TWI886755B - A system of intelligent patent type recommendation and draft generation and the method thereof - Google Patents

Abstract

The present invention provides a system and method for intelligent patent type recommendation and draft generation, which is implemented by providing a user-operated electronic device, wherein the electronic device includes a processor and a network interface controller, a server includes an application, and the processor connects to the server through the network interface controller to execute the application for the purpose of recommending intellectual property application types, patent types, and draft generation.

Description

An intelligent patent type recommendation and manuscript generation system and its execution method

The present invention relates to a system for intelligent recommendation and draft generation and its implementation method, in particular, a system and method for generating patent types for recommended applications, as well as schematic diagrams, block diagrams, flow charts and other diagrams generated in the draft.

In the traditional patent application process, whether domestically or internationally, paper documents need to be printed out and a large number of forms need to be filled out. Many documents are paper-based, which causes great trouble in management and classification. Not only is it not environmentally friendly and wastes a lot of paper, but if management errors are caused, the patent application process will go wrong or the patent will become invalid, which is more than worth it. In addition to traditional paperwork, if communication between people is required during the patent application process, due to their different professional backgrounds, different language usage, cultural differences and other unpredictable factors, inaccurate or misunderstood information may be conveyed, which leads to cognitive differences between applicants, firms and agents, and government agencies. Therefore, applicants may not get the results they originally wanted.

And the purpose of applying for a patent is by no means just to be used as a defensive weapon in patent infringement lawsuits or as a symbol of the company's image. In fact, patent rights can create value. In addition to authorizing others to implement the patent to obtain licensing income, the right to exclude infringement and damages can also obtain settlement income.

For enterprises, using patent applications to protect research and development results has long become a necessary and important part of the business process. Some companies believe that patent applications are the work of patent companies. As long as they invite contractors to discuss the technical content of the patent application, professional patent companies can write a patent document with correct technical description, complete content disclosure and a wide range of protection.

In fact, patent application is not only a collaborative work between enterprises and patent companies, but also requires communication between different departments of the enterprise, especially repeated communication between the R&D department and the intellectual property department. In the process of patent proposal within the enterprise, R&D engineers in the R&D department need to provide relevant technical content disclosure, which may include basic background description of the technology, deficiencies of existing technologies or problems that need to be improved, and features of new technologies. In addition, it is necessary to conduct preliminary searches of the content of the proposed technology in patent databases in multiple countries to search for similar previous technologies in order to promote patent proposal discussions within the enterprise. In many enterprises, the number of people in the R&D department is far greater than that of the internal intellectual property department, which increases the workload of the intellectual property department.

However, although most medium and large enterprises currently have relevant patent proposal systems, in actual implementation, due to the professional differences between the R&D department and the intellectual property department, huge communication time costs will be incurred, ranging from several days to several months. The R&D department is very familiar with the technology, but the content required for the patent proposal (patent disclosure) often cannot meet the requirements of the intellectual property department. On the contrary, when the intellectual property department discusses with the R&D department based on the previous technical search report, it is usually unable to describe the differences in detail so that the R&D department can see at a glance. Therefore, the progress of the discussion of patent proposals (patent disclosures) within the enterprise consumes a lot of labor and time costs. If the enterprise produces pictures and texts or even patent disclosures with claims, it usually takes at least several days to several weeks. Because most companies outsource to third-party patent and trademark firms or law firms to assist in writing patent specifications and patent submission services, if there is no patent disclosure, for R&D personnel, the second technical communication will only generate more communication and understanding costs, resulting in a delay in patent application time and affecting the rights and interests of corporate technology protection.

In addition, for small and medium-sized enterprises and start-ups that do not have intellectual property departments, patent searches are completely entrusted to third-party patent and trademark firms or law firms for assistance. However, since there are no patent proposals (patent prospectuses), most of the searches are conducted through presentations or interviews. This model often requires inventors to spend a lot of effort and time on technical communication. Compared with medium and large enterprises, the cost of communication and understanding between the two parties is even more staggering.

However, there are already many retrieval technologies and retrieval platforms that help reduce users' work time.

As disclosed in Chinese application number CN201610297330.0, the patent writing assistance system includes: a briefing template generation module, which is used to generate a template for writing a technical briefing book containing multiple columns; a briefing input module, which is used to input the content of writing a technical briefing book in the multiple columns of the template accordingly to generate a technical briefing book; a content identification and extraction module, which is used to identify and extract the technical briefing book (patent disclosure book) ) in each column; an associated retrieval module connected to an external database server, used to select the content extracted by the content identification extraction module and associate it with the external database server to retrieve data related to the extracted content from the external database; a copy storage module, used to copy and store the data from the external database; a file generation module, used to generate a file in a preset format from the copied data.

However, the above-mentioned Chinese application number CN201610297330.0 has several issues to be improved. Its main feature is that the writer inputs technical related content in the corresponding field, modularizes the input text, extracts keywords from the modularized text according to different fields, searches for each keyword separately, and finally generates search data for the writer's reference, reducing the time of manual keyword input and search. The benefits of reducing the time of actual manuscript writing, analysis and comparison, or communication time between different departments of the enterprise are quite limited, and it is limited to the text part and cannot assist in generating diagrams.

As disclosed in Taiwan application number TW097119308, a system for generating a patent specification mainly has a built-in method for generating a patent specification, which can be a software or an application. The system of the present invention can generate a written paper copy of a patent specification. The system includes a central processing unit, which is a processing unit for executing a patent specification generating method; a data storage unit, which is connected to the central processing unit, and the data storage unit stores the patent specification generating method and provides it to the central processing unit for executing the generation of the patent specification; an input unit, which is connected to the central processing unit, and is used to provide an interface for users to input relevant technical disclosure data; a control unit, which is connected to the central processing unit. A central processing unit is connected to the system to control the system to execute the processing of the data related to the patent specification; an output unit is connected to the central processing unit to serve as an interface for outputting the data related to the patent specification. The output unit can be connected to a printing unit to print the patent specification; finally, a display unit is connected to the central processing unit to serve as an output display interface for displaying the data of the patent specification.

However, the above Taiwan application number TW097119308 has several issues to be improved. It is necessary to enter the corresponding text in the corresponding fields. The main purpose is to generate the text of the specification by using some common terms in the patent specification, similar to the method of filling in missing words. However, the input text must meet the default specifications, otherwise the generated sentences will be grammatically incoherent, and it is not easy for ordinary users to get started. It is equivalent to users generating documents by themselves, which still takes a lot of time. It does not solve the problem of communication documents of patent disclosure books before application, and it is limited to the text part and cannot assist in generating diagrams.

As disclosed in U.S. Application No. US17745671, in a specific embodiment, an assistive system can assist a user in obtaining information or services. The assistive system can enable a user to interact with the assistive system through user input in various modes (e.g., audio, voice, text, image, video, gesture, motion, position, direction) in a stateful and multi-turn dialogue to receive help from the assistive system. As an example and not a limitation, the assistive system can support single-mode input (e.g., voice input only), multi-mode input (e.g., voice input and text input), mixed/multi-mode input, or any combination thereof. The user input provided by the user may be associated with a specific assistant-related task and may include, for example, a user request (e.g., a verbal request for information or the performance of an action), a user interaction with an assistant application associated with the assistant system (e.g., selecting a UI element by touch or gesture), or any other type of suitable user input that the assistant system can detect and understand (e.g., user movement detected by the user's client device). The assistant system can create and store a user profile that includes personal information and contextual information associated with the user. In a particular embodiment, the assistant system can use natural language understanding (NLU) to analyze the user input. The analysis can be based on the user's user profile to obtain a more personalized and context-aware understanding. The assistant system can resolve entities associated with the user input based on the analysis. In certain embodiments, the assistant system may interact with different agents to obtain information or services associated with the resolved entity. The assistant system may generate responses for the user regarding the information or services by using natural language generation (NLG).

However, the above-mentioned US application number US17745671 has several issues that need to be improved. The previous case uses natural language understanding and natural language generation to analyze the text input by the user and generate corresponding reaction actions or replies. This is achieved through a "voice assistant", but it is only applied in daily life. The language used or trained background data is very different from the legal terms of intellectual property rights (patents and trademarks) in terms of words and grammar. It is a horizontal training model. If it is applied to the vertical industry of intellectual property rights, the accuracy will be greatly reduced.

As revealed by Dr. Jiesheng Li of the Department of Computer Science and Engineering at National Taiwan University (Application of Deep Learning in Patent Field 10.6342/NTU202100999), the scale of language model training is not large enough and the amount of data is not large enough, which may lead to inaccurate training results. In addition, there is no vertical in-depth patent language training. Moreover, it generates a summary through input text and then generates technical content from the summary. It is very dependent on the accuracy of the input text and is prone to biased technical content text. The comparison method used in the paper is text mapping. Although text mapping has many advantages, such as being able to capture the semantic information and contextual relevance of the text, it also has some disadvantages, such as: (1) Dimensionality curse: When the text data is very large, the text Mapping may result in high-dimensional vector representations, which pose challenges for computation and storage. (2) Difficulty in semantic identification: Although text mapping can capture some semantic information, it may not be able to fully reflect the semantics of words with polysemy or context dependence. At the same time, there may also be difficulties in semantic similarity between different texts. (3) Training data requirements: Text mapping methods usually require a large amount of annotated training data to learn vector representations of words and texts. Obtaining large-scale and high-quality annotated data can be a time-consuming and expensive task. (4) Vocabulary maintenance: Text mapping methods map words into vector space, but in practical applications they may encounter problems with new vocabulary, spelling errors, or other word variants. This may require vocabulary maintenance and updating to ensure the correctness of the model and this paper is trained with Bert and 2.0, which has certain limitations. It still cannot solve the communication problem between inventors and patent practitioners.

In addition, the natural language processing (NLP) model Chat GPT 3.5 and 4.0 currently developed by OpenAI is a self-regressive language model trained based on human feedback reinforcement learning (RLHF). It is mainly used to process customer service dialogues, story creation, translation, grammar modification, poetry, lyrics, text editing, and even software program writing. If the inventor generates patent disclosure documents through a dialogue with GPT, it cannot directly produce accurate patent industry content, nor can it produce related diagrams such as patent processes. For the inventor, it can only handle consultations on patent law. The chatbot that does not conform to the rules, and the patent-related search through GPT will only generate the correct patent certificate number, but not the correct case name and patent diagram. Please refer to Figures 1 and 2. For patent practitioners, repeated proofreading is required, which wastes more time. In view of this, a language model that deeply learns the vertical field of intellectual property rights and targets industry accuracy is needed to assist in processing patent disclosure books and patent search systems to solve this problem.

Some existing patent writing tools, such as Patent theory, PowerPatent, Rowan, etc., can be used to assist in the writing of patent manuscripts. Professional practitioners can provide necessary assistance or editing functions during the writing process, so that the error rate of the completed patent specification manuscript can be reduced, such as maintaining the consistency of terms, and even providing writing suggestions or generating content as a reference, which also increases the efficiency of writing patent specifications.

However, the above existing tools still have several problems to be improved. The main reason is that these tools are developed for professionals in the intellectual property industry. That is, the main target users of these tools are engineers who already have certain experience in writing patent specifications. These tools only serve as auxiliary roles. Therefore, for relatively inexperienced corporate R&D personnel, using these tools cannot actually help them solve problems. The operation of these tools is not intuitive enough for them, and these tools also cannot generate diagrams. For non-professionals such as individuals or small and medium-sized enterprise personnel, these tools cannot generate technical disclosures and diagrams.

From the above explanation, it can be seen that it is necessary to improve or adjust the known technology to lower the threshold for ordinary people to apply for intellectual property rights, and enhance the selection of appropriate application types, production of patent drafts and drawings by R&D personnel. In order to avoid missing the opportunity for urgent patent applications, the inventor of this invention has made great efforts to research and create, and finally developed and completed the system and method of this invention.

The purpose of this invention is to propose an intelligent recommendation and manuscript generation system and its execution method to solve the problems existing in the above-mentioned existing technologies.

Therefore, in order to achieve the above-mentioned purpose of the present invention, the present invention provides an execution method of an intelligent patent type recommendation and draft generation system, in which a user operates an electronic device, and a processor of the electronic device connects to a server through a network interface controller and executes an application to perform patent type recommendation and draft generation, which at least includes the following steps: (S100) The user uses the user interface of the electronic device to make the processor execute an input module to receive input text content; (S200) The processor executes a semantic analysis module in the application to analyze the text content. Semantic analysis; (S300) the semantic analysis module is further connected to an industry classification module, a search module and a database module, and the analyzed text content is classified into industry technologies, and a comparison search is performed in the database module, and the compared data is transmitted to an intellectual property information disclosure module; (S400) the intellectual property information disclosure module analyzes and statistics the data and presents the information to the user through the user interface of the electronic device; (S500) the analyzed and statistical data is also transmitted to a recommendation module, and the recommendation module The types of intellectual property rights in the data are classified and counted, and the types of intellectual property rights recommended for application are displayed in a user interface in a sorted manner; (S600) the user selects a patent type through the user interface; (S700) the user causes the processor to execute the input module to receive the input technical description through the user interface; and (S800) the login is completed through a login module, causing the processor to execute the semantic analysis module and a content generation module to generate summary content and a simple diagram; wherein, after step (S800), the following further steps are included: (S8001) the processor The processor executes the recommendation module to analyze the generated summary content and simple diagrams and provide recommended patent types; (S8002) The intellectual property information disclosure module further evaluates each patent type and generates a patent risk assessment report, which is simultaneously presented in the information of the recommended patent types through the user interface; (S8003) The user selects the patent type to apply through the user interface, and the processor executes the semantic analysis module and the content generation module to generate patent disclosure documents, claim content and patent diagrams, and integrates them into a formatted template into a draft.

Another purpose of the present invention is to propose an intelligent recommendation and manuscript generation system to solve the problems existing in the above-mentioned existing technologies.

Therefore, in order to achieve another purpose of the present invention, the system provided by the present invention is implemented by providing an electronic device for user operation, wherein the electronic device includes a processor and a network interface controller, a server includes an application, the processor connects to the server through the network interface controller and executes the application to perform intellectual property diagnosis, and the system at least includes a login module, an order processing module, an input module, a semantic analysis module, an industry classification module, a search module, a content learning module, a content generation module, a database module, a natural language database, a recommendation module, and an intellectual property information disclosure module.

The content generation module also includes a disclosure document generation unit and a diagram generation unit.

The input module is used to receive the text content and technical description input by the user, convert the text content and technical description into strings for labeling, send a string of information and record the input language of the string information in a temporary memory.

The semantic analysis module receives the string information, analyzes and segment the words through a natural language database, and generates and sends the semantic analysis results.

The industry classification module performs industry classification code analysis on the semantic analysis results, connects to a database module to judge and generate at least one set of patent classification codes.

The search module performs a comparison search in a database module based on the at least one set of patent classification codes and generates data of the comparison search results.

The intellectual property information disclosure module receives the data and further performs statistical analysis to generate basic intellectual property information.

The recommendation module also receives the data and classifies and compiles the data according to the types of intellectual property rights in the data to generate the types of intellectual property rights recommended for application.

Login module, the user operates the electronic device to perform identity verification through the login module.

The content generation module further includes a disclosure document generation unit and a diagram generation unit, which receives and generates summary content and a simple diagram based on the semantic analysis result.

Specifically, when the user selects a patent in the recommended intellectual property type, the input module receives the technical description input by the user again and completes the identity verification through the login module. The semantic analysis module receives the string information about the technical description, performs analysis and generates a suffix and sends the analysis result of the technical description. The disclosure document generation unit generates a summary content based on the analysis result, and the diagram generation unit generates a summary content based on the summary content. Generate a simple diagram. The recommendation module analyzes the abstract content and the simple diagram to generate a recommended patent type. The intellectual property information disclosure module evaluates the recommended patent type and generates a patent risk assessment report. The disclosure document generation unit, the diagram generation unit and a claim content generation unit generate patent disclosure documents, claim content and patent diagrams based on the patent type, abstract content and simple diagram.

The following is a detailed description using only specific implementation examples and accompanying drawings.

100: Electronic devices

101:Processor

102: Network interface controller

200: Server

201: Applications

300: Login module

301: Order processing module

302: Input module

303:Semantic analysis module

304: Industry classification module

305: Content learning module

306: Content generation module

307: Disclosure document generation unit

308: Rights claim content generation unit

309: Schematic generation unit

310: Retrieval file generation module

312:Case processing module

335:Keyword extension module

320: Language judgment module

321:Translation module

322: Instruction manual generation unit

323: Application form generation unit

324: International conversion module

331:Keyword extraction module

400: Temporary Memory

500: Natural Language Database

600: Database module

702: Recommended modules

703: Intellectual Property Information Disclosure Module

(1), (2), (3), (4): Steps

(411), (412), (413), (414), (415): Steps

(412), (422), (423), (424), (425), (426): Steps

(4151), (4152): Steps

(4261), (4262), (4263): Steps

(31), (32), (33): Steps

(51), (52), (53): Steps

(6), (7): Steps

(71), (72), (73): Steps

S100, S200, S300, S400, S500, S600, S700, S800: Steps

S8001, S8002, S8003: Steps

FIG. 1 is a schematic diagram of the prior art; FIG. 2 is another schematic diagram of the prior art; FIG. 3 shows a flow chart of the system of the present invention; FIG. 4 shows a schematic diagram of the system of the present invention; FIG. 5 shows another schematic diagram of the system of the present invention; FIG. 6 shows another schematic diagram of the system of the present invention; FIG. 7 to FIG. 9 show schematic diagrams of application examples of the present invention; FIG. 10 shows another schematic diagram of the system of the present invention; FIG. 11 shows another schematic diagram of an implementation example of the system of the present invention; FIG. 12 to FIG. 18 show flow charts of the method of the present invention; and FIG. 19 to FIG. 22 show schematic diagrams of application examples of the present invention.

The inventive concept will now be more fully described below with reference to the accompanying drawings in which exemplary embodiments of the inventive concept are shown. The advantages and features of the inventive concept and the method of achieving the same will become apparent from the exemplary embodiments described in more detail below with reference to the accompanying drawings.

The terms used herein are used only to describe specific embodiments and are not intended to limit the present invention. Unless the context clearly indicates otherwise, the singular forms of the terms "a", "an" and "the" used herein are intended to include the plural forms as well. It should be understood that when an element is said to be "connected" or "coupled" to another element, the element may be directly connected or coupled to the other element or there may be intermediate elements.

Exemplary embodiments are described herein with reference to figures, which are idealized exemplary illustrations. Therefore, deviations from the shapes of the figures due to, for example, manufacturing techniques and/or tolerances are expected. Therefore, the regions shown in the figures are schematic, and their shapes are not intended to illustrate the actual shape of the device nor to limit the scope of the exemplary embodiments.

The present invention provides an execution method of an intelligent patent type recommendation and draft generation system, please refer to FIG3. The method of the present invention is performed by a user operating an electronic device, and a processor of the electronic device is connected to a server through a network interface controller and executes an application to perform patent type recommendation and draft generation. The method includes at least the following steps: (S100) the user uses the user interface of the electronic device to make the processor execute an input module to receive input text content; (S200) the processor executes a semantic analysis module in the application to perform semantic analysis on the text content. ; (S300) The semantic analysis module is further connected to an industry classification module, a search module and a database module, and the analyzed text content is classified into industry technologies, and a comparison search is performed in the database module, and the matched data is transmitted to an intellectual property information disclosure module; (S400) The intellectual property information disclosure module analyzes and statistics the data and presents the information to the user through the user interface of the electronic device; (S500) The analyzed and statistical data is also transmitted to a recommendation module, and the recommendation module recommends the user according to the data. Classify and count the types of intellectual property rights in the data, and sort and display the recommended types of intellectual property rights on the user interface; (S600) the user selects the patent type through the user interface; (S700) the user causes the processor to execute the input module to receive the input technical description through the user interface; and (S800) complete the login through a login module, causing the processor to execute the semantic analysis module and a content generation module to generate summary content and a simple diagram; Wherein, after step (S800), it further includes: (S8001) the processing The processor executes the recommendation module to analyze the generated summary content and simple diagram, and provides the recommended patent type; (S8002) the intellectual property information disclosure module further evaluates each patent type and generates a patent risk assessment report, which is simultaneously presented in the information of the recommended patent type through the user interface; (S8003) the user selects the patent type to apply through the user interface, and the processor executes the semantic analysis module and the content generation module to generate patent disclosure documents, claim content and patent diagrams, and inserts the formatted template to integrate them into the draft.

Please refer to Figures 4 and 5, which show the system schematic diagram of the present invention.

The system of the present invention is implemented by providing an electronic device 100 for user operation, wherein the electronic device 100 includes a processor 101 and a network interface controller 102, a server 200 includes an application 201, the processor 101 is connected to the server 200 through the network interface controller 102 and executes the application 201 to perform intellectual property rights diagnosis. The system at least includes a login module 300, an order processing module 301, an input module 302, a semantic analysis module 303, an industry classification module 304, a search module 700, a content learning module 305, a content generation module 306, a database module 600, a natural language database 500, a recommendation module 702, and an intellectual property information disclosure module 703.

The content generation module 306 also includes a disclosure document generation unit 307 and a diagram generation unit 309.

The input module is used to receive the text content and technical description input by the user, convert the text content and technical description into strings for labeling, send a string of information and record the input language of the string information in a temporary memory.

The semantic analysis module receives the string information, analyzes and segment the words through a natural language database, and generates and sends the semantic analysis results.

The industry classification module performs industry classification code analysis on the semantic analysis results, connects to a database module to judge and generate at least one set of patent classification codes.

The search module performs a comparison search in a database module based on the at least one set of patent classification codes and generates data of the comparison search results.

The intellectual property information disclosure module receives the data and further performs statistical analysis to generate basic intellectual property information.

The recommendation module also receives the data and classifies and compiles the data according to the types of intellectual property rights in the data to generate the types of intellectual property rights recommended for application.

Login module, the user operates the electronic device to perform identity verification through the login module.

The content generation module further includes a disclosure document generation unit and a diagram generation unit, which receives and generates summary content and a simple diagram based on the semantic analysis result.

Specifically, when the user selects a patent in the recommended intellectual property type, the input module receives the technical description input by the user again and completes the identity verification through the login module. The semantic analysis module receives the string information about the technical description, performs analysis and generates a suffix and sends the analysis result of the technical description. The disclosure document generation unit generates a summary content based on the analysis result, and the diagram generation unit generates a summary content based on the summary content. Generate a simple diagram. The recommendation module analyzes the abstract content and the simple diagram to generate a recommended patent type. The intellectual property information disclosure module evaluates the recommended patent type and generates a patent risk assessment report. The disclosure document generation unit, the diagram generation unit and a claim content generation unit generate patent disclosure documents, claim content and patent diagrams based on the patent type, abstract content and simple diagram.

Specifically, the user connects to the server 200 through the network interface controller 102 through the electronic device 100 and executes the application 201, and can input text or a URL link. The input text can be a description of the company's corporate brand or a technical introduction to the core product services of the company's corporate brand, or a URL link of the official website. If the input module receives the URL link, it will capture and read the website content through the network interface controller 102. Specifically, the semantic analysis module 303 can then perform semantic analysis on the website content and the input text, and transmit the analysis results to the industry classification module 304.

The industry classification module 304 performs industry classification code analysis on the semantic analysis result, connects to a database module to judge and generate at least one set of patent classification codes.

The search module connects to the database module 600 for comparison and analysis based on at least one set of patent classification codes, and compares the number and category of intellectual property applications in the same or similar industries in the database module 600 to generate data for the search results.

The intellectual property information disclosure module receives the data and further performs statistical analysis to generate basic intellectual property information and disclose basic intellectual property information, such as technology value, technology risk, intellectual property time cost and expenditure cost, etc.

The recommendation module also receives the data and classifies and compiles statistics based on the types of intellectual property rights in the data to generate recommended intellectual property types, such as recommended patent applications.

The present invention provides a system and method for intelligent patent type recommendation and draft generation, wherein the draft includes generating system block diagrams, flow charts and/or schematic diagrams. In this system, the draft can be automatically generated through the ideas input by the inventor; in addition, when the inventor publishes or inputs ideas, the system of the present invention can perform artificial intelligence semantic reading through the natural language processing (NLP) algorithm, convert the input ideas into programming language, and generate system block diagrams, flow charts or schematic diagrams; at the same time, the system uses the natural language generation (NLG) algorithm to generate drafts including previous technologies, core technology descriptions, technical advantages, etc.

In one embodiment of the present invention, the description focuses on the case where the user has previously completed the aforementioned intelligent recommendation and directly generates the draft. However, in this embodiment, the user may first perform the intelligent recommendation and then continue to generate the draft. Please refer to Figures 4 to 6. Figure 6 is another schematic diagram showing the system of the present invention. The system of the present invention is implemented by providing an electronic device 100 for user operation, wherein the electronic device 100 includes a processor 101 and a network interface controller 102, a server 200 includes an application 201, the processor 101 is connected to the server 200 through the network interface controller 102 and executes the application 201 to generate patent drawings and drafts. The system includes at least a login module 3 00, an order processing module 301, an input module 302, a semantic analysis module 303, an industry classification module 304, a content learning module 305, a graph learning unit 3051, a content generation module 306, a graph generation unit 309, a search document generation module 310, a graph search unit 311, a temporary memory 400, a database module 600 and a natural language database 500.

The server 200 may be a cloud server or a local server.

In this embodiment, the creator uses a server with the following specifications to perform the trademark product item conversion model training and inference execution of this technology; the processor 101 (CPU) uses a high-performance multi-core processor 101, especially for processing large amounts of data and performing complex calculations, at least one CPU with 16 cores or more (such as AMD Ryzen Threadripper or Intel Xeon series) is used. And the size of the memory (RAM) can be but not limited to 64GB or higher RAM specifications so that the size of the corpus and the size of the word vector model can be processed. The network interface controller 102 is a high-speed and stable network connection hardware, especially used in cloud computing resources or downloading/uploading large amounts of data. Among them, the most important graphics processing unit (GPU) uses a high-performance GPU (such as NVIDIA's RTX 30 series or Tesla series) to reduce the model training time; in the training architecture of the server 200 in this case, a higher-specification training server as above is used, and when the model training is completed and the inference processing is performed, a lower-specification server host can be used, and it can be a cloud or ground server host deployed in this system. In the architecture of this case, the adoption of the server host specification does not affect the technical features emphasized in this case, so any server host specification should still fall within the technical scope of this case.

Login module 300, the user operates the electronic device 100 to perform identity verification through the login module 300, confirming the identity of the user (loginer) and the identity data at the same time.

Order processing module 301 provides users with functions such as member registration/login, multinational patent case order creation and management, and multinational patent case import. It can also provide case order query, patent and trademark review status confirmation, and account query functions.

Specifically, users can register and log in to become members of this system, apply for patents, and manage case orders on the system platform, such as: document generation, case acceptance, case submission copies, official notifications, etc.

The user generates/creates a new case order in the order processing module 301, and after the user selects the first target country, the information in the case order is regularly updated to the system's temporary memory 400.

The input module 302 is used to receive the description text or graphics input by the user, convert the description text into a string for labeling, send the string information and record the input language of the string information in the temporary memory 400.

The specific implementation method can be described as follows: text preprocessing, label extraction, stringification and labeling. Text preprocessing is further divided into word segmentation, stop word removal, part-of-speech tagging and word form restoration. Label extraction first uses predefined rules or patterns to extract labels, and uses machine learning models to automatically extract labels. Stringification is to combine the extracted labels into a string, such as "電影#動作#劇". Labeling can further process the labels as needed, such as removing duplicate labels, sorting labels, etc. More specific methods can use natural language processing (NLP) toolkits: such as NLTK, Stanford NLP, HanLP, etc., or machine learning frameworks: such as TensorFlow, PyTorch, scikit-learn, etc.

The semantic analysis module 303 receives string information, analyzes and segment the information through a natural language database 500, and generates and sends semantic analysis results.

The industry classification module 304 performs industry classification code analysis on the semantic analysis results, and connects to the database module 600 to determine and generate at least one set of patent classification codes. The industry classification analysis can be International Patent Category, which classifies A: Human Necessities, B: Performing Operations, Transportation, C: Chemistry, Metallurgy, D: Textiles, Paper, E: Fixed Constructions, F: Mechanical Engineering, Lighting, Heating, Weapons, G: Physics, H: Electricity, or other classification codes such as USPC, JPC, European Classification, ECLA classification codes, etc.

To perform industrial classification analysis on a description text and generate a patent IPC classification number, the following methods can be used: Text preprocessing: First, preprocess the input description text. This may include removing punctuation, stop words and meaningless characters, and performing vocabulary normalization (such as stem extraction or word form restoration).

Feature extraction: Extract meaningful features from the pre-processed text. This can use common text feature representation methods such as Bag-of-Words or Word Embeddings. The Bag-of-Words model represents text as the frequency or presence of words, while Word Embeddings maps words into a continuous vector space.

Industry classification model: Build an industry classification model, which can be a machine learning-based classifier (such as support vector machines, decision trees, or deep learning models) or a rule-based approach. This model will be trained to classify text into corresponding industry categories.

IPC classification generation: Use the trained industry classification model to classify the input description text and generate the corresponding patent IPC classification number. IPC (International Patent Classification) is an international standard for patent document classification.

Suppose we have a description like this: "This company specializes in the research and development of new materials and technologies for solar power generation. They have designed a highly efficient solar cell that can convert solar energy into electricity while reducing energy costs and environmental impact. Their technology has great potential in the field of renewable energy and can be widely used for domestic and industrial purposes."

We can perform industry classification analysis and generate patent IPC classification numbers according to the following steps: Text preprocessing: preprocess the description text, remove punctuation, convert to lowercase letters, remove stop words, etc. For example, the above description text can be preprocessed as: "Company Specializes in research and development of solar power generation, new materials, technology, design, high-efficiency solar battery conversion, electricity reduction, energy cost, environmental impact, technology, renewable energy, potential, wide application, household, industrial use".

Feature extraction: Use the bag-of-words model to convert the preprocessed text into a feature vector representation. Each word can be regarded as a feature, and the frequency of the word in the description text can be used to indicate the importance of the word. For example, the following feature vector representation can be obtained: {"Company": 1,"Specialize": 1,"Research": 1,"Development": 1,"Solar": 2,"Power Generation": 1,"New": 1,"Material": 1,"Technology": 2,"Design": 1,"High Efficiency": 1,"Battery": 1,"Conversion": 1,"Electricity": 1,"Reduction": 1,"Energy": 1,"Cost": 1,"Environment": 1,"Impact": 1,"Renewable Energy": 1,"Potential": 1,"Wide": 1,"Application": 1,"Household": 1,"Industry": 1,"Use": 1}.

Industry classification model: Build a trained industry classification model, such as a support vector machine (SVM) or a deep learning model (such as a recurrent neural network or a convolutional neural network), and use training data with corresponding labels to train the model.

IPC classification generation: Use the trained industry classification model to input the pre-processed text feature vector into the model for prediction. Based on the prediction results, the corresponding patent IPC classification number can be generated. For example, the model may predict that the description text belongs to the solar-related industry category, and the corresponding IPC classification number may be H01L 31/02 (solar battery) or C09K 5/04 (solar material), etc.

The content learning module 305 is a large language model. For different patent classification codes, it learns the corresponding content in the patent specification from the database module 600. The schema learning unit 3051 further learns the schema in the patent specification. That is, the content learning module 305 can learn the writing methods of different types of patent specifications for different patent classification codes. The schema learning unit 3051 can also learn the writing methods of different types of patent specifications for different patent classification codes. 3051 can learn the correspondence between different types of patent drawings and specifications for different patent classification codes, such as system patents, information software patents, mechanism patents, method patents, etc. The specification content learned can at least include patent name, patent abstract, background technology, patent purpose, patent detailed description, patent claims, patent drawings and patent numbers, and even patent defense drafts or patent official documents.

The content generation module 306 also includes a disclosure document generation unit 307 and a claim content generation unit 308. Based on the semantic analysis result, the content learning module 305 and the database module 600 are combined to generate a patent disclosure document and claim content. The generated patent disclosure document at least includes a technical summary, a prior art content or a combination thereof.

Furthermore, the claim content generating unit 308 performs semantic analysis again based on the patent disclosure document through the semantic analysis module 303, and generates the claim content converted into patent words and terms after combining content learning.

The content generation module 306 also includes a component symbol generation unit 3091. The semantic analysis module analyzes the patent disclosure document, breaks down and decomposes nouns, and further completes the noun labels in the content of the patent disclosure document. Then, the patent diagram generated by the diagram generation unit is compared with the patent disclosure document with the completed label, and finally the patent diagram is labeled and the component symbol table is generated synchronously.

The main function of the component symbol generation unit 3091 is to analyze the semantic content in the patent disclosure document, identify and mark the component symbols on the patent diagram, and thus generate a complete component symbol table. This process includes several key steps: Semantic analysis: The semantic analysis module first performs an in-depth semantic analysis of the patent disclosure document. This includes using natural language processing (NLP) technology to segment and identify key terms. For example, if the document mentions terms such as "resistor" and "capacitor", the system will identify these terms as potential component symbols.

Noun Labeling: Next, the component symbol generation unit 3091 labels these identified nouns in the patent disclosure document. This step involves associating each component with a unique identifier. For example, a resistor may be labeled R1, a capacitor may be labeled C1, etc.

Drawing generation and comparison: At this time, the patent drawings generated by the drawing generation unit also include all the mentioned components and their arrangements. Then, the component symbol generation unit 3091 will compare these patent drawings with the patent disclosure documents that have completed the labeling.

Label mapping: After a detailed comparison, the component symbol generation unit 3091 will mark the corresponding labels on the patent drawings. This process is done by mapping the component names and labels in the patent disclosure document to their corresponding positions in the patent drawings. For example, if "resistor R1" in the document has a corresponding graphical representation on the drawing, the system will mark "R1" next to the graphic.

Simultaneous generation of component symbol table: The last step is to generate the component symbol table, which is a list that summarizes all component names, reference numbers, and their locations in the drawings in the patent disclosure document. This table makes it easy for readers to quickly locate the location of each component in the drawing and understand their functions and relationships.

To achieve these functions, the component symbol generation unit 3091 uses a series of complex algorithms and data structures. These include: Natural Language Processing (NLP) algorithms: used for text analysis, word segmentation, and noun recognition.

Image processing technology: used to identify and map component symbols from patent drawings.

Database management system: used to store and retrieve component names, designators and their locations in drawings in documents.

In addition, the component symbol generation unit 3091 may also use machine learning technology to improve its accuracy in identifying and mapping components. For example, by training the model to recognize different types of electronic component icons, the system can automatically number and locate these components.

D：t

d}｜)，N是語料庫中文件的總數，分母是包含詞語t的文件數量。 The formulas and algorithms to implement this process may include but are not limited to: Text analysis formulas: For example, TF-IDF (Term Frequency-Inverse Document Frequency) is used to evaluate the importance of a term to a document set or a document in a corpus. TF-IDF(t,d)=TF(t,d)×IDF(t,D), where TF(t,d) is the frequency of occurrence of term t in document d, and IDF(t,D) is the inverse document frequency, calculated as: IDF(t,D)=log(N/|{d

D:t

d}｜), N is the total number of documents in the corpus, and the denominator is the number of documents containing word t.

Image processing algorithms: For example, edge detection techniques are used to identify component boundaries in a diagram. A commonly used method is the Canny edge detector.

Machine learning models: For example, using convolutional neural networks (CNNs) to recognize and classify different component symbols in diagrams.

Database module 600 provides users with multiple databases required for searching for multinational trademark/patent prior cases in this system, such as official patent databases of various countries or public patent search databases, etc.

The diagram generation unit 309 analyzes the patent disclosure document or the content of the claim and translates it into a diagram generation language, generates a diagram code corresponding to the technical content or the content of the claim through the diagram generation language, and then generates a patent diagram corresponding to the technical content or the content of the claim through the compilation unit. The patent disclosure document, the content of the claim and the patent diagram are combined to form a draft.

Specifically, to translate the patent disclosure document or claim content into a schematic language, first, the translation module needs to analyze the extracted technical information and understand its meaning. Then, the translation module needs to generate a description of the schematic generation language according to the grammatical rules of the schematic generation language. In detail, the following steps can be used to generate the schematic generation language: identify technical concepts, identify technical features, and identify technical relationships.

Identify technical concepts: The translation module first identifies the technical concepts in the technical information. Technical concepts are the basic elements in the schema generation language. Identify technical features: The translation module then identifies the technical features in the technical information. Technical features are used to describe technical concepts. Identify technical relationships: The translation module finally identifies the technical relationships in the technical information. Technical relationships are used to describe the connection between technical concepts.

The generated patent diagram can be a flow chart, block diagram or schematic diagram. A flow chart represents a method step, a block diagram represents the structure of system components, and a schematic diagram represents the structure of a mechanism or machine.

For example, if the technical information mentions "having a touch screen", the translation module can identify it as the "having" technical feature and the "touch screen" technical concept. If the technical information mentions "containing a camera lens and a processor", the translation module can identify it as the "containing" technical relationship, the "camera lens" technical concept, and the "processor" technical concept.

Please refer to Figures 7 to 9. The top of Figure 7 is the technical summary in the patent disclosure document, and the bottom of Figure 7 is the corresponding schematic diagram generated; the top of Figure 8 is the technical summary in the patent disclosure document, and the bottom of Figure 8 is the corresponding system block diagram generated; the top of Figure 9 is the technical summary in the patent disclosure document, and the bottom of Figure 9 is the corresponding flow chart generated.

Specifically, the semantic analysis module 303 can perform keyword extraction, which is a natural language processing technology that aims to automatically extract important keywords or phrases from text. The method can be but not limited to statistical methods, frequency methods, statistical methods, text vectorization methods or machine learning methods.

The frequency method determines the importance of a word based on its frequency in the text. Common methods include TF-IDF (Term Frequency-Inverse Document Frequency) and Term Frequency. TF-IDF considers the frequency of a word in the text and its importance in the entire collection, while term frequency only considers the frequency of a word in the text.

Text statistics is based on statistical models to analyze the distribution and relevance of words in text. Common methods include mutual information, pointwise mutual information, and chi-squared test. These methods usually require the establishment of a statistical model between words and texts, and the calculation of word importance based on the model.

The text vectorization method converts the text into a vector representation and then uses the Vector Space Model to calculate the importance of words. Common methods include the Bag-of-Words Model, Word Embeddings, and text vectorization methods (such as TF-IDF vectorization).

Machine learning methods use machine learning algorithms to train models and learn the importance of words from text. Common methods include text classification, text clustering, and keyword extraction models. These methods require the use of labeled text data for model training.

The above can also be done by writing a program using a scripting language (such as Python).

Keyword extension is a further function of keyword extraction, which receives the generated keywords and extends them, that is, it searches for similar words or synonyms, connects to a vocabulary for comparison, and generates at least one synonym.

Next, please refer to Figure 6. The system also includes a language determination module 320 for determining whether the input language of the string information input by the user is the same as the official language of the first target country. If the language determination module 320 determines that the string information is not the same as the official language of the first target country, the string information is translated through a translation module 321, and after the patent diagram is finally generated, the translation module 321 translates the language of the patent diagram back to the input language of the string information.

The system of the present invention may further include a search document generation module 310 including a graphic search unit 311, a conversion unit 317 and a graphic comparison unit 318. The graphic search unit 311 receives the graphic input by the user and searches in the database module 600 to generate a search document.

After generating the patent disclosure document, claim content and patent diagram, the user can further search for previous cases through the search document generation module 310, especially searching for diagrams.

Specifically, after the image retrieval unit 311 receives the input image, the image is first converted by the conversion unit 317 and brought into the Vienna classification label, and the image is converted into a vector at the same time, and then the image comparison unit 318 compares it according to the database module 600 to generate a search document.

Convert graphics to vectors. For example, since graphics are basically composed of pixels, the brightest can be regarded as 1 and the darkest as 0. Assuming that the three primary colors are not considered, there are only 3 x 3 pixels, we arrange them from top left, middle left, top left, middle left, middle middle, middle right... to bottom right (1,0,1,0,1,0,1,0,1), this figure is black and white, black and white, black and white, black and white, considering the case of three primary colors, we expand the length of this vector three times, and it becomes R(1,0,1,0,1,0,1,0,1)+G(1,0,1,0,1,0,1,0,1)+B(1,0,1,0,1,0,1,0,1)=(1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1). And through the principle of local vector similarity = shape similarity, we can capture the approximate figure after comparison.

Specifically, in the comparison process, the similarity can be calculated in different ways, such as Edit Distance, Cosine Similarity or Jaccard Similarity, and the graphics with similarity exceeding a set value are filtered out. The set value can be adjusted by yourself.

Edit distance is usually used to measure the similarity between strings. It measures the minimum number of operations (insertion, deletion or replacement) required to transform one string into another. But it can also be used to approximate the similarity of vectors.

The following are the general steps to calculate the similarity of vectors using edit distance: Vector serialization: Convert each vector into a sequence, for example: a digital vector can be directly converted into a digital sequence. A feature vector can combine the name and value of each feature into a string sequence.

Calculate edit distance: Calculate the edit distance between two sequences using an edit distance algorithm (e.g. Levenshtein distance).

Normalization: Normalize the edit distance so that its value range is between 0 and 1 for easy comparison. A common normalization method is to divide by the maximum length of the sequence.

Approximation: (1-normalized edit distance), indicating the approximation between two vectors. Generally, the closer the value is to 1, the more similar the vectors are.

For example, suppose there are two vectors: vector1=[1,2,3,4], vector2=[2,3,1,4].

Serialization: vector 1->"1234", vector 2->"2314"

Edit distance: 2 (requires two operations: swapping "1" and "3")

Normalization: 2/4=0.5

Approximation: 1-0.5=0.5

Therefore, the approximation of these two vectors is 0.5, indicating that they have a certain degree of similarity.

Cosine similarity is a method to measure the directional similarity between vectors, and can be used to estimate the degree of similarity between vectors. The principle is to calculate the cosine value of the angle between two vectors. The closer the angle is to 0, the more similar the vectors are, and the closer the cosine value is to 1.

The following are the general steps to calculate the degree of vector similarity using cosine similarity: Vector preprocessing: Make sure the two vectors have the same dimension; Normalize the vectors (for example, make their length 1) to eliminate the impact of the difference in vector length.

Inner product calculation: Calculate the inner product between two vectors, that is, multiply the elements of corresponding dimensions pair by pair and then sum them up. The inner product reflects the correlation between the two vectors in direction.

Vector length calculation: Calculate the length of each vector, that is, square the elements of the corresponding dimension, sum them up, and then take the square root. The length of a vector reflects its overall size.

Cosine similarity calculation: Divide the inner product by the product of the lengths of the two vectors to get the cosine similarity of the two vectors.

Similarity analysis: The cosine similarity ranges from -1 to 1. When the cosine similarity is close to 1, it means that the directions of the two vectors are very similar and the similarity is very high. When the cosine similarity is 0, it means that the two vectors are orthogonal and completely dissimilar. When the cosine similarity is close to -1, it means that the directions of the two vectors are opposite and they are extremely dissimilar.

Suppose there are two vectors, vector 1 = [2,3,4], vector 2 = [4,6,8].

Vector preprocessing: These two vectors have the same dimensions and require no further processing.

Inner product calculation: 2*4+3*6+4*8=62.

Vector length calculation: sqrt(2^2+3^2+4^2)=sqrt(29), sqrt(4^2+6^2+8^2)=sqrt(100)

0.906 Cosine similarity calculation: 62/(sqrt(29) * sqrt(100))

0.906

Therefore, the cosine similarity of these two vectors is close to 0.9, indicating that the directions are very similar and the approximation is high.

Jaccard similarity is a method to measure the similarity between vectors, which can be used to estimate the approximation between vectors. The principle is to calculate the proportion of the same elements in two vectors. The higher the proportion, the more similar the vectors are, and the higher the Jaccard similarity is.

The following are the general steps to calculate the degree of vector similarity using Jaccard similarity: Vector preprocessing: Make sure the two vectors have the same dimension. Normalize the vectors (for example, make the sum of their elements equal to 1) to eliminate the effect of the difference in vector length.

Compute intersection: Count the number of identical elements in two vectors.

Compute the union: Calculate the number of all elements in two vectors.

Jaccard similarity calculation: Divide the intersection by the union to get the Jaccard similarity of the two vectors.

Approximation analysis: Jaccard similarity ranges from 0 to 1. If the Jaccard similarity is close to 1, it means that the two vectors are exactly the same and the approximation is very high. If the Jaccard similarity is 0, it means that the two vectors are completely different.

Suppose there are two vectors, vector 1 = [1,2,3,4], vector 2 = [1,3,4,5].

Vector preprocessing: These two vectors have the same dimensions and require no further processing.

Calculate the intersection: 1+3+4=8

Calculate the union: 1+2+3+4+5=15

0.533 Jaccard similarity calculation: 8/15

0.533

Therefore, the Jaccard similarity of these two vectors is close to 0.5, indicating a certain degree of similarity.

Furthermore, the system also includes a case processing module 312, which executes the case orders of the system, generates and converts them into patent application documents that meet the standards. The system further includes a cross-border conversion module 324 that can execute cross-border patent application document conversion, and even executes the function of connecting the system case orders with third-party electronic payment.

The case processing module 312 further includes a specification generating unit 322 and an application generating unit 323. The specification generating unit 322 receives the patent drawings and patent disclosure documents and generates implementation texts according to the content of the claims through the disclosure document generating unit 307, and then adjusts the content of the claims according to the implementation texts. Finally, the abstract content, prior art content, implementation texts, claim content and patent drawing content are further formatted using a template to generate a draft.

The application form generation unit 323 captures the user's login authentication identity information and brings it into the application information, or the user operates the electronic device 100 to directly input the application information in the field and the input module 302 receives it. The application information is then applied to the formatting template to generate the application form.

After generating the instruction document and application form, the user can operate the electronic device 100 and execute the case processing module 312 to submit or manage the case online. The submission can be submitted by the user to the official, or after selecting the submission function, the office providing the system will submit the application to the official, especially the different application documents and different file formats of different countries.

Please continue to refer to Figure 10, which is another schematic diagram showing the system of the present invention. The system also includes a transnational conversion module 324. After the user generates the specification document and the application form, the transnational conversion module 324 selects a second target country, thereby applying for patents in different countries to protect the idea.

After the user selects the second target country, the language determination module 320 will first determine whether the input language of the string information is the same as the official language of the second target country. If not, the translation module 321 will first translate the patent disclosure document and patent drawings into the official language of the second target country, and then translate the specification document and application into the official language of the second target country and perform format adjustment.

Alternatively, the language determination module 320 may directly determine whether the language of the specification document and the application is the same as the official language of the second target country. If they are the same, the specification document and the application are directly formatted to fit the official document format of the second target country. If they are not the same, the specification document and the application are first translated into the official language of the second target country through the translation module 321, and then the format is adjusted.

Please refer to FIG. 11 for a system in another embodiment of the present invention, which includes: a data receiving module 330, a semantic analysis module 303, a keyword extraction module 331, a search module 332, an IPC analysis module 333, a text generation module 334, a content learning module 305, a keyword extension module 335, a document classification module 336, an automatic writing module 337, a search report generation module 338, and a disclosure book generation module 339.

The data receiving module 330 may include a login module 300, an order processing module 301, and an input module 302. The login module 300 is used by the user to operate the electronic device 100 to perform identity verification, thereby confirming the identity of the user (loginer) and the identity data at the same time. The order processing module 301 provides the user with member registration/login, multinational patent and trademark Functions such as case order creation and management, import of multinational patent cases, etc. can also provide case order query, patent review status confirmation and account query functions; input module 302 is used to receive the description text or graphics input by the user, and convert the description text into a string for labeling, and send the string information and record the input language of the string information in the temporary memory 400.

The semantic analysis module 303 receives string information, analyzes and segmentation the information through a natural language database 500, and generates and sends semantic analysis results.

The keyword extraction module 331 performs word segmentation and keyword extraction based on the input content to generate multiple keywords.

Specifically, keyword extraction is a natural language processing technology that aims to automatically extract important keywords or phrases from text. The method can be but not limited to statistical methods, frequency methods, statistical methods, text vectorization methods or machine learning methods.

The frequency method determines the importance of a word based on its frequency in the text. Common methods include TF-IDF (Term Frequency-Inverse Document Frequency) and Term Frequency. TF-IDF considers the frequency of a word in the text and its importance in the entire collection, while term frequency only considers the frequency of a word in the text.

Text statistics is based on statistical models to analyze the distribution and relevance of words in text. Common methods include mutual information, pointwise mutual information, and chi-squared test. These methods usually require the establishment of a statistical model between words and texts, and the calculation of word importance based on the model.

The text vectorization method converts the text into a vector representation and then uses the Vector Space Model to calculate the importance of words. Common methods include the Bag-of-Words Model, Word Embeddings, and text vectorization methods (such as TF-IDF vectorization).

Machine learning methods use machine learning algorithms to train models and learn the importance of words from text. Common methods include text classification, text clustering, and keyword extraction models. These methods require the use of labeled text data for model training.

The above can also be done by writing a program using a scripting language (such as Python).

The search module 332 includes a graphic retrieval unit 311, a conversion unit 317 and a graphic comparison unit 318. The graphic retrieval unit 311 receives the graphic input by the user and searches in the database module 600 to generate a search document. After the graphic retrieval unit 311 receives the input graphic, the conversion unit 317 first converts the graphic and brings it into the Vienna classification label, and at the same time converts the graphic into a vector, and then the graphic comparison unit 318 compares it according to the database module 600 to generate a search document.

The IPC analysis module 333 can be the industry classification module 304, which performs industry classification code analysis on the input text, connects to the database module 600 to judge and generate at least one set of patent classification codes.

The text generation module 334 may include the content generation module 306, the search file generation module 310, and even further include the specification generation unit 322 and the application generation unit 323. Among them, the content generation module 306 includes the diagram generation unit 309, and the search file generation module 310 may be the above-mentioned search module 332.

The content generation module 306 receives and generates technical content based on the semantic analysis result in combination with the content learning module 305, wherein the diagram generation unit 309 forms a patent diagram.

The keyword extension module 335 receives the keywords generated by the keyword extraction module 331 and extends them, that is, it searches for similar words or synonyms, connects to a vocabulary for comparison and generates at least one synonym.

The document classification module 336 can classify all case files created by the user, such as attaching specific classification labels to the created case files, such as disclosure documents or search documents, patents or trademarks, etc.

Furthermore, the document classification module 336 can also be for the approximate graphics searched by the graphic retrieval unit 311 in the database module 600, and classify the approximate graphics according to the degree of similarity after calculating the degree of similarity, for example: the degree of similarity above the first risk value is classified into a high risk area, the degree of similarity between the first and second risk values is classified into a medium risk area, and the degree of similarity below the second risk value is classified into a low risk area.

The automatic writing module 337 can be integrated with the content generation module 306 and the content learning module 305, and can automatically write and generate the text required for the corresponding report in combination with the content learning module 305 and the content generation module 306.

The search report generation module 338 and the disclosure book generation module 339 further bring the generated text into a modular table format to generate a search document or patent disclosure document that meets the user's needs. They can also be replaced by the above-mentioned search document generation module 310 and disclosure document generation unit 307, claim content generation unit 308, and diagram generation unit 309.

For an implementation example of the execution method of the above-mentioned system, please refer to Figures 12 to 18, which show the flow chart of the method of the present invention. A user operates an electronic device 100, and a processor 101 of the electronic device 100 connects to a server 200 through a network interface controller and executes an application 201 to generate patent type recommendations and draft generation, which at least includes: (1) the user logs in through the electronic device 100 to perform identity verification and confirm the user's identity and identity information; (2) the user adds a case order through the electronic device 100 and selects the first target country. The information in the case order is The information is updated to the temporary memory 400 regularly; (3) after the user inputs the description text or graphic through the electronic device 100, the description text is converted into a string, and the string information is labeled, and the input language of the string information is recorded in the temporary memory 400; and (4) the semantic analysis module performs semantic analysis on the string information, the industry classification module performs industry classification on the string information, the content learning module performs content learning, and the content generation module generates a patent diagram.

Among them, step (4) further includes: (411) analyzing and segmenting the string information through a semantic analysis module to generate a semantic analysis result; (412) classifying the industry category through an industry classification module based on the semantic analysis result to generate at least one set of patent classification codes; (413) the content learning module learns the content and diagram in the patent specification from the database module based on the at least one set of patent classification codes; (414) combining the semantic analysis result with the learned content through the content generation module to generate a patent disclosure document; and (415) generating a patent diagram based on the patent disclosure document.

Wherein, step (414) further includes: (4141) understanding the semantics of the patent disclosure document and converting it into patent words and terms, and generating the content of the claim through a claim content generation unit; wherein, step (415) further includes: (4151) analyzing the patent disclosure document or the content of the claim and translating it into a diagram generation language, generating a diagram code corresponding to the patent disclosure document or the content of the claim through the diagram generation language, and then generating a patent diagram corresponding to the patent disclosure document or the content of the claim through a compilation unit.

Among them, step (3) further includes: (31) the language determination module determines whether the input language of the string information is the same as the official language of the first target country; (32) if so, directly perform semantic analysis; (33) if not, first translate the string information into the official language of the first target country through the translation module, then perform semantic analysis, and after step (4), translate the generated patent diagram back to the input language of the string information.

Among them, after step (4141), it further includes: (51) the disclosure document generation unit 307 generates the implementation method text according to the content of the claim; (53) the specification generation unit generates the draft after formatting the patent disclosure document, the content of the claim, the implementation method text and the patent diagram; and (6) the application form generation unit captures the identity information of the user login authentication and brings it into the application data, or the user directly enters the application data in the field to generate the application form.

Once generated, the case order can be completed and the patent application can be submitted.

Furthermore, after step (6), the method further includes: (7) the user selects a second target country; (71) the language determination module determines whether the language of the specification document and the application is the same as the official language of the second target country; (72) if not, the specification document and the application are first translated by the translation module, and the specification generation unit and the application generation unit then adjust the format of the translated specification document and the application; (73) if yes, the format adjustment is performed directly.

Once the format adjustment is completed, the multinational patent application conversion is completed, and the application is submitted in the system and handed over to the law firm for official delivery to the second target country.

Furthermore, step (4151) also includes analyzing and judging the patent disclosure document or claim by a semantic analysis module, judging that the description of the content of the patent disclosure document or claim is a process method, system architecture or mechanical mechanism, so that the graphic generation unit generates a corresponding flow chart, block diagram or schematic diagram.

Specifically, different compilation units are used to generate flow charts, system block diagrams, and schematic diagrams, because the training data of the compilation units are different.

The method of the present invention further comprises in step (4): (421) a graphic retrieval unit in a retrieval document generation module, receiving the graphic input by the user and searching in the database module 600 to generate a retrieval document.

Step (421) further includes: (422) after the image retrieval unit receives the input image, the image is first converted by the conversion unit and brought into the Vienna classification label, and the image is converted into a vector; and (423) the image comparison unit then compares the image in the database module to filter out the images whose similarity exceeds a set value.

Please refer to Figures 19 to 22, which show the application examples of the present invention, the actual operating system (https://aiplux.com/) inputs text or selects an industry to make recommendations and generate drafts.

As shown in Figure 19, select the industry category as AI.

As shown in FIG. 20 , the invention generates recommended intellectual property types for application after analysis and search in the database module 600 , including trademarks and patents, and explains what is suitable for trademarks and what is suitable for patents, providing examples.

As shown in Figure 21, after selecting the patent for the invention, enter a more detailed technical description and select the country (the first target country).

As shown in Figure 22, the abstract content and prior art content in the manuscript generated by the present invention.

Finally, the technical features of the present invention and the technical effects it can achieve are summarized as follows:

First, the invention provides an intelligent patent type recommendation and draft generation system and method to solve the difficulty of ordinary people in understanding the professional field of intellectual property rights, especially for those who have no idea which intellectual property rights can be applied for protection of their own technology, and provide an intelligent recommendation function.

Secondly, the intelligent patent type recommendation and draft generation system and method of the present invention can solve the problem that even if the type of intellectual property rights to be applied for has been determined, people still have difficulty in producing application documents (disclosure documents, drafts), and provide a system and method for rapid generation, which saves more time and money.

Thirdly, the invention provides an intelligent patent type recommendation and draft generation system and method to solve the problem of small and medium-sized enterprises and start-ups without intellectual property departments, and resolve the interview model without any patent proposal (patent disclosure book), thereby reducing the inventors' great efforts and time spent on technical communication, as well as the communication and understanding costs of each other in the process.

Fourthly, through the intelligent patent type recommendation and draft generation system and method of the present invention, a complete system and execution method is provided from the initial stage of technical content ideation to the recommendation of intellectual property protection type, and finally to the generation of draft and submission of application, which greatly reduces the professional threshold for ordinary people to enter the application of intellectual property rights to protect their own professional technology.

It must be emphasized that the above detailed description is a specific description of the feasible embodiments of the present invention, but the embodiments are not intended to limit the patent scope of the present invention. Any equivalent implementation or modification that does not deviate from the technical spirit of the present invention should be included in the patent scope of this case.

S100, S200, S300, S400, S500, S600, S700, S800: Steps

S8001, S8002, S8003: Steps

Claims (17)

A method for intelligent patent type recommendation and draft generation is provided. A user operates an electronic device, and a processor of the electronic device is connected to a server through a network interface controller and executes an application to perform patent type recommendation and draft generation. The method comprises at least the following steps: (S100) the user causes the processor to execute an input module to receive input text content through the user interface of the electronic device; (S200) the processor executes a semantic analysis module in the application to perform semantic analysis on the text content; (S300) The semantic analysis module is further connected to an industry classification module, a search module and a database module, and the analyzed text content is classified into industry technologies, and a comparison search is performed in the database module, and the matched data is transmitted to an intellectual property information disclosure module; (S400) The intellectual property information disclosure module analyzes and statistics the data and presents the information to the user through the user interface of the electronic device; (S500) The analyzed and statistical data is also transmitted to a recommendation module at the same time, and the recommendation module classifies and counts the types of intellectual property rights in the data, and sorts and displays the types of intellectual property rights recommended for application on the user interface; (S600) The user selects a patent type through the user interface; (S700) the user causes the processor to execute the input module to receive the input technical description through the user interface; and (S800) the user completes the login through a login module, causing the processor to execute the semantic analysis module and a content generation module to generate summary content and a simple diagram; wherein, after step (S800), the method further includes: (S8001) the processor executes the recommendation module to analyze the generated summary content and simple diagram, and provide the patent types recommended for application; (S8002) The intellectual property information disclosure module further evaluates the patent types and generates a patent risk assessment report, which is simultaneously presented in the information of the patent types recommended for application through the user interface; (S8003) The user selects the patent types to apply for through the user interface, and the processor executes the semantic analysis module and the content generation module to generate patent disclosure documents, claim contents and patent drawings, and integrates them into a formatted template into a draft. A system for intelligent patent type recommendation and draft generation is used to receive a user end operating an electronic device, and a processor of the electronic device is connected to a server through a network interface controller and executes an application program to perform patent type recommendation and draft generation. The system at least includes: an input module, which is used to receive text content and technical description input by the user, and convert the text content and the technical description into strings for labeling, and send a string information and record the input language of the string information in a temporary memory; a semantic analysis module, which receives the string information, analyzes and delimits the string through a natural language database, and generates and sends the semantic analysis result; an industry classification module, which performs industry classification code analysis on the semantic analysis result, connects to a database module to determine and generate at least one set of patent classification codes; a search module performs a comparison search in a database module based on the at least one set of patent classification codes, and generates data of the comparison search result; an intellectual property information disclosure module, which receives the data and further performs statistical analysis to generate basic intellectual property information; a recommendation module, which also receives the data and performs classification and statistics based on the types of intellectual property in the data to generate the types of intellectual property recommended for application; a login module, through which the user operates the electronic device to perform identity verification; and a content generation module, further comprising a disclosure document generation unit and a diagram generation unit, receiving and generating summary content and simple diagrams according to the semantic analysis result; wherein, when the user selects a patent in the recommended intellectual property right category, the input module then receives the technical description input by the user again, and completes identity verification through the login module, and the semantic analysis module then receives string information about the technical description, performs analysis and generates word divisions, and sends the analysis result of the technical description, the disclosure document generation unit generates summary content according to the analysis result, and the diagram generation unit generates summary content according to the summary content Generate a simple diagram, the recommendation module analyzes the summary content and the simple diagram to generate a recommended patent type, the intellectual property information disclosure module evaluates the recommended patent type and generates a patent risk assessment report, the disclosure document generation unit, the diagram generation unit and a claim content generation unit generate a patent disclosure document, claim content and patent diagram according to the patent type, the summary content and the simple diagram. An intelligent patent type recommendation and draft generation system is used to receive a user terminal operating an electronic device, and a processor of the electronic device is connected to a server through a network interface controller and executes an application program to generate patent type recommendation and draft generation. The system at least includes: a login module, and the user operates the electronic device to perform identity verification through the login module; an order processing module generates a new case order, and after the user selects the first target country, the information in the case order is regularly updated to a temporary memory of the system; An input module, for receiving the description text or graphic input by the user, and converting the description text into a string for labeling, and sending a string information and recording the input language of the string information in the temporary memory; a semantic analysis module, receiving the string information, and analyzing and delimiting the string information through a natural language database, generating and sending the semantic analysis result; an industry classification module, performing industry category classification code analysis on the semantic analysis result, connecting to a database module to determine and generate at least one set of patent classification codes; a content learning module, which is a large language model, further comprising a schema learning unit, for different at least one set of patent classification codes, learning the content and schema in the corresponding patent specification from the database module; a content generation module, further comprising a diagram generation unit, receiving and generating a patent diagram according to the semantic analysis result in combination with the content learning module; wherein the content generation module further comprises a disclosure document generation unit and a claim content generation unit, the disclosure document generation unit generates a patent disclosure document according to the semantic analysis result in combination with the content learning module, the generated patent disclosure document at least comprises a technical summary, a prior art content or a combination thereof, the claim content generation unit generates a claim content according to the patent disclosure document in combination with the content learning module, The patent diagram is generated by the diagram generation unit by analyzing the patent disclosure document or the content of the claim and translating it into a diagram generation language, generating a diagram code corresponding to the patent disclosure document or the content of the claim through the diagram generation language, and then generating a diagram content corresponding to the patent disclosure document or the content of the claim through a compilation unit. The system as described in claim 3 also includes a retrieval document generation module, which includes a graphic retrieval unit, a transformation unit and a graphic comparison unit. After the graphic retrieval unit receives the input graphic, it first transforms the graphic through the transformation unit and brings it into the Vienna classification label, and at the same time converts the graphic into a vector, which is then compared in the database module by the graphic comparison unit to produce a retrieval document. The system as described in claim 3, wherein the system further comprises a language determination module for determining whether the input language of the string information is the same as the official language of the first target country. A system as described in claim 5, wherein if the language determination module determines that the string information is different from the official language of the first target country, the string information is translated through a translation module, and after the patent diagram is finally generated, the translation module is used to translate the language in the patent diagram back into the input language of the string information. A system as described in claim 3, wherein the system further comprises a case processing module, which further comprises: a specification generating unit, which receives the patent disclosure document and patent drawings and generates implementation method text according to the content of the claims through the disclosure document generating unit, and further applies a formatting template to the summary content, the prior art content, the implementation method text, the content of the claims and the patent drawings to generate a draft; and an application generating unit, which captures the identity information of the user's login authentication and brings it in as application data, or the user operates the electronic device to directly input the application data in the field and the application data is received by the input module, and the application data is then generated into an application by applying a formatting template; wherein, after the specification document and the application are generated by the case processing module, the case order of the user is completed. A system as described in claim 4, wherein the conversion unit converts graphics from pixels to vectors. The system as described in claim 4, wherein the graphic comparison unit calculates the similarity using edit distance, cosine similarity or Jaccard similarity during the comparison process, and filters out graphics whose similarity exceeds a set value. The system as described in claim 3 further comprises a keyword extraction module, which performs word segmentation and keyword extraction based on the input content to generate a plurality of keywords, and uses the machine learning algorithm to train the model through the content learning module. The system as described in claim 6 further comprises a cross-country conversion module. After the user selects a second target country, the language determination module first determines whether the input language of the string information is the same as the official language of the second target country. If not, the translation module will first translate the patent disclosure document and patent diagram into the official language of the second target country and then perform format adjustment. A method for intelligent patent type recommendation and draft generation is provided, wherein a user operates an electronic device, a processor of the electronic device is connected to a server through a network interface controller, and an application is executed to generate a patent diagram, the method comprising at least the following steps: (1) the user logs in through the electronic device and performs identity verification, and confirms the user identity and identity information; (2) the user adds a case order through the electronic device and selects the first target country, and the information in the case order is regularly updated to a temporary memory; (3) the user inputs a descriptive text or graphic through an input module of the electronic device, and the descriptive text is converted into a string, and the string information is formed by labeling, and the input language of the string information is recorded in the temporary memory; and (4) A semantic analysis module performs semantic analysis on the string information, an industry classification module performs industry classification on the string information, a content learning module performs content learning, and a content generation module generates patent schemas; wherein, step (4) further includes: (411) analyzing and segmenting the string information through the semantic analysis module to generate a semantic analysis result; (412) performing industry classification based on the semantic analysis result through the industry classification module to generate at least one set of patent classification codes; (413) the content learning module learns the content and schema in the patent specification from the database module based on the at least one set of patent classification codes; (414) combining the semantic analysis result with the content generated by the content learning module through the content generation module to generate a patent disclosure document; and (415) generating a patent diagram based on the patent disclosure document; wherein, step (414) further includes: (4141) semantically understanding the patent disclosure document and converting it into patent words and terms, and generating claim content through a claim content generation unit; wherein, step (415) further includes: (4151) analyzing the patent disclosure document or the claim content and translating it into a diagram generation language, generating a diagram code corresponding to the patent disclosure document or the claim content through the diagram generation language, and then generating a patent diagram corresponding to the patent disclosure document or the claim content through a compilation unit. The method as described in claim 12, wherein step (3) further includes: (31) a language determination module determines whether the input language of the string information is the same as the official language of the first target country; (32) if so, directly performing semantic analysis; (33) if not, first translating the string information into the official language of the first target country through a translation module, then performing semantic analysis, and after step (4), translating the generated patent diagram back into the input language of the string information. The method as described in claim 12, wherein after step (4141), it further includes: (51) a disclosure document generation unit generates implementation method text according to the content of the claim; (53) a specification generation unit generates a draft after formatting the patent disclosure document, the content of the claim, the implementation method text and the patent drawing; and (6) an application form generation unit captures the identity information of the user's login authentication and brings it into the application data, or the user directly enters the application information in the field to generate the application form. The method as described in claim 14, wherein after step (6), the following further comprises: (7) the user selects a second target country; (71) a language determination module determines whether the language of the specification document and the application is the same as the official language of the second target country; (72) if not, first translate the specification document and the application through a translation module, and then the specification generation unit and the application generation unit perform format adjustment on the translated specification document and the application; (73) if so, perform format adjustment directly. The method as described in claim 12, wherein step (4) further includes: (421) a graphic retrieval unit in a retrieval document generation module, receiving the graphic input by the user and searching in the database module to generate a retrieval document. The method as described in claim 16, wherein step (421) further includes: (422) after the graphic retrieval unit receives the input graphic, the graphic is first transformed by the transformation unit and brought into the Vienna classification label, and the graphic is converted into a vector at the same time; and (423) the graphic comparison unit then compares the graphic in the database module to filter out the graphic whose similarity exceeds a set value.