Information search system

Abstract

An information search system 1000 comprises a technical information acquisition unit 112a1, a search unit 112c, a document set generation unit 112d, and a document set adjustment unit. The technical information acquisition unit 112a1 acquires technical information indicating an attribute of a target product. The search unit 112c searches a patent document and assesses whether wording that matches or approximates the technical information is present in the patent document. The document set generation unit 112d generates a set of patent documents in which wording that matches or approximates the technical information is present. The document set adjustment unit narrows down the generated set of patent documents on the basis of another narrowing condition different from the specific attribute.

Description

Information Search System

[0001] The present invention relates to an information retrieval system.

[0002] It is known that patent information that matches patent information search conditions set based on user input is searched for and displayed on the display unit of a client device (Patent Document 1). In Patent Document 1, the patent information is previously issued gazettes of industrial property rights such as patents. In addition, Patent Document 1 states that the patent information search conditions include, for example, IPC classification, F-terms, keywords, applicants or patentees, etc.

[0003] Japanese Patent Application Laid-Open No. 2019-067330

[0004] For example, a company may want to find others who would like to use a product that it is developing or has already developed. In such cases, one method is to search patent documents to find others who would like to use the product. Note that the term "others" here can be broadly interpreted to include not only competitors, but also other organizations or individuals within the same company.

[0005] However, the technology described in the above patent document simply searches for patent information that matches the patent information search conditions, and does not anticipate searching for potential users of the product. As a result, it is not possible to accurately determine the market needs for the product, and the search results may contain a lot of noise.

[0006] Therefore, it is necessary to provide an information retrieval system that can more accurately search for target users who are expected to use the product. Note that the target users may be interpreted in the same way as the other people described above.

[0007] For example, a company may be searching for a technology required for manufacturing a product or implementing an application, but may not have sufficient information about companies, other corporations, or individuals that possess that technology.

[0008] On the other hand, companies, other corporations, and individuals who possess the technology as seeds may not be aware of other companies, corporations, and individuals who are looking for that technology.In other words, companies, other corporations, and individuals who have the technology as seeds and want to provide it may not know where to sell their technology.

[0009] However, the technology described in the above patent documents does not anticipate matching between companies, other corporations, and individuals who have technology needs and companies, other corporations, and individuals who are looking for technology seeds. As a result, both companies, other corporations, and individuals who have technology needs and companies, other corporations, and individuals who are looking for technology seeds have difficulty finding a match.

[0010] In view of the above-mentioned problems, an object of the present disclosure is to provide an information search system that can search for a supplier of a target product or a supply destination of the target product with a higher degree of accuracy.

[0011] In view of the above-mentioned problems, an object of the present disclosure is to provide an information search system that can more accurately search for target users who are expected to use a product.

[0012] Another object of the present disclosure is to provide an information search system that can more accurately search for a target person who owns a product that the user wants to use.

[0013] The gist of the present disclosure is as follows.

[0014] (1) The information retrieval system includes a technical information acquisition unit, a search unit, a document set generation unit, and a document set adjustment unit. The technical information acquisition unit acquires technical information indicating attributes of a target product. The search unit searches for patent documents that contain descriptions that match or are similar to the technical information. The document set generation unit generates a set of the patent documents that contain descriptions that match or are similar to the technical information. The document set generation unit generates a set of the patent documents that are determined to contain descriptions that match or are similar to one or more specific attributes. The document set adjustment unit narrows down the generated set of patent documents based on other narrowing conditions that are different from the specific attributes.

[0015] (2) In the information search system described above in (1), the attributes include at least one of the names of substances constituting the target product, structural characteristics of the target product, functions that the target product can provide, problems that the target product solves, and uses of the target product.

[0016] (3) In the information search system described above in (1) or (2), the search unit searches a predetermined portion of the patent document corresponding to each attribute of the technical information intensively rather than other portions.

[0017] (4) The information retrieval system according to any one of (1) to (3) above further comprises a non-patent information acquisition unit and a provision unit. The non-patent information acquisition unit acquires non-patent information searched based on the technical information. The provision unit provides the non-patent information to a user together with the collection of patent documents.

[0018] (5) In the information retrieval system described above in (4), the document set adjustment unit narrows down the set of patent documents based on the non-patent information.

[0019] (6) The information retrieval system according to (4) or (5) further includes a priority setting unit, which, when the non-patent information corresponds to a predetermined patent document included in the set of patent documents, sets a high priority to the predetermined patent document or the non-patent information corresponding to the predetermined patent document.

[0020] (7) The information retrieval system according to any one of (1) to (6) above further includes a learning unit that learns an algorithm for the search unit to search for the patent documents based on a user's evaluation of the set of patent documents.

[0021] (8) The information retrieval system according to (7) above, wherein the search unit searches for patent documents containing descriptions that match or are similar to the technical information newly acquired by the technical information acquisition unit based on the algorithm learned by the learning unit, and further includes a providing unit that provides a user with the patent documents searched based on the algorithm.

[0022] (9) The information retrieval system according to any one of (1) to (8) above further includes a priority setting unit, which sets a higher priority for the patent document as the degree of coincidence between the problem and the description of the patent document increases.

[0023] (10) The information retrieval system according to any one of (1) to (9) further includes a providing unit. The document set adjusting unit adjusts the set of patent documents into a patent map according to the narrowing conditions. The providing unit provides the patent map to a user.

[0024] (11) The information retrieval system according to any one of (1) to (10) further includes a providing unit. The document set adjusting unit further narrows down the set of patent documents based on a user's evaluation of the set of patent documents. The providing unit provides the further narrowed down set of patent documents to the user.

[0025] (12) The information retrieval system according to any one of (1) to (11) above includes an information acquisition unit. The information acquisition unit acquires searcher information. The searcher information is unique information about a searcher searching for a supplier of the target technology or a candidate for a supplier of the target technology, and includes information about the searcher, a relationship between the searcher and the technical information, or a request from the searcher when searching for the candidate. The search unit searches for at least one of the patent literature and non-patent literature based on the technical information and the searcher information.

[0026] (13) The information retrieval system according to (12) above further comprises an evaluation unit that generates matching information. The evaluation unit generates matching information that indicates the degree of matching between the searcher and the candidate.

[0027] (14) In the information retrieval system of (13), the information acquisition unit further acquires external information. The external information includes information that serves as a reference when determining the degree of matching between the searcher and the candidate, information that supplements the unique information, information that supplements the technical information, and information that serves as a reference for document search and display processing. The evaluation unit creates the matching information based on the set of patent documents, the searcher information, and the external information.

[0028] (15) In the information retrieval system described above in (14), the retrieval unit retrieves the patent documents based on the technical information, the searcher's information, and the external information.

[0029] According to the present disclosure, an information search system is provided that can search for the supplier or destination of a target product with higher accuracy.

[0030] In particular, the present disclosure provides an information search system that can more accurately search for target users who are expected to use a product.

[0031] Furthermore, the present disclosure provides an information search system that can more accurately search for a target person who owns a product that the user wants to use.

[0032] 1 is a diagram showing an information retrieval system according to an embodiment of the present disclosure. FIG. 1 is a diagram showing the configuration of a search server. FIG. 2 is a diagram showing the configuration of a user's terminal. FIG. 3 is a diagram showing the functional configuration of the search server. FIG. 4 is a diagram showing the association between technical information and predetermined portions of patent documents. FIG. 5 is a diagram showing the functional configuration of a user's terminal. FIG. 6 is a diagram showing a screen through which a user inputs technical information. FIG. 7 is a diagram showing a collection of patent documents narrowed down by filtering conditions entered by a user into a terminal and displayed on a display device screen. FIG. 8 is a diagram showing an example in which a collection of patent documents is narrowed down by the uses of multiple different inventions and the problems that multiple different inventions aim to solve, and displayed in a bubble diagram. FIG. 9 is a diagram showing an example in which further filtering is performed based on filtering keywords entered by a user in FIG. 10. FIG. 11 is a diagram showing a state in which a user further narrows down the application year and applicant from the second screen of FIG. 8. FIG. 12 is a diagram showing an example in which the use of the invention is further limited as a filtering condition. FIG. 13 is a diagram showing an example of evaluation of search results by a user. FIG. 14 is a diagram showing a sequence of processing performed by a search system in a second embodiment. FIG. 15 is a diagram showing an example in which search results are displayed on a display device screen in a second embodiment. FIG. 16 is a diagram showing the functional configuration of a search server in a third embodiment. FIG. 17 is a diagram showing an input screen displayed on a display device in a third embodiment. FIG. 10 is a diagram showing an input screen displayed on a display device in the third embodiment. FIG. 11 is a diagram showing an input screen displayed on a display device in the third embodiment. FIG. 12 is a diagram showing data of a primary set generated by a document set generation unit. FIG. 13 is a diagram showing an input screen for the importance of individual documents. FIG. 14 is a diagram showing a screen on which a matching proposal is displayed. FIG. 15 is a diagram showing a screen on which a search set generated by instruction information for improving search accuracy is displayed as a search result. FIG. 16 is a diagram showing a sequence of processes performed by a search system in the third embodiment.

[0033] Hereinafter, several embodiments according to the present disclosure will be described with reference to the drawings. However, these descriptions are intended to merely exemplify preferred embodiments of the present disclosure and are not intended to limit the present disclosure to such specific embodiments. In the following description, similar components will be given the same reference numerals, and duplicate descriptions will be omitted as appropriate.

[0034] 1. First Embodiment An information retrieval system 1000 according to each embodiment of the present disclosure includes a search server 100 that searches for information, as shown in FIG. 1. The information retrieval system 1000 may further include a user terminal 200 and a generation AI server 300, and the functions of the information retrieval system 1000 may be realized by the search server 100, the terminal 200, and the generation AI server 300. On the other hand, the information retrieval system 1000 may be composed of only the search server 100, in which case the search server 100 may have the functions of the terminal 200 or the generation AI server 300. When the information retrieval system 1000 further includes a user terminal 200 and a generation AI server in addition to the search server 100, the search server 100, the terminal 200, and the generation AI server 300 may be able to communicate with each other via a communication network 500 composed of communication lines. That is, the communication network 500 relays communications between the search server 100, the terminal 200, and the generation AI server 300. The double-headed arrows in FIG. 1 indicate that information is transmitted and received via communication. The generation AI server 300 is configured to execute a generation AI process that generates content by utilizing relationships learned from data. The generation AI can generate content based on prompts provided by the user. The generation AI can learn patterns and relationships using, for example, a set of pre-collected data conditions and correct answers as training data, and generate content. Commercial generation AI may be used as the generation AI server 300. Such generation AIs include those that generate program code, images, videos, audio, etc. As the generation AI server 300, generation AIs that generate text data such as ChatGPT have become increasingly popular. Commercial text generation AIs include those that store large-scale language models such as Gemini (registered trademark), GPT-4 (registered trademark), Llama2, or PaLM2. Note that other foundation models may also be used as the large-scale language model. For example, the information retrieval system 1000 can be easily constructed by using a commercial generative AI, eliminating the need to collect training data separately.The generation AI server 300 may be provided separately from the search server 100. The functions of the generation AI server 300 may be integrated into the search server 100.

[0035] However, when it is assumed that technology X is sought as a need or a seed, the following problems arise. Note that searching for needs means searching for a supplier. Searching for seeds means searching for a supplier. For example, a company A may be searching for technology X as a seed that is necessary for manufacturing a product or implementing an application. In such a case, company A may have little knowledge of information about companies, other corporations, or individuals that possess technology X. Or company A may have little knowledge of the industry to which technology X belongs.

[0036] On the other hand, company b, which possesses technology X, may not be aware of the existence of companies, like company a, that are searching for technology X as a seed. In other words, company b, which has the seed and wants to provide technology X, may not be aware of the need for technology X and may not know where to sell technology X. As a result, the reality is that matching between company a, which is looking for the seeds of technology X, and company b, which has the need for technology X, does not necessarily go well.

[0037] Currently, companies and universities that are potential matchmakers are approached based on personal connections and information about past business partners. However, because the number of potential matchmakers is limited, the probability of finding someone who matches the needs or seeds is low. Furthermore, even when actually approaching potential matchmakers, there is a lot of confidential information, such as technology related to pending or unpublished patents, making it difficult to hold detailed discussions and reach a match.

[0038] The above applies not only to companies but also to universities. There is a huge amount of technology and market information about universities and companies. This makes it extremely difficult to match the technological seeds that universities possess with the needs of companies.

[0039] On the other hand, existing patent analysis databases are designed for experts. As a result, searchers may not have the necessary tool utilization and search skills to perform highly accurate matching. In particular, it is expected that university TLOs and corporate open innovation staff do not have sufficient skills in these areas.

[0040] To achieve highly accurate matching, it is necessary to individually search and combine patent analysis databases, market databases, company databases, and web information. However, the skill of utilizing various tools and hypothesis verification is left to the individual skills of the person in charge. Furthermore, because the amount of data handled by the person in charge is large, it takes a huge amount of time to achieve highly accurate matching.

[0041] According to this embodiment, an information retrieval system 1000 is provided that can narrow down matching candidates with high accuracy based on the user's needs search or seeds search, even if the user has little knowledge of the user's needs or seeds. The information retrieval system 1000 according to this embodiment does not require the user to have advanced search skills. The information retrieval system 1000 evaluates the possibility of matching based on evaluation criteria including industry trends, market size, and growth potential, and presents candidates for the user to contact.

[0042] As described above, a user may wish to explore a certain technology X as a need or seed. For example, a company may wish to search for other companies that would like to use a product, i.e., technology X, that the company is developing or has already developed. Furthermore, a company may wish to search for other companies that provide the underlying technology for a product, i.e., technology X, that the company is developing. In such cases, when the company sends technical information about the product from terminal 200 as a user, search server 100 receives the technical information via communication network 500. Search server 100 searches a literature database to find patent documents from other companies that match the product. In other words, search server 100 can use the literature database to search for information indicating the intended use or source of the product. Note that the term "product" here refers not only to goods processed and manufactured for sale, but may also be broadly interpreted to include raw materials such as seafood and agricultural products. Furthermore, products are not necessarily limited to those currently on sale, but may also include unsold samples planned for future sale. Furthermore, patent documents include patent gazettes, patent publications, or equivalent documents issued by regulatory authorities in various countries, as well as processed versions of these. Patent documents may be digitized information. Processed documents may also include abstracts and translations.

[0043] In particular, in this embodiment, the technical information includes information on one or more attributes. The one or more attributes include, for example, at least one of the names of materials constituting a product, structural characteristics of the product, functions that the product can provide, problems that the product solves, and uses of the product. For example, when technical information including these attributes is input, the terminal 200 transmits the technical information to the search server 100. For example, upon receiving the technical information, the search server 100 performs a focused search on a predetermined section of a patent document determined according to each attribute of the technical information. Here, "focusing on a predetermined section" refers to a search in which the search weighting for the predetermined section is higher than the search weighting for other sections other than the predetermined section. Furthermore, for example, "focusing on a predetermined section" refers to a search in which the search accuracy for the predetermined section is higher than the search accuracy for other sections other than the predetermined section. As an example, the search server 100 may search for the predetermined section using an AND condition with multiple keywords included in the attributes of the technical information, and for other sections other than the predetermined section using an OR condition with multiple keywords included in the attributes of the technical information. As another example, the search server 100 may use a wider range of synonyms for keywords of technical information attributes when searching a predetermined section than for searching sections other than the predetermined section. As another example, the search server 100 may use a higher threshold for determining the degree of match or similarity between the attributes of technical information and the descriptions in patent documents by quantifying them. For example, the search server 100 may focus its search on sections of the full text of each patent document that describe the problem to be solved by the invention and the uses of the invention. The search server 100 may also perform a focused search by weighting each item described in the patent document that is necessary for a person skilled in the art to understand the technical significance of the invention.

[0044] The search server 100 then transmits a set of patent documents containing descriptions related to the technical information, specifically a set of patent documents containing descriptions matching or similar to the technical information, as search results to the terminal 200. The search results are displayed on the screen of the terminal 200. The search server 100 can search for descriptions matching or similar to the technical information, for example, by utilizing highly co-occurring words, which indicate how frequently a certain word appears together with other words. More specifically, the search server 100 performs morphological analysis to divide the technical information, information described in patent documents, and non-patent information into the smallest meaningful morphemes. Next, the search server 100 generates word vectors by vectorizing the meanings of the morphologically analyzed words, and stores the vectors in a storage device. The search server 100 can search for matching or similar descriptions by extracting other words whose word vectors have a higher cosine similarity to the word vector in a specified location of the technical information.

[0045] In particular, when searching for potential product users from among a vast number of patent documents, it is preferable to perform a search based on the names of the substances that make up the product or the structural characteristics of the product as technical information attributes. In particular, when the set of patent documents returned by searching only the names of the substances that make up the product is large, it is preferable to perform a search based on the structural characteristics of the product as well as the names of the substances that make up the product as technical information attributes. Examples of names of substances that make up the product include magnesium oxide and hydrotalcite. Examples of structural characteristics of the product include spherical particles, particle size, and aspect ratio.

[0046] On the other hand, if a search is performed using the names of substances that make up a product or the structural characteristics of the product as attributes of technical information, it may not be possible to search for potential users of the product if the names of the substances or the structural characteristics of the product are not publicly known. In such cases, it is possible to search for potential users of the product by searching for the functions that the product can provide as attributes of technical information.

[0047] As described above, it is particularly preferable to perform a search based on a combination of the names of the substances that make up the product and the structural characteristics of the product as attributes of the technical information. Also, if the names of the substances that make up the product and the structural characteristics of the product are not considered to be publicly known, it is preferable to perform a search based on the functions that the product can provide as attributes of the technical information.

[0048] Furthermore, if the problem solved by the product described in the technical information matches or is similar to the problem that the invention described in the patent document attempts to solve, the product may be able to solve the problem that the inventor or applicant of the patent document is focusing on. Furthermore, if the use of the product in the technical information matches or is similar to the use of the invention described in the patent document, the product may be suitable for the use that the inventor or applicant of the patent document is focusing on. Therefore, the user, i.e., the person in charge of the company developing the product, can find others who may use the product by checking the search results displayed on terminal 200.

[0049] In particular, in this embodiment, the search server 100 focuses its search on specific sections of patent documents that are determined according to the individual attributes of the technical information. This allows the search server 100 to improve the accuracy of searching specific sections corresponding to the individual attributes, thereby reducing noise in the search results. This allows the user to find potential users of the user's product with a higher degree of accuracy.

[0050] Furthermore, the search server 100 can communicate with, for example, the generation AI server 300 and acquire non-patent information related to the technical information from the generation AI server 300. The search server 100 can also acquire non-patent information stored in other data servers (not shown) by communicating with such other data servers. The non-patent information may include non-patent literature or information posted on the Internet. Specifically, information posted on the Internet may include information on each company's website. Non-patent literature may also include technical papers. In the case of handwritten or printed text information, the non-patent information is converted into digital text data using optical character recognition before use. The search server 100 is configured to transmit the acquired non-patent information to the terminal 200 along with a collection of patent documents.

[0051] Alternatively, the search server 100 may narrow down the set of patent documents based on the acquired non-patent information. For example, the search server 100 narrows down the set of patent documents by extracting patent documents whose inventor names match the names of the authors of the non-patent information from the set of patent documents based on the names of the authors of the non-patent information. Furthermore, for example, the search server 100 narrows down the set of patent documents by extracting patent documents whose applicant names match the names of the organizations to which the authors of the non-patent information belong from the set of patent documents based on the names of the authors of the non-patent information. The search server 100 may weight patent documents whose inventor names match the names of the authors of the non-patent information so that they are displayed higher in the set of patent documents based on the names of the authors of the non-patent information.

[0052] Therefore, the search server 100 can provide the user with both the search results of patent documents and the search results of non-patent information. Furthermore, the search server 100 can improve the accuracy of the search results of patent documents by narrowing down the set of patent documents based on the non-patent information.

[0053] 2, the search server 100 includes a first control device 110 and a storage device 120. In other words, the search server 100 functions as an information search device.

[0054] The first control device 110 includes a first processor 112, a first memory 114, and a first communication interface 116. The first processor 112, the first memory 114, and the first communication interface 116 are connected via a first communication bus 118. The first processor 112 includes one or more central processing units (CPUs) and their peripheral circuits. The first processor 112 may further include other arithmetic circuits such as a logic unit, a numerical calculation unit, or a graphics processing unit. The first memory 114 includes, for example, a volatile semiconductor memory and a non-volatile semiconductor memory. The first communication interface 116 corresponds to the first communication I / F shown in FIG. 2 and includes an interface circuit for connecting the first control device 110 to a network within the search server 100 or a communication network 500. The first communication interface 116 is configured to be able to communicate with, for example, the terminal 200 and the generation AI server 300 via the communication network 500.

[0055] The storage device 120 includes, for example, a hard disk device or an optical recording medium and an access device therefor. In other words, the storage device 120 functions as a storage device. The storage device 120 may function as a patent database. The storage device 120 stores information on patent documents. The storage device 120 may also store non-patent information. The storage device 120 may also store synonyms of keywords that are associated with keywords constituting technical information. The storage device 120 may also store computer programs for executing processes that are executed on the first processor 112.

[0056] The hardware configuration of the generation AI server 300 may be the same as that of the search server 100 shown in Fig. 2. The generation AI server 300 stores a large-scale language model in its storage device.

[0057] As shown in FIG. 3 , the terminal 200 has a second control device 210, a display device 220, an input device 230, and a speaker 240. The terminal 200 may be a personal computer or a mobile device. The second control device 210 has a second processor 212, a second memory 214, and a second communication interface 216. The second processor 212, the second memory 214, and the second communication interface 216 are connected via a second communication bus 218. The second communication interface 216 corresponds to the second communication I / F shown in FIG. 3 . The terminal 200 may have the same configurations as the first processor 112, the first memory 114, and the first communication interface 116 of the search server 100, respectively.

[0058] Each function of the information retrieval system 1000 is realized by a functional block executed by a processor. As an example, the first processor 112 includes an information acquisition unit 112a, a search range designation unit 112b, a search unit 112c, a document set generation unit 112d, a document set adjustment unit 112e, and a priority setting unit 112f, as shown in Figure 4. The first processor 112 further includes a provision unit 112g, a prompt creation unit 112h, a prompt transmission unit 112i, a non-patent information acquisition unit 112j, and a learning unit 112k, as shown in Figure 4.

[0059] Each of these units of the first processor 112 is, for example, a functional module realized by a computer program running on the first processor 112. In other words, each of these units of the first processor 112 is composed of the first processor 112 and a program, i.e., software, for making the first processor 112 function. The program may be recorded in the first memory 114 of the first control device 110 or on a recording medium connected to an external device. Alternatively, each of these units of the first processor 112 may be composed of a dedicated arithmetic circuit provided in the first processor 112. The second processor 212 of the terminal 200 shown in FIG. 6 is realized by functional blocks executed by the second processor 212, similar to the first processor 112.

[0060] 4 and 6 are merely examples, and individual components of a certain processor may be provided in another processor. For example, at least some of the functional blocks of the processor of the search server 100 may be provided in the processor of the terminal 200. Furthermore, individual components of a certain processor may be provided in a duplicated manner in another processor.

[0061] Therefore, for example, by having the functions of the search server 100 on the generation AI server 300, it becomes possible to search, map, and generate a collection of patent documents using the generation AI server 300 with minimal instructions from the user, without using a conventional search system.The user can then train the search formula and search algorithm by inputting an evaluation of the search results.

[0062] The following describes the functions of the information retrieval system 1000 that are realized by the functional blocks of the processor.

[0063] First, the functional blocks of the first processor 112 of the search server 100 will be described. The information acquisition unit 112a shown in FIG. 4 acquires information transmitted from the terminal 200. Specifically, the information acquisition unit 112a includes a technical information acquisition unit 112a1. The technical information acquisition unit 112a1 acquires technical information transmitted from the terminal 200, the technical information including at least one or more attributes related to a product. The technical information acquisition unit 112a1 may acquire technical information indicating attributes of a target product related to a search for at least one of a supplier of the target product or a recipient of the target product. The technical information may consist of only one or more attributes related to the product. The one or more attributes include at least one of "names of materials constituting the product," "characteristics of the product's structure," "functions that the product can provide," "problems solved by the product," and "use of the product." The attributes of the technical information may further include "purpose of the product," "technical field of the product," and "properties of materials constituting the product." The information acquisition unit 112a acquires identification information indicating which of one or more attributes the technical information corresponds to. The information acquisition unit 112a may also acquire search range designation information sent from the terminal 200. The search range designation information is information that the user specifies for a specific portion of a patent document. The information acquisition unit 112a may also acquire narrowing down conditions sent from the terminal 200. Furthermore, the information acquisition unit 112a may acquire the user's evaluation of the search results sent from the terminal 200. When a prompt for acquiring new technical information from the technical information specified by the user is sent to the generation AI server 300, the information acquisition unit 112a acquires an answer to the prompt. The information acquisition unit 112a acquires new technical information generated from the technical information specified by the user from the generation AI server 300 as an answer to the prompt.

[0064] The search range designation unit 112b designates a predetermined location as a search range of a patent document based on the attribute of the technical information. An example of the correspondence between the attribute of the technical information and the predetermined location of a patent document is shown in FIG. 5 . For example, if the attribute of the technical information is the name of a substance constituting a product, the search range designation unit 112b may designate the "Mode for Carrying Out the Invention" or "Example" of the patent document as the predetermined location. For example, if the attribute of the technical information is the structural characteristics of a product, the search range designation unit 112b may designate the "Mode for Carrying Out the Invention" of the patent document as the predetermined location. For example, if the attribute of the technical information is the function that the product can provide, the search range designation unit 112b may designate the "Problem that the Invention Aims to Solve" of the patent document as the predetermined location. For example, if the attribute of the technical information is the problem that the product solves, the search range designation unit 112b may designate the "Problem that the Invention Aims to Solve" of the patent document as the predetermined location. Furthermore, for example, if the attribute of the technical information is the use of a product, the search range designation unit 112b may designate the "technical field" or "background art" of a patent document as the predetermined section. Furthermore, for example, if the attribute of the technical information is the properties of a material constituting a product, the search range designation unit 112b may designate the "mode for carrying out the invention" or "example" of a patent document as the predetermined section. The search range designation unit 112b may designate the predetermined section based on the search range designation information. Note that the correspondence shown in FIG. 5 is merely an example, and the predetermined section may be "claims," ​​"abstract," or "bibliographic information" corresponding to the application. The search range designation unit 112b may designate the predetermined section based on attribute identification information, which will be described later. Furthermore, if the patent document is an overseas document, predetermined sections of the overseas document that are recognized to correspond to each of the predetermined sections in FIG. 5 may be used.

[0065] The search unit 112c uses a patent database to search for patent documents containing descriptions related to the technical information. Specifically, the search unit 112c uses the patent database to search for patent documents containing descriptions that match or are similar to the technical information. The search unit 112c searches for patent documents containing descriptions that match or are similar to the technical information by determining whether the patent documents contain descriptions that match or are similar to the technical information. The search unit 112c may search a predetermined portion of a patent document corresponding to each attribute of the technical information more intensively than other portions to determine whether the patent document contains a description that matches or is similar to the technical information. Specifically, the search unit 112c accesses each of the patent documents stored in the storage device 120 and searches a predetermined portion of each patent document specified by the search range designation unit 112b to determine whether the patent document contains a description that matches or is similar to the technical information. The search unit 112c may also search only the predetermined portion and determine whether the predetermined portion contains a description that matches or is similar to the technical information. An external patent database connected via the communication network 500 may be used as the patent database. Patent databases include databases that contain patent information from gazettes issued by regulatory authorities in each country, and databases that compile patent information from published gazettes. Patent databases can be either paid or free.

[0066] For example, when determining whether the name of a substance constituting a product matches or is similar to a description in a patent document, the search unit 112c makes the determination by weighting the search for a description in the patent document specified as a predetermined portion more heavily than for other descriptions in the patent document. Specifically, the search unit 112c makes the determination by weighting the search for a description in the "Mode for Carrying Out the Invention" or the "Example" more heavily than for other descriptions in the patent document. As a result, a patent document in which the name of a substance constituting a product matches or is similar to a description in a predetermined portion is determined to contain a description in the patent document that matches or is similar to the technical information with a higher degree of accuracy than a patent document in which the name of a substance constituting a product matches or is similar to other descriptions.

[0067] Furthermore, for example, when determining whether the structural features of a product or the functions that the product can provide match or are similar to descriptions in patent documents, the search unit 112c weights the description in the patent document specified as the predetermined location more heavily than other descriptions in the search. Specifically, the search unit 112c weights the description in the "Mode for Carrying Out the Invention" in the patent document specified as the predetermined location more heavily than other descriptions in the patent document. As a result, patent documents in which the descriptions in the predetermined location match or are similar to these features or functions are determined to contain descriptions in the patent document that match or are similar to the technical information with a higher degree of accuracy than patent documents in which the descriptions in the predetermined location match or are similar to these features or functions.

[0068] Furthermore, for example, when determining whether the problem solved by the product and the description in a patent document match or are similar, the search unit 112c makes the determination by weighting the search for the description in the patent document specified as the predetermined portion more heavily than for other descriptions in the patent document. Specifically, the search unit 112c makes the determination by weighting the search for the description of "problem to be solved by the invention" more heavily than for other descriptions in the patent document. As a result, a patent document in which the description in the predetermined portion matches or is similar to the problem solved by the product is determined to contain a description in the patent document that matches or is similar to the technical information with a higher degree of accuracy than a patent document in which other descriptions match or are similar to the problem solved by the product.

[0069] Furthermore, for example, when determining whether a product's intended use and a description in a patent document match or are similar, the search unit 112c weights the search for descriptions in the patent document specified as a predetermined location more heavily than for other descriptions in the patent document. Specifically, the search unit 112c weights the search for descriptions in the "technical field" or "background art" more heavily than for other descriptions in the patent document. As a result, a patent document in which the description in the "technical field" or "background art" matches or is similar to the product's intended use is determined to contain a description in the patent document that matches or is similar to the technical information with a higher degree of accuracy than a patent document in which other descriptions match or are similar to the product's intended use. For example, when determining whether a product's intended use and a description in a patent document match or are similar, if there are two or more predetermined locations, the search unit 112c may weight the descriptions in the two or more specified locations.

[0070] The search unit 112c may use, for example, a full-text search technique to determine whether the patent document contains a description that matches or is similar to the technical information. Alternatively, the search unit 112c may use natural language processing to break down the technical information and the description in the patent document into phrases and then determine whether the patent document contains a description that matches or is similar to the technical information. Specifically, the search unit 112c may analyze the morphological elements of the technical information and the description in the patent document, vectorize them, and determine the cosine similarity to determine whether the patent document contains a description that matches or is similar to the technical information. In this case, the search unit 112c may set a threshold value for determining the similarity to determine whether the patent document contains a description that matches or is similar to the technical information. Alternatively, for example, the search unit 112c may determine whether the patent document contains a description that is similar to the technical information based on synonyms of individual keywords associated with the individual keywords constituting the technical information stored in the storage device 120.

[0071] The search unit 112c may obtain a patent classification from the technical information and use the patent classification to search for patent documents. The "patent classification" may include the International Patent Classification, FI, and F-terms. The search unit 112c may perform a search using a search formula or search algorithm that includes the patent classification and the technical information. The search unit 112c may obtain the patent classification based on, for example, a table that associates the technical information with the patent classification. These tables may be stored in the storage device 120 in advance. Alternatively, the search unit 112c may send a prompt to the generation AI server 300 to obtain the patent classification from the technical information, and obtain the patent classification from the response to the prompt.

[0072] Furthermore, the search unit 112c may acquire new technical information from the technical information specified by the user and determine, based on the new technical information, whether a patent document contains a description similar to the technical information. The new technical information may be technical information similar to the technical information specified by the user, or information with attributes different from the attributes of the technical information specified by the user. For example, the search unit 112c may acquire new technical information generated from the technical information specified by the user from the generation AI server 300 and determine, based on the acquired new technical information, whether a patent document contains a description similar to the technical information. Furthermore, for example, the search unit 112c may acquire new technical information from the technical information specified by the user from a trained model and, based on the acquired new technical information, determine whether a patent document contains a description similar to the technical information. In other words, the search unit 112c may use natural language processing to break down the technical information and the description in the patent document into phrases, and then use machine learning to extract technical information similar to the technical information from the patent document. Specifically, the search unit 112c may search for a different substance having similar properties to a substance determined from the name of a substance, based on a trained model created based on training data in which the names of substances are associated with the properties of the substances. More specifically, the search unit 112c may search for a different substance having similar properties to a substance determined from the name of a substance, based on a trained model created using the names of substances and their properties or properties similar to the properties of the substances as training data. In other words, descriptions similar to the technical information in patent documents searched by the search unit 112c may include not only descriptions similar in wording but also descriptions similar in properties. In this way, the search unit 112c may not only search based on technical information, but also search based on new technical information determined from the technical information.

[0073] That is, the search unit 112c uses a patent database to search for patent documents that contain descriptions related to the technical information.

[0074] Based on the search results of the search unit 112c, the document set generation unit 112d generates a set of patent documents containing descriptions related to the technical information, specifically a set of patent documents containing descriptions that match or are similar to the technical information. Specifically, the document set generation unit 112d generates a set of patent documents that are determined to contain descriptions that match or are similar to the attributes of the technical information. In other words, patent documents that match the technical information may be patent documents that are determined to contain descriptions that match or are similar to the attributes of the technical information. Each patent document included in the set of patent documents may include claims, specifications, abstracts, drawings, bibliographic information, and abstracts.

[0075] The document set adjustment unit 112e adjusts the patent documents included in the document set. For example, when the document set generation unit 112d generates a primary set of patent documents determined to contain descriptions matching or similar to one or more specific attributes, the document set adjustment unit 112e narrows down the generated set of patent documents based on other narrowing conditions than the specific attributes. In other words, the document set adjustment unit 112e narrows down the generated set of patent documents based on other narrowing conditions than the specific attributes. In this case, the search unit 112c further searches the primary set of patent documents generated based on the specific attributes based on other narrowing conditions than the specific attributes, and the document set adjustment unit 112e narrows down the set based on the search results. This generates a secondary set that is further narrowed down from the primary set. For example, when the primary set is generated based on the specific attribute "name of a substance constituting a product," the search unit 112c further searches the set of patent documents based on other narrowing conditions than the specific attributes. Then, the document set adjustment unit 112e further narrows down the primary set based on the search results. Examples of other narrowing conditions different from specific attributes include "patent classification," "year of application," "applicant," "problem to be solved by the invention," "use of the invention," "bibliographic information," "keywords specified by the user," and combinations thereof. "Patent classification" may include the International Patent Classification, FI, and F-terms. Furthermore, other narrowing conditions different from specific attributes may be other attributes that were not used when generating the primary set. In this case, as in the case of a search based on technical information, the other attributes that were not used when generating the primary set are not limited to the technical information itself but may also be information derived from the technical information. In particular, the other attribute that was not used when generating the primary set is preferably the "target field of the product." Furthermore, other narrowing conditions different from specific attributes may be a user's evaluation of the set of patent documents. The document set adjustment unit 112e may repeatedly narrow down the search results each time a narrowing condition is acquired. If the document set adjustment unit 112e receives an instruction from the terminal 200 to cancel the narrowing down based on the narrowing conditions, it may cancel the narrowing down and return to the original document set.

[0076] As described above, the "product target field" may be included as another narrowing-down condition different from a specific attribute. The "product target field" may be, for example, the healthcare field, the energy field, or the semiconductor field, and may be a field or area that the user company targets for business. In other words, the "product target field" may be, for example, a field in which the user hopes to expand their business currently or in the future. The "product target field" may be identified by a keyword or patent classification. By including the "product target field" in one or more attributes, it is possible to provide the user with a set of patents that includes, in particular, the field in which the product that the user ultimately wants to use will enter. The "product target field" may be entered into the terminal 200 as another narrowing-down condition different from a specific attribute when the user enters technical information.

[0077] For example, the document set adjustment unit 112e further narrows down the generated primary set of patent documents based on the problem to be solved by the invention and the use of the invention. When the search unit 112c further performs a search based on the use of the invention or the problem to be solved by the invention as a narrowing-down condition, the predetermined area to focus the search may be set in the same way as the use of the product or the problem to be solved by the product in the attributes of the technical information. Note that some patent databases may pre-link each patent document to the problem to be solved by the invention or the use of the invention. When the document set adjustment unit 112e further narrows down the search based on the problem to be solved by the invention and the use of the invention, it may use such a patent database to perform the further narrowing-down.

[0078] The document set adjustment unit 112e also narrows down the set of patent documents based on the non-patent information acquired by the non-patent information acquisition unit 112j. For example, the document set adjustment unit 112e narrows down the set of patent documents by extracting patent documents whose inventor names match the names of the authors of the non-patent information from the set of patent documents based on the names of the authors of the non-patent information. The document set adjustment unit 112e also narrows down the set of patent documents by extracting patent documents whose applicant names match the names of the organizations to which the authors of the non-patent information belong based on the names of the organizations to which the authors of the non-patent information belong. This improves the accuracy of the search results for patent documents.

[0079] The document set adjustment unit 112e may also adjust the set of patent documents into a patent map according to the narrowing-down conditions. The document set adjustment unit 112e may further narrow down the set of patent documents based on a user's evaluation of the set of patent documents.

[0080] The priority setting unit 112f sets priorities for patent documents determined by the search unit 112c to contain descriptions that match or are similar to the technical information. For example, the priority setting unit 112f sets a higher priority for a document when providing the patent document to a user, the higher the degree of match between the technical information and the description in the patent document. For example, the priority setting unit 112f sets a higher priority for a document that matches the technical information to a higher degree of match between the technical information and the description in a specified section corresponding to the attribute. For example, the priority setting unit 112f sets a higher priority for a document that matches the technical information to a higher degree of match between the technical information attribute and a description in a specified section corresponding to the attribute than for a document that matches the technical information attribute to a higher degree of match between the technical information attribute and other descriptions other than the specified section.

[0081] When the non-patent information acquired by the non-patent information acquisition unit 112j corresponds to a specific patent document included in the set of patent documents, the priority setting unit 112f sets a high priority for the specific patent document and the non-patent information corresponding to the specific patent document. For example, based on the name of the author of the non-patent information, if the name of the inventor of the patent document matches the name of the author of the non-patent information, the priority setting unit 112f sets a high priority for providing these patent documents and non-patent information to the user. Furthermore, for example, if the name of the organization to which the author of the non-patent information belongs matches the name of the applicant of the patent document, the priority setting unit 112f sets a high priority for providing these patent documents and non-patent information to the user.

[0082] Furthermore, the higher the value of the organization that is the applicant of the patent document or the organization to which the author of the non-patent information belongs, the higher the priority setting unit 112f sets the priority of the patent document or non-patent information when providing it to a user. The value of an organization can be, for example, its market capitalization. The value of an organization is not limited to market capitalization, but may also be its market share in the field of technical information. The value of an organization may also be determined from the patent score of the patents filed by the organization.

[0083] The providing unit 112g provides the set of patent documents generated by the document set generation unit 112d to the user. The providing unit 112g also provides the set of patent documents generated by the document set generation unit 112d, along with the non-patent information acquired by the non-patent information acquisition unit 112j, to the user. Specifically, the providing unit 112g transmits the set of patent documents generated by the document set generation unit 112d and the non-patent information acquired by the non-patent information acquisition unit 112j to the terminal 200 as search results. When the document set adjustment unit 112e adjusts the patent documents included in the document set, the providing unit 112g transmits the adjusted set of patent documents to the terminal 200 as search results to provide to the user. When the priority setting unit 112f sets priorities for patent documents or non-patent information, the providing unit 112g transmits the priorities along with the set of patent documents or non-patent information. Furthermore, when the document set adjustment unit 112e adjusts the set of patent documents into a patent map according to the narrowing conditions, the providing unit 112g transmits the patent map to the terminal 200 to provide it to the user.

[0084] The prompt creation unit 112h creates a prompt for searching for non-patent information based on the technical information. A prompt refers to instructions or information input to the generation AI. For example, if the name of a substance that constitutes a product represented by the technical information is substance A, the prompt creation unit 112h creates a prompt such as, "Please tell me non-patent information that describes substance A." The prompt creation unit 112h also creates a prompt for acquiring new technical information from the generation AI server 300 based on technical information specified by the user. For example, if the name of a substance that constitutes a product represented by the technical information is substance A, the prompt creation unit 112h creates a prompt such as, "Please tell me substances that have similar properties, structure, or function to substance A."

[0085] The prompt sending unit 112i sends the prompt created by the prompt creation unit 112h to the generation AI server 300.

[0086] The non-patent information acquisition unit 112j acquires non-patent information searched based on technical information. Specifically, the non-patent information acquisition unit 112j acquires non-patent information transmitted from the generation AI server 300 as a response to a prompt. The non-patent information acquisition unit 112j may acquire non-patent information stored in another data server by the search server 100 communicating with the other data server.

[0087] The learning unit 112k learns an algorithm for the search unit 112c to search for patent documents based on the user's evaluation of the set of patent documents. As an example, the learning unit 112k uses technical information and narrowing down conditions as input and machine-learns a neural network based on training data in which a set of patent documents excluding patent documents evaluated by the user as noise is output, thereby creating a trained model. Such a trained model has trained a search formula and search algorithm to exclude technical information and patent classifications related to patent documents evaluated by the user as noise. Therefore, by using the trained model in subsequent searches, patent documents evaluated by the user as noise will not be included in the set, and the user will be able to obtain search results that are more desirable.

[0088] Next, a description will be given of functional blocks included in the second processor 212 of the terminal 200. As an example, as shown in FIG. 6, the second processor 212 includes an input information acquisition unit 212a, a transmission unit 212b, a search result reception unit 212c, and a display processing unit 212d.

[0089] The input information acquisition unit 212a acquires information input by the user of the terminal 200 by operating the input device 230. Specifically, the input information acquisition unit 212a acquires technical information input by the user. The terminal 200 displays a screen on the display device 220 where the user inputs technical information. As shown in FIG. 7 , the display device 220 displays "Name of materials constituting the product," "Structural characteristics of the product," "Functions that the product can provide," and "Problems solved by the product" as one or more attributes of the technical information. As shown in FIG. 7 , the display device 220 further displays "Product use," "Product purpose," and "Product technical field" as one or more attributes of the technical information. As described above, when inputting technical information, the user may input a "target field of the product" into the terminal 200 as another narrowing condition different from a specific attribute. For this reason, the display device 220 may display the "target field of the product" as shown in FIG. 7 .

[0090] The user can operate the input device 230 while referring to the screen shown in Fig. 7 to input attributes of technical information relevant to the product in the input field 232. The user can also operate the input device 230 while referring to the screen shown in Fig. 7 to input the "target field of the product" in the input field 232. The user does not need to input all of the attributes shown in Fig. 7, and can input only those attributes shown in Fig. 7 that are relevant to the product.

[0091] Furthermore, the input information acquiring unit 212a acquires search range designation information when the user inputs search range designation information to the input device 230. By inputting the search range designation information, the user can set a desired predetermined location other than or instead of the predetermined location shown in Fig. 5 .

[0092] The input information acquiring unit 212a also acquires the conditions for narrowing down the set of patent documents input by the user, and also acquires the evaluation of the search results input by the user.

[0093] The transmitting unit 212b transmits the information acquired by the input information acquiring unit 212a to the search server 100. The transmitting unit 212b transmits the information entered in the input field 232 shown in FIG. 7 as technical information consisting of one or more attributes to the search server 100. The transmitting unit 212b also transmits identification information indicating which of the multiple input fields 232 the technical information was entered in to the search server 100. In other words, the identification information is information indicating which of one or more attributes the technical information corresponds to. Therefore, the search server 100 can specify a predetermined location that is the search range of patent documents based on the identification information.

[0094] The search result receiving unit 212c receives the search results sent from the search server 100. When a priority has been set for a patent document or non-patent information by the priority setting unit 112f of the search server 100, the search result receiving unit 212c receives the priority.

[0095] The display processing unit 212d performs processing to display the search results received from the search server 100 on the display device 220 of the terminal 200. When the search result receiving unit 212c receives a priority, the display processing unit 212d displays the search results on the display device 220 according to the priority. This allows the user to view patent documents or non-patent information included in the collection of patent documents according to the priority, making it easy to find the desired patent document or non-patent information. The search results may also be output as audio by the speaker 240.

[0096] In the process in which the user narrows down the primary set using the narrowing conditions and then further narrows down to the secondary and tertiary sets, the search results for each of the primary, secondary, and tertiary sets are displayed sequentially as they are generated on the display device 220. Therefore, the user can refer to these search results and set desired narrowing conditions to narrow down the set next.

[0097] The search server 100 or the terminal 200 narrows down the primary set 600 of patent documents generated by the search server 100 based on the narrowing conditions input by the user. On the terminal 200, the narrowed down primary set 600 of patent documents is displayed on the screen of the display device 220, as shown in Fig. 8. When the application year and patent classification are used as the narrowing conditions, the primary set 600 is narrowed down to a first secondary set 610 by the document set adjustment unit 112e of the search server 100. The narrowed down first secondary set 610 is then displayed on the first screen 220a of the display device 220.

[0098] Furthermore, when the application year and applicant are used as narrowing conditions, the primary set 600 is narrowed down to a second secondary set 620 by the document set adjustment unit 112e of the search server 100. The narrowed down second secondary set 620 is then displayed on the second screen 220b of the display device 220.

[0099] Furthermore, when the application year and the use of the invention are used as narrowing conditions, the primary set 600 is narrowed down to a third secondary set 630 by the document set adjustment unit 112e of the search server 100. The narrowed third secondary set 630 is then displayed on the third screen 220c of the display device 220. Note that the use of the invention or the problem to be solved by the invention as narrowing conditions is used to further narrow down the generated set of patent documents. Therefore, the use of the invention or the problem to be solved by the invention as narrowing conditions may correspond to the use of the invention or the problem to be solved by the invention described in the patent document. Therefore, the use of the invention or the problem to be solved by the invention as narrowing conditions may be different from the use of the product or the problem to be solved by the product in the attributes of the technical information.

[0100] When priorities are set for patent documents by the priority setting unit 112f of the search server 100, the display device 220 displays the patent documents in accordance with the priority on the first screen 220a, the second screen 220b, or the third screen 220c. For example, the patent documents are displayed in descending order of priority. Also, for example, the patent documents are displayed with labels indicating the priority. For example, the terminal 200 causes the display device 220 to display labels indicating the priority, such as "high priority," "medium priority," or "low priority."

[0101] The first screen 220a, the second screen 220b, or the third screen 220c may also display the description of the patent document, such as the claims or abstract, along with the patent number of the patent document. When the user clicks on the patent document number, the screen may transition to a link to a patent document database, where detailed information about the patent document is displayed. The information or the URL of the link displayed on the display device 220 is included in the collection of patent documents transmitted from the search server 100 to the terminal 200. The first screen 220a, the second screen 220b, or the third screen 220c can be scrolled by the user of the terminal 200 operating the input device 230.

[0102] As an example, Figure 9 shows a bubble diagram displaying a primary set 600 of patent documents generated by the search server 100, narrowed down by multiple different invention uses A-C and multiple problems A-D that the inventions aim to solve. In this example, the search server 100 acquires the invention uses A-C and problems A-D that the inventions aim to solve as the narrowing criteria. In Figure 9, the size of each bubble 700 indicates the number of patent documents. By displaying a bubble diagram with a matrix of invention uses A-C and problems A-D, as in Figure 9, the user can obtain search results that are a secondary set of patent documents corresponding to the product, further narrowing down the primary set of patent documents generated based on the product's technical information. In the example of Figure 9, the search can be further narrowed down to patent documents describing uses or problems that match the product's use or the problem that the product solves. In the example of Figure 9, the search can be further narrowed down to patent documents describing uses or problems similar to the product's use or the problem that the product solves. By selecting between uses A to C and problems a to d, users can discover that the product can be used for unexpected purposes or as a means of solving unexpected problems.

[0103] 10 shows an example in which the search results are further narrowed down based on the narrowing keywords entered by the user in FIG. 9 . The example shown in FIG. 10 shows an example in which the bubble diagram in FIG. 9 is narrowed down based on the keyword "decarbonization," which indicates a technology trend. In this case, the search server 100 acquires the keyword "decarbonization" as a narrowing condition. By further narrowing down the bubble diagram using keywords indicating technology trends as shown in FIG. 10 , the user can obtain search results that are a tertiary set of patent documents that are further narrowed down to patent documents corresponding to the technology trend.

[0104] A user viewing the first screen 220a, the second screen 220b, or the third screen 220c in FIG. 8 can further narrow down the set of patent documents. FIG. 11 shows the user further narrowing down the application year and applicant from the second screen 220b. More specifically, FIG. 11 shows the user narrowing down the application year to between 2021 and 2023. In FIG. 11, the "applicant," "patent classification," "use of the invention," and "problem to be solved by the invention" are displayed for the narrowed-down patent documents. As described above, some patent databases may pre-assign the problem to be solved by the invention or the use of the invention to each patent document. In FIG. 11, the "use of the invention" and "problem to be solved by the invention" may be displayed using such a patent database.

[0105] Similarly, when a user clicks on an individual bubble 700 in the bubble diagram of Figure 9 or 10, the "applicant" and "patent classification" may be displayed for patent documents narrowed down by the use of the invention and the problem that the invention is intended to solve. Furthermore, when a user clicks on an individual bubble 700, the "use of the invention" and the "problem that the invention is intended to solve" may be displayed. Furthermore, the entire "abstract" or "specification" may be displayed. In this case, if the user further narrows down the search by application year and applicant, a screen similar to Figure 11 may be displayed.

[0106] For example, assuming the present is 2024, the user can recognize that companies A, B, D, and E have been conducting research and development of patent documents that match the technical information of their company's products over the past three years based on the narrowed-down results in Figure 11. As an example, the search results in Figure 11 show patent documents in which the uses of the invention are "photovoltaic power generation," "solar panels," and "storage batteries" when the product use in the technical information is "power generation."

[0107] A user who has viewed the search results in Fig. 11 can further narrow down the patent documents, as shown in the box indicated by the arrow A1 in Fig. 11. An example in which the use of the invention is further limited as a narrowing-down condition is shown in Fig. 12. In Fig. 12, by limiting the use of the invention, documents in which the use of the invention shown in Fig. 11 is "storage battery," i.e., documents in which the applicant is Company e, are excluded from the search results.

[0108] A user who visually views the search results in FIG. 12 can evaluate the search results. As an example, FIG. 13 shows an example of evaluation in which patent documents for which the problems to be solved by the invention are "increased power generation" and "increased power generation efficiency" are useful to the user, while patent documents for which the problem to be solved by the invention is "cost reduction" are not useful to the user. In the example of FIG. 13, the user evaluates patent documents for which the problems to be solved by the invention are "increased power generation" and "increased power generation efficiency," i.e., documents by applicants A and B, as "useful." The user also evaluates patent documents for which the problem to be solved by the invention is "cost reduction," i.e., documents by applicant D, as "noisy." These evaluations are input by operating the input device 230 and acquired by the input information acquisition unit 212a. The evaluations are then transmitted to the search server 100 by the transmission unit 212b.

[0109] Next, a chronological flow of processing performed by the information retrieval system 1000 will be described. The basic sequence of processing performed by the information retrieval system 1000 is shown in Fig. 14. First, the input information acquisition unit 212a of the user's terminal 200 acquires technical information (step S200). Next, the transmission unit 212b of the user's terminal 200 transmits the technical information to the search server 100 (step S202).

[0110] Next, the technical information acquisition unit 112a1 of the search server 100 acquires the technical information transmitted from the terminal 200 (step S100). Next, the search range designation unit 112b of the search server 100 designates a predetermined portion of the patent document as the search range for the patent document based on the attributes of the technical information (step S102). Next, the prompt creation unit 112h of the search server 100 creates a prompt (step S104), and the prompt transmission unit 112i transmits the prompt to the generation AI server 300 (step S106).

[0111] In step S106, if a prompt is sent to obtain new technical information from the technical information specified by the user, the generation AI server 300 generates new technical information from the technical information specified by the user as a response (step S300).Then, the generation AI server 300 transmits the new technical information generated from the technical information specified by the user to the search server 100 (step S302).

[0112] Next, the search unit 112c of the search server 100 performs a focused search of predetermined sections of the patent documents based on the attributes of the technical information and determines whether the patent documents contain descriptions that match or are similar to the technical information (step S108). As described above, the search unit 112c may also perform a search based on information obtained from the technical information. Next, the document set generation unit 112d of the search server 100 generates a set of patent documents that contain descriptions that match or are similar to the technical information based on the search results of the search unit 112c (step S110). Next, the priority setting unit 112f sets priorities for the patent documents that the search unit 112c determines contain descriptions that match or are similar to the technical information (step S112).

[0113] If a prompt to obtain non-patent information is sent in step S106, the generation AI server 300 generates non-patent information as a response to the prompt (step S304), and then transmits the non-patent information to the search server 100 (step S306).

[0114] Next, the non-patent information acquisition unit 112j of the search server 100 acquires the non-patent information transmitted from the generation AI server 300 (step S114). Next, the document set adjustment unit 112e of the search server 100 adjusts the patent documents included in the document set by narrowing down the set of patent documents based on the non-patent information acquired by the non-patent information acquisition unit 112j (step S116). Next, the provision unit 112g of the search server 100 provides the set of patent documents generated by the document set generation unit 112d and adjusted by the document set adjustment unit 112e to the user (step S118). As a result, the set of patent documents is transmitted to the user's terminal 200. Note that when the document set adjustment unit 112e narrows down the set of patent documents based on the non-patent information, it is sufficient for the document set adjustment unit 112e to collect the non-patent information before adjusting the patent documents; the search server 100 does not necessarily have to create or send a prompt before searching for patent documents.

[0115] Next, the search result receiving unit 212c of the user's terminal 200 receives the collection of patent documents sent from the search server 100 as the search result, and the display processing unit 212d of the user's terminal 200 displays the search result on the screen of the display device 220 (step S204).

[0116] In step S116, the document set adjustment unit 112e may narrow down the search results based on other narrowing conditions. As described above, the narrowing conditions in this case may be other attributes that were not used when generating the primary set. If the other attribute that was not used when generating the primary set is the "target field of the product," in step S204, the set of patent documents narrowed down by the target field of the product is automatically displayed on the screen of the display device 220.

[0117] Next, the transmitting unit 212b of the user's terminal 200 transmits the conditions for narrowing down the set of patent documents that the user has input to the input device 230 to the search server 100 (step S206).

[0118] Next, the document set adjustment unit 112e of the search server 100 adjusts the patent documents included in the document set by narrowing down the set of patent documents based on the narrowing conditions transmitted from the user's terminal 200 (step S120). Next, the providing unit 112g of the search server 100 provides the set of patent documents adjusted again by the document set adjustment unit 112e to the user (step S122). As a result, the set of patent documents is transmitted to the user's terminal 200.

[0119] In step S120, the document set adjustment unit 112e may also narrow down the set of patent documents based on the "target field of the product." As a result, the set of patent documents narrowed down based on the target field of the product is automatically displayed on the screen of the display device 220.

[0120] Since the document set adjustment unit 112e has a function of adjusting the set of patent documents into a patent map according to the narrowing conditions, the document set adjustment unit 112e may adjust the set of patent documents into a patent map in step S116 or step S120. The patent map is transmitted to the terminal 200 in step S118 or step S122.

[0121] Next, the search result receiving unit 212c of the user's terminal 200 receives the set of patent documents transmitted from the search server 100 as the search result, and the display processing unit 212d of the user's terminal 200 displays the search result on the screen of the display device 220 (step S208). Next, the transmitting unit 212b of the user's terminal 200 transmits the evaluation of the search result input by the user to the input device 230 to the search server 100 (step S210).

[0122] After step S208, if the user inputs a narrowing-down condition into the input device 230, the processes of steps S206, S120, S122, and S208 may be repeated. As described above, the user can obtain desired search results by repeatedly narrowing down the search results by selecting and rejecting, for example, the uses of multiple different inventions and the problems that multiple different inventions aim to solve.

[0123] Next, the learning unit 112k of the search server 100 learns an algorithm to be used when the search unit 112c searches for patent documents based on the evaluation of the set of patent documents transmitted from the user terminal 200 (step S124). Note that this learning may be performed by the generation AI server 300.

[0124] After step S210, the document set adjustment unit 112e may adjust the patent documents included in the document set by narrowing down the set of patent documents based on the evaluation of the set of patent documents transmitted from the user's terminal 200. Then, the providing unit 112g of the search server 100 may provide the set of patent documents adjusted by the document set adjustment unit 112e to the user. Even in this case, if the user further inputs narrowing conditions into the input device 230, the processes of steps S206, S120, S122, and S208 may be repeated.

[0125] In the first embodiment, if the user inputs the product's intended use as a business policy in addition to technical information about the product, the information retrieval system 1000 may output a final patent list all at once without displaying the progress of the process. Furthermore, the information retrieval system 1000 may output the final patent list all at once and then return to each intermediate step to continue processing. In this case, the information retrieval system 1000 may, for example, output the final patent list all at once and then return to the intermediate step to display a patent document narrowing screen or a bubble diagram. That is, in the first embodiment, the user can input narrowing conditions to switch between the first screen 220a, the second screen 220b, and the third screen 220c shown in FIG. 8 and the bubble diagram shown in FIG. 9 on the display device 220. The user can then narrow down the desired patent documents by performing input operations while referring to the results displayed on the screen.

[0126] As described above, according to the first embodiment, a search is performed intensively on predetermined sections of patent documents determined according to the individual attributes of technical information related to a product, and it is determined whether the patent documents contain descriptions that match or are similar to the technical information. Therefore, market needs that match the product attributes can be easily searched for while minimizing noise, making it possible to extract companies that can use the product. In other words, the information search system of this embodiment can perform business matching to find candidates who are seeking the target product.

[0127] 2. Second Embodiment The second embodiment relates to a system that allows a user to easily search for useful information related to a product, for example, a product that a company is developing or has already developed. In this system, when a user inputs technical information about the product, the system can provide the user with the patent documents that the user desires as a final deliverable.

[0128] For example, if a user inputs "product use" as an attribute of technical information, patent documents related to uses in which the product could potentially be used are provided to the user. This product use may be an use that was not originally envisioned by the company that developed the product. The user can understand the uses in which the product could potentially be used based on the patent documents provided as the final deliverable. In addition to the uses in which the product could potentially be used, the user can further narrow down and output search results by inputting, for example, a new field in which the company that developed the product should enter or a field that the product is targeting in advance.

[0129] In the second embodiment, the first processor 112 is configured in the same manner as in the first embodiment shown in Fig. 4. In the second embodiment, when outputting the final product, the document set generation unit 112d does not generate a document set, and the document set adjustment unit 112e does not adjust the document set, and the search results by the search unit 112c are provided to the user as is.

[0130] In other words, in the second embodiment, when a final product is output, the document set generation unit 112d does not have to generate a document set, and the document set adjustment unit 112e does not have to adjust the document set. Also, in the second embodiment, when a final product is output, the document set generation unit 112d may generate a document set, and the document set adjustment unit 112e may adjust the document set. In the information retrieval system of the second embodiment, the technical information acquisition unit 112a1, the document set generation unit 112d, and the document set adjustment unit 112e may be configured using the generation AI server 300.

[0131] For this reason, in the second embodiment, a search algorithm that has been trained by the training unit 112k is used. In the trained search algorithm, a search algorithm desired by the user has been learned by previously generating and adjusting a document set. Therefore, when the search unit 112c performs a search based on the search algorithm, the patent document desired by the user is retrieved without generating or adjusting a document set, and the retrieved patent document can be provided to the user as a final product.

[0132] Specifically, when a user attempts to perform a search multiple times using the method of the first embodiment, the learning unit 112k learns the search algorithm. When outputting the final product, the search unit 112c performs a search using the search algorithm learned by the learning unit 112k.

[0133] As described in the first embodiment, learning by the learning unit 112k may be performed based on training data, which takes technical information and narrowing conditions as input and outputs a set of patent documents excluding patent documents that the user has evaluated as noise from the set of patent documents. When the user attempts multiple searches and evaluates the set of patent documents, multiple sets of training data are obtained. The neural network trained by machine learning based on the training data thus obtained reflects the user's evaluation, and therefore, when new technical information is input to the neural network, patent documents corresponding to the user's evaluation are output.

[0134] Therefore, the search unit 112c can search for patent documents based on new technical information using the search algorithm learned by the learning unit 112k. In this case, for example, if a product use that was not originally anticipated by the company that developed the product is input as technical information, the search unit 112c can search for patent documents related to such use. Because the generation and adjustment of the set of patent documents has already been reflected in the search algorithm through learning, the number of patent documents returned as a search result is reduced. Therefore, the search result is output as a list of relevant patent documents.

[0135] Next, a chronological flow of processing performed by the information retrieval system 1000 according to the second embodiment will be described. The sequence of processing performed by the information retrieval system 1000 according to the second embodiment is shown in Fig. 15. It is assumed here that the search algorithm has already been learned and stored in the storage device 120 by the processing of step S124 in Fig. 14 according to the first embodiment.

[0136] First, the input information acquisition unit 212a of the user terminal 200 acquires technical information (step S220), and then the transmission unit 212b of the user terminal 200 transmits the new technical information to the search server 100 (step S222).

[0137] Next, the technical information acquisition unit 112a1 of the search server 100 acquires the new technical information transmitted from the terminal 200 (step S130). Next, the search unit 112c of the search server 100 acquires the learned search algorithm stored in the storage device 120 (step S132).

[0138] Next, the search unit 112c of the search server 100 searches for patent documents using a search algorithm based on the attributes of the technical information, and searches for patent documents that contain descriptions that match or are similar to the technical information (step S134).

[0139] Next, the providing unit 112g of the search server 100 provides the patent documents searched by the searching unit 112c to the user (step S136), whereby a list of patent documents is sent to the user's terminal 200.

[0140] Next, the search result receiving unit 212c of the user's terminal 200 receives the list of patent documents sent from the search server 100 as the search result, and the display processing unit 212d of the user's terminal 200 displays the search results on the screen of the display device 220 (step S224).

[0141] FIG. 16 shows an example of search results displayed on the screen of the display device 220. In this example, the search results for a certain product are displayed by setting the product's use as "battery electrodes" as an attribute of technical information. The display is performed in the same manner as in the first embodiment. That is, the description of the patent document may be displayed together with the patent number of the patent document, such as the claims or abstract. Furthermore, when the user clicks on the number of the patent document, the screen may transition to a link to a patent document database, where detailed information about the patent document is displayed. The information displayed on the display device 220 or the URL of the link is included in the list of patent documents transmitted from the search server 100 to the terminal 200.

[0142] The search results in FIG. 16 indicate that the uses of the inventions related to the retrieved patent documents are "battery electrodes" and "batteries." By checking the descriptions in the patent documents, a user viewing the screen in FIG. 16 can determine whether the product can be used for "battery electrodes." Based on the results of this determination, the user can decide what new field the company that developed the product should enter.

[0143] As described above, according to the second embodiment, by inputting technical information about a product, patent documents containing useful information related to the product are provided to the user based on the trained search algorithm. Therefore, by carefully examining the descriptions in the provided patent documents, the user can, for example, consider and decide on new fields in which the company that developed the product should enter.

[0144] The information retrieval system 1000 may operate the processing of the first embodiment and the processing of the second embodiment separately, or may switch between the processing of the first embodiment and the processing of the second embodiment according to a user's request. A request to switch between the processing modes may be input by the user to the terminal 200 and transmitted from the terminal 200 to the search server 100. For example, in the first embodiment, if the user inputs the product's intended use as a business policy as technical information related to the product, switching to the processing of the second embodiment can switch to a processing mode that outputs a final patent list all at once. Also, as long as the user inputs the product's intended use as a final business policy, the processing of the second embodiment can output a final patent list all at once, and then switching to the processing of the first embodiment. By switching to the processing of the first embodiment, the user can review the bubble diagram and reassess the search results.

[0145] 3. Third Embodiment Next, a third embodiment will be described, which further embodies the user interface of the first or second embodiment. The third embodiment adds user-input information to the first or second embodiment. Furthermore, in the third embodiment, document search, document set generation, and document set adjustment are performed primarily using the large-scale language model of the generation AI server 300. In other words, in the information retrieval system of the third embodiment, the technical information acquisition unit 112a1, search unit 112c, document set generation unit 112d, and document set adjustment unit 112e may be configured using the generation AI server 300. Furthermore, in the third embodiment, matching information is created that represents the degree of match between a searcher searching for a source or potential recipient of the target technology and the candidate. In the third embodiment, the configurations of the first processor 112 and the second processor 212 are basically the same as those of the first embodiment shown in FIGS. 4 and 6 , and therefore, descriptions overlapping with those of the first embodiment will be omitted as appropriate. As shown in FIG. 17, the first processor 112 has the same configuration as that of FIG. 4, but further includes a candidate designator 112m and an evaluator 112n.

[0146] Input screens displayed on the display device 220 in the third embodiment are shown in FIGS. 18 to 20. The following description will be given taking as an example a case where a user searches for a partner related to "Mg hydroxide," a proprietary technology of the company. The user operates the input device 230 to click on a first tab 800, a second tab 810, or a third tab 820 shown in FIGS. 18 to 20. This switches between the screens shown in FIGS. 18 to 20. When the user clicks on the first tab 800, the screen shown in FIG. 18 is displayed. When the user clicks on the second tab 810, the screen shown in FIG. 19 is displayed. When the user clicks on the third tab 820, the screen shown in FIG. 20 is displayed.

[0147] A screen on which a user inputs searcher information, i.e., a persona, is shown in Fig. 18. The searcher information is unique information about a searcher who searches for a potential supplier of the target technology or a potential recipient of the target technology, and may include information about the searcher, the relationship between the searcher and the technical information, or the searcher's requests when searching for a candidate.

[0148] Specifically, unique information about the explorer may include the following information. 1. Information about the explorer 1.1. Basic profile information about the explorer Profile of the explorer as an organization (Examples) - Company / university / organization name, department name - Business field / research area: Affiliated industry, specialty, etc. - Organization size: Number of employees, number of bases, etc. - Business / research area: Geographical scope of activities - Financial situation: Sales / profit scale, investment capacity, fundraising status, etc. - Representative / responsible person information: (Especially in the case of research institutions) Professor name, laboratory head name, etc. 1.2. Strategic information about the explorer Information indicating the explorer's vision and how this match fits into that vision (Examples) - Vision / mission: The direction the organization as a whole is aiming for. - Priority technology fields and business fields and the reasons for them. - Plan: Milestones and roadmap for achieving goals. - Strategic positioning of this need: New business creation, strengthening existing businesses, cost reduction, research theme, etc. 1.3. External environment surrounding the explorer Information about external factors that will affect the explorer's strategy (Examples) ・Market trends: Growth potential and trends in related markets. ・Competitive environment: Trends of major competitors and the state of technological development. ・Policies and regulations: Legal reforms, industry standards, environmental regulations, and other factors that affect business and research. ・Social conditions: Events with wide-ranging impacts, such as trade friction, pandemics, and natural disasters. ・Technology trends: New technologies and paradigm shifts that are attracting attention across the industry. 2. Relationship between the explorer and technical information 2.1. Explorer's capability information Information about the explorer's own strengths, technologies, and activity history (Examples) ・Possessive technologies and research results: Technologies, patents, papers, research themes, products, etc. that have been developed to date. ・Expertise within the organization: Specialized personnel and research systems in specific technical fields. ・Past successes and failures: Lessons learned and achievements from past projects. ・Advantages over competitors: The explorer's own differentiating factors and core competencies.・Related party / customer information: major business partners, joint research partners, customer base, etc. ・Innovation initiatives and track record, etc.3. Requirements of the explorer when searching for candidates 3.1. Searcher's matching request information Information that specifically defines "what they want to solve" and "what they are looking for" (Examples) ・Search target: technology, people, organization, company, university, etc. ・Detailed information on the technology they wish to match with ・Issues / problems to solve: specific problems and bottlenecks they are currently facing. ・Expected effects / goals: what they want to achieve by solving the problem, quantitative and qualitative goals (examples: cost reduction rate, productivity improvement rate, new product development, etc.). ・Desired technology level / maturity: current level of the desired seeds, such as basic research stage, prototype, or already in practical use. ・Project background and importance: why they are tackling this issue now, and its strategic significance to the organization. ・Approximate related budget / time frame: range of resources that can be allocated. 3.2. Hint information for matching Information that serves as hints for efficient matching (Examples) ・Expectations for using the system: what they expect from the matching system.・Desired form of collaboration: What type of collaboration are you considering, such as joint research, technology introduction, or license agreement? ・Intentions regarding non-disclosure agreements (NDA): Your preferences regarding the handling of confidential information. ・Purpose of matching (want to license, transfer technology, sell, etc.) ・Conditions regarding the matching partner (give preference to companies with factories in Japan, want to increase exports to country X, etc.).

[0149] For example, a user can input "company name / organization name," "owned technology / owned business," and "search purpose" as searcher information. Furthermore, a user can input "research and development stage," "existing partner," "priority field," "priority partner," and "matching request" as searcher information. The example in FIG. 18 shows a situation in which "XX University" is input as the "company name / organization name" and "inorganic chemical material" is input as the "priority technology / owned business." The example in FIG. 18 also shows a situation in which "selection of technology licensee" is input as the "search purpose" and "basic research" is input as the "research and development stage." The example in FIG. 18 also shows a situation in which "XX Company" is input as the "existing partner" and "energy industry" is input as the "priority field." The example in FIG. 18 also shows a situation in which "XX Company" is input as the "priority partner," and "matching request" is input as "we would like to complete a match within two years. We would like to prioritize companies with factories in Japan."

[0150] A plurality of templates may be displayed as the searcher information input screen. For example, the user can input a template in advance by clicking the first template input button 802. The user can then select a desired input screen from the plurality of pre-input templates. Examples of templates include a template for when a university is searching for a technology transfer partner, and a template for when a company is searching for a company that owns seed technology. Another example of a template is a template for when a company is searching for a licensee. The search server 100 can also present proposed templates learned from past search conditions, evaluations, or search results. Newly input content can also be used as a template.

[0151] A screen for the user to input external information is shown in Fig. 19. The external information may include information that serves as a reference when determining the degree of matching between a searcher searching for a supplier of the target technology or a candidate recipient of the target technology and the candidate, information that supplements unique information about the searcher, information that supplements search conditions, and information that serves as a reference for document search and display processing.

[0152] External information may include the following: Information that supplements the searcher's own information (e.g., detailed specification data on the company's technology, customer list, etc.) Information that supplements the search conditions (e.g., a list of NG companies, a list of important documents and URLs) Information used as reference for document searches and display processing (e.g., a list for matching patent rights holders and company names for patent documents)

[0153] A user can load, for example, a table-format file into the terminal 200 and specify the loaded file as external information. An example of the file is a CSV file. For example, a user can load a table-format file into the terminal 200 by inserting a recording medium on which the table-format file is recorded into the terminal 200 and clicking the file input button 812. An example of a table-format file is a file that records data such as sales, profits, and CAGR for each company. Another example of a table-format file is a file that records a collection of homepage links for each company. Another example of a table-format file is a file that records a list of a company's customer companies. Another example of a table-format file is a file that records a list of technologies that are subject to technology transfer.

[0154] An example of a case where a user loads a file containing a list of their company's client companies is shown in Figure 19. In this case, the user can enter further external information into the input screen, such as "This list contains our company's client companies. Please use this as a reference when determining whether or not a match is possible."

[0155] FIG. 20 shows a screen where a user inputs search conditions. The search conditions correspond to the technical information described in the first embodiment. The user can input "Technology Description," "Shape," "Function / Performance," and "Known Uses" as search conditions. The user can also input "Patent Classification Priority," "Source of Information Priority," "Country / Period Priority," and "Maximum Number of Results" as search conditions. The example in FIG. 20 shows "Mg Hydroxide" input as "Technology Description" and "Nanoparticles" input as "Shape." The example in FIG. 20 also shows "High Dispersibility / High Flame Retardancy" input as "Function / Performance" and "Resin Additive" input as "Known Uses." The example in FIG. 20 also shows "C08****" input as "Patent Classification Priority" and "Corporate Press Releases, Newspapers, News, Papers, and Policy Information" input as "Source of Information Priority." In addition, the example of Figure 20 shows that "Global, 2020 and beyond" is entered as the "Country and period of importance" and "2000" is entered as the "Maximum number of documents" to be searched.

[0156] Multiple templates may be displayed as the search condition input screen. For example, the user can input a template in advance by clicking the second template input button 822. The user can then select a desired input screen from the multiple pre-input templates. Examples of templates include templates for chemical material manufacturers and templates for universities searching for transfer recipients of pharmaceutical research results. Another example of a template is a template for companies searching for licensees. The search server 100 can also present proposed templates learned from past search conditions, evaluations, or search results. It is also possible to save the input information as a new template.

[0157] Furthermore, the user can select the purpose of the search by operating the input device 230 and clicking a check box 823. Specifically, the user can select whether he / she wants to search for elemental technologies, or whether he / she wants to search for applications, licenses, or purchasers of the technology. In other words, the user can select whether he / she wants to search for seeds or needs. In FIG. 20, items with black check boxes indicate items selected by the user. Items with white check boxes indicate items not selected by the user.

[0158] After completing the above input, the user can generate a primary set by operating the input device 230 to operate the search set generation button 824. The user can also select whether or not to use an inference model by operating the inference model use button 825. Specifically, the user can switch ON / OFF whether or not to use an inference model by operating the inference model use button 825.

[0159] When the user inputs natural language, the inference model is used by turning on the inference model use button 825. On the other hand, when the user inputs a set itself, such as specifying a patent number list, the inference model is not used by turning off the inference model use button 825.

[0160] 18 to 20 is acquired by the input information acquisition unit 212a and transmitted to the search server 100. The information transmitted to the search server 100 is then acquired by the information acquisition unit 112a. In the third embodiment, in addition to the search conditions in FIG. 20, which correspond to the technical information in the first embodiment, the searcher information in FIG. 19 and the external information in FIG. 20 are transmitted to the search server 100. This makes it possible to further improve the accuracy of document search and matching search.

[0161] In the third embodiment, the search unit 112c searches for literature related to the information input by the user on the screens shown in Figures 18 to 20. That is, the search unit 112c searches for literature based on the searcher's unique information, external information, and search conditions. The search unit 112c may search for literature based on the search conditions and either the searcher's unique information or external information. In the third embodiment, the search unit 112c may search for literature using the generation AI server 300. When the search unit 112c searches for literature using the large-scale language model of the generation AI server 300, the prompt creation unit 112h creates a prompt such as the following. In this case, the prompt creation unit 112h creates a prompt for searching for literature. For example, the prompt creation unit 112h creates a prompt such as, "Please collect literature related to the input searcher information, external information, and search conditions." The search unit 112c acquires literature sent from the generation AI server 300 as a response to the prompt. Therefore, the processing of the search unit 112c may be performed by the generation AI server 300.

[0162] The document set generation unit 112d generates a set of documents related to the input information based on the search results of the search unit 112c. The document set generation unit 112d may generate a subset of documents related to the input information. For example, the document set generation unit 112d may generate a subset by sampling the primary set. Then, after the document set adjustment unit 112e adjusts the document search range on the screen of FIG. 22, the document set generation unit 112d may generate the entire set in response to clicking the search execution button 910. Data of the primary set generated by the document set generation unit 112d is shown in FIG. 21. The primary set may include both patent and non-patent documents. In FIG. 21, the information source, document number, title, date information, and company name are shown for each document in the primary set. Note that the information source indicates whether each document is a patent document or a non-patent document.

[0163] An input screen for the importance of each document is shown in FIG. 22. The importance is instruction information for increasing the accuracy of the search. The instruction information for increasing the accuracy of the search may be, but is not limited to, equivalent to the narrowing down conditions described in the first embodiment. FIG. 22 shows a state in which the user has input importance levels "A", "B", and "X" for each document in the primary set shown in FIG. 21. "A" is the highest importance level, followed by "B" and "C" in decreasing order. "X" indicates that the document is unimportant, i.e., noise.

[0164] 22 also shows a column 900 for entering the reason for determining the importance. In column 900, the user has entered "content related to the production of magnesium hydroxide" as the reason for assigning importance level "A." In column 900, the user has entered "content related to users of magnesium hydroxide" as the reason for assigning importance level "B." In column 900, the user has entered "not an essential description, but a footnote-level description" and "only a description related to magnesium oxide" as reasons for assigning importance level X. The reason for determining the importance is also instruction information for improving the accuracy of the search.

[0165] The user does not have to input the importance level for all documents in the primary set. The user may also input only the reason for determining the importance level of each document without inputting the importance level. In this case, the document set adjustment unit 112e may adjust the document set based on the reason for determining the importance level. For example, consider the case where the reasons for determination, "The document contains information about the production of magnesium hydroxide" and "The document contains information about users of magnesium hydroxide," shown in FIG. 22, are input. In this case, the document set adjustment unit 112e may generate a search set based on these reasons for determination.

[0166] As described above, in this embodiment, the "instruction information for improving accuracy" includes the following: Information specifying a search period Information specifying importance Information indicating the reason for determining the importance Information instructing re-execution of a search

[0167] On the screen of Fig. 22 , the user can generate a search set based on the importance or the reason for determining the importance by operating the search execution button 910 using the input device 230. The user can also select a learning method by operating the first button 826 and the second button 828. Specifically, the user can switch ON / OFF whether or not to use the input information of Fig. 22 for learning by operating the first button 826. The user can also switch ON / OFF whether or not to use a unique learning model for search by operating the second button 828.

[0168] The learning unit 112k learns a search algorithm based on the input information in FIG. 22. The learning unit 112k creates a trained model using the primary set shown in FIG. 21 and the importance levels input by the user for the primary set as training data. That is, the learning unit 112k uses the primary set and the importance levels input by the user for each document in the primary set as training data, and trains a neural network to create a trained model. As a result, this trained model is one that has learned a search algorithm based on the user's evaluation of the importance levels of the primary set.

[0169] When the input information is used for learning, that is, when the user turns on the first button 826, the input information in Fig. 22 is used for learning. When a search is performed using a unique learning model, that is, when the user turns on the second button 828, a search is performed based on a learned model that has been learned based on the input information that the user has input up to now.

[0170] In the trained search algorithm, the search algorithm desired by the user is learned based on the evaluation of importance that has already been performed. Therefore, when the search unit 112c performs a search based on the trained search algorithm, the user can perform the search desired. As an example, the learning unit 112k trains a neural network based on training data that outputs documents evaluated as having high importance, thereby creating a trained model. Such a trained model has trained a search formula and search algorithm to extract documents evaluated as having high importance. Therefore, by using the trained model in subsequent searches, search results that are more desired by the user can be obtained.

[0171] Specifically, when the user performs evaluations multiple times, the learning unit 112k learns the search algorithm. When outputting the final product, the search unit 112c performs a search using the search algorithm learned by the learning unit 112k.

[0172] The learning unit 112k learns a search algorithm based on the input information in Fig. 22. The learning unit 112k creates a trained model using the primary set shown in Fig. 21, the importance input by the user for the primary set, and the reason for determining the importance as training data. As a result, this trained model is one that has learned a search algorithm based on the user's evaluation of the primary set.

[0173] When the input information is used for learning, that is, when the user turns on the first button 826, the input information in Fig. 22 is used for learning. When a search is performed using a unique learning model, that is, when the user turns on the second button 828, a search is performed based on a learned model that has been learned based on the input information that the user has input up to now.

[0174] The information acquisition unit 112a acquires instruction information for improving search accuracy, which is input to the terminal 200 by the user operating the input device 230. The document set adjustment unit 112e adjusts the generated document set. Specifically, the document set adjustment unit 112e adjusts the generated document set to generate a search set according to the instruction information for improving search accuracy. The search set may correspond to, but is not limited to, the secondary set described in the first embodiment. The document set adjustment by the document set adjustment unit 112e may include "narrowing the document set" and "expanding the document set" based on the instruction information for improving search accuracy. Furthermore, the document set adjustment by the document set adjustment unit 112e may include "re-searching the document set" based on the instruction information for improving search accuracy.

[0175] The document set adjustment unit 112e may adjust the document set and generate a search set as a result of adjusting the document set by sending a predetermined prompt to the generation AI server 300. Therefore, the processing of the document set adjustment unit 112e may be performed by the generation AI server 300. Note that the prompt may be input by the user from the terminal 200.

[0176] When the document set adjuster 112e adjusts the document set using a large-scale language model, the prompt creator 112h creates the following prompts. In this case, the prompt creator 112h creates a prompt for adjusting the document set. For example, the prompt creator 112h creates a prompt such as, "Please narrow down the document set based on the specified importance." Also, for example, the prompt creator 112h creates a prompt such as, "Please narrow down the document set based on the reason for determining the importance."

[0177] When the user operates the search execution button 910 on the screen of FIG. 22, the screen transitions to the screen of FIG. 23, which displays a matching proposal. The matching proposal is created by the evaluation unit 112n of the processor 112 based on the input information of FIGS. 18 to 20 and a set of documents. The set of documents may be a primary set, a search set, or a search set regenerated by repeatedly executing a search. Specifically, the evaluation unit 112n may create matching information indicating the degree of match between a searcher searching for a candidate who will be a supplier or a recipient of the target technology and the candidate based on a set of patent documents and searcher information. The evaluation unit 112n may also create matching information based on a set of patent documents, searcher information, and external information. The matching information may include information about the candidate itself, related information about the candidate, the degree of match, a reason for determining the degree of match, and matching advice information. The matching proposal is included in the matching information. The matching information is transmitted to the terminal 200 by the providing unit 112g and displayed on the display device 220.

[0178] As shown in Fig. 23, the matching information includes a plurality of candidates as a proposed matching destination. In Fig. 23, "XX Industry" and "YY Chemical" are shown as candidates for the matching destination. The candidates for the matching destination are determined by the candidate determination unit 112m.

[0179] The candidate determination unit 112m determines candidates that may be suppliers or recipients of the technology that is the target of needs discovery or seeds discovery, based on each patent document included in the collection of patent documents. Here, the candidates may be organizations that the user should contact, such as corporations, companies, universities, organizations, individuals, etc. The candidate determination unit 112m determines candidates for each patent document based on the name of the corporation, company, university, or individual that is the applicant of each patent document. More specifically, the candidate determination unit 112m determines candidates corresponding to the applicant by applying the name of the corporation, company, university, or individual that is the applicant to a candidate database. The candidate determination unit 112m may also determine candidates corresponding to the applicant by applying the name of the corporation, company, university, or individual that is the applicant to a candidate trained model. The determined candidates may be the corporation, company, university, or individual that is the applicant. The determined candidates may also be organizations related to the applicant that are derived from the applicant. The candidate designator 112m may use external information to designate candidates based on the names of corporations, companies, universities, or individuals who are applicants.

[0180] The storage device 120 may store a candidate database that associates the applicants of individual patent documents in the patent database with the names of corporations, companies, universities, and individuals associated with the applicants. The storage device 120 may also store a candidate trained model that has been trained by machine learning by associating the applicants of individual patent documents in the patent database with the names of corporations, companies, organizations, individuals, and the like associated with the applicants. The candidate trained model may be a model trained to predict the names of associated companies, other corporations, or individuals from the name of an applicant.

[0181] For example, if a corporation that is an applicant of an individual patent document has been transferred or merged, the candidate designator 112m may designate the transferee or merger as a candidate. Alternatively, the candidate designator 112m may designate the parent company of the corporation that is an applicant of an individual patent document as a candidate.

[0182] The processing by the candidate determination unit 112m may be performed using a large-scale language model. That is, this processing may be performed by the generation AI server 300. When the candidate determination unit 112m determines candidates using a large-scale language model, the prompt creation unit 112h creates a prompt such as the following. In this case, the prompt creation unit 112h creates a prompt for determining candidates based on individual patent documents included in the patent document collection. An example of the prompt is, "Based on the applicants of individual patent documents included in the patent document collection, please determine candidates that may utilize or supply the technology related to the patent documents." In this case, the prompt sending unit 112i may send the set of patent documents together with the prompt to the generation AI server 300.

[0183] FIG. 23 shows the relevance of each matching candidate with the search criteria. FIG. 23 also shows a state in which a user operates an input device to click on the "XX Industry" row on the screen. When the user clicks on the "XX Industry" row, the reasons for determining the relevance between the user's company and "XX Industry" are displayed at the bottom of the screen. This information is included in the matching proposal provided to the user. The matching proposal is created by the evaluation unit 112n based on the user's input information, the contents of each document in the document set, and information on "XX Industry," which is a candidate determined from the document set. The evaluation unit 112n may obtain the matching proposal from the generation AI server 300. In this case, the prompt creation unit 112h creates a prompt such as: "Please determine the relevance between the user's company and "XX Industry" based on the searcher's unique information and external information, and provide the reason for determining the relevance." Therefore, the processing of the evaluation unit 112n may be performed by the generation AI server 300. The prompt may be input by the user from the terminal 200 .

[0184] 23 also shows how patents held by matching candidates and non-patent information related to the matching candidates are displayed. When the user clicks on the row for "XX Industry," a list of patents held by "XX Industry" and non-patent information related to "XX Industry" is displayed on the right side of the screen.

[0185] The screen of FIG. 23 also shows a third tab 912 and a fourth tab 914. When the user operates the input device 230 to click the fourth tab 914, a search set generated by instruction information for improving the search accuracy of FIG. 22 is displayed, as shown in FIG. 24. In FIG. 24, for each document in the search set of FIG. 23, the proposed companies, i.e., candidates, determined from each document by the candidate designator 112m are displayed. Also, FIG. 24 shows a state in which the user has input importance levels "A," "B," and "X" for each proposed company. As in FIG. 22, "A" is the highest importance level, followed by "B" and "C" in decreasing order. "X" indicates that the document is unimportant, i.e., noise.

[0186] 24 also shows a column 900 for entering the reason for determining the importance level, similar to FIG. 23. The user inputs the reason for determining the importance level "A" or "X" in column 900. In FIG. 24, if the user operates input device 230 to click on third tab 912, the screen returns to the state shown in FIG. 23. In this way, the user can check each screen and, if necessary, return to another screen to continue the search. Therefore, the user can search for a desired partner while successively checking the search results.

[0187] The user does not need to input importance for all documents. In this case, the document set adjustment unit 112e may adjust the search set only for documents for which the user has input importance.

[0188] Alternatively, the user may input only the reason for determining the importance of each document without inputting the importance level. In this case, the document set adjustment unit 112e may adjust the document set based on the reason for determining the importance level. For example, assume that the reasons for the determination, "appropriate as a candidate" and "this company is negative about open innovation and will be excluded this time," shown in FIG. 24, are input. In this case, the document set adjustment unit 112e may regenerate the search set based on these reasons for the determination.

[0189] Furthermore, the user can regenerate a search set based on the importance or the reason for determining the importance by operating the re-search button 920 using the input device 230. Similarly to FIG. 22 , the user can select a learning method by operating the first button 826 and the second button 828. Specifically, the user can switch ON / OFF whether or not to use the input information in FIG. 24 for learning by operating the first button 826. Furthermore, the user can switch ON / OFF whether or not to use a unique learning model for search by operating the second button 828.

[0190] The learning unit 112k learns a search algorithm based on the input information of FIG. 24. The learning unit 112k may create a trained model using the search set, the importance level input by the user for the search set, and the reason for determining the importance level as training data. That is, the learning unit 112k may use these as training data to train a neural network and create a trained model. As a result, this trained model will have learned a search algorithm based on the user's evaluation of the search set.

[0191] When the input information is used for learning, that is, when the user turns on the first button 826, the input information in Fig. 24 is used for learning. When a search is performed using a unique learning model, that is, when the user turns on the second button 828, a search is performed based on a learned model that has been learned based on the input information that the user has input up to now.

[0192] In the third embodiment, the search results may also be provided to the user in the form of various maps, such as a map showing the transition of the number of applications and a problem-solving map.

[0193] Next, the chronological flow of processing performed by the information retrieval system 1000 in the third embodiment will be described. The processing performed by the information retrieval system 1000 using the large-scale language model of the generation AI server 300 is shown in FIG. 25. First, the input information acquisition unit 212a of the user's terminal 200 acquires input information entered from the screens shown in FIGS. 18 to 20 (step S230). Next, the transmission unit 212b of the user's terminal 200 transmits the input information to the search server 100 (step S232).

[0194] Next, the information acquisition unit 112a of the search server 100 acquires the input information transmitted from the terminal 200 (step S140). Next, the search unit 112c of the search server 100 searches for documents based on the input information, and the document set generation unit 112d generates a primary set of documents related to the input information (step S142). Next, the provision unit 112g of the search server 100 transmits the primary set to the terminal 200 (step S143).

[0195] Next, the search result receiving unit 212c of the user's terminal 200 receives the primary set sent from the search server 100 as the search result, and the display processing unit 212d of the user's terminal 200 displays the primary set on the screen of the display device 220 (step S234).

[0196] Next, the input information acquisition unit 212a of the user's terminal 200 acquires instruction information for improving search accuracy that the user has input into the input device 230 (step S236). Next, the transmission unit 212b of the user's terminal 200 transmits the instruction information for improving search accuracy to the search server 100 (step S238). The instruction information for improving search accuracy is acquired by the information acquisition unit 112a of the search server 100. Next, the document set adjustment unit 112e of the search server 100 generates a search set by adjusting the primary set based on the instruction information for improving search accuracy transmitted from the user's terminal 200 (step S146). Next, the provision unit 112g of the search server 100 transmits the search set to the terminal 200 (step S148).

[0197] Next, the search result receiving unit 212c of the user's terminal 200 receives the search set sent from the search server 100 as the search result, and the display processing unit 212d of the user's terminal 200 displays the search set on the screen of the display device 220 (step S240).

[0198] After step S148, the evaluation unit 112n of the search server 100 creates a matching proposal based on the search set (step S150). Next, the provision unit 112g of the search server 100 transmits the matching proposal to the terminal 200 (step S152).

[0199] Next, the user's terminal 200 receives the matching proposal, and the display processing unit 212d of the terminal 200 displays the matching proposal on the screen of the display device 220 (step S242).

[0200] As described above, according to the third embodiment, documents are searched for based on searcher information, external information, and search conditions, thereby enabling improved document search accuracy. Furthermore, according to the third embodiment, matching information is created based on the searcher information, external information, and search conditions, representing the degree of matching between a searcher searching for a candidate supplier or recipient of the target technology and the candidate. Therefore, the searcher, who is a user, can determine the validity of the candidate related to the target technology based on the matching information.

[0201] 100 Search server 110 First control device 112 First processor 112a Information acquisition unit 112a1 Technical information acquisition unit 112b Search range designation unit 112c Search unit 112d Document set generation unit 112e Document set adjustment unit 112f Priority setting unit 112g Provision unit 112h Prompt creation unit 112i Prompt transmission unit 112j Non-patent information acquisition unit 112k Learning unit 112m Candidate determination unit 112n Evaluation unit 114 First memory 116 First communication interface 118 First communication bus 120 Storage device 200 Terminal 210 Second control device 212 Second processor 212a Input information acquisition unit 212b Transmission unit 212c Search result reception unit 212d Display processing unit 214 Second memory 216 Second communication interface 218 Second communication bus 220 Display device 220a First screen 220b Second screen 220c Third screen 230 Input device 232 Input field 240 Speaker 300 Generation AI server 500 Communication network 600 Primary set 610 First secondary set 620 Second secondary set 630 Third secondary set 700 Bubble 1000 Information retrieval system

Claims

1. An information retrieval system comprising: a technical information acquisition unit that acquires technical information indicating the attributes of a target product; a search unit that searches for patent documents that contain descriptions that match or are similar to the technical information; and a document set generation unit that generates a set of the patent documents that contain descriptions that match or are similar to the technical information, wherein the document set generation unit generates a set of the patent documents that are determined to contain descriptions that match or are similar to one or more specific attributes, and further comprising a document set adjustment unit that narrows down the set of the generated patent documents based on other narrowing conditions that are different from the specific attributes.

2. The information search system of claim 1, wherein the attributes include at least one of the names of substances that make up the target product, the structural characteristics of the target product, the functions that the target product can provide, the problems that the target product solves, and the uses of the target product.

3. An information search system according to claim 1 or 2, wherein the search unit searches intensively predetermined portions of the patent documents corresponding to individual attributes of the technical information, rather than other portions.

4. An information search system as described in claim 1 or 2, further comprising: a non-patent information acquisition unit that acquires non-patent information searched based on the technical information; and a provision unit that provides the non-patent information to a user together with the collection of patent documents.

5. An information retrieval system according to claim 4, wherein the document set adjustment unit narrows down the set of patent documents based on the non-patent information.

6. An information retrieval system as described in claim 4, further comprising a priority setting unit that sets a high priority to the specified patent document or the non-patent information corresponding to the specified patent document when the non-patent information corresponds to a specified patent document included in the set of patent documents.

7. An information retrieval system according to claim 1 or 2, further comprising a learning unit that learns an algorithm for the search unit to search the patent documents based on a user's evaluation of the set of patent documents.

8. The information retrieval system of claim 7, wherein the search unit searches for patent documents containing descriptions that match or are similar to the technical information newly acquired by the technical information acquisition unit based on the algorithm learned by the learning unit, and further comprises a provision unit that provides the patent documents searched for based on the algorithm to a user.

9. An information search system according to claim 1 or 2, further comprising a priority setting unit that sets a higher priority for a patent document the higher the degree of match between the technical information and the description of the patent document.

10. An information search system as described in claim 1 or 2, wherein the document set adjustment unit adjusts the set of patent documents into a patent map according to the narrowing conditions, and further includes a provision unit that provides the patent map to a user.

11. An information retrieval system as described in claim 1 or 2, wherein the document set adjustment unit further narrows down the set of patent documents based on a user's evaluation of the set of patent documents, and further comprises a provision unit that provides the user with the further narrowed down set of patent documents.

Images