysis target information and the related technology information using a first large-scale language model; a negative opinion generation means for generating a negative opinion regarding the possibility of obtaining rights based on the analysis target information and the related technology information using a second large-scale language model; and a conclusion generation means for generating a conclusion regarding the possibility of obtaining rights based on the positive opinion and the negative opinion using a third large-scale language model.

[0007] An information processing method according to an exemplary aspect of the present disclosure includes a related technology acquisition process in which at least one processor acquires related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence; a positive opinion generation process in which the at least one processor uses a first large-scale language model to generate a positive opinion regarding the possibility of obtaining rights to the intellectual property to be analyzed based on the analysis target information and the related technology information; a negative opinion generation process in which the at least one processor uses a second large-scale language model to generate a negative opinion regarding the possibility of obtaining rights based on the analysis target information and the related technology information; and a conclusion generation process in which the at least one processor uses a third large-scale language model to generate a conclusion regarding the possibility of obtaining rights based on the positive opinion and the negative opinion.

[0008] An information processing program according to an exemplary aspect of the present disclosure is an information processing program that causes at least one processor to function as an information processing device, and functions as: a related technology acquisition means that acquires related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in natural language sentences; a positive opinion generation means that uses a first large-scale language model to generate a positive opinion regarding the possibility of obtaining rights for the intellectual property to be analyzed based on the analysis target information and the related technology information; a negative opinion generation means that uses a second large-scale language model to generate a negative opinion regarding the possibility of obtaining rights based on the analysis target information and the related technology information; and a conclusion generation means that uses a third large-scale language model to generate a conclusion regarding the possibility of obtaining rights based on the positive opinion and the negative opinion. [Effects of the Invention]

[0009] According to an exemplary aspect of the present disclosure, an exemplary effect is achieved in that a technology can be provided that increases a user's sense of satisfaction with the determination result of the possibility of acquiring rights related to intellectual property. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a block diagram illustrating a configuration of an information processing device according to the present disclosure. [Figure 2] FIG. 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 3] 1 is a diagram schematically illustrating an overview of an information processing system according to the present disclosure. [Figure 4] 1 is a block diagram illustrating a configuration of an information processing system according to the present disclosure. [Figure 5] FIG. 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 6] FIG. 10 is a diagram schematically illustrating an example of an analysis target input screen according to the present disclosure. [Figure 7] FIG. 10 is a diagram schematically illustrating an example of a generated keyword screen according to the present disclosure. [Figure 8]FIG. 10 is a diagram schematically illustrating an example of a prior art screen according to the present disclosure. [Figure 9] FIG. 10 is a diagram schematically illustrating an example of a difference screen according to the present disclosure. [Figure 10] FIG. 10 is a diagram schematically illustrating an example of a dual opinion screen according to the present disclosure. [Figure 11] FIG. 10 is a diagram schematically illustrating another example of the two opinions screen according to the present disclosure. [Figure 12] FIG. 10 is a diagram schematically illustrating an example of a conclusion screen according to the present disclosure. [Figure 13] FIG. 10 is a diagram schematically illustrating an example of a final proposal screen according to the present disclosure. [Figure 14] FIG. 2 is a block diagram showing the hardware configuration of a computer that functions as each device according to the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0011] The following are examples of embodiments of the present invention. However, the present invention is not limited to the exemplary embodiments shown below, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technologies (part or all of the products or methods) employed in the exemplary embodiments shown below may also be included in the scope of the present invention. Furthermore, embodiments obtained by appropriately omitting some of the technologies employed in the exemplary embodiments shown below may also be included in the scope of the present invention. Furthermore, the effects mentioned in the exemplary embodiments shown below are examples of effects expected in the exemplary embodiments, and do not define the scope of the present invention. In other words, embodiments that do not exhibit the effects mentioned in the exemplary embodiments shown below may also be included in the scope of the present invention.

[0012] First Exemplary Embodiment A first exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. This exemplary embodiment is the basic form of each exemplary embodiment described later. The scope of application of each technology employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise. Furthermore, each technology shown in the drawings referenced to explain this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise.

[0013] (Configuration of information processing device 1) The configuration of the information processing device 1 will be described with reference to FIG. 1. FIG. 1 is a block diagram showing the configuration of the information processing device 1. As shown in FIG. 1, the information processing device 1 includes a related technology acquisition unit 11, a positive opinion generation unit 12, a negative opinion generation unit 13, and a conclusion generation unit 14. The related technology acquisition unit 11 is an example of a configuration that realizes related technology acquisition means. The positive opinion generation unit 12 is an example of a configuration that realizes positive opinion generation means. The negative opinion generation unit 13 is an example of a configuration that realizes negative opinion generation means. The conclusion generation unit 14 is an example of a configuration that realizes conclusion generation means.

[0014] The related technology acquisition unit 11 acquires related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in natural language. Here, the intellectual property to be analyzed is intellectual property that can be described in natural language. For example, the intellectual property to be analyzed may be an invention, a device, or a paper, but is not limited to these. Furthermore, the intellectual property to be analyzed may be intellectual property in any state, such as under consideration, before application, after application, before grant, after grant, etc. Furthermore, for example, the analysis target information may include information indicating the scope of rights (e.g., claims), a summary, and part or all of a detailed description. The intellectual property to be analyzed may be input, for example, by a user operation or by being read from an arbitrary storage medium.

[0015] Furthermore, the related technology acquisition unit 11 may select related acquisition information from among a plurality of candidates of related technology information based on the analysis target information. Furthermore, the related technology acquisition unit 11 may acquire related technology information designated by a user according to the analysis target information.

[0016] The positive opinion generation unit 12 uses the first large-scale language model to generate a positive opinion regarding the possibility of obtaining rights for the intellectual property to be analyzed, based on the analysis target information and related technical information. For example, the possibility of obtaining rights for the intellectual property to be analyzed may include the possibility that one or both of novelty and inventive step are recognized. Furthermore, for example, the possibility of obtaining rights may include the possibility of satisfying other requirements in addition to novelty and / or inventive step. For example, the positive opinion is a natural language sentence indicating an opinion that the intellectual property to be analyzed has novelty and / or inventive step relative to related technology. For example, the positive opinion may include a natural language sentence indicating that the intellectual property to be analyzed has novelty and / or inventive step and the basis for this.

[0017] For example, when the first large-scale language model receives the analysis target information and related technical information, it outputs a positive opinion regarding the possibility of obtaining the right. Note that the information input to the first large-scale language model includes at least the analysis target information and related technical information, and may or may not include other information.

[0018] For example, the first large-scale language model may be a general-purpose large-scale language model fine-tuned using positive case information. The positive case information may include, for example, cases of the analysis target information, cases of related technical information, and cases of positive opinions regarding the possibility of obtaining rights for intellectual property indicated by the cases of the analysis target information. Such positive case information may include information obtained regarding other intellectual property whose possibility of obtaining rights has actually been confirmed, or may include information generated for training purposes.

[0019] Furthermore, for example, the first large-scale language model does not necessarily have to be fine-tuned and may be a general-purpose large-scale language model. In this case, for example, a positive opinion may be output by in-context learning in which the analysis target information, related technical information, and the above-mentioned positive case information are input to the first large-scale language model.

[0020] The negative opinion generation unit 13 uses the second large-scale language model to generate a negative opinion regarding the possibility of obtaining rights for the intellectual property to be analyzed, based on the analysis target information and related technology information. Specific examples of the possibility of obtaining rights for the intellectual property to be analyzed are as described above. For example, the negative opinion is a natural language sentence indicating an opinion based on the premise that the intellectual property to be analyzed does not have novelty and / or inventive step compared to related technology. For example, the negative opinion may include a natural language sentence indicating that the intellectual property to be analyzed does not have novelty and / or inventive step, and the basis for this.

[0021] For example, when the second large-scale language model receives the analysis target information and related technical information, it outputs a negative opinion regarding the possibility of obtaining the right. Note that the information input to the second large-scale language model includes at least the analysis target information and related technical information, and may or may not include other information.

[0022] For example, the second large-scale language model may be a general-purpose large-scale language model fine-tuned using negative case information. The negative case information may include, for example, cases of the analysis target information, cases of related technical information, and cases of negative opinions regarding the obtainability of rights to intellectual property indicated by the cases of the analysis target information. Such negative case information may include information obtained regarding other intellectual property whose obtainability has actually been denied, or may include information generated for training purposes.

[0023] Furthermore, for example, the second large-scale language model does not necessarily have to be fine-tuned and may be a general-purpose large-scale language model. In this case, negative opinions may be output by in-context learning, in which the analysis target information, related technical information, and the above-mentioned negative case information are input to the second large-scale language model.

[0024] The conclusion generation unit 14 uses the third large-scale language model to generate a conclusion regarding the possibility of obtaining rights for the intellectual property to be analyzed based on the affirmative and negative opinions. For example, the conclusion is a natural language sentence indicating whether the affirmative or negative opinion regarding the presence or absence of novelty and / or inventive step of the intellectual property to be analyzed relative to related technology is appropriate. For example, the conclusion may include a natural language sentence indicating either the affirmative or negative opinion and the reason for it.

[0025] For example, the third large-scale language model outputs a conclusion regarding the possibility of obtaining the right when the analysis target information, related technical information, positive opinions, and negative opinions are input. Note that the information input to the third large-scale language model includes at least positive opinions and negative opinions, and may or may not include other information.

[0026] For example, the third large-scale language model may be a general-purpose large-scale language model fine-tuned using case information for conclusion. The case information for conclusion may include, for example, business information related to a business related to the intellectual property to be analyzed and / or a patent portfolio related to the intellectual property to be analyzed. This allows a conclusion to be generated taking into account the business information and / or the patent portfolio. Furthermore, the case information for conclusion may include, for example, cases of the information to be analyzed, cases of related technical information, cases of positive opinions, cases of negative opinions, and cases of conclusion. Such case information for conclusion may be generated based on information about other intellectual property whose possibility of obtaining rights has actually been confirmed or denied, so as to include positive opinions and negative opinions generated by the positive opinion generation unit 12 and the negative opinion generation unit 13 regarding the other intellectual property. Furthermore, such case information for conclusion may be information generated for training.

[0027] Furthermore, for example, the third large-scale language model does not necessarily have to be fine-tuned and may be a general-purpose large-scale language model. In this case, a conclusion may be output by in-context learning, in which the analysis target information, related technical information, positive opinions, negative opinions, and the above-mentioned case information for conclusions are input to the third large-scale language model.

[0028] Note that, when at least two of the first, second, and third large-scale language models are fine-tuned models, the at least two models are different from each other. Also, for example, when at least two of the first, second, and third large-scale language models are general-purpose large-scale language models, the at least two models may be the same or different.

[0029] (Effects of information processing device 1) As described above, the information processing device 1 is configured to include a related technology acquisition unit 11 that acquires related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in natural language sentences, a positive opinion generation unit 12 that generates a positive opinion regarding the possibility of acquiring rights to the intellectual property to be analyzed based on the analysis target information and the related technology information using a first large-scale language model, a negative opinion generation unit 13 that generates a negative opinion regarding the possibility of acquiring rights based on the analysis target information and the related technology information using a second large-scale language model, and a conclusion generation unit 14 that generates a conclusion regarding the possibility of acquiring rights based on the positive and negative opinions using a third large-scale language model. Therefore, the information processing device 1 generates a conclusion based on both the positive and negative opinions regarding the possibility of acquiring rights to the intellectual property, thereby achieving the effect of increasing the user's satisfaction with the conclusion regarding the possibility of acquiring rights.

[0030] (Flow of information processing method S1) The flow of information processing method S1 will be described with reference to Fig. 2. For example, when information processing device 1 has at least one processor, information processing device 1 executes information processing method S1. Fig. 2 is a flow diagram showing the flow of information processing method S1. As shown in Fig. 2, information processing method S1 includes related technology acquisition processing S11, positive opinion generation processing S12, negative opinion generation processing S13, and conclusion generation processing S14.

[0031] In the related technology acquisition process S11, at least one processor (for example, the related technology acquisition unit 11) acquires related technology information indicating related technologies related to the intellectual property to be analyzed based on the analysis target information in which the intellectual property to be analyzed is described in natural language. Details of the related technology acquisition process S11 will be explained in the same manner as the details of the related technology acquisition unit 11 described above.

[0032] In the positive opinion generation process S12, at least one processor (for example, the positive opinion generation unit 12) uses the first large-scale language model to generate a positive opinion regarding the possibility of obtaining rights to the intellectual property to be analyzed, based on the analysis target information and related technical information. Details of the positive opinion generation process S12 will be explained in the same manner as the details of the positive opinion generation unit 12 described above.

[0033] In the negative opinion generation process S13, at least one processor (for example, the negative opinion generation unit 13) uses the second large-scale language model to generate a negative opinion regarding the possibility of obtaining rights to the intellectual property to be analyzed, based on the analysis target information and related technical information. Details of the negative opinion generation process S13 will be explained in the same manner as the details of the negative opinion generation unit 13 described above.

[0034] The positive opinion generation process S12 and the negative opinion generation process S13 are not limited to being executed in the order described above, but may be executed in the reverse order, or some or all of the processes may be executed in parallel.

[0035] In the conclusion generation process S14, at least one processor uses the third large-scale language model to generate a conclusion regarding the possibility of obtaining rights for the intellectual property to be analyzed based on the positive and negative opinions. Details of the conclusion generation process S14 will be described in the same manner as the details of the conclusion generation unit 14 described above.

[0036] (Effect of information processing method S1) As described above, the information processing method S1 includes a related technology acquisition process S11 in which at least one processor acquires related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence, a positive opinion generation process S12 in which at least one processor generates a positive opinion regarding the possibility of obtaining rights to the intellectual property to be analyzed based on the analysis target information and the related technology information using a first large-scale language model, a negative opinion generation process S13 in which at least one processor generates a negative opinion regarding the possibility of obtaining rights to the intellectual property to be analyzed based on the analysis target information and the related technology information using a second large-scale language model, and a conclusion generation process S14 in which at least one processor generates a conclusion regarding the possibility of obtaining rights to the intellectual property to be analyzed based on the positive and negative opinions using a third large-scale language model. Therefore, the information processing method S1 can achieve the same effects as the information processing device 1.

[0037] Second Exemplary Embodiment A second exemplary embodiment, which is one example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same functions as those described in the above exemplary embodiment will be assigned the same reference numerals, and their description will be omitted as appropriate. The scope of application of each technology employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical hindrance occurs. Furthermore, each technology shown in each drawing referenced to explain this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical hindrance occurs.

[0038] (Overview of information processing system 100A) The information processing system 100A presents a positive opinion, a negative opinion, and a conclusion regarding the novelty and inventive step (an example of the possibility of obtaining a right) of the subject invention (an example of intellectual property) indicated by the claim to be analyzed, based on the claim to be analyzed (an example of information to be analyzed) and prior art documents (an example of related technical information). In addition, if the conclusion does not satisfy a predetermined condition, the information processing device 1A generates an improvement proposal for the claim (an example of improvement proposal information), and by repeating the operation using the improvement proposal as a new claim to be analyzed, presents to the user an improvement proposal whose conclusion satisfies the predetermined condition as a final proposal.

[0039] FIG. 3 is a diagram illustrating a schematic overview of an information processing system 100A. As shown in FIG. 3, in the information processing system 100A, keywords are generated from a claim to be analyzed using a large-scale language model LLM1, or an abstract is generated using a large-scale language model LLM2. While both keywords and abstracts may be generated, the following description focuses on an example in which only one of them is generated. A classification, such as an International Patent Classification (IPC) classification, is identified from the claim to be analyzed using a large-scale language model LLM3. Next, multiple prior art documents are retrieved from Database 3 (described below) using the keywords or abstract and the classification. Next, a prior art document to be compared with the claim to be analyzed is identified from the multiple prior art documents using a large-scale language model LLM4. Next, differences and commonalities between the claim to be analyzed and the prior art document are analyzed using a large-scale language model LLM5. Next, a positive opinion regarding the novelty and inventive step of the subject invention indicated by the claim to be analyzed is generated using a large-scale language model LLM6. A negative opinion regarding the novelty and inventive step of the subject invention is generated using a large-scale language model LLM7. Next, a conclusion as to whether a positive or negative opinion is appropriate is generated using a large-scale language model LLM8. Next, if the conclusion does not satisfy a predetermined condition (for example, the conclusion is not positive), a proposed improvement to the claim is generated using a large-scale language model LLM9. The proposed improvement to the claim is then used as the new claim to be analyzed, and the above-described series of processes are repeated.

[0040] (Configuration of information processing system 100A) The configuration of the information processing system 100A will be described with reference to FIG. 4. FIG. 4 is a block diagram showing the configuration of the information processing system 100A. As shown in FIG. 4, the information processing system 100A includes an information processing device 1A, a large-scale language model storage device 2, a database 3, an input device 4, and a display device 5. The information processing device 1A is communicatively connected to the large-scale language model storage device 2, the database 3, the input device 4, and the display device 5 via a network, a peripheral device connection interface, or the like. Some or all of the information stored in the large-scale language model storage device 2 and the database 3 may be stored in a storage unit 120 of the information processing device 1A. Alternatively, one or both of the input device 4 and the display device 5 may be built into the information processing device 1A instead of being connected to the information processing device 1A. Alternatively, the input device 4 and the display device 5 may be connected to or built into a user terminal (not shown), and the user terminal may be communicatively connected to the information processing device 1A via a network. Although FIG. 4 shows one each of the large-scale language model storage device 2, database 3, input device 4, and display device 5, the information processing system 100A may include multiple copies of some or all of these devices.

[0041] (Large-scale language model storage device 2) The large-scale language model storage device 2 stores large-scale language models LLM1 to LLM9. Each of the large-scale language models LLM1 to LLM9 is a deep learning model generated to perform a natural language processing task. For example, the large-scale language models LLM1 to LLM9 are models that perform a sentence generation task, and input a prompt in natural language sentences to output generated natural language sentences. Each of the large-scale language models LLM1 to LLM9 may be a fine-tuned model of a general-purpose large-scale language model, or may be a general-purpose large-scale language model. If at least one of the large-scale language models LLM1 to LLM9 is a general-purpose large-scale language model, in-context learning may be performed using that large-scale language model. Furthermore, if at least two of the large-scale language models LLM1 to LLM9 are general-purpose large-scale language models, the two may be the same model or different models.

[0042] The large-scale language model LLM1 is an example of a fifth large-scale language model used to generate keywords. For example, when a claim to be analyzed is input, the large-scale language model LLM1 outputs keywords related to the claim to be analyzed. For example, the large-scale language model LLM1 may be a general-purpose large-scale language model fine-tuned for the technical field expected for the claim to be analyzed. Alternatively, the large-scale language model LLM1 may be a general-purpose large-scale language model. In this case, keywords may be output by inputting knowledge related to the field indicated by the claim to be analyzed in addition to the claim to be analyzed into the large-scale language model LLM1.

[0043] The large-scale language model LLM2 is an example of a large-scale language model used to generate a summary of information to be analyzed. For example, the large-scale language model LLM2 receives a claim to be analyzed and outputs the summary. For example, the large-scale language model LLM2 may be fine-tuned for the technical field expected for the claim to be analyzed. Alternatively, the large-scale language model LLM2 may be a general-purpose large-scale language model. In this case, the large-scale language model LLM2 may receive knowledge related to the field indicated by the claim to be analyzed in addition to the claim to be analyzed, and output a summary.

[0044] The large-scale language model LLM3 is an example of a seventh large-scale language model used to identify the classification of intellectual property indicated by the information to be analyzed. For example, when the analysis target is an invention, an example of the "intellectual property classification" may include, but is not limited to, the aforementioned IPC. For example, the large-scale language model LLM3 outputs a classification when a claim to be analyzed is input. For example, the large-scale language model LLM3 may be a general-purpose large-scale language model fine-tuned using case information for classification identification. The case information for classification identification includes cases of the claim to be analyzed and cases of the classification. The case information for classification identification may be generated, for example, based on published patent documents. Alternatively, the large-scale language model LLM3 may be a general-purpose large-scale language model. In this case, the large-scale language model LLM3 may output a classification through in-context learning, in which the claim to be analyzed and the above-mentioned case information for classification identification are input.

[0045] The large-scale language model LLM4 is an example of a sixth large-scale language model used to select relevant technical information from multiple candidates. For example, when a claim to be analyzed and multiple patent documents are input to the large-scale language model LLM4, it outputs patent documents from the multiple patent documents that are relevant to the claim to be analyzed as prior art documents. For example, the large-scale language model LLM4 may be a general-purpose large-scale language model fine-tuned for the technical field expected for the claim to be analyzed. Alternatively, the large-scale language model LLM4 may be a general-purpose large-scale language model. In this case, in addition to the claim to be analyzed and multiple patent documents, knowledge related to the field indicated by the claim to be analyzed may be input to the large-scale language model LLM4, thereby outputting the most relevant prior art document.

[0046] The large-scale language model LLM5 is used to generate differences and commonalities between the information to be analyzed and related technical information. For example, when the large-scale language model LLM5 receives the claims to be analyzed and prior art documents, it outputs the differences and commonalities between the claims to be analyzed and the prior art documents. For example, the large-scale language model LLM5 may be a general-purpose large-scale language model fine-tuned using case information including differences and commonalities. The case information including differences and commonalities includes cases of the claims to be analyzed, cases of the prior art documents, and cases of differences and commonalities between the two cases. The case information including differences and commonalities may be generated based on, for example, historical information of published patent documents or may include information generated for training. Alternatively, the large-scale language model LLM5 may be a general-purpose large-scale language model. In this case, the large-scale language model LLM5 may output the differences and commonalities through in-context learning, in which case information including the above-mentioned differences and commonalities is input to the large-scale language model LLM5 in addition to the claims to be analyzed and prior art documents.

[0047] The large-scale language model LLM6 is an example of a first large-scale language model used to generate a positive opinion. For example, when the large-scale language model LLM6 receives the differences and commonalities between the analyzed claim and the prior art document, it outputs a positive opinion regarding the novelty and inventive step of the subject invention indicated by the analyzed claim. The positive opinion includes an opinion on the novelty and inventive step and the reasons for the opinion. For example, the large-scale language model LLM6 may be a general-purpose large-scale language model fine-tuned using positive case information. The positive case information includes cases of differences and commonalities between the analyzed claim case and the prior art document case, as well as cases of positive opinions regarding the novelty and inventive step of the subject invention indicated by the analyzed claim case. The positive case information may be generated based on, for example, historical information of published patent documents or may include information generated for training. The large-scale language model LLM6 may also be a general-purpose large-scale language model. In this case, a positive opinion may be output by in-context learning, in which the large-scale language model LLM6 is input with the above-mentioned positive case information in addition to the differences and similarities between the claims being analyzed and the prior art documents.

[0048] The large-scale language model LLM7 is an example of a second large-scale language model used to generate a negative opinion. For example, when the large-scale language model LLM7 receives the differences and commonalities between the claims to be analyzed and the prior art documents, it outputs a negative opinion regarding the novelty and inventive step of the subject invention indicated by the claims to be analyzed. The negative opinion may include an opinion that the invention lacks novelty and inventive step and the reasons for that opinion. Alternatively, the negative opinion may include an opinion that the invention is novel but lacks inventive step and the reasons for that opinion. For example, the large-scale language model LLM7 may be a general-purpose large-scale language model fine-tuned using negative case information. The negative case information includes cases of differences and commonalities between the claims to be analyzed and the prior art documents, as well as cases of negative opinions regarding the novelty and inventive step of the claims to be analyzed. The negative case information may be generated based on, for example, historical information from published patent documents or may include information generated for training. Alternatively, the large-scale language model LLM7 may be a general-purpose large-scale language model. In this case, a negative opinion may be output by in-context learning, in which the large-scale language model LLM7 is input with the above-mentioned negative case information as well as the differences and similarities between the claims being analyzed and the prior art documents.

[0049] The large-scale language model LLM8 is an example of a third large-scale language model used to generate a conclusion. For example, the large-scale language model LLM8 receives the claims to be analyzed, prior art documents, affirmative opinions, and negative opinions, and outputs a conclusion. The conclusion includes whether the affirmative or negative opinion is appropriate and the reasons for the conclusion. For example, the large-scale language model LLM8 may be a general-purpose large-scale language model fine-tuned using case information for the conclusion. As described above, the case information for the conclusion may include, for example, business information related to the business related to the intellectual property to be analyzed and / or a patent portfolio related to the intellectual property to be analyzed. This allows a conclusion to be generated taking into account the business information and / or the patent portfolio. The case information for the conclusion includes cases for the claims to be analyzed, cases for the prior art documents, cases for affirmative opinions, cases for negative opinions, and cases for the conclusion. The case information for the conclusion may be generated based on, for example, historical information on published patent documents, or may include information generated for training. Furthermore, for example, the positive opinion cases and negative opinion cases in the case information for the conclusion may be generated using the large-scale language models LLM6 and LLM7 with the claims to be analyzed and the prior art documents as input. The large-scale language model LLM8 may be a general-purpose large-scale language model. In this case, the conclusion may be output by in-context learning, in which the claims to be analyzed, the prior art documents, the positive opinions, and the negative opinions as well as the above-mentioned case information for the conclusion are input to the large-scale language model LLM8.

[0050] The large-scale language model LLM9 is an example of a fourth large-scale language model used to generate improvement proposal information. For example, the large-scale language model LLM9 outputs improvement proposals for the analyzed claims when it receives input of the claims to be analyzed, prior art documents, differences and commonalities, positive opinions, negative opinions, and conclusions. For example, the large-scale language model LLM9 may be a general-purpose large-scale language model fine-tuned using improvement case information. The improvement case information may include, for example, business information related to the business related to the analyzed intellectual property, a patent portfolio related to the analyzed intellectual property, and / or strategic information related to the patent portfolio. This allows important components of the business information, patent portfolio, and / or strategic information to be included in the improvement proposal information. The improvement case information may also include examples of the claims to be analyzed, examples of prior art documents, examples of differences and commonalities, examples of positive opinions, examples of negative opinions, examples of conclusions, and examples of improvement proposals. The improvement case information may be generated based on, for example, historical information on published patent documents. For example, the improvement proposal examples may be generated based on amendments to claims in the historical information. The improvement proposal example information may also include information generated for training. The large-scale language model LLM9 may be a general-purpose large-scale language model. In this case, the large-scale language model LLM9 may output improvement proposals through in-context learning, where the claims to be analyzed, prior art documents, differences and commonalities, positive opinions, negative opinions, and conclusions, as well as the improvement proposal example information, are input.

[0051] (Database 3) The database 3 stores search targets for related technical information. For example, the database 3 may store multiple patent documents as search targets. Furthermore, for example, the database 3 may store each of the multiple patent documents in a manner that enables searching based on the similarity of features. For example, each patent document may be associated with a vector representation indicating the features of the patent document. The vector representation may be information in which at least a portion of each patent document (e.g., claims and abstracts) is converted into a vector format using an embedding model. In this case, each patent document is indexed using the vector format. The processes of conversion to vector format and indexing may be performed in advance by the related technology acquisition unit 11 or by a device external to the information processing device 1A. The search targets for related technical information stored in the database 3 are not limited to patent documents, but may also include non-patent documents.

[0052] (Input device 4 and display device 5) The input device 4 is configured to receive input to the information processing device 1A, and may include, for example, input devices such as a keyboard, a mouse, a touch panel, a camera, and a microphone. The display device 5 is configured to display a screen output from the information processing device 1A, and may include, for example, a display. The input device 4 and the display device 5 may also be formed integrally as a touch panel or the like.

[0053] (Configuration of information processing device 1A) 4, information processing device 1A includes a control unit 110 and a storage unit 120. Control unit 110 controls each unit of information processing device 1A in an integrated manner. Storage unit 120 stores various data and programs referenced by control unit 110.

[0054] The control unit 110 includes an improvement unit 15, a dual opinion presentation unit 16, a final proposal presentation unit 17, an analysis target acquisition unit 18, and a difference generation unit 19 in addition to the related technology acquisition unit 11, the positive opinion generation unit 12, the negative opinion generation unit 13, and the conclusion generation unit 14 provided in the information processing device 1. The improvement unit 15 is an example of a configuration that realizes an improvement means. The dual opinion presentation unit 16 is an example of a configuration that realizes a dual opinion presentation means. The final proposal presentation unit 17 is an example of a configuration that realizes a final proposal presentation means.

[0055] The analysis target acquisition unit 18 acquires a claim (an example of analysis target information) to be analyzed. The claim to be analyzed may be acquired based on a user's operation using the input device 4, for example.

[0056] The related art acquisition unit 11 is configured similarly to the first exemplary embodiment, and is also configured as follows. The related art acquisition unit 11 may use the large-scale language model LLM1 or LLM2 to generate keywords or summaries from claims to be analyzed (an example of information to be analyzed), and may use the generated keywords or summaries to acquire prior art documents (an example of related art information) from the database 3. The related art acquisition unit 11 may also acquire multiple candidates for prior art documents (an example of related art information), and use the large-scale language model LLM4 to select one of the multiple candidates as the prior art document (an example of related art information). The related art acquisition unit 11 may also use the large-scale language model LLM3 to identify a classification of the invention to be analyzed (an example of intellectual property to be analyzed), and use the identified classification to acquire prior art documents (an example of related art information).

[0057] For example, the related technology acquisition unit 11 includes a keyword generation unit 111 , a summary generation unit 112 , a category specification unit 113 , a candidate acquisition unit 114 , and a related technology selection unit 115 .

[0058] The keyword generation unit 111 generates keywords from the claims to be analyzed using the large-scale language model LLM1. The details of the large-scale language model LLM1 have been described above. This makes it possible to acquire multiple candidates of prior art documents from the database 3 based on the similarity to the keywords. Note that the keyword generation unit 111 may acquire keywords input by a user instead of or in addition to generating keywords from the claims to be analyzed using the large-scale language model LLM1.

[0059] The abstract generation unit 112 generates an abstract from the claim to be analyzed using the large-scale language model LLM2. The details of the large-scale language model LLM2 are as described above. This makes it possible to retrieve multiple candidates of prior art documents from the database 3 based on the similarity with the abstract.

[0060] In addition, the user may be able to select whether to use keywords or abstracts to obtain multiple candidates for prior art documents. Furthermore, if the user has not selected whether to use keywords or abstracts, a pre-defined one (e.g., keywords) may be used, and the other (e.g., abstract) may be optionally selectable based on the user's operation. Furthermore, whether to use keywords or abstracts may be selected based on predetermined conditions, independent of the user's operation. For example, if the length of the input claim to be analyzed is equal to or greater than a threshold, a summary may be generated, and if not, keywords may be generated.

[0061] The classification identification unit 113 identifies the classification of the claim to be analyzed using the large-scale language model LLM3. The details of the large-scale language model LLM3 are as described above. This makes it possible to narrow down multiple candidates of prior art documents based on classification.

[0062] The candidate acquisition unit 114 acquires multiple candidates for prior art documents from the database 3 using the generated keywords or abstracts and the identified classification. For example, the candidate acquisition unit 114 generates a vector representation indicating the characteristics of the generated keywords or abstracts using an embedding model. The candidate acquisition unit 114 also identifies multiple patent documents stored in the database 3, for which the similarity between the vector representation of the keywords or abstracts and the vector representation of the patent documents is equal to or greater than a threshold. The candidate acquisition unit 114 also acquires, from the multiple patent documents, those that match the identified classification as multiple candidates for prior art documents. This makes it possible to acquire patent documents that are more appropriate as multiple candidates for prior art documents from the database 3 than when simply using keywords or abstracts. Note that if the candidate acquisition unit 114 acquires only one candidate (for example, one patent document whose similarity with the keywords or abstract is equal to or greater than a threshold, or one patent document that matches the identified classification), the processing by the related technology selection unit 115, described below, can be omitted.

[0063] The related art selection unit 115 uses the large-scale language model LLM4 to select one of multiple candidates for prior art documents as a prior art document. The number of selected prior art documents may be one or multiple. The details of the large-scale language model LLM4 have been described above. This allows for identifying a prior art document appropriate for comparison with the claim being analyzed from multiple candidates for prior art documents acquired based on similarity with keywords or abstracts and classification. The related art selection unit 115 may select a candidate selected by a user as a prior art document. Furthermore, when a user instructs a computer to select a prior art document, the related art selection unit 115 may select a prior art document from multiple candidates using the large-scale language model LLM4.

[0064] The difference generation unit 19 generates differences and commonalities between the claims to be analyzed and the prior art documents using the large-scale language model LLM5. Details of the large-scale language model LLM5 are as described above.

[0065] The positive opinion generation unit 12 is configured similarly to the first exemplary embodiment, and is further configured as follows: The positive opinion generation unit 12 uses the large-scale language model LLM6 to generate a positive opinion regarding the novelty and inventive step of the analyzed invention based on the differences and commonalities between the analyzed claims and the prior art documents. The large-scale language model LLM6 has been described above in detail.

[0066] The negative opinion generation unit 13 is configured similarly to the first exemplary embodiment, and is further configured as follows: The negative opinion generation unit 13 uses the large-scale language model LLM7 to generate a negative opinion regarding the novelty and inventive step of the invention to be branched, based on the differences and commonalities between the claims to be analyzed and the prior art documents. The large-scale language model LLM7 has been described above in detail.

[0067] The conclusion generation unit 14 is configured similarly to the first exemplary embodiment, but is also configured as follows: The conclusion generation unit 14 uses the large-scale language model LLM8 to generate a conclusion regarding the novelty and inventive step of the subject invention based on the analyzed claims, prior art documents, positive opinions, and negative opinions.

[0068] The improvement unit 15 uses the large-scale language model LLM9 to generate a claim improvement proposal (an example of improvement proposal information) that indicates an improvement proposal for the claim being analyzed (an example of analysis target information). Details of the large-scale language model LLM9 have been described above. The related art acquisition unit 11, the affirmative opinion generation unit 12, the negative opinion generation unit 13, and the conclusion generation unit 14 operate again with the claim improvement proposal as the new claim being analyzed. For example, the improvement unit 15 may generate a claim improvement proposal when the conclusion does not satisfy the predetermined conditions described below. Furthermore, for example, the improvement unit 15 may generate a claim improvement proposal when the user instructs a claim improvement, regardless of whether the conclusion satisfies the predetermined conditions. This recursively repeats the generation of claim improvement proposals, allowing claim improvement proposals to be created while gradually increasing the likelihood that novelty and inventive step will be affirmed.

[0069] The two-opinion presenting unit 16 presents the positive and negative opinions to the user, which increases the user's sense of satisfaction with the conclusion compared to when the conclusion is simply presented.

[0070] The final proposal presenting unit 17 presents the claim improvement proposal (an example of improvement proposal information) to the user as a final claim proposal (an example of final proposal information) when a conclusion generated using the claim improvement proposal as the target claim satisfies a predetermined condition. The predetermined condition may be, for example, that a positive opinion regarding novelty and inventive step is deemed appropriate. The predetermined condition may also be, for example, that a positive opinion regarding at least novelty is deemed appropriate. In this case, in other words, the predetermined condition may be satisfied when, for example, a negative opinion that the claim has novelty but not inventive step is deemed appropriate. However, the predetermined condition is not limited to this. This makes it possible to present to the user a final claim proposal that is more likely to be affirmed for novelty and inventive step, thereby improving the user's sense of satisfaction. (Flow of information processing method S1A) The information processing device 1A configured as above executes an information processing method S1A. Fig. 5 is a flow diagram showing the flow of the information processing method S1A. As shown in Fig. 5, the information processing method S1A includes steps S101 to S113.

[0071] In step S101, the analysis target acquisition unit 18 acquires a claim to be analyzed.

[0072] FIG. 6 is a diagram schematically illustrating an example of an analysis target input screen displayed on the display device 5 in step S101. As shown in FIG. 6, the screen example G1 includes a claim input area G11 and an operation object G12. The claim input area G11 accepts input of a natural language sentence indicating a claim to be analyzed. The input natural language sentence is displayed in the claim input area G11. Note that the screen example G1 shows an example in which one claim is input, but multiple claims may also be input. The multiple claims may be in a parallel relationship or may be in a citation relationship. For example, when an operation on the operation object G12 is accepted, the next step S102 is executed.

[0073] Steps S102 to S104 are an example of related art acquisition processing. In step S102, the related art acquisition unit 11 generates keywords or abstracts from the claims to be analyzed. For example, when generating keywords, the keyword generation unit 111 generates keywords from the claims to be analyzed using a large-scale language model LLM1. Also, for example, when generating abstracts, the summary generation unit 112 generates abstracts from the claims to be analyzed using a large-scale language model LLM2. Whether keywords or abstracts are generated is as described above, and therefore a detailed description will not be repeated. Screen example G1 shows an example in which keywords are generated using a natural language sentence entered in the claim input area G11 as the claim to be analyzed.

[0074] FIG. 7 is a diagram illustrating an example of a generated keyword screen displayed on the display device 5 in step S102. As shown in FIG. 7, the example screen G2 includes a keyword area G21, a search count setting area G22, a search target setting area G23, and an operation object G24. The keyword area G21 indicates keywords generated from the claims to be analyzed. In this example, four keywords are generated. The search count setting area G22 accepts an operation to set the number of prior art document candidates to be acquired. In this example, the number is set to five. That is, five candidates are acquired as prior art document candidates. The search target setting area G23 accepts an operation to set a search target for searching the prior art document candidates. In this example, it is possible to select whether to set patent documents recorded in the database 3 as search targets by disclosure year, and 2022 and 2023 are set as search targets. For example, when an operation on the operation object G24 is accepted, the following steps S103 to S105 are executed. If a summary is generated instead of the keywords, the example screen G2 includes an area where the summary is displayed instead of the keyword area G21.

[0075] In step S103, the category identification unit 113 uses the large-scale language model LLM3 to identify the category of the invention indicated by the claim to be analyzed.

[0076] In step S104, the candidate acquisition unit 114 acquires a plurality of patent documents from the database 3 based on the similarity with the keywords or abstracts. The candidate acquisition unit 114 also acquires, from the plurality of patent documents, those that match the specified classification as a plurality of candidates for prior art documents. Note that if only one candidate is acquired in step S104, the candidate is considered to be the prior art document, and the next step S105 is omitted.

[0077] In step S105, the related technology selection unit 115 selects prior art documents from among multiple candidates using the large-scale language model LLM4. As described above, the related technology selection unit 115 may select prior art documents from among multiple candidates based on a user operation instead of using the large-scale language model LLM4. Prior art documents not selected from among the multiple candidate prior art documents may be used for further fine-tuning of some or all of the large-scale language models LLM1 to LLM9, in-context learning, or the like.

[0078] FIG. 8 is a diagram schematically illustrating an example of a prior art screen displayed on the display device 5 in step S105. As shown in FIG. 8, the screen example G3 includes a prior art document area G31 and an operation object G32. The prior art document area G31 shows an outline of the prior art document selected by the related art selection unit 115. In the screen example G3, bibliographic items are displayed as the outline, but the prior art document area G31 may also include other information (e.g., an abstract, an independent claim, etc.). For example, when an operation on the operation object G32 is accepted, the next step S106 is executed.

[0079] In step S106, the difference generation unit 19 generates differences and commonalities between the claims to be analyzed and the prior art documents using the large-scale language model LLM 5. The difference generation unit 19 also presents the generated differences and commonalities to the user, for example, by displaying them on the display device 5. Note that when multiple claims are input as claims to be analyzed, the difference generation unit 19 may generate differences and commonalities for each claim.

[0080] 9 is a diagram schematically illustrating an example of a difference screen displayed on the display device 5 in step S106. As shown in FIG. 9, the example screen G4 includes a common point area G41, a difference area G42, and an operation object G43. The common point area G41 includes common points between the claim to be analyzed and the prior art document. The difference area G42 includes differences between the claim to be analyzed and the prior art document. For example, when an operation on the operation object G43 is accepted, the following steps S107 to S109 are executed.

[0081] Step S107 is an example of a positive opinion generation process. In step S107, the positive opinion generation unit 12 generates a positive opinion regarding the novelty and inventive step of the subject invention indicated by the claim to be analyzed, based on the claim to be analyzed and prior art documents, using the large-scale language model LLM6. Note that when multiple claims are input as the claims to be analyzed, the positive opinion generation unit 12 may generate a positive opinion for each claim.

[0082] Step S108 is an example of a negative opinion generation process. In step S108, the negative opinion generation unit 13 generates a negative opinion regarding the novelty and inventive step of the subject invention indicated by the claim to be analyzed, based on the claim to be analyzed and prior art documents, using the large-scale language model LLM7. Note that when multiple claims are input as the claims to be analyzed, the negative opinion generation unit 13 may generate a negative opinion for each claim.

[0083] The execution order of steps S107 and S108 is not limited to the above-mentioned order, and they may be executed in the reverse order, or some or all of them may be executed in parallel.

[0084] Step S109 is an example of a two-opinion presentation process. In step S109, the two-opinion presentation unit 16 presents the positive opinion and the negative opinion to the user by displaying them on the display device 5, for example.

[0085] FIG. 10 is a diagram illustrating an example of a screen displaying both opinions displayed on the display device 5 in step S109. As shown in FIG. 10, the screen example G9 includes a positive opinion area G91 and operation objects G92 and G93. The positive opinion area G91 includes a positive opinion regarding novelty and a positive opinion regarding inventive step. The positive opinion includes a sentence indicating the opinion, "The subject invention has novelty," and a sentence indicating the basis for that opinion, "...(omitted)...not disclosed." The positive opinion also includes a sentence indicating the opinion, "The subject invention has inventive step," and a sentence indicating the basis for that opinion, "...(omitted)...not easily conceivable." Note that if multiple claims are input as claims to be analyzed, the positive opinions may be displayed categorized by claim. The operation object G92 accepts an operation to display the negative opinion. When an operation on the operation object G92 is accepted, the screen example G9 transitions to a screen example G10, which will be described next.

[0086] FIG. 11 is a diagram illustrating another example of the two-opinion screen displayed on the display device 5 in step S109. As shown in FIG. 11, the screen example G10 includes a negative opinion area G101 and operation objects G102 and G93. The negative opinion area G101 includes a negative opinion regarding novelty and a negative opinion regarding inventive step. The negative opinion includes a sentence indicating the opinion that "the subject invention lacks novelty" and a sentence indicating the basis for that opinion, "Element A: ~ (omitted) ~...." The negative opinion also includes a sentence indicating the opinion that "the subject invention lacks inventive step" and a sentence indicating the basis for that opinion, "It is considered that A and B are ~ (omitted) ~." Note that if multiple claims are input as claims to be analyzed, the negative opinions may be categorized and displayed by claim. The operation object G102 accepts an operation to display the affirmative opinion. When an operation on the operation object G102 is accepted, the screen example G10 transitions to the above-described screen example G9.

[0087] In this way, by being able to switch between screen examples G9 and G10, the user can confirm both the affirmative and negative opinions regarding the novelty and inventive step of the subject invention. Note that instead of switching between the affirmative and negative opinions as in screen examples G9 and G10, both opinions may be displayed on a single screen.

[0088] When an operation on the operation object G93 is accepted in the example screen G9 or G10, the next step S110 is executed.

[0089] Step S110 is an example of a conclusion generation process. In step S110, the conclusion generation unit 14 uses the large-scale language model LLM8 to generate a conclusion regarding the novelty and inventive step of the subject invention indicated by the analyzed claim based on the analyzed claim, prior art documents, positive opinions, and negative opinions. The conclusion generation unit 14 also presents the generated conclusion to the user by, for example, displaying it on the display device 5.

[0090] In step S111, the control unit 110 determines whether the conclusion satisfies a predetermined condition. As described above, the predetermined condition may be that a positive opinion is deemed appropriate with respect to novelty and inventive step, or that a positive opinion is deemed appropriate with respect to at least novelty. The case where a "Yes" decision is made in step S111 will be described later. If a "No" decision is made in step S111, the next step S112 is executed.

[0091] Step S112 is an example of the improvement process. In step S112, the improvement unit 15 generates a claim improvement proposal based on the analyzed claim, prior art documents, differences and commonalities, negative opinions, positive opinions, and conclusions using the large-scale language model LLM9. The improvement unit 15 then presents the generated claim improvement proposal to the user, for example, by displaying it on the display device 5.

[0092] Next, the control unit 110 repeats the process from step S102, using the claim improvement proposal as a new claim to be analyzed.

[0093] If the determination in step S111 is Yes, step S113 is executed. In step S113, the final proposal presenting unit 17 presents the latest claim improvement proposal, which is the claim under analysis whose conclusion satisfies the predetermined condition, to the user as the claim final proposal, for example, by displaying it on the display device 5.

[0094] 12 is a diagram schematically illustrating an example of a conclusion screen that displays a conclusion determined to satisfy a predetermined condition in step S111. Note that this conclusion screen is also an example of a conclusion screen displayed on the display device 5 in step S110. As shown in FIG. 12, the example screen G13 includes a conclusion area G131 and a rationale area G132. In this example, the conclusion area G131 includes a conclusion that the positive opinion is valid. The rationale area G132 includes text that shows the rationale.

[0095] 13 is a diagram showing an example of a final proposal screen displayed on the display device 5 in step S113. As shown in FIG. 13, the example screen G14 includes a final proposal area G141. In the final proposal area G141, changes to the claims that were originally the subject of analysis are displayed in a recognizable display format (in this example, underlined display format). Note that the recognizable display format for the changes is not limited to underlined display format, and may also be, but is not limited to, a highlighted display format.

[0096] (Effects of information processing device 1A and information processing method S1A) As described above, the information processing device 1A further includes an improvement unit 15 that generates improvement suggestion information indicating improvement suggestions for the information to be analyzed using the fourth large-scale language model, and the related technology acquisition unit 11, the positive opinion generation unit 12, the negative opinion generation unit 13, the conclusion generation unit 14, and the improvement unit 15 are configured to further function using the improvement suggestion information as the information to be analyzed. Therefore, according to the information processing device 1A, in addition to the effects achieved by the information processing device 1, the repeated generation of improvement suggestion information can gradually increase the possibility of obtaining rights to the intellectual property indicated by the information to be analyzed.

[0097] Furthermore, in the information processing device 1A, the related technology acquisition unit 11 is configured to generate keywords from the information to be analyzed using a fifth large-scale language model, and acquire related technology information from a database using the generated keywords. Therefore, in addition to the effects of the information processing device 1, the information processing device 1A can also acquire more appropriate related technology information.

[0098] Furthermore, in the information processing device 1A, the related technology acquisition unit 11 acquires a plurality of candidates for related technology information and selects one of the plurality of candidates as the related technology information using a sixth large-scale language model. Therefore, in addition to the effect achieved by the information processing device 1, the information processing device 1A can obtain an effect of being able to acquire more appropriate related technology information.

[0099] Furthermore, in the information processing device 1A, the related technology acquisition unit 11 is configured to identify the classification of the intellectual property to be analyzed using the seventh large-scale language model and acquire related technology information using the identified classification, thereby achieving the effect of acquiring more appropriate related technology information.

[0100] Furthermore, the information processing device 1A employs a configuration further including an opinion presenting unit 16 that presents both positive and negative opinions to the user. Therefore, in addition to the effects achieved by the information processing device 1, the information processing device 1A has the effect of further increasing the user's sense of satisfaction with the conclusion by allowing the user to recognize both positive and negative opinions.

[0101] Furthermore, the information processing device 1A employs a configuration further including a final proposal presenting unit 17 that presents improvement proposal information to a user as final proposal information when a conclusion generated using the improvement proposal information as analysis target information satisfies a predetermined condition. Therefore, in addition to the effects of the information processing device 1, the information processing device 1A can provide the user with an effect of being able to provide improvement proposals for claims to increase the possibility of obtaining rights.

[0102] (Variation) In the second exemplary embodiment, the intellectual property to be analyzed is not limited to inventions. For example, other intellectual property described in natural language sentences, such as ideas, may be applied as the intellectual property to be analyzed. Furthermore, the possibility of obtaining rights is not limited to both novelty and inventive step, but either one may be applied. Furthermore, other requirements may be applied to the possibility of obtaining rights in addition to novelty and / or inventive step. Furthermore, the information to be analyzed is not limited to claims. For example, information including summaries, detailed descriptions, ideas, etc. may be applied as the information to be analyzed instead of or in addition to claims. Furthermore, non-patent literature may be applied as the related technical information instead of patent literature. Furthermore, the number of prior art documents is not limited to one, but may be multiple.

[0103] [Software implementation example] Some or all of the functions of the information processing devices 1, 1A and each device constituting the information processing system 100A (hereinafter also referred to as "each of the above-mentioned devices") may be realized by hardware such as an integrated circuit (IC chip), or by software.

[0104] In the latter case, each of the above devices is realized by, for example, a computer that executes instructions of a program, which is software that realizes each function. An example of such a computer (hereinafter referred to as computer C) is shown in Figure 14. Figure 14 is a block diagram showing the hardware configuration of computer C that functions as each of the above devices.

[0105] The computer C includes at least one processor C1 and at least one memory C2. The memory C2 stores a program P for causing the computer C to operate as each of the above-mentioned devices. In the computer C, the processor C1 reads and executes the program P from the memory C2, thereby realizing the functions of each of the above-mentioned devices.

[0106] The processor C1 may be, for example, a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, a microcontroller, or a combination thereof. The memory C2 may be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0107] The computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and for temporarily storing various data. The computer C may also include a communication interface for transmitting and receiving data to and from other devices. The computer C may also include an input / output interface for connecting input / output devices such as a keyboard, mouse, display, and printer.

[0108] Furthermore, the program P can be recorded on a non-transitory tangible recording medium M that can be read by the computer C. Such a recording medium M can be, for example, a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer C can acquire the program P via such a recording medium M. The program P can also be transmitted via a transmission medium. Such a transmission medium can be, for example, a communication network or broadcast waves. The computer C can also acquire the program P via such a transmission medium.

[0109] Furthermore, the functions of each of the devices may be realized by a single processor provided in a single computer, by multiple processors provided in a single computer working in cooperation, or by multiple processors provided in each of multiple computers working in cooperation. Furthermore, the programs for causing each of the devices to realize the functions may be stored in a single memory provided in a single computer, or may be distributed and stored in multiple memories provided in a single computer, or may be distributed and stored in multiple memories provided in each of multiple computers.

[0110] [Appendix A] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0111] (Appendix A1) related technology acquisition means for acquiring related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence; a positive opinion generating means for generating a positive opinion regarding the possibility of obtaining rights for the intellectual property of the analysis target based on the analysis target information and the related technical information by using a first large-scale language model; a negative opinion generating means for generating a negative opinion regarding the possibility of obtaining the right based on the analysis target information and the related technical information by using a second large-scale language model; a conclusion generating means for generating a conclusion regarding the possibility of obtaining the right based on the affirmative opinion and the negative opinion by using a third large-scale language model; An information processing device comprising:

[0112] (Appendix A2) further comprising an improvement means for generating improvement proposal information indicating an improvement proposal for the analysis target information using a fourth large-scale language model; the related technology acquisition means, the positive opinion generation means, the negative opinion generation means, the conclusion generation means, and the improvement means further function as the improvement plan information as the analysis target information; 10. The information processing device according to claim 1,

[0113] (Appendix A3) the related technology acquisition means generates keywords or summaries from the analysis target information using a fifth large-scale language model, and acquires the related technology information from a database using the generated keywords or summaries; An information processing device according to appendix A1 or A2.

[0114] (Appendix A4) the related technology acquisition means acquires a plurality of candidates for the related technology information, and selects one of the plurality of candidates as the related technology information using a sixth large-scale language model; An information processing device according to any one of appendices A1 to A3.

[0115] (Appendix A5) the related technology acquisition means identifies a classification of the intellectual property to be analyzed using a seventh large-scale language model, and acquires the related technology information using the identified classification; 10. The information processing device according to claim 1, wherein the first and second information processing units are connected to each other.

[0116] (Appendix A6) further comprising opinion presenting means for presenting the positive opinion and the negative opinion to a user; An information processing device according to any one of appendices A1 to A5.

[0117] (Appendix A7) and a final proposal presenting means for presenting the improvement proposal information to a user as final proposal information when the conclusion generated using the improvement proposal information as the analysis target information satisfies a predetermined condition. 10. The information processing device according to claim 9, wherein the information processing device is a

[0118] [Appendix B] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0119] (Appendix B1) a related technology acquisition process in which at least one processor acquires related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence; a positive opinion generation process in which the at least one processor uses a first large-scale language model to generate a positive opinion regarding the possibility of obtaining rights to the intellectual property of the analysis target based on the analysis target information and the related technical information; a negative opinion generation process in which the at least one processor generates a negative opinion regarding the possibility of obtaining the right based on the analysis target information and the related technical information using a second large-scale language model; a conclusion generation process in which the at least one processor generates a conclusion regarding the possibility of obtaining the right based on the positive opinions and the negative opinions using a third large-scale language model; An information processing method comprising:

[0120] (Appendix B2) The at least one processor further includes an improvement process for generating improvement suggestion information indicating an improvement suggestion for the analysis target information using a fourth large-scale language model; the at least one processor further functions such that the related technology acquisition process, the positive opinion generation process, the negative opinion generation process, the conclusion generation process, and the improvement process use the improvement proposal information as the analysis target information; 1. The information processing method described in Appendix B1.

[0121] (Appendix B3) In the related technology acquisition process, the at least one processor generates keywords or summaries from the analysis target information using a fifth large-scale language model, and acquires the related technology information from a database using the generated keywords or summaries. 1. An information processing method according to Appendix B1 or B2.

[0122] (Appendix B4) In the related technology acquisition process, the at least one processor acquires a plurality of candidates of the related technology information, and selects one of the plurality of candidates as the related technology information using a sixth large-scale language model. 1. An information processing method according to any one of appendices B1 to B3.

[0123] (Appendix B5) In the related technology acquisition process, the at least one processor identifies a classification of the intellectual property to be analyzed using a seventh large-scale language model, and acquires the related technology information using the identified classification. 1. An information processing method according to any one of appendices B1 to B4.

[0124] (Appendix B6) The at least one processor further includes a process for presenting both the positive opinion and the negative opinion to a user. 1. An information processing method according to any one of Appendices B1 to B5.

[0125] (Appendix B7) The method further includes a final proposal presentation process, performed by the at least one processor, for presenting the improvement proposal information to a user as final proposal information when the conclusion generated using the improvement proposal information as the analysis target information satisfies a predetermined condition. 1. The information processing method described in Appendix B2.

[0126] [Appendix C] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0127] (Appendix C1) A program that causes a computer to function as an information processing device, The computer related technology acquisition means for acquiring related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence; a positive opinion generating means for generating a positive opinion regarding the possibility of obtaining rights for the intellectual property of the analysis target based on the analysis target information and the related technical information by using a first large-scale language model; a negative opinion generating means for generating a negative opinion regarding the possibility of obtaining the right based on the analysis target information and the related technical information by using a second large-scale language model; a conclusion generating means for generating a conclusion regarding the possibility of obtaining the right based on the affirmative opinion and the negative opinion by using a third large-scale language model; An information processing program that functions as a

[0128] (Appendix C2) The computer further functioning as an improvement means for generating improvement proposal information indicating an improvement proposal for the analysis target information using a fourth large-scale language model; the related technology acquisition means, the positive opinion generation means, the negative opinion generation means, the conclusion generation means, and the improvement means further function as the improvement plan information as the analysis target information; An information processing program as described in Appendix C1.

[0129] (Appendix C3) the related technology acquisition means generates keywords or summaries from the analysis target information using a fifth large-scale language model, and acquires the related technology information from a database using the generated keywords or summaries; An information processing program according to appendix C1 or C2.

[0130] (Appendix C4) the related technology acquisition means acquires a plurality of candidates for the related technology information, and selects one of the plurality of candidates as the related technology information using a sixth large-scale language model; An information processing program according to any one of appendices C1 to C3.

[0131] (Appendix C5) the related technology acquisition means identifies a classification of the intellectual property to be analyzed using a seventh large-scale language model, and acquires the related technology information using the identified classification; An information processing program according to any one of appendices C1 to C4.

[0132] (Appendix C6) The computer and further functioning as an opinion presenting means for presenting the positive opinion and the negative opinion to a user. An information processing program according to any one of appendices C1 to C5.

[0133] (Appendix C7) The computer and further functioning as a final proposal presenting means for presenting the improvement proposal information to a user as final proposal information when the conclusion generated using the improvement proposal information as the analysis target information satisfies a predetermined condition. An information processing program as described in Appendix C2.

[0134] [Appendix D] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0135] (Appendix D1) at least one processor, a related technology acquisition process for acquiring related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in natural language sentences; a positive opinion generation process for generating a positive opinion regarding the possibility of obtaining rights for the intellectual property of the analysis target based on the analysis target information and the related technical information using a first large-scale language model; a negative opinion generation process for generating a negative opinion regarding the possibility of obtaining the right based on the analysis target information and the related technical information using a second large-scale language model; a conclusion generation process for generating a conclusion regarding the possibility of obtaining the right based on the affirmative opinions and the negative opinions using a third large-scale language model; An information processing device that executes the above.

[0136] The information processing device may further include a memory, and the memory may store a program for causing the at least one processor to execute each of the processes.

[0137] (Appendix D2) the at least one processor: further performing an improvement process using a fourth large-scale language model to generate improvement proposal information indicating an improvement proposal for the analysis target information; the related technology acquisition process, the positive opinion generation process, the negative opinion generation process, the conclusion generation process, and the improvement process further function using the improvement plan information as the analysis target information; 10. The information processing device according to claim 9, wherein the information processing device is an information processing device according to claim 1.

[0138] (Appendix D3) In the related technology acquisition process, the at least one processor generates keywords or summaries from the analysis target information using a fifth large-scale language model, and acquires the related technology information from a database using the generated keywords or summaries. An information processing device according to appendix D1 or D2.

[0139] (Appendix D4) In the related technology acquisition process, the at least one processor acquires a plurality of candidates of the related technology information, and selects one of the plurality of candidates as the related technology information using a sixth large-scale language model. An information processing device according to any one of appendices D1 to D3.

[0140] (Appendix D5) In the related technology acquisition process, the at least one processor identifies a classification of the intellectual property to be analyzed using a seventh large-scale language model, and acquires the related technology information using the identified classification. An information processing device according to any one of appendices D1 to D4.

[0141] (Appendix D6) the at least one processor: further performing a process of presenting both the positive opinion and the negative opinion to a user; An information processing device according to any one of appendices D1 to D5.

[0142] (Appendix D7) the at least one processor: and further executing a final proposal presentation process for presenting the improvement proposal information to a user as final proposal information when the conclusion generated using the improvement proposal information as the analysis target information satisfies a predetermined condition. 10. The information processing device according to claim 9, wherein the information processing device is an information processing device according to claim 1, wherein

[0143] [Appendix E] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0144] (Appendix E1) A program that causes a computer to function as an information processing device, The computer, a related technology acquisition process for acquiring related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in natural language sentences; a positive opinion generation process for generating a positive opinion regarding the possibility of obtaining rights for the intellectual property of the analysis target based on the analysis target information and the related technical information using a first large-scale language model; a negative opinion generation process for generating a negative opinion regarding the possibility of obtaining the right based on the analysis target information and the related technical information using a second large-scale language model; a conclusion generation process for generating a conclusion regarding the possibility of obtaining the right based on the affirmative opinions and the negative opinions using a third large-scale language model; A non-transitory recording medium on which an information processing program that causes the program to be executed is recorded. [Explanation of symbols]

[0145] 1, 1A Information processing equipment 100A Information Processing System 2 Large-scale language model storage 3 Database 4 Input Devices 5 Display device 11 Related Technology Acquisition Department 12 Positive opinion generation section 13 Negative opinion generation section 14 Conclusion generation part 15 Improvement Department 16. Opinion Section 17 Final proposal presentation section 18 Analysis target acquisition section 19 Difference generation part 110 control section 111 Keyword Generation Unit 112 Summary generator 113 Classification Specification Department 114 Candidate acquisition section 115 Related Technology Selection Division 120 Storage section C1 processor C2 Memory

Claims

1. related technology acquisition means for acquiring related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence; a positive opinion generating means for generating a positive opinion regarding the possibility of obtaining rights to the intellectual property of the analysis target based on the analysis target information and the related technical information by using a first large-scale language model; a negative opinion generating means for generating a negative opinion regarding the possibility of obtaining the right based on the analysis target information and the related technical information by using a second large-scale language model; a conclusion generating means for generating a conclusion regarding the possibility of obtaining the right based on the affirmative opinion and the negative opinion using a third large-scale language model; It is equipped with The generated conclusion includes a natural language sentence indicating whether the positive opinion or the negative opinion is valid. Information processing device.

2. further comprising an improvement means for generating improvement plan information indicating an improvement plan for the analysis target information using a fourth large-scale language model; the related technology acquisition means, the positive opinion generation means, the negative opinion generation means, the conclusion generation means, and the improvement means further function as the improvement plan information as the analysis target information; The information processing device according to claim 1 .

3. the related technology acquisition means generates keywords or summaries from the analysis target information using a fifth large-scale language model, and acquires the related technology information from a database using the generated keywords or summaries.

3. The information processing device according to claim 1 or 2.

4. the related technology acquisition means acquires a plurality of candidates for the related technology information, and selects one of the plurality of candidates as the related technology information using a sixth large-scale language model; 3. The information processing device according to claim 1 or 2.

5. the related technology acquisition means identifies a classification of the intellectual property to be analyzed using a seventh large-scale language model, and acquires the related technology information using the identified classification; 3. The information processing device according to claim 1 or 2.

6. further comprising opinion presenting means for presenting the positive opinion and the negative opinion to a user; 3. The information processing device according to claim 1 or 2.

7. and a final proposal presenting means for presenting the improvement proposal information to a user as final proposal information when the conclusion generated using the improvement proposal information as the analysis target information satisfies a predetermined condition. The information processing device according to claim 2 .

8. a related technology acquisition process in which at least one processor acquires related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence; a positive opinion generation process in which the at least one processor uses a first large-scale language model to generate a positive opinion regarding the possibility of obtaining rights to the intellectual property of the analysis target based on the analysis target information and the related technical information; a negative opinion generation process in which the at least one processor generates a negative opinion regarding the possibility of obtaining the right based on the analysis target information and the related technical information using a second large-scale language model; a conclusion generation process in which the at least one processor generates a conclusion regarding the possibility of obtaining the right based on the positive opinions and the negative opinions using a third large-scale language model; It contains The generated conclusion includes a natural language sentence indicating whether the positive opinion or the negative opinion is valid. Information processing methods.

9. An information processing program that causes at least one processor to function as an information processing device, related technology acquisition means for acquiring related technology information indicating related technologies related to the intellectual property to be analyzed based on analysis target information in which the intellectual property to be analyzed is described in a natural language sentence; a positive opinion generating means for generating a positive opinion regarding the possibility of obtaining rights to the intellectual property of the analysis target based on the analysis target information and the related technical information by using a first large-scale language model; a negative opinion generating means for generating a negative opinion regarding the possibility of obtaining the right based on the analysis target information and the related technical information by using a second large-scale language model; a conclusion generating means for generating a conclusion regarding the possibility of obtaining the right based on the affirmative opinion and the negative opinion using a third large-scale language model; It functions as The generated conclusion includes a natural language sentence indicating whether the positive opinion or the negative opinion is valid. Information processing program.

Images