Technical information classification system and method for classifying technical information

Abstract

A technical information classification system (1) comprises an acquisition unit that acquires technical information in which technical matters are included, a classification unit that classifies the technical information in accordance with a classification criterion, and a display device. The classification unit includes: a language processing model (430); and a retrieval-augmented generation (RAG) data storage unit (420) in which a data source for causing the language processing model to execute language processing on the basis of RAG is stored. The RAG data storage unit (420) includes a classification criterion data storage unit (422) in which a data source relating to the classification criterion is stored in a format conforming to the RAG. The data source relating to the classification criterion includes a plurality of pieces of classification information established according to the classification criterion. The classification unit determines classification information corresponding to the technical information using the language processing model (430) and the data source relating to the classification criterion, and displays the classification information on the display device in the manner in which the technical information and the classification information are associated with one another.

Description

Technical information classification system and method for classifying technical information

[0001] TECHNICAL FIELD The present disclosure relates to a technical information classification system and method for classifying technical information.

[0002] Various classification information is known for classifying technical information such as papers and patents. For example, IPC classification and F-terms are known as classification information related to patents. By classifying technical information according to the classification information and managing the technical information in association with the classification information, users can appropriately search for the technical information they need.

[0003] For example, Patent Document 1 (JP 2019-207610 A) describes a technology that uses artificial intelligence to assign patent classifications to papers and the like.

[0004] Japanese Patent Application Laid-Open No. 2019-207610

[0005] As described in Patent Document 1, by automating technology classification using artificial intelligence, the burden of classification work can be reduced and technical information can be uniformly classified.

[0006] Recently, there has been a trend toward utilizing large-scale language models, which are rapidly advancing, for text analysis. Therefore, it is conceivable to apply large-scale language models to the classification of technical information. However, large-scale language models have the characteristic that it is difficult to obtain precise answers because they make inferences during the process of generating answers. Therefore, there is a problem that it is difficult to appropriately classify technical information by simply using general large-scale language models.

[0007] The present disclosure has been made to solve the above-mentioned problems, and its purpose is to appropriately classify technical information.

[0008] A technical information classification system relating to a first aspect of the present disclosure includes an acquisition unit that acquires technical information including technical matters, a classification unit that classifies the technical information in accordance with classification criteria, and a display device, wherein the classification unit includes a language processing model and a RAG (Retrieval-Augmented Generation) data storage unit that stores a data source for causing the language processing model to perform language processing based on RAG, the RAG data storage unit includes a classification criteria data storage unit that stores the data source related to the classification criteria in a format that complies with RAG, and the data source related to the classification criteria includes a plurality of classification information defined by the classification criteria, and the classification unit determines classification information corresponding to the technical information using the language processing model and the data source related to the classification criteria, and displays the technical information on the display device in a state where it is associated with the classification information.

[0009] A method according to a second aspect of the present disclosure is a method for classifying technical information, the method causing a computer to execute steps of acquiring technical information including technical matters and classifying the technical information in accordance with classification criteria, the computer being configured to access a Retrieval-Augmented Generation (RAG) data storage unit, the RAG data storage unit storing a data source for causing a language processing model to perform language processing based on the RAG, the RAG data storage unit including a classification criteria data storage unit in which the data source regarding the classification criteria is stored in a format conforming to the RAG, the data source regarding the classification criteria including a plurality of pieces of classification information defined by the classification criteria, the classifying step including steps of determining classification information corresponding to the technical information using the language processing model and the data source regarding the classification criteria, and displaying the technical information on a display device in a state where it is associated with the classification information.

[0010] According to the present disclosure, technical information can be appropriately classified.

[0011] FIG. 1 is a block diagram showing an overview of a technical information classification system. FIG. 2 is a diagram for explaining the concept of a classification method by the technical information classification system. FIG. 3 is a block diagram showing the configuration of a technical information database. FIG. 4 is a block diagram showing the configurations of a classification unit, an AI construction unit, a data import unit, an acquisition unit, a technical information search unit, and a classification information search unit. FIG. 5 is a diagram showing the processing procedure of the acquisition unit. FIG. 6 is a diagram showing the processing procedure of the data import unit. FIG. 7 is a diagram showing the processing procedure of the AI ​​construction unit. FIG. 8 is a diagram showing the processing procedure of a chat unit (summarization). FIG. 9 is a diagram showing the processing procedure of a chat unit (classification). FIG. 10 is a diagram showing the processing procedure of a classification storage unit. FIG. 11 is a diagram showing the processing procedure of a technical information search unit. FIG. 12 is a block diagram showing an overview of a technical information classification system relating to a modified example.

[0012] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are designated by the same reference numerals, and description thereof will not be repeated.

[0013] 1 is a block diagram showing an overview of a technical information classification system 1. The technical information classification system 1 includes a generation server 400 and a user device 500. The generation server 400 and the user device 500 are connected via a communication network such as the Internet 50.

[0014] A user accesses the generation server 400 using a user device 500. The user device 500 includes a processor 501, a memory 502, a communication interface 503, an input / output interface 504, a display 505, and an operation unit 506. The operation unit 506 is configured with a mouse, a keyboard, etc.

[0015] The memory 502 may include a random access memory (RAM), a read only memory (ROM), a flash memory, or any other suitable memory system. The memory 502 stores programs required for the arithmetic processing of the processor 501, temporary data calculated in the arithmetic processing, and the like.

[0016] The processor 501 connects to the Internet 50 via a communication interface 503 in accordance with a program stored in a memory 502. The processor 501 connects to the Internet 50 and communicates with the generation server 400.

[0017] The generating server 400 includes a microcomputer 410 and a storage 404 referenced by the microcomputer 410. The microcomputer 410 is an example of a "computer."

[0018] The microcomputer 410 includes a processor 401, a memory 402, and a communication interface 403. The memory 402 includes a random access memory (RAM), a read only memory (ROM), a flash memory, or any other suitable memory system. The memory 402 stores programs necessary for the arithmetic processing of the processor 401, temporary data calculated in the arithmetic processing, and the like. The processor 401 operates based on the programs stored in the memory 402 and references the storage 404.

[0019] The storage 404 is configured with a hard disk drive, a solid state drive, etc. The storage 404 stores the technical information database 125, a large language model (LLM) 430, and an RAG database 420. The large language model 430 is an example of a language processing model.

[0020] The large-scale language model 430 is a language model (trained language model) that has been trained in advance by machine learning. A huge amount of training data is used to train the large-scale language model 430. The large-scale language model 430 is formed as an autoregressive model that uses a transformer such as a GPT (Generative Pre-trained Transformer). The large-scale language model 430 according to this embodiment may be configured using a GPT (GPT-2, GPT-3, GPT-4), Bard, or the like.

[0021] Technical information is registered in the technical information database 125. The technical information is, for example, the content of technical documents. More specifically, the technical information is file data including text information. The technical information database 125 is an example of a technical information storage unit.

[0022] The technical information may be not only technical documents but also moving image data such as educational videos and scanner data obtained by scanning paper books.

[0023] The RAG database 420 is, for example, a vector database. In addition to the data source of the classification information, the RAG database 420 defines instructions (behavior information) and embedding parameters (chunk size, etc.). The RAG database 420 is an example of a RAG data storage unit. The RAG database 420 includes multiple related information databases in which various data related to RAG (Retrieval-Augmented Generation) is registered. FIG. 1 illustrates related information databases 421 and 422 as examples of the RAG database 420.

[0024] RAG is a technology that allows the large-scale language model 430 to generate answers based on a knowledge source containing the most recent and accurate information. Generally, large-scale language models may occasionally output inaccurate or misleading information, which is called hallucination. RAG can compensate for such imperfect behavior of the large-scale language model and improve the quality of the answers generated by the large-scale language model.

[0025] The related information databases 421 and 422 are examples of knowledge sources containing the latest, accurate information. In this embodiment, a method is proposed for appropriately classifying technical information in the large-scale language model 430 using RAG technology. A large amount of technical information is registered in the related information database 421. A large amount of classification information for classifying technical information based on classification criteria is registered in the related information database 422. The related information database 421 is an example of a first data storage unit. The related information database 422 is an example of a second data storage unit. FIG. 1 shows an example in which the technical information database 125 and the RAG database 420 are configured in the storage 404. However, the storage in which the technical information database 125 is configured may be different from the storage in which the RAG database 420 is configured. Furthermore, within the RAG database 420, the storage in which the related information database 421 is configured may be different from the storage in which the related information database 422 is configured. The data contained in each of technical information database 125, related information database 421, and related information database 422 may be stored in table format instead of database format. A single database may have a table area for storing data contained in related information database 421 and a table area for storing data contained in related information database 422.

[0026] As classification information, for example, the International Patent Classification (IPC) can be used. The IPC is an internationally standardized technology classification for classifying inventions for which patent applications have been filed. By using the IPC, technologies can be classified in detail by dividing them into hierarchical levels such as "section," "subsection," "class," "subclass," "main group," and "subgroup."

[0027] The IPC is originally an index used to classify patent documents. However, by using the IPC as an index to classify various technical information including patent documents (such as papers, in-house technical reports, books, in-house documents, technical training materials, equipment manuals, and educational videos), it is believed possible to classify a wide variety of technical information using a uniform standard. In other words, the technical information assumed in this embodiment includes patents, inventions (including registered utility models), papers, in-house technical reports, and educational videos.

[0028] As classification information, patent classifications such as F-terms, CPC, or FI may be used instead of IPC. Alternatively, the Japanese Decimal Classification (NDC) classification table for books and the International Decimal Classification may be used. Alternatively, if the technical information classification system 1 is used exclusively within a specific company, classification information used within the company may be used. Also, product classifications such as equipment may be applied. In other words, classification information may be any of IPC, F-terms, CPC, FI, book classifications, and in-house classifications.

[0029] The microcomputer 410 accesses the RAG database 420 as needed, and provides information based on the RAG database 420 to the large-scale language model 430 in a dialogue between the user and the large-scale language model 430. As a result, the large-scale language model 430 responds to the user using accurate information based on the RAG database 420.

[0030] 1, an AI (Artificial Intelligence) 450 is constructed by a microcomputer 410 and a storage 404. A user categorizes and searches for technical information through a chat dialogue with the AI ​​450.

[0031] In particular, in this embodiment, the technical information is classified through a dialogue between the AI ​​450 and the user. A method for classifying the technical information through a dialogue will be described below with reference to FIG.

[0032] 2 is a diagram illustrating the concept of the classification method used by the technical information classification system 1. A user uses a large amount of technical information and a large amount of classification information to construct an RAG database 420 (related information databases 421, 422) in advance (see (i) in FIG. 2). The user here is, for example, an administrator of the technical information classification system 1. Once the RAG database 420 is constructed, preparations are complete for classifying each piece of technical information.

[0033] A user who wishes to classify technical information interacts with AI 450 using a chat screen provided by user device 500. In particular, in this embodiment, two types of interactions are assumed for classifying technical information. One is an interaction for summarizing the technical information (see (ii) in FIG. 2). The other is an interaction for determining classification information corresponding to the technical information based on the summarized technical information (see (iii) in FIG. 2).

[0034] AI 450 summarizes technical information specified by the user through dialogue with the user. When summarizing the technical information, AI 450 refers to the related information database 421 as necessary. AI 450 presents the summarized technical information to the user. AI 450 re-summarizes the technical information in response to user instructions. AI 450 determines the summary content of the technical information in response to user instructions.

[0035] Next, AI 450 classifies the summarized technical information through dialogue with the user. When classifying the technical information, AI 450 refers to related information database 422 as necessary. AI 450 presents the user with classification information that is estimated to be appropriate, along with an overview of the technology encompassed by the classification information. AI 450 re-estimates the classification information in response to a user instruction. AI 450 determines the classification information corresponding to the summary of the technical information in response to a user instruction.

[0036] When the AI ​​450 determines appropriate classification information corresponding to the summarized technical information, it registers the technical information in association with the classification information in the technical information database 125 (see (iv) in FIG. 2). As a result, the technical information is stored in the technical information database 125 in a state where it is classified by the classification information.

[0037] In response to the request of the user (illustrated as an administrator in Figure 2), AI 450 analyzes the registered technical information and classification information and displays it on the screen of the user device 500 in a specified format (see (v) in Figure 2).

[0038] Note that unclassified technical information may be stored in the technical information database 125. In this case, the AI ​​450 may call up the unclassified technical information in response to a user's instruction and carry out the two dialogues described above with respect to the called-up technical information. Alternatively, the AI ​​450 may accept technical information that is not stored in the technical information database 125 and carry out the two dialogues described above.

[0039] 3 is a block diagram showing the configuration of the technical information database 125. A large number of pieces of technical information are registered in the technical information database 125. Each piece of technical information is assigned a technical information ID. The generating server 400 identifies each piece of technical information based on the technical information ID.

[0040] The technical information includes a title and content. As shown in Fig. 3, the technical information is registered in the technical information database 125 in association with classification information. The microcomputer 410 constituting a part of the AI ​​450 registers the classification information determined through a dialogue with the user in the technical information database 125.

[0041] FIG. 4 is a block diagram showing the configuration of the classification unit 800, AI construction unit 902, data import unit 903, acquisition unit 904, technical information search unit 905, and classification information search unit 906. The classification unit 800, acquisition unit 904, technical information search unit 905, and classification information search unit 906 are realized by the microcomputer 410 and data stored in the storage 404. The classification unit 800 is an example of a classification unit that classifies technical information acquired by the acquisition unit according to classification criteria. The classification unit 800 includes a chat unit (summary) 803, a chat unit (classification) 804, and a classification storage unit 805. The processing procedures of each unit will be described below with reference to FIGS. 5 to 12.

[0042] 5 is a diagram showing the processing procedure of the acquisition unit 904. The acquisition unit 904 acquires technical information to be classified from the user and stores the acquired technical information in the technical information database 125 (step S1). At this stage, the technical information stored in the technical information database 125 is not associated with classification information.

[0043] The acquiring unit 904 may use the technical information acquired from the user as a target for classification processing by the classifying unit 800 without storing it in the technical information database 125. Step S1 is an example of an acquiring unit that acquires technical information including technical matters. Such an acquiring unit may be configured by the data import unit 903.

[0044] 6 is a diagram showing the processing procedure of the data import unit 903. The data import unit 903 generates information sources necessary to construct the RAG database 420 according to the following procedure.

[0045] First, the data import unit 903 receives a large amount of technical information to be analyzed from a user such as an administrator (step S11). Next, the data import unit 903 saves the collected technical information in a specific area of ​​the storage 404 as an information source of the technical information (step S12). Next, the data import unit 903 receives a large amount of classification information to be analyzed from a user such as an administrator (step S13). Next, the data import unit 903 saves the collected classification information in a specific area of ​​the storage 404 as an information source of the classification information (step S14).

[0046] 7 is a diagram showing the processing procedure of the AI ​​construction unit 902. First, the AI ​​construction unit 902 acquires an information source (technical information) from the storage 404 (step S21). Next, the AI ​​construction unit 902 uses the acquired information source (technical information) to construct a related information database 421 related to the RAG (step S22). This generates the related information database 421 in which data sources related to technical information are stored in a format conforming to the RAG.

[0047] Next, the AI ​​construction unit 902 acquires the information source (classification information) from the storage 404 (step S23). Next, the AI ​​construction unit 902 constructs the related information database 422 related to the RAG (step S24). This generates the related information database 422 in which the data source related to the classification information is stored in a format conforming to the RAG.

[0048] 8 is a diagram showing the processing procedure of the chat unit (summary) 803. First, the chat unit 803 calls an AI algorithm (step S31). Next, the chat unit 803 refers to the technical information database 125 to identify technical information to be classified (step S32). Note that after the processing of step S31, the chat unit (summary) 803 may obtain the technical information to be classified from the user through a chat with the user.

[0049] Next, the chat unit 803 refers to the related information database 421 related to the technical information as necessary (step S33). Next, the chat unit 803 summarizes the technical information while chatting with the user using the RAG (step S34). Next, the chat unit 803 passes the summary to the chat unit (classification) 804 (step S35).

[0050] Step S34 includes the steps of having the user confirm the summary of the technical information created by the chat unit 803, revising the summary of the technical information in response to the user's instructions, and finalizing the summary of the technical information with the user's approval. That is, the chat unit 803 determines the content of the summary after interacting with the user regarding the summary of the technical information. Note that the chat unit 803 may determine the content of the summary without obtaining explicit approval from the user.

[0051] 9 is a diagram showing the processing procedure of the chat unit (classification) 804. First, the chat unit 804 calls an AI algorithm (step S41). Next, the chat unit 804 references the related information database 422 related to the classification information as needed (step S42). Next, the chat unit 804 determines classification information corresponding to the technical information using a summary of the technical information while chatting with the user using the RAG (step S43).

[0052] In this way, the chat unit (classification) 804 uses the large-scale language model 430 and the RAG database 422 to determine classification information corresponding to the technical information acquired by the acquisition unit. In particular, the chat unit (classification) 804 determines classification information corresponding to the technical information based on a summary of the technical information. Next, the chat unit 804 displays the classification information corresponding to the technical information (step S44). More specifically, the chat unit 804 transmits data for displaying an image in which the technical information and the classification information are associated to the user device 500. Based on the received data, the user device 500 displays the technical information on the display device in association with the classification information. In this way, the chat unit 804 displays the technical information on the display device in association with the classification information. Next, the chat unit 804 passes the determined content to the classification storage unit 805 (step S45).

[0053] Step S43 includes a step of having the user confirm the classification information provisionally determined by the chat unit 804, a step of changing the classification information in response to a user instruction, and a step of finalizing the classification information with the user's approval. That is, the chat unit 804 determines the classification information corresponding to the technical information after having a dialogue with the user regarding the classification information. Note that the classification unit 800 may determine the classification information without having a dialogue with the user using the chat unit 804. For example, the classification unit 800 may collectively determine the classification information corresponding to each of a large number of pieces of pre-registered technical information without having a dialogue with the user.

[0054] 10 is a diagram showing the processing procedure of the classification storage unit 805. The classification storage unit 805 stores technical information in association with classification information in the technical information database 125 (step S51).

[0055] 11 is a diagram showing the processing procedure of the technical information search unit 905. First, the technical information search unit 905 receives classification information of the technical information and a display format of the technical information from the user (step S61). The technical information search unit 905 may receive multiple classification information. By receiving the display format, the technical information search unit 905 can identify the type of graph to be used when displaying the technical information on the screen.

[0056] Next, the technical information search unit 905 refers to the graph model and acquires a graph display method corresponding to the display format specified by the user (step S62). The graph model may be stored in the storage 404, for example. Next, the technical information search unit 905 searches the technical information database 125 for the technical information to be displayed (step S63). At this time, the technical information search unit 905 searches for technical information associated with the classification information specified by the user. Step S63 is an example of a search unit.

[0057] Next, the technical information search unit 905 uses the technical information found by the search in step S63 to generate a graph in a format in accordance with the user's instructions, and then transmits the generated graph (including the technical information) to the user device 500 (step S64). Step S64 is an example of a transmission unit.

[0058] The user device 500 displays the received graph. For example, if multiple pieces of classification information are accepted in step S61, the technical information search unit 905 may generate a graph showing the number of pieces of technical information for each piece of classification information. When the technical information search unit 905 accepts classification information from the user, it may transmit a list of technical information corresponding to the accepted classification information to the user device 500.

[0059] 12 is a diagram showing the processing procedure of the classification information search unit 906. First, the classification information search unit 906 accepts technical information from a user (step S71). Next, the classification information search unit 906 searches the accepted technical information using the technical information database 125 (step S72). Next, the classification information search unit 906 identifies classification information associated with the discovered technical information (step S73).

[0060] Next, the classification information search unit 906 transmits the identified classification information to the user device 500 (step S74). The user device 500 displays the technical information received from the user on the display 505 in association with the identified classification information.

[0061] As described above, according to this embodiment, technical information can be uniformly and appropriately classified. Furthermore, by using RAG, confidential information can also be classified. Furthermore, by storing data related to the classification process in advance in the RAG database 420, the error rate of the large-scale language model 430 can be reduced. In this disclosure, "storing" includes the concept of continuously receiving and processing data. A transmitting unit that transmits data to a storage unit such as a memory may transmit all data to the storage unit in a lump. Alternatively, the transmitting unit may divide the data to be transmitted into packets or chunks and then transmit the divided data to the storage unit sequentially.

[0062] [Modification] As described above, the technical information classification system 1 determines the summary content of the technical information and the classification information corresponding to the technical information through a dialogue with the user. The dialogue between the technical information classification system 1 and the user is realized by a chat function provided in the technical information classification system 1. Such a chat function may not be necessary.

[0063] The technical information classification system 1 may determine classification information corresponding to technical information without going through a process of summarizing the technical information. When determining classification information corresponding to technical information without going through a process of summarizing the technical information, the technical information classification system 1 may determine the classification information through a dialogue with a user, or may determine the classification information without a dialogue with a user. In addition to RAG, Low-Rank Adaptation of Large Language Models (LoRA) or the like may be used.

[0064] The user device 500 need not necessarily be equipped with all of the processor 501, memory 502, communication interface 503, and input / output interface 504, but may also be a thin client system using VDI (Virtual Desktop Infrastructure). A thin client system using VDI is a system in which a desktop environment on a server is transferred to a terminal in a remote location for use. The user device 500 and the generation server 400 do not necessarily have to be independent devices. When using such a thin client system, the functions of the user device 500 and the generation server 400 can be provided on the same aggregation server. Note that the technical information classification system 1 may be configured without including the user device 500.

[0065] The generation server 400 is an example of a computing device. The computing device may be configured by a server (an on-premise server, a cloud server, etc.), a serverless system, etc. Here, an on-premise server is a server installed and managed in facilities managed within a company. A cloud server is a server (a rented server) provided by another business via a network. A serverless system is a system in which computing and memory functions can be used only when necessary, without being aware of the existence of a server. Computing devices include servers and serverless systems. Servers include on-premise servers and cloud servers.

[0066] In the technical information classification system according to the present disclosure, the RAG database 420 may be located outside the generating server 400. Fig. 13 is a block diagram showing an overview of a technical information classification system 1A according to a modified example. As shown in Fig. 13, in the technical information classification system 1A, the RAG database 420 is located outside the generating server 400. In the technical information classification system 1A, the RAG database 420 is provided in an external server 600. The external server 600 is communicably connected to the generating server 400 via a communication network such as the Internet 50.

[0067] The external server 600 includes a microcomputer 610 that accesses the RAG database 420 and a storage 604 in which the RAG database 420 is stored. Like the microcomputer 410, the microcomputer 610 includes a processor 601, a memory 602, and a communication interface 603. The microcomputer 610 accesses the RAG database 420 in response to a request from the generating server 400. The microcomputer 610 transmits information in the RAG database 420 to the generating server 400 in response to a request from the generating server 400, or updates the RAG database 420.

[0068] In the technical information classification system 1A, the Model Context Protocol (MCP) may be adopted for communication between the generation server 400 and the external server 600. MCP is a standard protocol used for large-scale language models to communicate with data sources, external tools, etc. By adopting MCP, the generation server 400 can provide context to the external server 600 using a standardized protocol.

[0069] In the present disclosure, the acquisition unit 904 is an example of an acquisition unit that acquires technical information including technical matters. The acquisition unit may be configured by the data import unit 903. The acquisition unit 904 and the data import unit 903 are realized, for example, by the microcomputer 410 of the generating server 400. The display 505 is an example of a display device.

[0070] The classification unit 800 is an example of a classification unit that classifies technical information according to classification criteria. The classification unit 800 is realized, for example, by the microcomputer 410 and storage 404 of the generating server 400. In the modification of FIG. 13 , the classification unit 800 is realized by the microcomputer 410 of the generating server 400, the microcomputer 610 of the external server 600, and the storage 604 of the external server 600.

[0071] [Aspects] Aspects of the present disclosure are listed below.

[0072] (Clause 1) The technical information classification system described in Clause 1 comprises an acquisition unit that acquires technical information including technical matters, a classification unit that classifies the technical information in accordance with classification criteria, and a display device, wherein the classification unit includes a language processing model and a RAG data storage unit that stores a data source for causing the language processing model to perform language processing based on RAG (Retrieval-Augmented Generation), the RAG data storage unit includes a classification criteria data storage unit that stores a data source related to the classification criteria in a format that conforms to RAG, and the data source related to the classification criteria includes multiple pieces of classification information defined by the classification criteria, and the classification unit uses the language processing model and the data source related to the classification criteria to determine classification information corresponding to the technical information and displays the technical information on the display device in association with the classification information.

[0073] (Section 2) The technical information classification system described in Section 2 is the same as the technical information classification system described in Section 1, except that the technical information includes any one of patents, inventions, papers, in-house technical reports, and educational videos.

[0074] (Section 3) The technical information classification system described in Section 3 is a technical information classification system described in Section 1 or 2, and in addition, the classification information is one of IPC, F-term, CPC, FI, book classification, and in-house classification.

[0075] (4) The technical information classification system described in paragraph 4 is, in addition to the technical information classification system described in any one of paragraphs 1 to 3, characterized in that the classification unit generates summary information of the technical information, determines classification information corresponding to the technical information based on the generated summary information, and generates the summary information while interacting with the user.

[0076] (Section 5) In addition to the technical information classification system described in Section 4, the technical information classification system described in Section 5 has the following configuration: the classifier determines classification information corresponding to the technical information while interacting with the user.

[0077] (Clause 6) The technical information classification system described in Clause 6, in addition to the technical information classification system described in Clause 4 or 5, further comprises a technical information storage unit, the classification unit stores the technical information in the technical information storage unit in association with the classification information, and the technical information classification system further comprises a user device operated by a user, a search unit that uses the classification information to search the technical information storage unit for technical information in accordance with the user's instructions, and a transmission unit that transmits the technical information found by the search unit to the user device.

[0078] (Clause 7) The method described in clause 7 is a method for classifying technical information, the method causing a computer to execute steps of acquiring technical information including technical matters and classifying the technical information in accordance with classification criteria, the computer being configured to access a Retrieval-Augmented Generation (RAG) data storage unit, the RAG data storage unit storing a data source for causing a language processing model to perform language processing based on RAG, the RAG data storage unit including a classification criteria data storage unit in which the data source regarding the classification criteria is stored in a format conforming to RAG, the data source regarding the classification criteria including a plurality of classification information defined by the classification criteria, the classifying step including steps of determining classification information corresponding to the technical information using the language processing model and the data source regarding the classification criteria, and displaying the technical information on a display device in association with the classification information.

[0079] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the description of the above embodiments, and is intended to include all modifications within the meaning and scope of the claims.

[0080] 1, 1A Technical information classification system, 50 Internet, 125 Technical information database, 400 Generation server, 401 Processor, 402 Memory, 403 Communication interface, 404 Storage, 410 Microcomputer, 420 RAG database, 421, 422 Related information database, 430 Large-scale language model (LLM), 450 AI, 500 User device, 501 Processor, 502 Memory, 503 Communication interface, 504 Input / output interface, 505 Display 505, 506 Operation unit, 600 External server, 601 Processor, 602 Memory, 603 Communication interface, 604 Storage, 800 Classification unit, 903 Data import unit, 902 AI construction unit, 803, 804 Chat unit, 805 Classification storage unit, 904 Acquisition unit, 905 Technical information search unit, 906 Classification information search section.

Claims

1. A technical information classification system comprising: an acquisition unit that acquires technical information containing technical matters; a classification unit that classifies the technical information in accordance with classification criteria; and a display device, wherein the classification unit includes a language processing model; and a RAG data storage unit that stores a data source for causing the language processing model to perform language processing based on RAG (Retrieval-Augmented Generation), wherein the RAG data storage unit includes a classification criteria data storage unit that stores a data source related to the classification criteria in a format that conforms to the RAG, and the data source related to the classification criteria includes a plurality of pieces of classification information defined by the classification criteria, and wherein the classification unit uses the language processing model and the data source related to the classification criteria to determine classification information corresponding to the technical information, and displays the technical information on the display device in a state where it is associated with the classification information.

2. The technical information classification system of claim 1, wherein the technical information includes one of patents, inventions, papers, in-house technical reports, and instructional videos.

3. A technical information classification system according to claim 1 or claim 2, wherein the classification information is one of IPC, F-term, CPC, FI, book classification, and in-house classification.

4. A technical information classification system as described in any one of claims 1 to 3, wherein the classification unit generates summary information of the technical information, and determines classification information corresponding to the technical information based on the generated summary information, and the classification unit generates the summary information while interacting with a user.

5. The technical information classification system according to claim 4, wherein the classification unit determines classification information corresponding to the technical information while interacting with the user.

6. A technical information classification system as described in claim 4 or claim 5, further comprising a technical information storage unit, wherein the classification unit stores the technical information in the technical information storage unit in association with the classification information, and the technical information classification system further comprises: a user device operated by the user; a search unit that uses the classification information to search the technical information storage unit for technical information corresponding to the user's instructions; and a transmission unit that transmits the technical information found by the search unit to the user device.

7. A method for classifying technical information, the method causing a computer to execute the steps of: acquiring technical information including technical matters; and classifying the technical information in accordance with classification criteria; the computer is configured to access a Retrieval-Augmented Generation (RAG) data storage unit, the RAG data storage unit stores a data source for causing a language processing model to perform language processing based on RAG, the RAG data storage unit includes a classification criteria data storage unit in which a data source related to the classification criteria is stored in a format conforming to the RAG, the data source related to the classification criteria includes a plurality of pieces of classification information defined by the classification criteria, and the classifying step includes the steps of: determining classification information corresponding to the technical information using the language processing model and the data source related to the classification criteria; and displaying the technical information on a display device in a state where it is associated with the classification information.

Images