Patent search support method and patent search support system

Abstract

A patent search support method according to the present invention is for supporting an infringement search to determine whether or not technology being searched infringes an existing patent, and involves a computer executing: a first accepting step (S5) for accepting search target text information written in the form of a patent claim relating to the technology being searched; a second accepting step (S5) for accepting comparison target text information, which comprises the claims of an existing patent; a determination step (S6 to S7) for determining the possibility that the technology being searched falls within the technical scope of the invention set forth in the existing patent claims by using a generative AI to compare the search target text information with the comparison target text information; and an output step (S9) for outputting the determination result obtained by the determination step.

Description

Patent research support method and patent research support system

[0001] The present invention relates to a method and system for assisting in patent research, and in particular in patent infringement research.

[0002] If a company launches a new product or technology on the market and it is discovered that it infringes on a patent held by another company, the company could suffer significant losses. To avoid such risks, it is important to conduct a patent infringement investigation to determine whether the new product or technology infringes on a patent held by another company at the early stages of new product development or design, or at the latest, before launching the new product or technology on the market.

[0003] A patent infringement investigation is a process of examining whether the structure, configuration, technology, etc. of a new product, etc., fall within the scope of claims of an existing patent. As is well known, there are a huge number of existing patents whose rights remain valid, so a patent infringement investigation usually involves extracting, or narrowing down, potentially infringing patents from a fairly large number of existing patents, although this depends on the technical field, and then conducting a detailed investigation of each patent to determine whether or not there is an infringement.

[0004] Toshiyuki Ando, ​​"Efficient Patent Research Methods Using Machine Learning - Data Science-Based Patent Research for Utilizing AI Research Tools," Japan Patent Information Organization, [Online], [Retrieved July 17, 2025], Internet <URL: https: / / japio.or.jp / 00yearbook / files / 2016book / 16_2_10.pdf>

[0005] When extracting patents in a target technical field from a vast number of patents, the International Patent Classification (IPC) or the Japanese Patent Classification (known as F-terms) is generally used. However, even when using such classifications to narrow down the search, dozens, hundreds, or even thousands of patents remain. Conventional patent infringement investigations involve manual reading and comparison of the claims of each patent with the target technology, written in natural language, to determine whether the target technology is likely to fall within the technical scope of the invention described in the claims of existing patents. However, this process is extremely time-consuming and burdensome, and there has long been a need for improved efficiency.

[0006] With the rapid advancement of artificial intelligence (AI) technology in recent years, AI is being actively introduced into the field of patent work (see, for example, Non-Patent Document 1). For example, systems have been developed that use AI to create keywords for patent searches or to extract patents with high similarity based on invention content entered in free-form natural language. Furthermore, attempts have been made to use more general-purpose generative AI for patent searches, rather than AI using patent-specific models. However, according to the inventors' findings, even when using these conventional methods to narrow down patents for patent infringement investigations, it is difficult to obtain reliable results and they do not contribute to improving work efficiency.

[0007] The present invention has been made in consideration of these problems, and its main purpose is to provide a patent research support method and system that can accurately and efficiently extract existing patents that may or may not be infringed by the technology being researched.

[0008] One aspect of the patent research support method of the present invention is a patent research support method for supporting an infringement investigation to determine whether a technology under investigation infringes an existing patent, and includes using a computer to execute the following steps: a first reception step of receiving text information to be investigated, which is written in the form of a patent claim, for the technology under investigation; a second reception step of receiving comparison text information, which is a claim of an existing patent; a determination step of using a generation AI to compare the text information to be investigated received in the first reception step with the comparison text information received in the second reception step, to determine the possibility that the technology under investigation falls within the technical scope of the invention described in the claim of the existing patent; and an output step of outputting the determination result from the determination step.

[0009] One aspect of the patent research support system of the present invention is a computer-based patent research support system that supports an infringement investigation to determine whether a technology to be investigated infringes an existing patent, and includes: a first memory unit that stores search target text information written in the form of a patent claim for the technology to be investigated; a second memory unit that stores at least patent identifying information and claims for one or more existing patents; a determination unit that acquires the search target text information from the first memory unit and acquires a claim of one existing patent from the second memory unit as comparison target text information, and uses a generation AI to compare the search target text information with the comparison target text information to determine the possibility that the technology to be investigated falls within the technical scope of the invention described in the claim of the existing patent; and an output unit that outputs the determination result by the determination unit.

[0010] According to the present invention, by comparing the search target text information written in patent claim format with the comparison target text information, which is the claims of an existing patent, rather than the search target technology written in natural language, it is possible to determine whether the search target technology is likely to fall within the technical scope of the invention of the existing patent. This highly accurate determination allows for accurate and efficient extraction of existing patents that are likely or highly likely to infringe the search target technology. Furthermore, when rewriting the search target technology written in natural language into patent claim format, variations in terminology and language occur, and the claims of the existing patents themselves also contain variations in terminology and language. However, such variations can be absorbed by the generation AI, resulting in highly accurate determination results that eliminate or reduce the influence of such variations. Furthermore, when comparing the search target technology written in natural language with the claims of the existing patent, the corresponding configurations are clearly defined, resulting in highly accurate determination results.

[0011] In this specification, "text information" refers to a wide range of information, including not only data in text format (with a .txt extension) commonly used in computer systems, but also data in various files containing documents (e.g., .docx files used in Microsoft Word, .xlsx files used in Microsoft Excel, and portable document format (.pdf) files containing text data), as well as image files containing documents as images (e.g., portable document format (.pdf) files containing text images, JPEG files, etc.). Furthermore, document information extracted from such various text information is referred to as text. Because various forms of text information can easily be converted into text using existing OCR (Optical Character Recognition / Reader) software or generation AI, hereinafter, text information and text may be considered the same.

[0012] 1 is a block diagram of an example of a patent research support system according to one embodiment of the present invention; a flowchart showing the steps of work and processing in a patent research support method using the system of this embodiment; an explanatory diagram of a conversion processing work for a technology to be searched; a diagram showing an example of a question in a prompt to be input to a generation AI; a diagram showing an example of an instruction for an output format in a prompt to be input to a generation AI; a diagram showing an example of the relationship between a summary description of a provisional claim to be searched and an independent claim of an existing patent, and a determination result by a generation AI.

[0013] [Supplementary explanation for the above aspect] The "search target technology" in the above aspect of the present invention may include various methods such as manufactured products, products, systems, manufacturing / production methods, work methods, and usage methods, etc. They are not limited to so-called inventions.

[0014] Furthermore, the term "existing patents" as used here includes patents that have been granted rights (i.e., patents for which patent gazettes have been published), as well as patents that have not been granted rights and for which only public publications have been published. In addition to patents, it also includes utility models.

[0015] Additionally, the term "generative AI" as used herein refers to artificial intelligence that generates various types of text based on input data, constraints, instructions, etc. Such generative AI is called text generation AI or generative AI. Currently widely known text generation AIs include "ChatGPT" provided by Open AI, Inc., USA; "Copilot" provided by Microsoft, Inc., USA; "Gemini" provided by Google, Inc., USA; and "Llama" provided by Meta, Inc., USA, but are not limited to these.

[0016] An embodiment of a patent search support method and system according to the present invention will be described in detail below with reference to the drawings. [System Configuration] Fig. 1 is a schematic block diagram of a patent search support system according to this embodiment. Fig. 2 is a flowchart showing an example of the procedure for work and processing in a patent infringement search using this patent search support system.

[0017] As shown in Figure 1, this patent research support system includes a terminal 1, a patent database system 3, and a generation AI system 4, all of which are connected to the Internet 2. The terminal 1 is typically a personal computer (PC), and software (computer program) installed on the PC runs on the PC, and in some cases, also uses functions provided by a system (not shown) present on the Internet 2 to execute processing in each of the functional blocks described below.

[0018] The patent database system 3 may include widely available databases (systems) such as the patent information platform "J-PlatPat" provided by the National Center for Industrial Property Information and Training, a Japanese independent administrative institution, the patent search tool "Patent Public Search" provided by the United States Patent and Trademark Office, and the patent search service "Espacenet" provided by the European Patent Office, as well as various patent information search databases provided by private businesses that are available for a fee, and databases independently accumulated and constructed by users. Furthermore, the generation AI system 4 may be all or part of a system constructed by a business providing a generation AI service such as those exemplified above.

[0019] The generative AI used here is generally called text generation AI. Text generation AI is a model that uses large amounts of web text data available on the Internet as training data and predicts the next word following an input word to complete the sentence. "ChatGPT," one of the well-known GPT (Generative Pre-trained Transformer) algorithms, is a model tuned to output appropriate responses to sentences input by the user.

[0020] Terminal 1 comprises, as functional blocks, a patent search section 10, a provisional claim creation support section 11, an independent claim extraction section 12, a prompt generation support section 13, a comparison processing section 14, a result processing section 15, and a memory section 16, and memory section 16 includes a prompt template storage section 160, a patent list storage section 161, a provisional claim storage section 162, an independent claim storage section 163, a search result storage section 164, etc. An input section 17 and a display section 18 are connected to this terminal 1 as a user interface.

[0021] [Outline of the Work and Processing Procedures for Patent Research] Next, we will briefly explain an example of the work and processing procedures when conducting a patent infringement research using the above system. As an example, the technology to be researched is a document written in natural language that describes the details of a technology or product that a certain company has newly developed or is planning to develop.

[0022] First, the person in charge (user) creates a patent search query based on the target technology to broadly extract patents related to the target technology. The search query typically combines patent classifications, such as IPC and F-terms, with keywords. Then, when the person in charge performs a predetermined operation on the input unit 17, the patent search unit 10 accesses the patent database system 3 via the Internet 2 and displays a search query input screen provided by the system 3 on the display unit 18. When the person in charge inputs the search query on this input screen and instructs the system 3 to execute a search, the patent search unit 10 accesses the patent database system 3 and retrieves patent information matching the search query from the system 3. When searching from the perspective of infringement prevention, hundreds to tens of thousands of patents are retrieved to ensure that high-risk cases are not missed. The patent search unit 10 receives this information, creates a patent list, which is a list of patents that meet the search query, and stores it in the patent list storage unit 161 (Step 1).

[0023] If the database system 3 can be accessed appropriately and the necessary information can be obtained, the patent list created at this point only needs to include information that can identify each patent (typically the patent number), but for accurate determination, it must also include claims. Of course, the patent list may include the full text of the unexamined patent publication (registered publication) of each patent, or the full text information may be linked to the information in the patent list and stored in a location separate from the patent list.

[0024] Next, the person in charge uses the provisional claim drafting support unit 11 to create a provisional claim to be searched, which describes the above-mentioned technology to be searched in a patent claim format, and stores the created provisional claim in the provisional claim storage unit 162 (step 2). As will be described in detail later, in this embodiment, the provisional claim to be searched can be in principle in a two-part format, consisting of a preamble portion that describes elements, etc., corresponding to what is called a "preamble" in a general claim, and a characteristic portion that corresponds to new elements.

[0025] Next, when the person in charge performs a predetermined operation on the input unit 17, the independent claim extraction unit 12 selects one patent from the patent list, reads the entire text of the patent (or the contents of the claims column), and extracts only the independent claims from it in accordance with predetermined rules.The extracted independent claims are then temporarily stored in the independent claim storage unit 163 (step 3).If one patent includes multiple independent claims, all of the multiple independent claims are extracted and stored.

[0026] A method for mechanically extracting independent claims from the claims section can be achieved, for example, by using a text manipulation algorithm to exclude claims that contain phrases indicating dependent claims, such as "as defined in claim X," in the body of the claim. Note that if the researcher limits the scope of the search (e.g., category), it is not necessary to extract all independent claims; only some independent claims, such as apparatuses and methods, may be extracted. In this embodiment, only independent claims are extracted to determine whether the technology under search falls within the technical scope of the inventions described in the independent claims of existing patents. However, independent claims and some or all of the dependent claims may also be extracted.

[0027] Next, the person in charge performs a predetermined operation on the input unit 17, and uses the prompt generation support unit 13 to generate a prompt that requests the generation AI to compare the independent claim extracted in step 3 with the provisional claim to be searched created in step 2, and to respond with the results of an evaluation of the likelihood that the technology to be searched falls within the technical scope of the invention described in the independent claim of the existing patent (step 4). This prompt includes two texts to be compared, i.e., search target text information containing the provisional claim to be searched and comparison target text information containing the independent claim of the existing patent. Here, a simple example of a prompt is given.

[0028] <Example of a prompt> Please answer whether {# technology under investigation} falls within the technical scope of {# existing patent}. # Technology under investigation An apparatus comprising A, and in the apparatus comprising B, an apparatus characterized by comprising C. # Existing patent An apparatus comprising X, and in the apparatus comprising Y, an apparatus characterized by comprising Z.

[0029] Next, the person in charge performs a predetermined operation on the input unit 17 and inputs the generated prompt as an instruction to the generation AI system 4 through the comparison processing unit 14 (step 5). The comparison processing unit 14 manages the exchange of information between the user and the system 4, for example, using an API (Application Programming Interface) provided by the generation AI system 4. The prompt is sent to the generation AI system 4 via the Internet 2, and the generation AI system 4 performs an evaluation of the possibility that the technology being investigated falls within the technical scope of the invention described in the independent claim of the existing patent (step 6). The comparison processing unit 14 then receives a response to the prompt from the generation AI system 4 (step 7). This response is stored in the search result storage unit 164 in association with information identifying the existing patent.

[0030] Next, the person in charge determines whether the processing of steps 3 to 7 has been completed for all existing patents listed in the patent list created in step 1 (step 8). If there are any unprocessed patents (if the determination is "No"), the person in charge designates one of them as the next patent to be processed and returns to step 3. Steps 3 to 7 are repeatedly executed for each patent in the patent list until the determination is "Yes" in step 8. As a result, the generation AI system 4 sends to the terminal 1 a response including an evaluation result of the likelihood that the technology being investigated falls within the technical scope of the invention described in the independent claim of each existing patent listed in the patent list.

[0031] It is desirable to execute the above repetitive process sequentially by a program. For example, this can be achieved by incorporating a command into the prompt that instructs the generation AI to refer to the independent claims of existing patents included in the patent list. In this case, the work of the person in charge can be simplified to just preparing the technology to be searched in the form of claims.

[0032] If the answer in step 8 is Yes, the person in charge performs a predetermined operation on the input unit 17, and in response, the result processing unit 15 aggregates the responses for each existing patent from the generation AI system 4 and outputs them as patent search results (step 9). The output format of the patent search results should be able to be specified in advance by the person in charge as appropriate. This output format should preferably include the "basis for judgment." By outputting the basis for judgment, the person in charge can confirm the validity of the judgment made by the generation AI system 4. The specification of the output format can be included in the prompt. An example of a prompt in this case is shown below.

[0033] <Example of a prompt including an output format> Please answer whether {# technology under investigation} falls within the technical scope of {# existing patent}. # Technology under investigation An apparatus comprising A, and in said apparatus comprising B, an apparatus characterized by comprising C. # Existing patent An apparatus comprising X, and in said apparatus comprising Y, an apparatus characterized by comprising Z. # Output format - Answer whether it falls within the technical scope. - Answer the basis for whether it falls within the technical scope.

[0034] In the above explanation, the generation AI system 4 makes a judgment on each of the existing patents included in the patent list created by the patent search, but it is also possible to perform steps 2 to 7 on a single existing patent from the beginning without creating a patent list, and obtain an answer from the generation AI system 4 that includes the evaluation results for that existing patent.

[0035] [Detailed Description of Processing at Each Step] Next, processing at each step in the flow shown in FIG. 2 will be described in detail.

[0036] <Step 1> Patent searches can be performed using conventional methods such as patent classification and keywords.

[0037] <Step 2> According to the inventor's research, when the generation AI compares the target technology with the claims of existing patents, inputting the target technology into the generation AI system in a rewritten (or converted) form rather than in natural language is crucial for obtaining accurate answers from the generation AI system. The reason for this improved accuracy is likely due to the fact that the existing patents to be compared and the target technology share the same sentence structure. Furthermore, it is desirable for the text written in claim format to include configurations that demonstrate technical significance. For example, if the target technology is a machine, the structure, arrangement, and function of parts and devices would correspond to the configuration; if the target technology is a process, the individual steps and their procedures would correspond to the configuration; and if the target technology is a chemical substance, the components and chemical composition would correspond to the configuration. The accuracy of the generation AI system's answers can be improved by excluding elements that represent the effects of the invention, such as the selling points of the target technology, from the text written in claim format. Furthermore, the accuracy of the generation AI system's answers can be improved by excluding ambiguous expressions such as "big" and "small" from the text.

[0038] Figure 3 is an explanatory diagram of the conversion process for converting the search target technology into text written in claim format. Note that the format of the search target provisional claim in Figure 3 is merely an example. The claim format here refers to a two-part format, as shown in Figure 3, consisting of a preamble section that explains the prior art and background art, and a characteristic section that lists the components related to the new technology. In another embodiment, a linking section may be added to connect the preamble section and the characteristic section.

[0039] In the example of Figure 3, the template for a provisional claim to be searched has a preamble section that reads "In (YYY) comprising (A1) is (a1), (A2) is (a2), (A3) is (a3), ...," and a feature section that reads "comprises (B1) is (b1), (B2) is (b2), (B3) is (b3), ...," with a perforated sentence written in it. For example, one or more templates for provisional claims to be searched are prepared, and in response to a predetermined operation by the person in charge, the provisional claim drafting support unit 11 selects an appropriate template and displays it on the screen of the display unit 18. After reading the natural language description of the technology to be searched, the person in charge enters appropriate explanations in (A1), (A2), (A3), ... in the preamble section, appropriate component names in (a1), (a2), (a3), ..., and an apparatus name, etc., in (YYY). In addition, appropriate explanations are entered for each of the characteristic parts (B1), (B2), (B3), etc., and appropriate component names are entered for each of the characteristic parts (b1), (b2), (b3), etc. Of course, explanations such as (A1), (A2), (A3), etc. can be omitted if unnecessary. In this way, the person in charge can complete the provisional claim to be searched by filling in the blanks in the template. The text information of the provisional claim to be searched (search target text information) created in this way is stored in the provisional claim storage unit 162 when the person in charge performs a predetermined operation on the input unit 17.

[0040] Depending on the content and field of the technology being investigated, it may be easier to create a provisional claim by using the content of the processing and the names of the steps rather than the names of the components and their descriptions, so it is a good idea to prepare multiple templates with different blank sentences.

[0041] In another embodiment, the provisional claims to be investigated may be created using another generation AI. For example, the provisional claims to be investigated may be output by inputting technical documents such as specifications or plans into a trained generation AI or a generation AI given predetermined instructions. In yet another embodiment, the provisional claims to be investigated may be manually created by an engineer with high intellectual property skills, an intellectual property department member, a patent attorney, or the like.

[0042] Note that the term "new technology" here refers only to what the developer or other party recognizes as new technology, and does not necessarily mean that it is an absolute new technology. For example, a new configuration or added function may be considered new when comparing the company's existing product with the research target technology being developed. Ideally, the research target provisional claim should describe as comprehensively as possible the functions necessary for the product or method in which the research target technology is positioned, but determining how detailed the description should be can be difficult. However, by clearly indicating the field to which the product or method belongs and the general configuration and processing of the product or method in the preamble, the generating AI can understand the field of the product or method in which the research target technology is positioned and can expand its understanding of the configurations and functions typically found in that field, i.e., those considered to be within the scope of prior art, to include those not explicitly stated in the research target provisional claim.

[0043] For example, if the target technology is a liquid chromatograph, the preamble of the target claim may read, "A chromatograph comprising a sample introduction unit for introducing an analytical sample into the device, a column for separating the sample, and a detection unit for detecting the separated sample..." In this case, if an independent claim of the existing patent includes a "control unit for controlling the flow rate of the sample passing through the column," the generated AI may determine that the flow rate control unit is a different configuration from the target technology. However, in this embodiment, by explicitly stating the general configuration of a liquid chromatograph in the preamble, the generated AI can identify a representative device associated with the preamble and infer that the general functions of the representative device should be excluded from the differences.

[0044] <Step 3> Extracting independent claims from patent publications and the like is relatively easy. In other words, in principle, it is sufficient to extract claims that do not have the phrase "recited in claim XX" in the text among the claims described in the claims section of the patent publication. Therefore, the independent claim extraction unit 12 can generally extract independent claims by extracting claims from the claims section of the patent publication and then extracting claims that do not have the phrase "recited in claim XX." Note that, because this task is relatively simple, it may be performed manually by a person in charge or may be implemented by a program using an existing text manipulation algorithm. The text information of the extracted independent claims (comparison text information) is stored in the patent list storage unit 161.

[0045] <Step 4> Prompts are instructions or questions to the generated AI, and it is well known that prompts are important for obtaining accurate answers. In this case, the prompt includes a question requesting a determination of whether the provisional claim under investigation falls within the technical scope of the independent claim of the existing patent. Generally, determining whether a patent infringes or not is difficult for anyone other than those highly familiar with intellectual property practice. This is primarily because it requires a multi-stage process involving multiple factors before reaching the final determination of whether or not there is infringement, and the thought process for making such a determination is generally not clearly articulated or verbalized. For this reason, even if the generated AI is trained using collective knowledge from the Internet, simply creating a prompt such as, "Compare A and B and determine whether or not there is patent infringement," when using that generated AI to determine the possibility of patent infringement, is unlikely to produce a valid result.

[0046] In contrast, the patent search support method of this embodiment not only inputs the text information of the independent claims of the existing patent and the provisional claims to be searched, but also generates prompts that include instructions and questions that specify the thought process leading up to an infringement determination. Additionally, a tool is also provided that allows for the easy and reliable generation of such prompts.

[0047] It is common to instruct the text generation AI not to output unnecessary data for reasons such as speeding up output, reducing costs, and reducing energy consumption. However, in this embodiment, it is preferable to have the generation AI sequentially output the thought results of each study step that constitutes the thought process. In the system of this embodiment, the ultimate goal is to determine whether the technology under investigation falls within the technical scope of an existing patent. However, the accuracy of the determination is improved by outputting the thought results each time each study step that constitutes the thought process is performed (i.e., the result of the first study, the result of the second study, ..., the determination of whether it falls within the technical scope).

[0048] The reason is as follows: Transformer algorithms, which are often used in text generation AI, infer sentences from inputs and (intermediate) outputs. By including the results of each study in the intermediate output, the algorithm performs inference by referring to the results of the previous study steps when performing the next study step. Therefore, by outputting the thought results of each study step rather than simply outputting the determination result of whether or not something falls within the technical scope, the accuracy of the determination is improved. However, outputting the thought results of each study step does not necessarily improve the accuracy of the determination in any thought process. This method can be used precisely because the inventor discovered that the thought process for determining patent infringement is a sequential thought process in which the study steps are inseparable. However, in thought processes other than determining whether or not there is patent infringement, the study steps may be independent of each other. In such cases, it is important to note that the output of the thought results of each study step may reduce the accuracy of the determination.

[0049] Fig. 4 is a diagram showing instructions for the task execution procedure (review procedure) in a prompt for determining the possibility of patent infringement, and Fig. 5 is a diagram showing instructions for the output format written after the instructions for the task execution procedure in the same prompt. An example of a prompt will be specifically described with reference to Figs. 4 and 5.

[0050] The instructions at the beginning of the prompt are a brief description of the task to be performed. The task here is to output answers to the following instructions with the aim of examining the possibility that the developed technology (the technology under investigation) may infringe on existing patents, so it is sufficient to write that.

[0051] Next, the search target text information created in step 1 is set as the development technology, and the comparison target text information extracted in step 2 is set as the existing patent in the prompt. Since both the search target text information and the comparison target text information are stored in a predetermined area of ​​the storage unit 16, such text information may be set by, for example, specifying the file name of the stored text information, reading out the text information, and pasting the text into the sentence of the prompt being created. Alternatively, as described below, the text information may be automatically read out by specifying the file name, and the text may be inserted into the sentence of the prompt.

[0052] The instructions for the first review stage are to compare the developed technology with the existing patent, extracting elements and processes unique to the existing patent as differences, and provide a comprehensive response. Conversely, even if there are elements or processes unique to the developed technology, they will not be considered differences. This instruction is intended to have the generative AI imitate the thought process of determining whether a technology falls within the technical scope, namely, that if the existing patent contains a configuration that the technology under investigation clearly does not implement, it can be determined to be non-infringing. Furthermore, when determining whether a difference exists, focusing solely on the differences in the terminology itself may result in a mistaken recognition of a difference when in fact there is no substantial difference. Therefore, instructions are also added to not consider differences to be present when the terms are inclusive of each other.

[0053] The instructions for the second review stage are to consider whether or not the differences identified in the first review stage will disappear in accordance with predetermined rules, if any. Examples of predetermined rules include the following: - Functions and processes that are common in the technical field (product) of the developed technology will not be considered differences. - Functions and processes that are novel in the technical field (product) of the developed technology will be considered differences. - Related terms will be considered to be the same (not considered differences) to the extent that the functions of the terms are not exceeded.

[0054] Without this second review stage, the generative AI may identify features that are not differences as differences. For example, "analytical device" and "measuring instrument" are distinct terms in the narrow sense. However, the terms in claims are interpreted based on the content of the specification, and in many cases, the difference between the terms "analytical device" and "measuring instrument" is not considered a difference. To address such cases, the system broadly interprets the terms and prompts the user to carefully examine the differences. In addition to including the above rules in the prompt, it is also recommended to include at least one specific example of a case where a difference is determined according to the rule and a case where a difference is not determined according to the rule. For example, it is recommended to include a specific example such as, "When the target technology is XX (where XX is a product classification), △△ is not a difference because △△ is a function or configuration that XX generally possesses."

[0055] The instructions in the third review stage are to ask the examinee to respond to the significance of each difference if there are one or more components or processes identified as differences in the instructions in the second review stage. This is the step to identify the technical scope of the differences by identifying the purpose and significance of the identified differences. The technical scope of the configuration extracted as a difference may be identified from the functional description or the description before and after the difference. For example, if there is a statement such as "a detection unit that detects abnormal conditions based on a sensor," the term "detection unit" can be interpreted as a detection unit that detects various abnormal conditions. However, if there are statements such as "a sensor that detects temperature" before and after it, the "detection unit" can be interpreted as "a detection unit that detects abnormal conditions based on temperature."

[0056] The instructions for the fourth review stage are to reexamine whether each difference can truly be considered a difference if there is one or more components or steps determined to be differences in the instructions for the second review stage. Specifically, for example, it is advisable to reexamine whether a difference is a difference from the following perspectives and to clarify the reasons for that determination. - Determine whether the difference is related to a component or step included in the characteristic part, rather than the preamble part, of the provisional claim under search, and if so, determine that it is not a difference. - When determining the relevance of the difference to a component or step included in the characteristic part of the provisional claim under search, use the commonality between the significance of the difference identified in the third review stage and the significance of that component or step as one of the criteria for making the determination.

[0057] As mentioned above, it is desirable to describe the searched technology comprehensively when describing it in claim format, but this is not realistic. Even if it is a new element, there is a possibility that the more detailed functions that make up the new element will not be fully described. Therefore, even if there is a difference in the existing patent that the searched technology does not implement, if there is a commonality between the technical significance of that difference and the technical significance of the new element of the searched technology, the existing patent will be deemed to be a highly relevant case.

[0058] Here, as in the second examination stage, it is preferable to provide at least one concrete example in the prompt for each of the cases where the difference is judged to be related to the elements or processes included in the characterizing portion of the provisional claim under investigation, and the cases where the difference is judged to be unrelated. For example, if the characterizing portion of the provisional claim under investigation includes an element that performs "temperature control inside the housing," and the difference (i.e., an element that exists only in the independent claim of the existing patent) is a temperature control method that is different from that implemented in the technology under investigation, then although the control methods are different, both are technologies related to temperature control, and therefore it is recommended to provide an example such that the two are judged to be closely related.

[0059] The instructions for the final review stage are to review whether the technology under investigation falls within the technical scope of the existing patent based on the review results up to the fourth review stage. Specifically, if there are no differences at the end of the fourth review stage, the examiner should conclude that the technology is likely to fall within the technical scope. Even if there are differences, if the review in the fourth review stage determines that the differences are related to the elements or steps included in the characterizing portion of the provisional claim under investigation, the examiner should conclude that the technology is likely to fall within the technical scope. On the other hand, if the review in the fourth review stage determines that there are differences that are not clearly related to the elements or steps included in the characterizing portion of the provisional claim under investigation, the examiner should conclude that the technology is unlikely to fall within the technical scope.

[0060] The prompt includes instructions for the procedure for executing the task, as described above, followed by instructions for the output format shown in Figure 5. Simply put, this output format outputs the determination result of whether the researched technology falls within the technical scope of the invention described in the claims of the existing patent, along with the reason for that determination. Outputting not only the determination result but also the reason for the determination makes it easier for humans to confirm or verify the validity of the determination result made by the generative AI. Additionally, it can improve the determination accuracy of algorithms that infer sentences from both input and output sentences, such as transformer algorithms. Of course, if the determination accuracy of the generative AI is high or depending on the application, human reconfirmation is not necessarily required, so outputting the reason for the determination is not required.

[0061] The prompts shown in Figure 5 prompt the user to output the results of each stage, from the first review stage to the final review stage. That is, the first instruction is to output all differences identified in the first review stage. The second instruction is to output the results of the reexamination of the differences in the second review stage, including whether the differences have been eliminated (i.e., what differences have been eliminated and what differences remain), and to explain the reasons for that determination. The third instruction is to explain the significance of each difference examined in the third review stage. The fourth instruction is to output the results of the reexamination of the differences in the fourth review stage, including whether the differences are related to the research target technology, and to explain the reasons for that determination. Finally, the fifth instruction is to output the final determination, whether the research target technology is likely or unlikely to fall within the technical scope of the existing patent, and the reasons for that determination. Of course, this is just one example, and some of the output format instructions for each review stage can be omitted as appropriate.

[0062] As mentioned above, the prompts used here are quite detailed and lengthy. While such prompts could be created by the person in charge each time a patent infringement investigation is conducted for the technology under investigation, this process is tedious and time-consuming, and is prone to errors such as typos. On the other hand, even if the technical fields are different, many of the instructions and questions that should be included in prompts for patent infringement investigations are standardized. Therefore, in the patent investigation system of this embodiment, one or more prompt templates are stored in advance in the prompt template storage unit 160, and during patent investigations, the prompt generation support unit 13 uses the prompt templates stored in the prompt template storage unit 160 to support the person in creating prompts.

[0063] As is clear from the above explanation, many of the questions and instructions in prompts can be included in templates. Furthermore, the survey target text information and comparison target text information can be automatically imported from a file with the same file name by entering the file name, for example, without having to enter the text itself. Other content that the person in charge should enter includes, for example, specific examples of rules used in the second review stage and specific examples of relevance determinations used in the fourth review stage. For these specific examples, general examples can be included in templates, allowing the person in charge to add more appropriate specific examples tailored to the technical field, etc., as needed. This significantly reduces the workload of the person in charge of generating prompts.

[0064] <Step 5> In step 5, the comparison processing unit 14 in the terminal 1 inputs the prompt generated in step 4 to the generation AI system 4 using the API provided by the generation AI system 4, requesting an answer. As described above, the prompt includes the search target text information and the comparison target text information, so the processing of step 5 includes the comparison processing unit 14 accepting both the search target text information and the comparison target text information. Using an API to input the prompt enables automated processing of a large number of existing patents.

[0065] In addition, the API allows you to specify whether to refer to the results of the previous review in the next review, i.e., to specify the behavior of some of the context windows. When reviewing multiple existing patents, as in this embodiment, it is best to review one existing patent without referencing the review results or process before reviewing the next existing patent. This is because if the previous review results or process are referenced, the interpretation of terms contained in the previous existing patent may be referenced in the interpretation of terms in the next existing patent, resulting in a restrictive interpretation being output.

[0066] Furthermore, in general commercial generative AI systems, confidentiality is guaranteed when the API provided by the system itself is used, but confidentiality may not be guaranteed when the API is not used. This can be an obstacle when aiming to obtain a patent for the technology under investigation. However, using an API to input prompts is not required; prompts can also be input through a chat-style user interface, for example.

[0067] <Steps 6 and 7> The generative AI system 4 compares the search target text information with the comparison target text information in accordance with the given prompts to determine whether the search target technology infringes an existing patent. Because the search target technology is described in the form of a two-part claim consisting of a preamble and a characteristic part, and the prompts are written to lead to a conclusion through the multiple review stages described above and to output the reason for the judgment, the generative AI system 4 performs the following particularly characteristic processing.

[0068] Generally, patent claims for a certain device do not comprehensively describe the device's general components; in fact, they are often missing. In contrast, when describing a target technology in claim format, the device's product field and at least some of its general configuration are explicitly specified. By providing this information to the generation AI system 4, the generation AI can understand the device field in which the target technology falls. As a result, even if the general components are not comprehensively described, the generation AI can determine differences and similarities by assuming that the target technology contains such general components. Furthermore, when comparing a target technology with existing patents using interactive text generation AI, variations in the description of the target technology from one person to another can occur, leading to reduced accuracy. In contrast, by standardizing the input format to the claim format, the impact of variations in the description from one person to another can be reduced, leading to improved accuracy.

[0069] Furthermore, text generation AI has the ability to absorb variations in word expressions and terminology through extensive language learning. Therefore, even if the notation of the same component differs between the search text information and the comparison text information, the notational differences can be absorbed as variations and a determination of differences or matches can be made. Furthermore, if the search target technology described in the search target provisional claim is included in the independent claim of the existing patent, the generation AI determines that there are no differences. Therefore, even if the independent claim of the existing patent is written in a generic concept of the search target technology, it can determine that there are no differences and accurately determine that there is a high possibility of infringement.

[0070] Furthermore, even if a difference is found that exists only in the independent claim of an existing patent, if that difference is related to the preamble portion of the provisional claim under investigation, i.e., the background art, it is often the case that the difference is simply not specified when the provisional claim under investigation was created, but is in fact a configuration that is naturally provided in the field of the device. In contrast, the method of this embodiment determines whether the difference is "related" to the components and steps included in the characteristic portion of the provisional claim under investigation, thereby reducing the possibility of mistakenly identifying a configuration that is generally considered to be provided in the technology under investigation as a difference, and preventing the overlooking of potentially infringing existing patents.

[0071] Figure 6 shows an example of the relationship between the outline descriptions of the provisional claim (the technology under investigation) and the independent claim of the existing patent based on the above process and the determination results by the AI ​​generation. In Figure 6, the preambles of the provisional claim under investigation and the independent claim of the existing patent are very simplified, with A through G in the characteristic section below each representing a component. In Figures 6(a) through 6(d), components A, B, and C are common in the characteristic section of the claim. In Figure 6(a), component D is present only in the independent claim of the existing patent, so component D is extracted as a difference in the first review stage. If component D is determined to be a novel function in the product field of the developed technology and unrelated to the characteristic section of the provisional claim under investigation, it will remain a difference until the fifth review stage. As a result, it is concluded that this is unlikely to fall within the technical scope of the existing patent.

[0072] In Figure 6(b), elements A, B, and C in the independent claims of the existing patent are all included in the elements of the provisional claim under investigation. Therefore, there are no differences, and the second to fourth review stages are essentially skipped. In the fifth review stage, it is concluded that there is a high probability that the provisional claim under investigation falls within the technical scope of the existing patent.

[0073] In Figure 6(c), because element F exists only in the independent claim of the existing patent, element F is extracted as a difference in the first review stage. However, if element F is a means for realizing a component or function commonly found in the product field described in the preamble of the provisional claim under investigation, it will not be determined as a difference in the second review stage. As a result, the third and fourth review stages are essentially skipped, and in the fifth review stage, it is concluded that there is a high probability that the proposed invention falls within the technical scope of the existing patent.

[0074] In Figure 6(d), because element F exists only in the independent claim of the existing patent, element F is extracted as a difference in the first review stage. However, if, for example, in the fourth review stage, it is determined that the meaning of element F is substantially the same as the meaning of element G in the characterizing part of the provisional claim under investigation and that they are related, and if it is a means for realizing a component or function that is commonly found in the product field, it can be concluded in the fifth review stage that there is a high possibility that element F falls within the technical scope of the existing patent.

[0075] The generation AI system 4 evaluates the possibility of infringement or non-infringement of each existing patent in accordance with the given prompts, for example, as described above, and outputs a response. The terminal 1 associates the response output from the generation AI system 4 (for example, the response according to the instructions shown in FIG. 5 ) with information identifying the existing patent (usually the patent application number or patent number) and stores it in a pre-designated database. Furthermore, in the patent research support system of this embodiment, in addition to the response output from the generation AI, the claims and prompts of the existing patent can also be stored in the same database. This allows the user to reconfirm the generation AI's judgment results at any time.

[0076] However, it goes without saying that the data fields to be stored may be changed as appropriate, as long as the information identifying the existing patent to be judged and the response from the generation AI are associated and stored in a referable manner. Furthermore, the format of the database for storing these results is not particularly limited, and examples include CSV files, spreadsheets such as Microsoft Excel, and databases such as SQL Server. Furthermore, the data may be stored in separate databases for each data field.

[0077] As described above, the patent research support system of this embodiment allows for a simple evaluation of the numerous existing patents extracted through a patent search based on the target technology to determine whether the target technology falls within the technical scope of the invention described in the claims of the existing patents, thereby narrowing down the list to existing patents that may be infringing. This narrowing process using generative AI eliminates existing patents that are highly unlikely to be infringed by the target technology, while extracting almost all existing patents that may be infringed by the target technology. Therefore, by having the person in charge determine whether the existing patents narrowed down by the system are infringing or not, a patent infringement investigation can be conducted with fewer omissions and with a reduced burden on the person in charge.

[0078] It should be noted that the above embodiment is merely an example of the present invention, and it goes without saying that any appropriate modifications, alterations, additions, etc. made within the spirit of the present invention will also fall within the scope of the claims of the present application.

[0079] Aspects It will be appreciated by those skilled in the art that the exemplary embodiments described above are examples of the following aspects.

[0080] (Item 1) One aspect of the patent research support method of the present invention is a patent research support method for supporting an infringement investigation to determine whether a technology under investigation infringes an existing patent, and includes using a computer to execute the following steps: a first reception step of receiving text information to be investigated that is written in the form of a patent claim for the technology under investigation; a second reception step of receiving comparison text information that is a claim of an existing patent; a determination step of using a generation AI to compare the text information to be investigated received in the first reception step with the comparison text information received in the second reception step to determine the possibility that the technology under investigation falls within the technical scope of the invention described in the claim of the existing patent; and an output step of outputting the determination result from the determination step.

[0081] (Item 8) One aspect of the patent research support system of the present invention is a computer-based patent research support system that supports an infringement investigation to determine whether a technology under investigation infringes an existing patent, and includes: a first memory unit that stores search target text information written in the form of patent claims for the technology under investigation; a second memory unit that stores at least patent identifying information and claims for one or more existing patents; a determination unit that acquires the search target text information from the first memory unit and acquires a claim of one existing patent from the second memory unit as comparison target text information, and uses generation AI to compare the search target text information with the comparison target text information to determine the possibility that the technology under investigation falls within the technical scope of the invention described in the claim of the existing patent; and an output unit that outputs the determination result by the determination unit.

[0082] According to the patent search support method described in paragraph 1 or the patent search support system described in paragraph 8, by comparing the search target text information written in patent claim format with the comparison target text information, which is the claim of an existing patent, rather than the search target technology written in natural language, it is determined whether the search target technology is likely to fall within the technical scope of the invention of the existing patent, so the accuracy of the determination is high and existing patents that are likely or highly likely to infringe the search target technology can be accurately and efficiently extracted.In addition, when the search target technology written in natural language is rewritten into patent claim format, variations in terminology and language occur, and there is also variation in the terminology and language used in the claims of the existing patent themselves, but such variations can be absorbed by the generation AI, so highly accurate determination results can be obtained that eliminate or reduce the influence of such variations.

[0083] (Section 2) The patent research support method described in Section 2 is the patent research support method described in Section 1, wherein the judgment step includes a plurality of steps, and the plurality of steps can include a first judgment step of comparing the search target text information with the comparison target text information to extract and record differences, and a second judgment step of judging the possibility that the differences fall within the technical scope of the invention based on the recorded differences.

[0084] (Item 9) The patent research support system described in item 9 may be the patent research support system described in item 8, wherein the judgment unit includes a difference extraction unit that compares the search target text information with the comparison target text information, extracts and records differences, and a main judgment unit that judges the possibility of the information falling within the technical scope of the invention based on the recorded differences.

[0085] According to the patent search support method described in paragraph 2 or the patent search support system described in paragraph 9, the structural differences between the technology described in the search target text information and the technology described in the comparison target text information are clarified, and then a determination is made based on those differences as to whether the technology falls within the technical scope of the invention, thereby improving the accuracy of the determination. Furthermore, while determinations using generative AI tend to have a black box-like process, according to the method described in paragraph 2 or the system described in paragraph 9, regardless of whether the determination result is that the technology falls within the technical scope of the invention or not, the basis for the determination can be traced, making it easy to evaluate the appropriateness of the determination.

[0086] (Clause 3) The patent research support method described in paragraph 3 may be the patent research support method described in paragraph 2, wherein the judgment step includes a step of outputting the results of at least one step other than the second step.

[0087] In the method described in Section 3, not only the final result of the decision on whether or not something belongs to the group is output, but also the result of thinking during the process of the decision. This makes it possible to improve the accuracy of the decision.

[0088] (4) The patent research support method described in paragraph 4 is the patent research support method described in paragraph 1, wherein the text information to be searched can include a preamble portion indicating components of the conventional technology that is the premise of the technology to be searched, and a characteristic portion indicating components that are new in the technology to be searched.

[0089] According to the patent research support method described in paragraph 4, the components described in the preamble can be processed as falling within the category of prior art, narrowing down the components to be compared and improving the accuracy of the determination. In addition, since the sentence structures of the two pieces of text information to be compared are almost the same, the comparison becomes easier, leading to improved accuracy of the determination.

[0090] (5) The patent research support method described in paragraph 5 is the patent research support method described in paragraph 4, wherein the difference is a difference due to the existence of a component only in the comparison target text information, and in the second determination step, it is determined whether the difference is closely related to the characteristic part of the search target text information, and if it is determined that the difference is closely related, it can be concluded that the search target technology is highly likely to fall within the technical scope of the invention described in the claims of the existing patent.

[0091] (Item 6) The patent research support method described in item 6 is the patent research support method described in any one of items 1 to 5, further comprising a patent extraction step of extracting a patent list by using a computer to input a search formula created based on the technology to be searched into a database, wherein the second receiving step sequentially receives comparison text information for existing patents included in the patent list, and the determination step sequentially performs a determination on the invention regarding the comparison text information received in the second receiving step for one piece of search target text information received in the first receiving step.

[0092] According to the patent research support method described in Section 6, even when determining the possibility of infringement of a large number of existing patents listed in a patent list, the provisional claims to be researched and input into the generation AI only need to be written once, which helps to reduce work time.

[0093] (Item 7) The patent research support method described in item 7 is a patent research support method described in any one of items 1 to 6, which further executes a prompt creation step using a computer to create a prompt including questions and / or instructions for the generated AI, and in the prompt creation step, a prepared prompt template can be retrieved from a memory unit and displayed on a display screen, and the prompt can be completed in response to user operations including additions and modifications to the prompt template.

[0094] (Item 10) The patent research support system described in item 10 is the patent research support system described in item 8 or 9, further comprising a prompt creation support unit that supports the creation of prompts including questions and / or instructions for the generation AI, and the prompt creation support unit can retrieve a prepared prompt template from a memory unit, display it on a display screen, and complete the prompt in response to user operations including additions and modifications to the prompt template.

[0095] When using generative AI, it is known that the content of prompts used to provide questions and instructions to the generative AI is extremely important for obtaining a more appropriate answer. On the other hand, questions and instructions for obtaining appropriate answers in specific narrow fields such as patent infringement investigations can be largely standardized, and conversely, standardizing questions and instructions is advantageous for ensuring appropriate answers. In contrast, the method described in Section 7 or the system described in Section 10 can standardize many of the questions and instructions by using prompt templates, thereby enabling more accurate judgments and reducing the workload on users when creating prompts, thereby improving efficiency.

[0096] (Item 11) The patent research support method described in Item 11 is a patent research support method described in any one of Items 1 to 7, wherein in the determination step, a generation AI is used to compare the search target text information with the comparison target text information, and if it is determined that a configuration corresponding to a constituent element described in a claim of the existing patent exists in the search target technology and that the configuration is described in the search target text information as a higher-level concept of the constituent element, it can be determined that the search target technology neither falls within the technical scope of the invention described in the claim of the existing patent nor does it not fall within the technical scope.

[0097] (Clause 12) The patent research support method described in paragraph 12 is the patent research support method described in paragraph 11, and in the output step, the judgment result of the judgment step can be output together with constituent elements that are described in the claims of the existing patent, and that have been determined to have a configuration corresponding to those constituent elements in the technology being searched, and that are represented in the text information being searched as a superordinate concept of those constituent elements.

[0098] When creating search target text information in the form of patent claims, it is difficult to accurately describe specific embodiments of the search target technology in the search target text, considering the constituent features described in all existing patent claims. When the configuration described in the search target text information is unclear, making it impossible to determine whether the target technology contains a configuration corresponding to the constituent features described in the existing patent claims. To avoid misjudgment, a prompt is created to determine whether the target technology belongs to the technical scope. However, it is impossible to distinguish whether the corresponding configuration actually exists in the target technology or whether the determination could be overturned by confirming the specific embodiment of the target technology. The method described in Section 12 makes it possible to determine whether further investigation of specific embodiments in the target technology is necessary, thereby facilitating patent infringement investigations using generative AI, even if the corresponding configuration is described in the search target text information as a generic concept.

[0099] (Item 13) The patent research support method described in item 13 is the patent research support method described in item 11 or 12, wherein the judgment step uses a generation AI to compare the search target text information with the comparison target text information, and if a configuration corresponding to one constituent element described in a claim of the existing patent exists in the search target technology and a configuration corresponding to another constituent element does not exist in the search target technology, and it is determined that the configuration is written in the search target text information as a higher concept of the constituent element, it can be determined that the search target technology does not fall within the technical scope of the invention described in the claim of the existing patent.

[0100] If a so-called missing element exists, there is no need to further investigate specific embodiments of the technology under investigation, which makes it possible to reduce unnecessary efforts.

[0101] (Item 14) The patent research support method described in item 14 is a patent research support method described in any one of items 1 to 13, in which the model of the generative AI is used as an inference model.

[0102] As a result of verification by the inventors, it was confirmed that the inference model can accurately judge superordinate and subordinate concepts, whereas the non-inference model often made erroneous judgments. This makes it possible to achieve an extremely high level of judgment accuracy in the judgment step.

[0103] DESCRIPTION OF SYMBOLS 1...Terminal (PC) 10...Patent search unit 11...Provisional claim creation support unit 12...Independent claim extraction unit 13...Prompt generation support unit 14...Comparison processing unit 15...Result processing unit 16...Memory unit 160...Prompt template storage unit 161...Patent list storage unit 162...Provisional claim storage unit 163...Independent claim storage unit 164...Search result storage unit 17...Input unit 18...Display unit 2...Internet 3...Patent database system 4...Generation AI system

Claims

1. A patent research support method for supporting an infringement investigation to determine whether a technology under investigation infringes an existing patent, the method using a computer to execute the following steps: a first reception step of receiving text information to be investigated, which is written in the form of a patent claim, for the technology under investigation; a second reception step of receiving comparison text information, which is a claim of an existing patent; a determination step of using generation AI to compare the text information to be investigated received in the first reception step with the comparison text information received in the second reception step, to determine the possibility that the technology under investigation falls within the technical scope of the invention described in the claim of the existing patent; and an output step of outputting the determination result from the determination step.

2. The patent research support method according to claim 1, wherein the judgment step includes a plurality of steps, the plurality of steps including a first judgment step of comparing the search target text information with the comparison target text information to extract and record differences, and a second judgment step of judging the possibility that the search target text information falls within the technical scope of the invention based on the recorded differences.

3. The patent research support method according to claim 2, wherein the determination step includes a step of outputting the result of at least one step other than the second step.

4. The patent research support method described in claim 1, wherein the text information to be searched includes a preamble portion indicating components of the conventional technology that is the premise of the technology to be searched, and a characteristic portion indicating components that are new in the technology to be searched.

5. The patent research support method described in claim 4, wherein the difference is a difference due to the existence of a component only in the comparison target text information, and in the second judgment step, it is determined whether the difference is closely related to the characteristic part of the search target text information, and if it is determined that the difference is closely related, it is concluded that the search target technology is likely to fall within the technical scope of the invention described in the claims of the existing patent.

6. The patent research support method according to claim 1, further comprising: a patent extraction step of extracting a patent list by using a computer to input a search formula created based on the technology to be researched into a database; wherein in the second reception step, comparison text information is sequentially received for existing patents included in the patent list; and in the judgment step, judgments regarding inventions regarding the comparison text information received in the second reception step are sequentially made for one piece of research target text information received in the first reception step.

7. A patent research support method as claimed in claim 1, further comprising a prompt creation step of using a computer to create a prompt including a question and / or instruction for the generating AI, wherein in the prompt creation step, a prepared prompt template is retrieved from a storage unit and displayed on a display screen, and the prompt is completed in response to user operations including additions and modifications to the prompt template.

8. A computer-based patent research support system that supports infringement investigations to determine whether a technology under investigation infringes an existing patent, comprising: a first memory unit that stores text information to be investigated for the technology under investigation, written in the form of a patent claim; a second memory unit that stores at least information identifying each patent and claims for one or more existing patents; a determination unit that acquires the text information to be investigated from the first memory unit and acquires the claims of one existing patent from the second memory unit as comparison text information, and uses generation AI to compare the text information to be investigated with the comparison text information to determine the possibility that the technology under investigation falls within the technical scope of the invention described in the claims of the existing patent; and an output unit that outputs the determination result made by the determination unit.

9. The patent research support system of claim 8, wherein the judgment unit includes a difference extraction unit that compares the search target text information with the comparison target text information, extracts and records differences, and a main judgment unit that judges the possibility of the information falling within the technical scope of the invention based on the recorded differences.

10. A patent research support system as described in claim 8, further comprising a prompt creation support unit that supports the creation of prompts including questions and / or instructions for the generation AI, said prompt creation support unit retrieving prepared prompt templates from the memory unit and displaying them on the display screen, and completing the prompt in response to user operations including additions and modifications to the prompt templates.

11. A patent research support method as described in claim 1, wherein in the judgment step, a generation AI is used to compare the search target text information with the comparison target text information, and if it is determined that a configuration corresponding to a constituent element described in a claim of the existing patent exists in the search target technology and that the configuration is described in the search target text information as a higher-level concept of the constituent element, it is determined that the search target technology neither falls within the technical scope of the invention described in the claim of the existing patent nor does it not fall within the technical scope.

12. A patent research support method as described in claim 11, wherein the output step outputs the judgment result of the judgment step together with constituent elements described in the claims of the existing patent, which have a corresponding configuration in the technology under search and which are determined to be represented in the text information under search as a superordinate concept of the constituent elements.

13. A patent research support method as described in claim 11, wherein in the judgment step, a generation AI is used to compare the search target text information with the comparison target text information, and if a configuration corresponding to one constituent element described in a claim of the existing patent exists in the search target technology and a configuration corresponding to another constituent element does not exist in the search target technology, and it is determined that the configuration is expressed in the search target text information as a higher concept of the constituent element, it is determined that the search target technology does not belong to the technical scope of the invention described in the claim of the existing patent.

14. A patent research support method as described in claim 1, wherein the model of the generative AI is an inference model.

Images