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

Abstract

To provide an information processing device that enables even a user without specialized knowledge to create a search formula, reduces the time and effort required for creating a search formula, and improves the accuracy of search results. [Solution] An outline of the invention is accepted, and a patent classification code or keyword is predicted using a trained generative model based on the outline of the invention obtained from the acceptance unit. A search formula is generated based on the patent classification code or keyword predicted by the prediction unit. Patent documents are searched based on the search formula generated by the search formula generation unit. The search results obtained by the search unit are input to the prediction unit, and the patent classification code or keyword is predicted again using the outline of the invention and the search results as input.

Description

[Technical Field]

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

[0002] In recent years, the massive increase in patent information has created a demand for highly accurate patent searches when filing patent applications and verifying existing technologies. Creating traditional patent search formulas requires specialized knowledge and advanced skills to appropriately select patent classifications and keywords. Furthermore, patent searches using AI require the interpretation and adjustment of search results, raising challenges in search efficiency and accuracy.

[0003] Patent Document 1 provides a device and program for classifying patent information, and discloses that patent classification information is used to achieve classification prediction and accuracy improvement using an optimal algorithm. This technology includes an algorithm determination device for classifying patent information, and is equipped with the function of generating a learning device and obtaining predicted classification information. It also discloses a process for evaluating the accuracy of the predicted classification information and obtaining algorithm information based on the predicted classification information with the highest accuracy. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2021-036427 Summary of the Invention [Problem to be solved by the invention]

[0005] However, although Patent Document 1 discloses the classification of patent information and the optimization of algorithms, it does not disclose a specific method for generating search queries for actual patent searches, including keywords and patent classifications.

[0006] Therefore, the present invention aims to provide an information processing device that allows even users without specialized knowledge to create search queries, thereby reducing the effort required to create search queries and improving the accuracy of search results. [Means for solving the problem]

[0007] In order to solve the above problems, the information processing device of the present invention includes a receiving unit that receives an outline of the invention, and a processing unit that receives the outline of the invention from the receiving unit. of Trained generative model Enter a prediction unit that predicts a patent classification code or a keyword; a search expression generation unit that generates a search expression based on the patent classification code or the keyword predicted by the prediction unit; and a search unit that searches for patent documents based on the search expression generated by the search expression generation unit. before The outline of the invention and the search results obtained by the search section The contents of the patent document as input to the trained generative model a re-prediction unit that re-predicts the patent classification code or keyword; The search query generation unit regenerates a search query based on the patent classification code or keyword re-predicted by the re-prediction unit. It is characterized by: [Effects of the Invention]

[0008] According to the present invention, even users without specialized knowledge can create search queries, thereby reducing the effort required to create search queries and providing an information processing device that can improve the accuracy of search results. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a diagram illustrating an overall configuration including a re-prediction unit, showing an example of an information processing device according to the present invention; [Figure 2] 1 is a diagram illustrating an overall configuration of an example of an information processing device according to the present invention, excluding a re-prediction unit. [Figure 3] 1 is a process flow diagram showing an example of an information processing apparatus according to the present invention; [Figure 4] 1 is a diagram showing an example of a data flow of an information processing device according to the present invention, including a re-prediction unit. [Figure 5]FIG. 10 is a diagram showing an example of a data flow of an information processing device according to the present invention, which does not include a re-prediction unit. [Figure 6] FIG. 1 is a diagram showing an example of an outline input and search query display screen of the invention in the first embodiment. [Figure 7] FIG. 10 is a diagram showing an example of a search result display screen in the first and second embodiments. [Figure 8] FIG. 10 is a diagram showing an example of an outline input and search query display screen of the invention in the second embodiment. [Figure 9] FIG. 10 is a diagram showing an example of a search result display screen in the second embodiment. [Figure 10] FIG. 10 is a diagram showing an example of a regenerated query formula display screen in Example 2. DETAILED DESCRIPTION OF THE INVENTION

[0010] The following describes embodiments of the present invention. The present invention is not limited to the embodiments described below, and may be modified, added, modified, deleted, or otherwise altered within the scope of what one skilled in the art can conceive. Any embodiment is within the scope of the present invention as long as it exhibits the functions and effects of the present invention.

[0011] The information processing device of the present invention is characterized by comprising: a reception unit that receives an outline of an invention; a prediction unit that predicts a patent classification code or keyword using a trained generative model based on the outline of the invention obtained from the reception unit; a search formula generation unit that generates a search formula based on the patent classification code or keyword predicted by the prediction unit; a search unit that searches for patent documents based on the search formula generated by the search formula generation unit; and a re-prediction unit that re-predicts a patent classification code or keyword using a trained generative model based on the outline of the invention and the search results obtained by the search unit.

[0012] The information processing device of the present invention comprises a reception unit that receives an outline of an invention, a prediction unit that predicts a patent classification code or keyword using a trained generative model based on the outline of the invention obtained from the reception unit, a search formula generation unit that generates a search formula based on the patent classification code or keyword predicted by the prediction unit, and a search unit that searches for patent documents based on the search formula generated by the search formula generation unit, and is characterized in that the prediction unit re-predicts the patent classification code or keyword using the trained generative model based on the outline of the invention and the search results obtained by the search unit.

[0013] It is preferable that the search query generation unit regenerates a search query based on the patent classification code or keyword re-predicted by the re-prediction unit, and the search unit re-searches patent documents based on the re-generated search query.

[0014] It is preferable that the processes by the re-prediction unit, the search query generation unit, and the search unit are repeatedly executed multiple times.

[0015] It is preferable that the search query generation unit regenerates a search query based on the patent classification code or keyword re-predicted by the prediction unit, and the search unit re-searches patent documents based on the regenerated search query.

[0016] It is preferable that the processes by the prediction unit, the search formula generation unit, and the search unit are executed multiple times.

[0017] It is preferable that the information processing device includes a display unit, which displays the search formula generated by the search formula generation unit on a screen and allows the user to edit the search formula.

[0018] The information processing method of the present invention is an information processing method performed by a computer, and is characterized by including: a reception step for receiving an outline of an invention; a prediction step for predicting a patent classification code or keyword using a trained generative model based on the outline of the invention obtained from the reception step; a search formula generation step for generating a search formula based on the patent classification code or keyword predicted by the prediction step; a search step for searching patent documents based on the search formula generated by the search formula generation step; and a re-prediction step for re-predicting a patent classification code or keyword using a trained generative model based on the outline of the invention and the search results obtained in the search step.

[0019] The information processing method of the present invention is an information processing method performed by a computer, and includes: a reception step for receiving an outline of an invention; a prediction step for predicting a patent classification code or keyword using a trained generative model based on the outline of the invention obtained from the reception step; a search formula generation step for generating a search formula based on the patent classification code or keyword predicted by the prediction step; and a search step for searching patent documents based on the search formula generated by the search formula generation step, wherein the prediction step is characterized in that the patent classification code or keyword is re-predicted using the trained generative model based on the outline of the invention and the search results obtained in the search step.

[0020] The information processing program of the present invention is characterized in that it causes a computer to execute a reception process for receiving an outline of an invention; a prediction process for predicting a patent classification code or keyword using a trained generative model based on the outline of the invention obtained from the reception process; a search expression generation process for generating a search expression based on the patent classification code or keyword predicted by the prediction process; a search process for searching patent documents based on the search expression generated by the search expression generation process; and a re-prediction process for re-predicting a patent classification code or keyword using a trained generative model based on the outline of the invention and the search results obtained by the search process.

[0021] In addition, the information processing program of the present invention is an information processing program that causes a computer to execute a reception process that receives an outline of an invention, a prediction process that predicts a patent classification code or keyword using a trained generative model based on the outline of the invention obtained from the reception process, a search formula generation process that generates a search formula based on the patent classification code or keyword predicted by the prediction process, and a search process that searches for patent documents based on the search formula generated by the search formula generation process, wherein the prediction process is characterized in that the patent classification code or keyword is re-predicted using the trained generative model based on the outline of the invention and the search results obtained by the search process.

[0022] Fig. 1 is a block diagram showing the overall configuration of an information processing device according to one embodiment of the present invention. As shown in Fig. 1, information processing device 10 is a device for predicting patent classification codes and keywords, searching patent documents, etc., and includes a reception unit 11, a prediction unit 12, a search formula generation unit 13, a search unit 14, a re-prediction unit 15, and a display unit 16.

[0023] The reception unit 11 receives an outline of an invention input by a user 40 via a terminal 41. The reception unit 11 can receive, as the outline of an invention, information such as a description of the invention, a technical field, a problem that the invention is intended to solve, a means for solving the problem, and effects of the invention.

[0024] Examples of text related to an invention include the title of the invention, the purpose of the invention, the structure of the invention, the function of the invention, examples, experimental data, and comparative examples. Examples of technical fields include mechanical, electrical, chemical, information and communications, biotechnology, and other fields, as well as information indicating more detailed technical categories. Examples of problems include issues with existing technology, market needs, technical constraints, and cost issues. Examples of means for solving problems include novel device configurations, control methods, material compositions, and system configurations. Examples of benefits of the invention include improved performance, cost reductions, improved work efficiency, and reduced environmental impact. In addition to entering a summary of the invention, users 40 can also enter supplementary information such as drawing data related to the invention and information on known patent documents.

[0025] The prediction unit 12 predicts a patent classification code or a keyword using a trained generative model based on the outline of the invention received from the reception unit 11. Examples of trained generative models used by the prediction unit 12 include generative AI and large-scale language models (LLM). The trained generative model may also be referred to as a trained model, etc. Furthermore, the generative AI may also be referred to as generative AI, etc.

[0026] Examples of generative AI that can be used include the GPT model, PaLM model, and LLaMA model. It is also possible to use a patent-specific generative AI model that has been pre-trained on a dataset of patent documents. Generative AI can not only directly predict patent classification codes from an invention summary, but also explain the features of the invention in natural language and derive classification codes from that explanation.

[0027] Examples of large-scale language models include models using Transformer-type architecture, the T5 model, BERT, etc. In particular, large-scale language models that include patent documents as training data are suitable for understanding technical documents and predicting patent classifications.

[0028] The trained generative model is trained using, for example, a combination of patent documents and their patent classification codes as training data, which may include, for example, text data of the patent documents, patent classification codes, keywords, technical terms, invention titles, abstracts, claims, detailed descriptions, etc.

[0029] The prediction unit 12 uses the trained generative model to output, as prediction results, patent classification codes such as the International Patent Classification (IPC), File Index (FI), and F-terms, as well as keywords that represent the characteristics of the invention, technical terms, synonyms, hypernyms, and hyponyms. The predicted classification codes and keywords can also be assigned scores of prediction confidence and similarity. When predicting patent classification codes or keywords using a trained generative model, for example, a prompt is used as input data for the trained generative model, and the prompt includes instructions for the trained generative model. Examples of instructions include an instruction to read an invention summary and predict a patent classification code or keyword. Examples of instructions include an instruction to read an invention summary and predict a patent classification code or keyword for searching patent documents. For example, the trained generative model is input with instructions such as "Predict the related FI (file index) from the invention summary. The invention content is as follows. Invention summary: An invention in which a drone holds an umbrella" or "Extract related keywords from the invention summary. The invention summary is as follows. Invention content: An invention in which a drone holds an umbrella." The prediction unit 12 may predict both the patent classification code and the keyword.

[0030] The search formula generation unit 13 generates a search formula based on the patent classification codes or keywords predicted by the prediction unit 12. The search formula generation unit 13 generates a search formula, for example, by combining patent classification codes or keywords with logical operators (AND, OR, NOT, etc.).

[0031] Methods for generating search queries by the search query generation unit 13 include, for example, a method based on predetermined rules, a method using generative AI, etc. In rule-based generation, combination patterns of patent classification codes and keywords are defined in advance, and search queries are constructed accordingly. In methods using generative AI, optimal search queries are automatically generated using past search queries and search results as learning data. Although we refer to the "generation" of search queries, this concept includes the creation of search queries. When generating queries based on rules, it is possible to use a generation AI, but not necessarily.

[0032] As a specific example, when using search expressions that only use keywords, it is possible to combine them using simple logical operations, such as "drone AND umbrella." It is also possible to create search expressions that include synonyms, such as "(drone OR unmanned aerial vehicle) AND (umbrella OR parasol)." When including patent classification codes, it is also possible to create search expressions that combine patent classification codes and keywords, such as "(B64C39 / 02 OR B64D1 / 00) AND umbrella."

[0033] The search formula generation unit 13 can generate search formulas in various formats, for example, by combining patent classification codes and keywords, such as hierarchical search formulas, parallel search formulas, nested search formulas, etc. It can also automatically add higher and lower classifications of patent classification codes, synonyms of keywords, and related words.

[0034] When generating a search query, the structure of the search query and the use of operators can be optimized, taking into consideration, for example, the importance of the patent classification code, the frequency of keyword appearance, the specialization of the term, the characteristics of the technical field, etc. In addition, the search query generation rules can be adjusted based on feedback from past search history and search results.

[0035] The generated search formula can be converted into a format commonly used in patent information databases or a format specialized for a specific database system, for example. The generated search formula is also output to the display unit 16 in a format that can be edited or modified by the user 40.

[0036] The search unit 14 searches for patent documents based on the search formula generated by the search formula generation unit 13. The search unit 14 applies the search formula to, for example, a patent information database, and extracts patent documents that match the conditions of the search formula.

[0037] As patent information databases, for example, patent information databases provided by the Japan Patent Office, patent information databases provided by the World Intellectual Property Organization (WIPO), patent information databases provided by patent offices of various countries, patent information databases provided by private companies, etc. It is also possible to perform a cross-sectional search across multiple databases. Patent documents include, for example, patent applications, patents, utility model registration applications, utility model registrations, and the like.

[0038] The search unit 14 can perform searches, for example, from bibliographic information of patent documents (application number, publication number, filing date, publication date, applicant, inventor, etc.), abstracts, claims, full specifications, drawing descriptions, etc. Searches can also be performed by specifying various conditions such as the language, country / region, and filing year of the patent document to be searched.

[0039] The search results may include, for example, a list of patent documents that match the search query, bibliographic information, summaries, representative drawings, etc. Search results may also be assigned a relevance score, or may be classified and organized based on the technical field, application year, etc. The search results are presented to the user 40 via the display unit 16.

[0040] The re-prediction unit 15 re-predicts the patent classification code or keyword using the trained generative model based on the invention summary and the search results obtained by the search unit 14. For example, the re-prediction unit 15 combines the content of the patent document obtained as the search result with the invention summary that was initially input, and inputs this into the trained generative model to predict a more appropriate patent classification code or keyword.

[0041] The trained generative model used in the re-prediction unit 15 may be, for example, a generative AI or a large-scale language model (LLM). The trained generative model may be the same as the model used in the prediction unit 12, or may be a different model.

[0042] The re-prediction and search process can be repeated multiple times until a predetermined condition is met, such as when the change in the prediction result is equal to or less than a predetermined threshold, when the number of re-predictions reaches a predetermined number, or when the relevance of the search result meets a predetermined standard.

[0043] The display unit 16 displays the search formula generated by the search formula generation unit 13 on a screen and allows the user 40 to edit the search formula. The display unit 16 displays, for example, patent classification codes, keywords, logical operators, etc. that make up the search formula in an editable format. The display unit 16 can also display a list of search results by the search unit 14.

[0044] Fig. 2 is a block diagram showing the overall configuration of an information processing device according to another embodiment of the present invention. As shown in Fig. 2, information processing device 10 is a device for predicting patent classification codes and keywords, searching patent documents, etc., and includes reception unit 11, prediction unit 12, search formula generation unit 13, search unit 14, and display unit 16.

[0045] The prediction unit 12 of this embodiment has a function of predicting a patent classification code or keyword using a trained generative model based on the invention summary obtained from the reception unit 11, as well as a function of re-predicting a patent classification code or keyword using a trained generative model based on the invention summary and the search results obtained by the search unit 14. In other words, the prediction unit 12 has both an initial prediction function and a re-prediction function.

[0046] In the re-prediction by the prediction unit 12, for example, the initially input invention summary and the contents of the patent documents obtained as search results are combined and input into the trained generative model to predict a more appropriate patent classification code or keyword. The prediction and search process can be repeated multiple times until, for example, predetermined conditions are met. When re-predicting patent classification codes or keywords using a trained generative model, for example, a prompt is used as input data for the trained generative model, and the prompt includes instructions for the trained generative model. Examples of instructions include an instruction to read the invention summary and search results and predict a patent classification code or keyword. Examples of instructions include an instruction to read the invention summary and search results and predict a patent classification code or keyword for searching patent documents. For example, an instruction such as "Predict related FIs (file indexes) from the invention summary and patent search results. The invention details are as follows. Invention summary: Invention of holding an umbrella on a drone. Patent search results: Invention title ~, claim ~" or an instruction such as "Extract related keywords from the invention summary and patent search results. The invention summary is as follows. Invention details: Invention of holding an umbrella on a drone. Patent search results: Invention title ~, claim ~" is input to the trained generative model.

[0047] 3 is a flowchart showing the processing flow in an information processing device according to one embodiment of the present invention. Each step will be described below.

[0048] In step S1, the reception unit 11 receives an outline of the invention from the user 40.

[0049] In step S2, the prediction unit 12 predicts a patent classification code or a keyword using a trained generative model based on the outline of the invention.

[0050] In step S3, the search query generation unit 13 generates a search query based on the predicted patent classification code or keyword.

[0051] In step S4, the search unit 14 searches patent documents based on the generated search formula.

[0052] In step S5, the patent classification code or keyword is re-predicted. There are two embodiments of this re-prediction. In the first embodiment, the re-prediction unit 15 re-predicts the patent classification code or keyword using a trained generative model based on the invention summary and search results. In the second embodiment, the prediction unit 12 re-predicts the patent classification code or keyword using a trained generative model based on the invention summary and search results.

[0053] In step S6, it is determined whether or not to perform a search again. For example, if any of the following conditions is met, it is determined that the search again is to be terminated (NO).

[0054] The first condition is when the change in patent classification codes or keywords obtained by re-prediction falls below a predetermined threshold, such as when the number of newly added or deleted patent classification codes falls below a predetermined number, or when the update rate of keywords falls below a predetermined rate.

[0055] The second condition is when the AI ​​evaluates the comprehensiveness of the search results and determines that sufficient comprehensiveness has been achieved. The AI ​​analyzes, for example, whether the patent documents in the search results sufficiently cover the technical field and whether they contain sufficient documents related to the essential features of the invention, and determines the degree of completeness of the search.

[0056] The third condition is when the number of re-prediction and re-search iterations reaches a predetermined number, which functions as a termination condition taking into account processing efficiency.

[0057] If any of the above conditions are not met (YES), the process returns to step S3, a new search formula is generated based on the re-predicted patent classification code or keywords, and the subsequent processes are repeated. If any of the conditions are met (NO), the process ends. Note that these conditions may be used alone or in combination.

[0058] This processing flow uses a trained generative model to predict patent classification codes or keywords from the outline of an invention, and generates search queries based on these, allowing even users without specialized knowledge to create search queries, reducing the effort required to create search queries. Furthermore, repeated re-prediction and re-search using search results enables more appropriate searches for patent documents, improving the accuracy of search results.

[0059] 4 is a diagram showing the flow of data in an information processing device according to one embodiment of the present invention. The flow of data will be explained below.

[0060] A user 40 inputs an outline of the invention via a terminal 41. The input outline of the invention is sent to the reception unit 11.

[0061] The receiving unit 11 sends the received invention summary to the prediction unit 12. The prediction unit 12 sends the invention summary to the trained model 42 and instructs it to predict a patent classification code or keyword. The trained model 42 returns a generated result a (code / keyword) to the prediction unit 12 as the prediction result.

[0062] The prediction unit 12 sends the generated result a to the search expression generation unit 13. The search expression generation unit 13 sends the received patent classification code or keyword to the trained model 43 and instructs it to generate a search expression. The trained model 43 returns the generated result b (search expression) to the search expression generation unit 13.

[0063] The search formula generation unit 13 sends the generated result b to the search unit 14. The search unit 14 uses the received search formula to search the database 44 and extracts patent documents that match the conditions of the search formula. The extracted search results are returned to the search unit 14.

[0064] The search unit 14 sends the search results to the re-prediction unit 15. The re-prediction unit 15 sends the invention summary and search results to the trained model 45 and instructs it to re-predict the patent classification code or keyword. The trained model 45 returns the generated result c (code / keyword) to the re-prediction unit 15 as the prediction result. The re-prediction unit 15 sends the generated result c to the search expression generation unit 13.

[0065] This data flow allows a series of processes to be carried out, starting with an outline of the invention, followed by prediction, search query generation, search, and re-prediction. This allows even users without specialized knowledge to create search queries, reducing the effort required to create search queries.

[0066] 5 is a diagram showing the flow of data in an information processing device according to another embodiment of the present invention. The flow of data will be described below.

[0067] A user 40 inputs an outline of the invention via a terminal 41. The input outline of the invention is sent to the reception unit 11.

[0068] The receiving unit 11 sends the received invention summary to the prediction unit 12. The prediction unit 12 sends the invention summary to the trained model 42 and instructs it to predict a patent classification code or keyword. Note that, as indicated by the * mark in the figure, when making a second prediction, the search results are also used in addition to the invention summary. The trained model 42 returns the generated result a (code / keyword) to the prediction unit 12 as the prediction result.

[0069] The prediction unit 12 sends the generated result a to the search expression generation unit 13. The search expression generation unit 13 sends the received patent classification code or keyword to the trained model 43 and instructs it to generate a search expression. The trained model 43 returns the generated result b (search expression) to the search expression generation unit 13.

[0070] The search formula generation unit 13 sends the generated result b to the search unit 14. The search unit 14 uses the received search formula to search the database 44 and extracts patent documents that match the conditions of the search formula. The extracted search results are returned to the search unit 14.

[0071] The search unit 14 sends the search results to the prediction unit 12. This allows the prediction unit 12 to predict patent classification codes or keywords again based on the invention summary and the search results.

[0072] This data flow allows for a series of processes, starting with an outline of the invention, followed by prediction, search formula generation, and search. By linking with a trained model and searching a database at each stage of the process, even users without specialized knowledge can create search formulas, reducing the effort required to create search formulas.

[0073] Regarding the transmission and reception of data between each unit, for example, although it is stated that the reception unit 11 sends the outline of the invention to the prediction unit 12, this may be achieved by the prediction unit 12 requesting the data from the reception unit 11, or by the exchange of data via another unit. In other words, the data flow in each embodiment does not limit the method of data transmission and reception, and as long as data is appropriately transferred between each unit, the specific method is not particularly limited. Trained model 42, trained model 43, and trained model 45 may be the same or different.

[0074] Next, an example of a preferred embodiment will be described. When the re-prediction unit 15 performs re-prediction, the search formula generation unit 13 re-generates a search formula based on the re-predicted patent classification code or keyword. For example, if a patent classification code or keyword is added, a new search formula is generated by combining them with an appropriate logical operator. Also, if a patent classification code or keyword is deleted, a new search formula is generated that excludes the patent classification code or keyword. Furthermore, if a patent classification code or keyword is changed, a new search formula that reflects the changes is generated.

[0075] The search unit 14 performs a re-search of the patent documents using the re-generated search formula. In the re-search, patent documents that match the conditions of the re-generated search formula are extracted. This makes it possible to obtain more appropriate patent documents that reflect the re-prediction results.

[0076] In one preferred embodiment, the series of processes performed by the re-prediction unit 15, search query generation unit 13, and search unit 14 can be repeated multiple times. That is, the process starts with the re-prediction of patent classification codes or keywords by the re-prediction unit 15, followed by the re-generation of a search query by the search query generation unit 13, and then the re-search of patent documents by the search unit 14. For example, based on the search results obtained by the re-search, the re-prediction unit 15 can perform re-prediction again, a new search query can be re-generated, and a further re-search can be performed, and this cycle can be repeated multiple times. This allows for more accurate search results to be obtained.

[0077] Next, an example of a preferred embodiment will be described. When the prediction unit 12 performs re-prediction, the search formula generation unit 13 re-generates a search formula based on the re-predicted patent classification code or keyword. For example, if a patent classification code or keyword is added, a new search formula is generated by combining them with an appropriate logical operator. Also, if a patent classification code or keyword is deleted, a new search formula is generated that excludes the patent classification code or keyword. Furthermore, if a patent classification code or keyword is changed, a new search formula that reflects the changes is generated.

[0078] The search unit 14 performs a re-search of the patent documents using the re-generated search formula. In the re-search, patent documents that match the conditions of the re-generated search formula are extracted. This makes it possible to obtain more appropriate patent documents that reflect the re-prediction results.

[0079] In one preferred embodiment, the series of processes performed by the prediction unit 12, search query generation unit 13, and search unit 14 can be repeated multiple times. That is, the process starts with the prediction unit 12 re-predicting a patent classification code or keyword, followed by the search query generation unit 13 re-generating a search query, and then the search unit 14 re-searching patent documents. For example, based on the search results obtained from the re-search, the prediction unit 12 can perform re-prediction again, re-generating a new search query, and performing another re-search. This cycle can be repeated multiple times. This allows for more accurate search results to be obtained.

[0080] In one preferred embodiment, the information processing device 10 includes a display unit 16. The display unit 16 displays on a screen the search formula generated by the search formula generation unit 13. For example, the search formula is displayed in an editable format as a combination of patent classification codes, keywords, and logical operators.

[0081] The user 40 can edit the search formula displayed on the display unit 16. For example, the user 40 can add, delete, or change patent classification codes and keywords, as well as change logical operators and modify the structure of the search formula. In this way, the automatically generated search formula can be adjusted according to the user's intentions.

[0082] The search unit 14 executes a search of patent documents based on the search formula edited by the user 40. This allows a search that reflects the intentions and knowledge of the user 40 in the automatically generated search formula.

[0083] Next, an information processing method according to the present invention will be described. The information processing method according to this embodiment is an information processing method executed by a computer.

[0084] Specifically, it includes a reception process for receiving an outline of the invention, a prediction process for predicting a patent classification code or keyword using a trained generative model based on the outline of the invention obtained from the reception process, a search formula generation process for generating a search formula based on the patent classification code or keyword predicted by the prediction process, a search process for searching patent documents based on the search formula generated by the search formula generation process, and a re-prediction process for re-predicting a patent classification code or keyword using a trained generative model based on the outline of the invention and the search results obtained in the search process.

[0085] In another embodiment, the method may include a reception process, a prediction process, a search formula generation process, and a search process, and in the prediction process, a trained generative model is used to re-predict patent classification codes or keywords based on the summary of the invention and the search results obtained in the search process.

[0086] Next, an information processing program according to the present invention will be described. The information processing program according to this embodiment is a program for causing a computer to execute each step of the information processing method.

[0087] Specifically, the computer is made to perform a reception process for receiving an outline of the invention, a prediction process for predicting a patent classification code or keyword using a trained generative model based on the outline of the invention obtained from the reception process, a search expression generation process for generating a search expression based on the patent classification code or keyword predicted by the prediction process, a search process for searching patent documents based on the search expression generated by the search expression generation process, and a re-prediction process for re-predicting a patent classification code or keyword using the trained generative model based on the outline of the invention and the search results obtained by the search process.

[0088] In another embodiment, a computer can be configured to perform reception processing, prediction processing, search query generation processing, and search processing, and in the prediction processing, re-predict patent classification codes or keywords using a trained generative model based on the summary of the invention and the search results obtained in the search processing. [Example]

[0089] The information processing device of the present invention will be described in more detail below with reference to examples, but the present invention is not limited to these examples.

[0090] A first embodiment will be described. As shown in Fig. 6, the user interface of the information processing device includes a search execution screen. The search execution screen displays a message saying "Please enter an outline of the invention," and is provided with an invention outline input field 51. In this embodiment, "a drone holding an umbrella" is entered as the outline of the invention.

[0091] When a user presses the search query generation button 52, the invention summary is first sent to the reception unit 11. The prediction unit 12 predicts a patent classification code or keyword using a trained generative model, and the search query generation unit 13 generates a first search query. The first search query is a basic keyword combination, such as "drone AND umbrella" (not shown). The search unit 14 then executes a search based on this first search query, obtaining first search results. Here, the first search query and the first search results are used for internal system processing and are not displayed to the user. The re-prediction unit 15 then re-predicts the patent classification code or keyword based on the invention summary and the obtained first search results, and the search query generation unit 13 generates a second search query: "(drone or UAV) and (umbrella or parasol) and (automatic control or weather sensing)." This second search query is displayed in the search query display area 53. The displayed second search query is editable, allowing the user to modify it as needed.

[0092] When the user presses the search execution button 54, a search of patent documents is executed based on the displayed second search formula. The search results are displayed on the search result display screen shown in FIG.

[0093] The search results display screen is displayed in a list format including items such as the name of the invention, applicant, filing date, and abstract. In this example, patent documents such as "Aerial Parasol Drone" (Applicant A) are displayed as search results. The display change button 55 allows other items such as claims and detailed descriptions to be displayed, and the report creation button 56 allows a report based on the search results to be created.

[0094] Although the present embodiment illustrates a configuration including a re-prediction unit 15, similar results can be obtained with a configuration in which the prediction unit 12 performs re-prediction. Furthermore, re-prediction, re-generation of a search formula, and re-searching may be performed multiple times between pressing the search formula generation button 52 and displaying the search formula. The first search formula is the search formula generated the first time and has not been re-predicted. The second search formula is a search formula generated after the second time, and is a search formula that reflects the re-prediction.

[0095] Next, a description will be given of Example 2. As shown in Fig. 8, the user interface of the information processing device in Example 2 includes a search execution screen, similar to Example 1. In this example, "a drone holding an umbrella" is entered in the invention summary input field 51.

[0096] When the user presses the search expression generation button 52, the outline of the invention is sent to the reception unit 11, the prediction unit 12 predicts the patent classification code or keywords using the trained generative model, and the search expression generation unit 13 generates the search expression "drone AND umbrella." This search expression is displayed in the search expression display area 53.

[0097] When the user presses the search execution button 54, a search of patent documents is executed based on the displayed search formula. The search results are displayed on the search result display screen shown in Figure 9. The search results are displayed in a list format including items such as the title of the invention, applicant, filing date, and abstract.

[0098] When the user presses the re-execute button 57 on the screen of Fig. 9, the re-prediction unit 15 re-predicts the patent classification code or keyword based on the invention summary and search results. Then, the search formula generation unit 13 generates a new search formula, and the search formula based on the re-prediction, "(drone or UAV) and (umbrella or parasol) and (automatic control or weather sensing)", is displayed in the search formula display area 53, as shown in Fig. 10. When the user presses the search execution button 54 of Fig. 10, the same search results as those in Fig. 7 of Example 1 are obtained.

[0099] Note that re-prediction, re-generation of the search formula, and re-search may be repeated multiple times between pressing the re-execute button 57 and displaying a new search formula. Also in this embodiment, as in the first embodiment, a configuration including the re-prediction unit 15 is illustrated, but similar results can be obtained even if the prediction unit 12 performs re-prediction.

[0100] Furthermore, a report based on the search results can be created by pressing the report creation button 56. This report includes information on the patent documents obtained as search results. [Explanation of symbols]

[0101] 10. Information processing equipment 11 Reception 12 Prediction Department 13 Search Expression Generation Unit 14 Search section 15 Re-prediction section 40 users 41 terminals 42 trained models 43 trained models 44 databases 51 Invention summary input field 52 Search expression generation button 53 Search expression display area 54 Search execution button 55 Display change button 56 Report creation button 57 Retry button

Claims

1. a reception unit for receiving an outline of the invention; a prediction unit that inputs the outline of the invention acquired from the reception unit into a trained generative model to predict a patent classification code or a keyword; a search query generation unit that generates a search query based on the patent classification code or keyword predicted by the prediction unit; a search unit that searches for patent documents based on the search formula generated by the search formula generation unit; a re-prediction unit that inputs the outline of the invention and the contents of the patent documents obtained as search results by the search unit into a trained generative model and re-predicts patent classification codes or keywords; Equipped with The information processing device is characterized in that the search formula generation unit regenerates a search formula based on the patent classification code or keyword re-predicted by the re-prediction unit.

2. a reception unit for receiving an outline of the invention; a prediction unit that inputs the outline of the invention acquired from the reception unit into a trained generative model to predict a patent classification code or a keyword; a search query generation unit that generates a search query based on the patent classification code or keyword predicted by the prediction unit; a search unit that searches for patent documents based on the search formula generated by the search formula generation unit; Equipped with the prediction unit inputs the outline of the invention and the contents of the patent documents obtained as search results by the search unit into a trained generative model to re-predict patent classification codes or keywords; The information processing device is characterized in that the search formula generation unit regenerates a search formula based on the patent classification code or keyword predicted again by the prediction unit.

3. 2. The information processing apparatus according to claim 1, wherein the search unit searches again for patent documents based on the regenerated search query.

4. 4. The information processing apparatus according to claim 3, wherein the processes by the re-prediction unit, the search query generation unit, and the search unit are repeated multiple times.

5. 3. The information processing apparatus according to claim 2, wherein the search unit searches again for patent documents based on the regenerated search query.

6. 6. The information processing device according to claim 5, wherein the prediction unit, the search query generation unit, and the search unit execute processes multiple times.

7. the information processing device includes a display unit, the display unit displays the search formula generated or regenerated by the search formula generation unit on a screen; 7. The information processing device according to claim 1, wherein the user can edit the search formula displayed on the screen.

8. An information processing method performed by a computer, comprising: a receiving step of receiving an outline of the invention; a prediction step of inputting the outline of the invention received in the receiving step into a trained generative model to predict a patent classification code or a keyword; a search query generation step of generating a search query based on the patent classification code or keyword predicted in the prediction step; a search step of searching for patent documents based on the search formula generated in the search formula generation step; a re-prediction step of inputting the outline of the invention and the contents of the patent documents obtained as search results in the search step into a trained generative model and re-predicting patent classification codes or keywords; Including, the computer executes the search formula generating step again after the re-prediction step, An information processing method characterized in that in the search formula generation step, a search formula is regenerated based on the patent classification code or keyword re-predicted in the re-prediction step.

9. An information processing method performed by a computer, comprising: a receiving step of receiving an outline of the invention; a prediction step of inputting the outline of the invention received in the receiving step into a trained generative model to predict a patent classification code or a keyword; a search query generation step of generating a search query based on the patent classification code or keyword predicted in the prediction step; a search step of searching for patent documents based on the search formula generated in the search formula generation step; Including, the computer executes the prediction step and the search formula generation step again after the search step; In the prediction step, the outline of the invention and the contents of the patent documents obtained as search results in the search step are input to a trained generative model to predict patent classification codes or keywords again; An information processing method characterized in that in the search formula generation step, a search formula is regenerated based on the patent classification code or keyword predicted again in the prediction step.

10. On the computer, a reception process for receiving an outline of the invention; a prediction process in which the outline of the invention received in the reception process is input into a trained generative model to predict a patent classification code or a keyword; a search query generation process for generating a search query based on the patent classification code or keyword predicted in the prediction process; a search process for searching patent documents based on the search formula generated in the search formula generation process; a re-prediction process in which the outline of the invention and the contents of the patent documents obtained as search results in the search process are input into a trained generative model to re-predict patent classification codes or keywords; and, after the re-prediction process, executes the search query generation process again, An information processing program characterized in that, in the search formula generation process, a search formula is regenerated based on the patent classification code or keyword re-predicted in the re-prediction process.

11. On the computer, a reception process for receiving an outline of the invention; a prediction process in which the outline of the invention received in the reception process is input into a trained generative model to predict a patent classification code or a keyword; a search query generation process for generating a search query based on the patent classification code or keyword predicted in the prediction process; a search process for searching patent documents based on the search formula generated in the search formula generation process; and, after the search process, executes the prediction process and the search formula generation process again, In the prediction process, the outline of the invention and the contents of the patent documents obtained as search results in the search process are input into a trained generative model to predict patent classification codes or keywords again; The information processing program is characterized in that, in the search formula generation process, a search formula is regenerated based on the patent classification code or keyword predicted again in the prediction process.

Images