Information processing method, program, and information processing device

Abstract

This invention provides an information processing method that enables the appropriate classification of patent documents. [Solution] The computer acquires multiple patent documents according to a program. The computer then extracts sentences from each of the acquired patent documents that are of interest among the items contained in the patent documents, and uses a language model to generate a summary of the sentences extracted from each patent document. Then, based on the generated summary, the computer uses a language model to classify the multiple patent documents into multiple categories.

Description

Technical Field

[0001] The present disclosure relates to an information processing method, a program, and an information processing apparatus.

Background Art

[0002] When developing new products and services, or creating new themes such as new businesses, patent documents are read in order to grasp the patent application status and technological trends of other companies in the technical field under consideration. However, since the task of reading a huge number of patent documents requires time, there is a limit to the number of documents that can be read. In Patent Document 1, morphological analysis is performed on information data, which is divided into a plurality of terms, and based on a coding rule that assigns a plurality of terms according to the appearance frequency, the plurality of terms are coded into a first-axis code (category) and the second-axis code (category), and an information map generation technique for generating a similarity matrix with the first-axis code and the second-axis code as rows and columns is disclosed.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Regarding the terms extracted by morphological analysis as in Patent Document 1, since general terms may have a higher appearance frequency, there is a risk that the content of the document cannot be appropriately grasped. Therefore, there is a problem that it is difficult to appropriately classify information data (documents) and create an information map with high accuracy.

[0005] An object of the present disclosure is to provide an information processing method and the like that can appropriately classify patent documents.

Means for Solving the Problems

[0006] An information processing method according to one aspect of the present disclosure involves a computer that acquires a plurality of patent documents, extracts sentences described in items of interest from each of the plurality of patent documents, generates a summary of the extracted sentences using a language model, and then uses the language model to classify the plurality of patent documents into a plurality of categories based on the generated summary. [Effects of the Invention]

[0007] In one aspect of this disclosure, patent documents can be appropriately classified. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing an example of the configuration of an information processing device. [Figure 2] This is a flowchart showing an example of the procedure for creating a patent map. [Figure 3] This is a flowchart showing an example of the procedure for creating a patent map. [Figure 4] This is a flowchart showing an example of the procedure for creating a patent map. [Figure 5] This is an explanatory diagram showing an example screen. [Figure 6] This is an explanatory diagram showing an example of a list generated when creating a patent map. [Figure 7] This is an explanatory diagram showing an example of a list generated when creating a patent map. [Figure 8] This is an explanatory diagram showing an example of a prompt. [Figure 9] This is an explanatory diagram showing an example screen. [Figure 10] This flowchart shows an example of the procedure for creating a patent map according to Embodiment 2. [Figure 11] This is an explanatory diagram showing an example of a patent map. [Modes for carrying out the invention]

[0009] The information processing method, program, and information processing apparatus of this disclosure will be described in detail below with reference to the drawings illustrating their embodiments.

[0010] (Embodiment 1) This embodiment describes an information processing device that categorizes multiple patent documents based on the content of their items and creates a patent map representing the categorized results. In this embodiment, the configuration using published patent gazettes and patent publications as patent documents is described as an example, but a configuration using technical documents and invention descriptions relating to pre-patent application technology and inventions may also be used.

[0011] Figure 1 is a block diagram showing an example configuration of an information processing device. The information processing device 10 is a computer capable of various information processing and information transmission and reception, and is composed of a personal computer, a server computer, etc. The information processing device 10 includes a control unit 11, a storage unit 12, a communication unit 13, an input unit 14, a display unit 15, a reading unit 16, etc., and each of these units is connected via a bus. The control unit 11 includes one or more processors (arithmetic processing units) such as a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), a GPU (Graphics Processing Unit), or an AI chip (AI semiconductor). The control unit 11 executes the processing that the information processing device 10 should perform by appropriately executing the program P stored in the storage unit 12. Note that if the control unit 11 includes multiple processors, each process may be executed by the same processor, or each process may be executed by different processors.

[0012] The memory unit 12 includes RAM (Random Access Memory), flash memory, hard disk, SSD (Solid State Drive), etc. The memory unit 12 stores the program P (program product, computer program) executed by the control unit 11 and various data. The memory unit 12 also temporarily stores data generated when the control unit 11 executes program P. The memory unit 12 also stores the language model M. The language model M is a general-purpose large language model (LLM: Large Language Model) constructed by pre-training with a large set of documents, etc. The language model M is trained to, for example, when input data including text data written in natural language is input, perform calculations to generate output data according to the content of the input data based on the input data, and output the generated output data. The language model M can be constructed using algorithms such as GPT (Generative Pre-trained Transformer)-3, GPT-3.5, GPT-4, etc., and may be constructed by combining multiple algorithms. Note that the language model M is not limited to the transformer-based model described above. The language model M is intended to be used as a program module constituting artificial intelligence software. The language model M performs predetermined calculations on input data and outputs the calculation results. The memory unit 12 stores data such as the coefficients and thresholds of the function that defines these calculations as the language model M. Instead of storing the language model M in the memory unit 12, the information processing device 10 may access and read the language model M from a language processing server that stores it. The language model M may also be one that has been fine-tuned using patent documents. In addition to fine-tuning the entire language model M, the parameters of only some layers may be fine-tuned using technologies such as LoRA (Low-Rank Adaptation). The memory unit 12 may be composed of multiple storage devices, and part of the memory unit 12 may be other storage devices connected to the information processing device 10, or other storage devices that the information processing device 10 can communicate with.

[0013] The communication unit 13 is a communication module for processing wired or wireless communication, and transmits and receives information with other devices via the network N. The network N may be the Internet or a public telephone network, or it may be a LAN (Local Area Network) built in a facility where the information processing device 10 is installed. The input unit 14 receives operation input from the user and sends control signals corresponding to the operation content to the control unit 11. The display unit 15 is a liquid crystal display or an organic EL (Electro Luminescence) display, etc., and displays various information according to instructions from the control unit 11. The input unit 14 and the display unit 15 may be a touch panel configured as an integrated unit.

[0014] The reading unit 16 reads information stored on a portable storage medium 10a such as a CD (Compact Disc), DVD (Digital Versatile Disc), USB (Universal Serial Bus) memory, SD card, microSD card, or CompactFlash (registered trademark). The program P and various data to be stored in the storage unit 12 may be read by the control unit 11 from the portable storage medium 10a via the reading unit 16 and stored in the storage unit 12. Alternatively, the program P and various data may be written to the storage unit 12 during the manufacturing stage of the information processing device 10, or the control unit 11 may download them from another device via the communication unit 13 and store them in the storage unit 12.

[0015] In this embodiment, the information processing apparatus 10 may be not limited to a single computer but a multi-computer including a plurality of computers, or may be a virtual machine virtually constructed by software within one device. Further, the information processing apparatus 10 may be a local server installed in the facility where the information processing apparatus 10 is provided, or may be a cloud server communicatively connected via the network N. Hereinafter, the information processing apparatus 10 will be described as being a single computer. Also, the program P may be arranged and executed on a single computer or at one site, or may be distributed over a plurality of sites and executed in a distributed manner on a plurality of computers interconnected via the network N. Furthermore, the input unit 14 and the display unit 15 are not essential for the information processing apparatus 10, and it may be configured to receive operations through the connected computer, or may be configured to output the information to be displayed to an external display device.

[0016] The information processing apparatus 10 of this embodiment categorizes a plurality of patent documents based on the description contents for two items of interest, and creates a patent map showing the distribution of the patent documents categorized for each item of interest. The patent map uses a two-dimensional chart showing each category of the two items of interest on the horizontal axis and the vertical axis (the first axis and the second axis). Note that the number of items of interest is not limited to two, and may be three or more, or may be one.

[0017] Hereinafter, the process in which the information processing apparatus 10 of this embodiment creates a patent map will be described. FIGS. 2 to 4 are flowcharts showing an example of the procedure for creating a patent map, FIG. 5 is an explanatory diagram showing an example of a screen, FIGS. 6 and 7 are explanatory diagrams showing examples of lists generated when creating a patent map, FIG. 8 is an explanatory diagram showing an example of a prompt, and FIG. 9 is an explanatory diagram showing an example of a screen. The following processes are executed by the control unit 11 of the information processing apparatus 10 according to the program P stored in the storage unit 12.

[0018] The control unit 11 of the information processing device 10 acquires multiple patent documents to be used in creating a patent map (S11). For example, the control unit 11 acquires patent documents included in the technical field specified by the user, or patent documents obtained by a document search based on keywords specified by the user, from a server that publishes patent documents, such as the Japan Patent Information Platform (J-PlatPat). The control unit 11 may also pre-download multiple patent documents and store them in the storage unit 12 (e.g., a hard disk or SSD), in which case the patent documents stored in the hard disk or SSD can be read into memory (RAM or flash memory).

[0019] The control unit 11 displays a focus item input screen on the display unit 15, as shown in Figure 5 (S12). The focus item input screen accepts the designation of focus items shown on the horizontal axis (first axis) and vertical axis (second axis) of the patent map, and is provided with input fields for focus items for the first axis and focus items for the second axis. Each input field is provided with a pull-down menu that allows the user to select any one item from the items included in the patent document, and accepts the selection of any item as a focus item via the pull-down menu. In the example in Figure 5, the pull-down menu displays the options for the item: problem, solution, use, filing date, and applicant, but other items may be displayed. The focus item input screen may be configured to accept the designation of only one focus item (a focus item for the first axis), or it may be configured to accept the designation of three or more focus items.

[0020] The control unit 11 receives the designation of a focus item via the focus item input screen (S13), and if the map generation button is operated, it performs the processing from step S14 onwards. As shown in Figure 5, the focus item input screen is provided with a detailed settings button, and when the detailed settings button is operated, it is configured to accept the setting of various thresholds to be used in subsequent processing.

[0021] The control unit 11 reads one patent document to be processed from the patent documents obtained in step S11 (S14). The control unit 11 identifies the item of interest to be processed from the specified item of interest (S15) and determines whether the identified item of interest is the filing date or the applicant (S16). If it is determined that the item of interest is not the filing date or the applicant (S16: NO), the control unit 11 extracts the text corresponding to the item of interest to be processed from the patent document to be processed (S17). For example, if the item of interest is specified as a problem, the control unit 11 extracts the text described in the "Problem to be Solved by the Invention" section of the specification in the patent document (published patent gazette or patent gazette), the text described in the "Problem" section of the abstract, etc. Also, if the item of interest is specified as a solution, the control unit 11 extracts the text described in the "Means for Solving the Problem" section of the specification in the patent document, the text described in the "Solution" section of the abstract, the text described in the claims, etc. Furthermore, if an application is specified as the item of interest, the control unit 11 extracts text, etc., described in the "Industrial Applicability" or "Technical Field" section of the specification in the patent document. The control unit 11 may be configured to extract all text described in each section, or to extract only some of it. The sections (items) read from the patent document in response to the item of interest are not limited to the examples described above. The control unit 11 stores the extracted text in the storage unit 12.

[0022] The control unit 11 generates a summary (summary) of the text of the items of interest extracted from the patent document (S18). For example, the control unit 11 generates a prompt to instruct the generation of a summary by inserting a command such as "Generate a summary of the following text in approximately 200 to 300 characters" at the beginning of the extracted text. The control unit 11 may include all of the extracted text in the prompt, or it may include only a portion of the extracted text in the prompt. The control unit 11 then inputs the generated prompt into the language model M and obtains the summary generated by the language model M. The control unit 11 registers the generated summary in a list prepared in the storage unit 12, for example (hereinafter referred to as the selection element list) (S19). Figure 6A shows an example of the selection element list. The selection element list stores patent document information for identifying the patent document, corresponding to the element ID, as well as summaries generated for the items of interest on the first axis and summaries generated for the items of interest on the second axis. The patent document information is identification information assigned to the patent document, and for example, the publication number and patent number assigned by the Japan Patent Office can be used. The control unit 11 issues an element ID for the patent document read in step S14, and stores the element ID, the identification information of the patent document, and the summary generated for the items of interest to be processed in a selection element list in association with each other.

[0023] If the control unit 11 determines that the item of interest is the filing date or applicant (S16: YES), it extracts the filing date or applicant information from the patent document to be processed (S20). Here, the control unit 11 extracts the contents of the filing date or applicant item (the item of interest to be processed) in the patent document (published patent gazette or patent gazette). The control unit 11 registers the extracted filing date or applicant information in the selection element list (S21). Here, if an element ID has already been assigned to the patent document to be processed, the control unit 11 stores the extracted filing date or applicant information in the selection element list in association with this element ID. If an element ID has not been assigned to the patent document to be processed, the control unit 11 issues an element ID and stores the filing date or applicant information in the selection element list in association with the issued element ID.

[0024] After processing in step S19 or step S21, the control unit 11 determines whether the processing described above has been completed for all the items of interest specified in step S13 (S22). If it determines that the processing has not been completed (S22: NO), it returns to step S15 and executes the processing in steps S15 to S21 for the unprocessed items of interest. As a result, for the patent documents selected as processing targets in step S14, information or summaries of each item of interest specified in step S13 are extracted or generated and stored in the selection element list.

[0025] If the control unit 11 determines that processing has been completed for all items of interest (S22: YES), it determines whether the above processing has been completed for all patent documents obtained in step S11 (S23). If it determines that processing has not been completed for all patent documents (S23: NO), the control unit 11 returns to step S14 and executes the processing in steps S14 to S22 for the unprocessed patent documents. As a result, information or summaries of each item of interest specified in step S13 are extracted or generated for all patent documents and stored in the selection element list.

[0026] If the control unit determines that processing has been completed for all patent documents (S23:YES), that is, after registering the information of all patent documents in the selection element list, the control unit 11 identifies the items of interest to be processed from the items of interest specified in step S13 (S24). The control unit 11 determines whether the identified items of interest are the filing date or the applicant (S25), and if it determines that the items of interest are neither the filing date nor the applicant (S25:NO), it classifies each patent document into one of several categories based on the summary of the items of interest to be processed in each patent document (S26). Here, as shown in Figure 8A, the control unit 11 generates a prompt to instruct the execution of category classification for each patent document, inputs the generated prompt to the language model M, and obtains the result of the category classification of the patent document by the language model M. The example in Figure 8A includes the instruction statement for the processing to be executed, information on the patent document to be processed, and information specifying the output format. The instruction statement may also include the number of categories to be specified for classification. The information of the patent documents to be processed includes the patent document information of each patent document registered in the selection element list, information on the item of interest to be processed (in Figure 8A, "Solution"), and a summary of the item of interest. The information instructing the output format includes, for example, information indicating the classified category, information on the patent documents classified into each category, and a sentence containing instructions to display a score representing the likelihood of outputting the information of each patent document. Note that the information of the patent documents to be processed may be loaded into the language model M in advance, in which case the information of the patent documents to be processed does not need to be included in the prompt. In the case of the prompt in Figure 8A, a prompt template containing the command statement and information instructing the output format is stored in the storage unit 12, and the control unit 11 generates it by reading the information of each patent document from the selection element list and inserting it into the prompt template.

[0027] The control unit 11 generates a post-classification list, such as the one shown in Figure 6B, as a result of categorizing the patent documents, and stores it in the storage unit 12. In the example in Figure 6B, the contents of the categories generated by the language model M, the information of the patent documents classified into each category, the summaries of the items of interest in each patent document, and the scores identified by the language model M are stored in association with each other. Figure 6B shows the post-classification list for the items of interest in "solutions".

[0028] The control unit 11 counts the number of patent documents classified into each category based on the post-classification list showing the category classification results for the items of interest to be processed (S27), and calculates the ratio of the counted number of patent documents to the total number of patent documents (the ratio of patent documents classified into each category to the total number of patent documents) (S28). As a result of the calculation, the control unit 11 generates a classification result table, such as the one shown in Figure 6C, and stores it in the storage unit 12. In the example in Figure 6C, for the item of interest "Solution," the content of each category, the number of patent documents classified into each category, and the ratio of patent documents classified into each category to the total number of patent documents are stored in association with each other.

[0029] The control unit 11 counts the number of categories whose proportion is less than a predetermined value based on the classification result table (S29), and determines whether the counted number of categories is greater than a predetermined number (first threshold) (S30). Categories whose proportion is less than a predetermined value are categories in which the number of documents classified into that category is small. If the number of such categories is greater than the predetermined number, it is possible that the classification has been performed at too fine a level. Therefore, in this case, the control unit 11 reduces the number of categories and performs the category classification again. Accordingly, if the control unit 11 determines that the counted number of categories is greater than the predetermined number (S30: YES), it identifies the number of categories to be classified (S31), returns to step S26, and classifies each patent document into the identified number of categories based on the summary of the items of interest to be processed in each patent document (S26). The control unit 11 performs steps S27 to S29 for the newly performed category classification and repeats the processing of steps S26 to S31 until it determines that the counted number of categories is less than or equal to the predetermined number. In step S31, the control unit 11 identifies a number of categories that is less than the number of categories that were classified in the most recent step S26, for example, by subtracting a predetermined number (e.g., 1, 2, etc.). Alternatively, the control unit 11 may identify the number of categories to be classified in the new category classification by subtracting the number obtained by subtracting a predetermined number, which is a first threshold, from the number of categories counted in the most recent step S29 (the number of categories that exceeded the first threshold), from the number of categories that were classified in the most recent step S26.

[0030] If the control unit 11 determines that the number of counted categories is less than or equal to a predetermined number (S30: NO), it determines whether there are any categories whose proportion is greater than a predetermined value (second threshold) (S32). Categories whose proportion is greater than a predetermined value are categories in which there are many documents classified, and it is possible that a coarse-grained classification has been performed. Therefore, in this case, the control unit 11 further classifies the patent documents classified into categories whose proportion is greater than a predetermined value into multiple categories. Accordingly, if the control unit 11 determines that there are categories whose proportion is greater than a predetermined value (S32: YES), it reads summaries of notable items of the patent documents classified into categories whose proportion is greater than a predetermined value (target categories), for example, from a post-classification list (S33). Then, based on the summaries of notable items in each patent document read out, the control unit 11 further classifies each patent document into multiple categories (S34). Here, the control unit 11 generates a prompt in the prompt shown in Figure 8A that includes, for example, an instruction statement for a process to classify into two categories, and information on the patent documents classified into the target categories. The generated prompt is input to the language model M to obtain the category classification result. In the classified list shown in Figure 6B, the control unit 11 updates the information on the target categories to each further categorized category and the information on the patent documents classified into each category. For example, as shown in Figure 6C, if it is determined that the proportion of the category "Means related to transcription recording" is greater than a predetermined value (e.g., 50%), the patent documents classified into this category are classified into two categories. As a result, the category "Means related to transcription recording" in the classified list shown in Figure 6B is updated to the categories "Means related to temperature control" and "Means related to photographic paper," as shown in Figure 7A.

[0031] The control unit 11 counts the number of patent documents classified into each category based on the updated classified list (S35), and calculates the ratio of the counted number of patent documents to the total number of patent documents (the ratio of patent documents classified into each category to the total number of patent documents) (S36). Based on the calculation results, the control unit 11 updates the classification result table as shown in Figure 6C. Then the control unit 11 returns to step S32 and determines whether there are any categories whose ratio is greater than a predetermined value (S32), and repeats steps S32 to S36 until it determines that there are no categories whose ratio is greater than a predetermined value.

[0032] If the control unit 11 determines that there are no categories whose proportion is greater than a predetermined value (S32: NO), it reads the contents of the categories registered in the post-classification list obtained at this point and generates a category list of items of interest to be processed (S37). For example, it reads the categories registered in the post-classification list shown in Figure 7A and generates the category list shown in Figure 7B. Then, based on the summary of items of interest to be processed in each patent document, the control unit 11 classifies each patent document into the category in the generated category list (S38). Here, the control unit 11 generates a prompt as shown in Figure 8B, inputs the generated prompt to the language model M, and obtains the category classification result for the patent document from the language model M. The prompt shown in Figure 8B further includes the category list to be classified as shown in Figure 7B in the category classification, in addition to the prompt shown in Figure 8A. The control unit 11 generates a post-classification list as shown in Figure 7A as the classification result to each category in the category list and stores it in the storage unit 12 (S39).

[0033] If step S25 determines that the item of interest is the filing date or applicant (S25:YES), the control unit 11 classifies each patent document by filing date or by applicant based on the filing date or applicant registered in the selection element list (S40). The control unit 11 may also classify the filing dates by predetermined periods (e.g., every 6 months, every 1 year, etc.). The control unit 11 generates a classified list as the classification result by filing date or by applicant and stores it in the storage unit 12 (S41). In the classified list of filing dates and applicants shown in Figure 6B or Figure 7A, the information of each filing date or each applicant is registered as a category, the filing date or applicant is registered instead of an abstract, and no score is registered.

[0034] After processing in step S39 or step S41, the control unit 11 determines whether the above processing has been completed for all of the items of interest specified in step S13 (S42). If it determines that the processing has not been completed (S42: NO), it returns to step S24 and executes the processing in steps S24 to S41 for the unprocessed items of interest. As a result, category classification at an appropriate level of granularity is performed for all patent documents based on the description of each item of interest specified in step S13, and a post-classification list can be generated for each item of interest.

[0035] If the control unit determines that processing has been completed for all items of interest (S42: YES), the control unit 11 determines, based on the classified list for each item of interest, which of the combinations of the categories of the items of interest on the first axis and the categories of the items of interest on the second axis each patent document belongs to, and identifies the patent documents that belong to each combination (S43). The control unit 11 counts the number of patent documents that belong to each combination of categories of each item of interest (S44). As a result of the counting, the control unit 11 generates a classification result table as shown in Figure 7C and stores it in the storage unit 12. The classification result table shown in Figure 7C registers the number of patent documents that belong to each combination of the categories classified for the items of interest on the first axis (problems) and the categories classified for the items of interest on the second axis (solutions).

[0036] The control unit 11 creates a patent map as shown in Figure 9A based on the classification result table shown in Figure 7C (S45). Figure 9A shows an example of the patent map screen. In the patent map shown in Figure 9A, each category of the items of interest on the first axis in the classification result table of Figure 7C is shown on the horizontal axis, and each category of the items of interest on the second axis is shown on the vertical axis. The number of patent documents corresponding to each category combination is plotted at the intersection of the vertical line representing each category on the first axis and the horizontal line representing each category on the second axis. By plotting a larger circle when there are more documents, the number of documents corresponding to each category and each combination of categories can be intuitively grasped. The shape of the plotted points is not limited to the example in Figure 9A, and the configuration can be such that points corresponding to the number of patent documents classified into each category combination are plotted.

[0037] The patent map shown in Figure 9A represents a problem-solution map, with the problem on the horizontal axis and the solution on the vertical axis. Each plot point in the patent map is linked to a patent document classified into each combination of categories. When any plot point is manipulated, the control unit 11 displays a list screen of patent documents as shown in Figure 9B. This allows the user to identify the patent documents included in the specified plot point. The screen in Figure 9B displays information on seven patent documents classified into the "Ink Concentration Unevenness Countermeasures" category for the "Problem" item on the horizontal axis, and into the "Ink-Related Means" category for the "Solution" item on the vertical axis. The screen in Figure 9B is generated when the control unit 11 reads the patent document information and score of the patent documents classified into each category from the classified list generated in step S39. By displaying such a list screen, the user can understand the likelihood that each patent document is classified into each category. Alternatively, the list screen may also display summaries of each patent document obtained from the selection element list in Figure 6A.

[0038] The above-described process enables categorization at an appropriate granularity when classifying patent documents using the language model M for each item of interest. For example, if the range of categories is narrow, the number of documents classified into a category will be small, and if the range of categories is wide, the number of documents classified into a category will be large. In such classifications, it becomes difficult to grasp the differences and trends between each patent document. Also, categorization based on the number and frequency of terms extracted by morphological analysis, for example, may not adequately grasp the content of the documents, making it highly likely that each document cannot be classified appropriately. On the other hand, in this embodiment, categorization is performed based on a summary of the content of the item of interest, so appropriate categorization is possible according to the content of each document. Furthermore, categorization is possible such that the number of patent documents classified into each category is within an appropriate range. In other words, pinpoint categories with too few patent documents and ambiguous categories with too many patent documents are suppressed, so by using the results of such categorization, it is possible to create a patent map that allows for appropriate comparison between each patent document. In this embodiment, the language model M is used when generating a summary of the text and when classifying each patent document into categories. However, separate language models may be provided for summary generation and category classification.

[0039] In this embodiment, each patent document is classified into one of several combinations of the categories of the items of interest on the first axis and the categories of the items of interest on the second axis, but it may be classified into multiple combinations. When classifying patent documents into categories for each item of interest, it is not limited to being classified into one category, but may be classified into multiple categories. In this case, each patent document may fall into one of several combinations of the categories of the items of interest on the first axis and the categories of the items of interest on the second axis, and a single patent document can be classified into multiple combinations.

[0040] In this embodiment, a patent map was created that represents two items of interest on the horizontal and vertical axes. However, a configuration that creates a patent map representing three items of interest on three axes is also possible. In this case, the item of interest input screen may be configured to accept the specification of three items of interest. Alternatively, the system may accept the specification of three or more items of interest via the item of interest input screen, process the three or more specified items of interest to create a classified list for each item of interest, and then accept the specification of two items of interest to be used for the horizontal and vertical axes when creating the patent map. In this case, by using the classified lists of the two specified items from the classified lists created for each item of interest, a patent map like the one shown in Figure 9A can be created and presented. Alternatively, a configuration that generates a one-axis patent map representing one item of interest is also possible.

[0041] (Embodiment 2) This embodiment describes an information processing device that creates a patent map using a target patent document (hereinafter referred to as the target document) and multiple patent documents to be compared with the target document as processing targets, and adds a mark to the created patent map indicating the category into which the target document is classified. The information processing device 10 of this embodiment has the same configuration as the information processing device 10 of Embodiment 1 shown in Figure 1, so a description of the configuration will be omitted.

[0042] Figure 10 is a flowchart showing an example of the patent map creation process procedure in Embodiment 2, and Figure 11 is an explanatory diagram showing an example of a patent map. The process shown in Figure 10 is the same as the process shown in Figures 2 to 4, but with step S51 added before step S11 and steps S52 to S53 added after step S45. The explanation of the same steps as in Figures 2 to 4 is omitted. Also, in Figure 10, the illustration of steps other than steps S11 and S45 in Figures 2 to 4 is omitted.

[0043] In the information processing device 10 of this embodiment, the control unit 11 acquires the target patent document (target document) (S51). For example, the control unit 11 may acquire the target document via the input unit 14 or via the communication unit 13, or it may have previously downloaded the target document and stored it in the storage unit 12. The control unit 11 acquires a plurality of patent documents to be compared with the target document as a plurality of patent documents to be used in creating a patent map (S11). For example, the control unit 11 acquires patent documents obtained by a document search based on keywords extracted from the target document from a server such as the Patent Information Platform (J-PlatPat). The control unit 11 may also acquire patent documents that are the same as the target document in terms of International Patent Classification (IPC), File Index (FI), or technical field (e.g., F-term (File Forming Term)), or it may acquire patent documents that have been narrowed down from the acquired patent documents by a document search based on keywords extracted from the target document. The control unit 11 may also acquire patent documents whose similarity to the target document is above a predetermined value. The similarity between the target document and each patent document can be measured using similarity metrics such as correlation coefficient or cosine similarity. Alternatively, a machine learning model may be used to estimate the similarity between the target document and each patent document. For example, a machine learning model, such as a Convolutional Neural Network (CNN), can be used, which is trained to output the similarity between two documents when the target document and patent documents are input. Furthermore, if the applicant of the patent document to be compared with the target document is known, the control unit 11 may receive the applicant's information and retrieve the patent documents of the received applicant. In this case, the control unit 11 may also narrow down the selection to retrieve patent documents that share the same International Patent Classification or FI as the target document.

[0044] The control unit 11 processes the target document acquired in step S51 and the multiple patent documents acquired in step S11 as documents to be processed in steps S12 to S45, and creates a patent map as shown in Figure 9A. After processing in step S45, the control unit 11 acquires the categories of the items of interest on the first axis and the items of interest on the second axis to which the target document is classified (S52). Here, the control unit 11 identifies the categories to which the target document is classified for the items of interest on the first axis and the items of interest on the second axis, based on the classified list shown in Figures 6B and 7A. The control unit 11 then adds a mark S as shown in Figure 11 to the plot points corresponding to the combination of the identified categories of the items of interest on the first axis and the items of interest on the second axis (S53). In the example in Figure 11, the mark S is added to the plot points of patent documents classified into the category of "improvement of color expression" in the items of interest (problems) on the horizontal axis and the category of "image processing" in the items of interest (solutions) on the vertical axis.

[0045] In this embodiment, a patent map can be created using multiple patent documents that are used for comparison with the target document. Furthermore, by indicating the classification result of the target document with a mark S on the patent map, the position of the target document on the patent map can be easily grasped. Note that the classification result of the target document is not limited to the configuration shown by the mark S, but can be highlighted in a manner different from other plot points.

[0046] The information processing device 10 of this embodiment performs the same processing as in Embodiment 1, except for the process of adding a mark S indicating the location of the target document to the created patent map, and therefore the same effects as in Embodiment 1 can be obtained. In this embodiment as well, the modifications described as appropriate in Embodiment 1 described above can be applied.

[0047] The following additional information is disclosed regarding embodiments including the above embodiments 1 and 2.

[0048] (Note 1) Obtained multiple patent documents, From each of the aforementioned multiple patent documents, extract the sentences described in the items of interest among the items contained in the patent documents, Using a language model, a summary of the extracted text is generated. Based on the generated abstract, a language model is used to classify the multiple patent documents into multiple categories. An information processing method in which a computer performs the processing.

[0049] (Note 2) Obtained multiple patent documents, We accept the designation of multiple items of interest from among the items included in the patent document. From each of the aforementioned multiple patent documents, extract the sentences described in each designated item of interest, Using a language model, a summary of each extracted sentence is generated. For each item of interest, the multiple patent documents are classified into multiple categories using a language model based on the generated summary text. An information processing method in which a computer performs the processing.

[0050] (Note 3) The number of patent documents classified into each category of the first of the aforementioned multiple items of interest and each category of the second item of interest is counted. A graph is generated in which each category in the first area of ​​interest is shown on the first axis, each category in the second area of ​​interest is shown on the second axis, and the number of patent documents classified into each category in the first area of ​​interest and each category in the second area of ​​interest are plotted. The information processing method described in Appendix 2, wherein the computer performs the processing.

[0051] (Note 4) The aforementioned multiple patent documents are classified according to the patent application date or patent applicant for each patent document, The first axis represents each patent application period or each patent applicant, and the second axis represents each category in the items of interest. A graph is generated that plots the number of patent documents for each patent application period or each patent applicant that are classified into each category in the items of interest. An information processing method described in any one of the appendices 1 to 3, wherein the computer performs the processing.

[0052] (Note 5) The proportion of patent documents classified into each category of the items of interest among the aforementioned multiple patent documents is calculated. The number of categories in which the proportion of patent documents classified into each category is less than a predetermined value is counted. If the number of categories counted is greater than a predetermined value, the process of classifying the multiple patent documents into multiple categories is performed until the number of categories in which the proportion of patent documents classified into each category is less than a predetermined value, until the number of categories in which the proportion of patent documents classified into each category is less than a predetermined value is less than or equal to a predetermined value. An information processing method described in any one of the appendices 1 to 4, wherein the computer performs the processing.

[0053] (Note 6) The proportion of patent documents classified into each category of the items of interest among the aforementioned multiple patent documents is calculated. If the proportion of patent documents classified into each category exceeds a predetermined value, the category is divided into multiple categories based on the summary of the items of interest in the patent documents classified into that category. The aforementioned multiple patent documents are classified into multiple categories, including the multiple categories into which they were divided. An information processing method described in any one of the appendices 1 to 5, wherein the computer performs the processing.

[0054] (Note 7) The chart is generated by plotting points in a manner corresponding to the number of patent documents classified into each category of the first item of interest and each category of the second item of interest. An information processing method described in any one of the appendices 3 to 6, wherein the computer performs the processing.

[0055] (Note 8) Retrieve the target document, Obtain the aforementioned multiple patent documents to be compared with the acquired target document. An information processing method described in any one of the appendices 1 to 7, wherein the computer performs the processing.

[0056] (Note 9) We accept the designation of multiple items of interest from among the items included in the patent document. From each of the acquired target documents and the multiple patent documents, extract the sentences described in each of the specified items of interest. For each item of interest, based on the summary generated from the extracted text, a language model is used to classify the target document and the multiple patent documents into multiple categories. The number of target documents and patent documents classified into each category of the first of the multiple items of interest and each category of the second item of interest is counted. A graph is generated in which each category in the first item of interest is shown on the first axis, each category in the second item of interest is shown on the second axis, and the number of target documents and patent documents classified into each category in the first item of interest and each category in the second item of interest are plotted. Highlight the plot points in the generated chart that include the target document. The information processing method described in Appendix 8, wherein the computer performs the processing.

[0057] (Note 10) Obtained multiple patent documents, From each of the aforementioned multiple patent documents, extract the sentences described in the items of interest among the items contained in the patent documents, Using a language model, a summary of the extracted text is generated. Based on the generated abstract, a language model is used to classify the multiple patent documents into multiple categories. A program that instructs a computer to perform a process.

[0058] (Note 11) In an information processing device having a control unit, The control unit, Obtained multiple patent documents, From each of the aforementioned multiple patent documents, extract the sentences described in the items of interest among the items contained in the patent documents, Using a language model, a summary of the extracted text is generated. Based on the generated abstract, a language model is used to classify the multiple patent documents into multiple categories. Information processing device.

[0059] The technical features (constituent elements) described in each embodiment are interchangeable, and by combining them, new technical features can be formed. Furthermore, the independent and dependent claims described in the claims can be combined with each other in any combination, regardless of the form of reference. In addition, the claims use a multi-claim format in which claims refer to two or more other claims (multi-claim format), but are not limited to this. A multi-claim format in which at least one multi-claim is referenced (multi-multi-claim) may also be used.

[0060] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the invention is indicated by the claims, not in the sense described above, and all modifications are intended to be in the sense and scope equivalent to the claims. [Explanation of Symbols]

[0061] 10 Information Processing Devices 11 Control Unit 12 Storage section 13 Communications Department 14 Input section 15 Display M language model

Claims

1. Obtained multiple patent documents, From each of the aforementioned multiple patent documents, extract the sentences described in the items of interest among the items contained in the patent documents, Using a language model, a summary of the extracted text is generated. Based on the generated abstract, a language model is used to classify the multiple patent documents into multiple categories. An information processing method in which a computer performs the processing.

2. Obtained multiple patent documents, We accept the designation of multiple items of interest from among the items included in the patent document. From each of the aforementioned multiple patent documents, extract the sentences described in each designated item of interest, Using a language model, a summary of each extracted sentence is generated. For each item of interest, the multiple patent documents are classified into multiple categories using a language model based on the generated summary text. An information processing method in which a computer performs the processing.

3. The number of patent documents classified into each category of the first of the aforementioned multiple items of interest and each category of the second item of interest is counted. A graph is generated in which each category in the first area of ​​interest is shown on the first axis, each category in the second area of ​​interest is shown on the second axis, and the number of patent documents classified into each category in the first area of ​​interest and each category in the second area of ​​interest are plotted. The information processing method according to claim 2, wherein the processing is performed by the computer.

4. The aforementioned multiple patent documents are classified according to the patent application date or patent applicant for each patent document, The first axis represents each patent application period or each patent applicant, the second axis represents each category in the items of interest, and a graph is generated that plots the number of patent documents for each patent application period or each patent applicant that are classified into each category in the items of interest. The information processing method according to claim 1 or 2, wherein the computer performs the processing.

5. The proportion of patent documents classified into each category of the items of interest among the aforementioned multiple patent documents is calculated. The number of categories in which the proportion of patent documents classified into each category is less than a predetermined value is counted. If the number of categories counted is greater than a predetermined value, the process of classifying the multiple patent documents into multiple categories is performed until the number of categories in which the proportion of patent documents classified into each category is less than a predetermined value, until the number of categories in which the proportion of patent documents classified into each category is less than a predetermined value is less than or equal to a predetermined value. The information processing method according to any one of claims 1 to 3, wherein the computer performs the processing.

6. The proportion of patent documents classified into each category of the items of interest among the aforementioned multiple patent documents is calculated. If the proportion of patent documents classified into each category exceeds a predetermined value, the category is divided into multiple categories based on the summary of the items of interest in the patent documents classified into that category. The aforementioned multiple patent documents are classified into multiple categories, including the multiple categories into which they were divided. The information processing method according to any one of claims 1 to 3, wherein the computer performs the processing.

7. The chart is generated by plotting points in a manner corresponding to the number of patent documents classified into each category of the first item of interest and each category of the second item of interest. The information processing method according to claim 3, wherein the computer performs the processing.

8. Retrieve the target document, Obtain the aforementioned multiple patent documents to be compared with the acquired target document. The information processing method according to any one of claims 1 to 3, wherein the computer performs the processing.

9. We accept the designation of multiple items of interest from among the items included in the patent document. From each of the acquired target documents and the multiple patent documents, extract the sentences described in each of the specified items of interest. For each item of interest, based on the summary generated from the extracted text, a language model is used to classify the target document and the multiple patent documents into multiple categories. The number of target documents and patent documents classified into each category of the first of the multiple items of interest and each category of the second item of interest is counted. A graph is generated in which each category in the first item of interest is shown on the first axis, each category in the second item of interest is shown on the second axis, and the number of target documents and patent documents classified into each category in the first item of interest and each category in the second item of interest are plotted. Highlight the plot points in the generated chart that include the target document. The information processing method according to claim 8, wherein the computer performs the processing.

10. Obtained multiple patent documents, From each of the aforementioned multiple patent documents, extract the sentences described in the items of interest among the items contained in the patent documents, Using a language model, a summary of the extracted text is generated. Based on the generated abstract, a language model is used to classify the multiple patent documents into multiple categories. A program that instructs a computer to perform a process.

11. In an information processing device having a control unit, The control unit, Obtained multiple patent documents, From each of the aforementioned multiple patent documents, extract the sentences described in the items of interest among the items contained in the patent documents, Using a language model, a summary of the extracted text is generated. Based on the generated abstract, a language model is used to classify the multiple patent documents into multiple categories. Information processing device.

Images