A layered intelligent collaborative patent full-process auxiliary drafting system and method

Abstract

The application discloses a layered intelligent collaborative patent whole-process auxiliary compiling system and method, and relates to the technical field of artificial intelligence multi-agent collaboration. The system comprises a text auxiliary writing intelligent module, a special line drawing auxiliary drawing module, an existing technology auxiliary retrieval module, a compliance auxiliary audit verification module, a format standardization auxiliary processing module and a dynamic feedback unit. Each module is interconnected through a standardized data interaction interface, and forms a closed-loop collaborative mechanism through a preset time sequence. The method comprises receiving technical disclosure materials, generating structured patent text and drawing requirements, automatically drawing patent specification line drawings, existing technology semantic retrieval and risk assessment, compliance audit verification, format standardization conversion and dynamic feedback correction. The application realizes intelligent auxiliary of the whole process of patent compilation, greatly improves the writing efficiency, reduces the compliance and rejection risk, and is suitable for automatic auxiliary processing of large-scale patent declaration files.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence multi-agent collaborative processing technology, and in particular to an auxiliary compilation system and method for patent application scenarios, which realizes layered intelligent collaboration throughout the entire process of text compilation, figure generation, risk retrieval, compliance review, and format standardization. Background Technology

[0002] Patent application is a core part of intellectual property strategy, and the complete process includes several key steps such as technical document drafting, patent figure drawing, prior art search, document compliance review, and standardization of application formats. Traditional operating models heavily rely on the manual participation of experienced intellectual property staff throughout the entire process, which has significant shortcomings: fragmented and scattered process steps, inefficient data flow between procedures, large workload, long cycle, and high labor costs for manual drafting; drafting documents is prone to omissions in technical feature descriptions and logical inconsistencies; inconsistent figure drawing formats make it difficult to conform to patent application specifications; prior art search efficiency is low, and novelty risk prediction is inaccurate; document compliance review standards are highly subjective, easily creating potential rejection risks; and post-processing formatting is cumbersome, with a high rate of standardization errors. Existing related auxiliary solutions mostly use single intelligent models operating independently, only achieving simple text assistance in a single step, failing to form a multi-step, layered, collaborative closed-loop operating system. The degree of automation is limited, and a large amount of repetitive manual labor remains, making it difficult to meet the actual needs of batch, efficient, and highly compliant patent application drafting. Summary of the Invention

[0003] Purpose of the invention: To address the shortcomings of existing technologies, such as fragmented processes, limited auxiliary capabilities, high reliance on manual labor, and significant compliance risks, this invention provides a layered intelligent collaborative patent full-process auxiliary compilation system and method. Based on a multi-agent layered division of labor and collaborative architecture, and with the core creative technical concepts provided by humans, it realizes intelligent auxiliary processing of each stage of the patent process, significantly reducing repetitive manual workload, improving compilation efficiency, unifying compliance standards, and reducing the risk of application rejection.

[0004] Technical Solution: A layered intelligent collaborative patent drafting system includes a text-assisted writing module, a dedicated line drawing module, a prior art retrieval module, a compliance review and verification module, a format standardization processing module, and a dynamic feedback unit. The text-assisted writing module receives externally inputted technical disclosure and inventiveness information, intelligently assists in generating initial drafts of the patent text, and outputs structured technical data with semantic tags. The dedicated line drawing module, communicatively connected to the text-assisted writing module, parses the figure requirement tags in the structured technical data and, based on a vector graphics automatic drawing algorithm, assists in generating two-dimensional black-and-white line drawings that conform to patent application specifications. The prior art retrieval module is connected to both the text-assisted writing module and the compliance review and verification module, and extracts core data from the structured technical data. The system identifies key technical features and conducts prior art comparison searches to assist in outputting novelty risk assessment results with risk level indicators. The risk level is generated based on a comparison of semantic similarity calculation results with a preset threshold. The compliance auxiliary review and verification module, combined with the risk assessment results, performs auxiliary compliance verification on the patent text and corresponding figures, focusing on high-risk technical features and outputting optimization and modification suggestions. The format standardization auxiliary processing module, communicatively connected to the compliance auxiliary review and verification module, assists in converting the optimized patent documents into the official standard application format. The dynamic feedback unit connects to all modules, automatically triggering the corresponding module's correction process based on the review results, forming a closed-loop collaborative mechanism. Each module flows sequentially according to a preset business timeline, achieving interconnection through standardized data interaction interfaces, forming an intelligent auxiliary closed loop for the entire patent application process.

[0005] Furthermore, the standardized data interaction interface uses JSON format for data transmission.

[0006] Furthermore, the intelligent text-assisted drafting module possesses capabilities for continuous processing of long contexts, automatic output of structured content, and low-latency, high-frequency response, enabling intelligent assistance in drafting complete patent specifications, claims, and embodiments. Furthermore, the dedicated line drawing assistance module incorporates the State Intellectual Property Office's patent drawing format verification rules, generating only pure two-dimensional black-and-white engineering line drawings without fillers or shadows, with neat lines and conforming to standard size proportions, automatically filtering artistic embellishments, and outputting drawings that directly meet patent application format requirements. Furthermore, the prior art assistance retrieval module employs semantic similarity retrieval to perform a full-dimensional prior art comparison of technical solutions, automatically identifying duplicate technical features and generating a quantitative risk report to assist in predicting application risks. Furthermore, the compliance assistance review and verification module includes at least a claim logic verification unit, a technical feature integrity verification unit, and a graphic content consistency verification unit, enabling multi-dimensional assistance in identifying potential patent application rejection risks and providing optimization reference directions. Furthermore, the dynamic feedback unit supports dynamic adjustment of business timing, automatically adjusting the execution order and priority of each module according to the risk level.

[0007] A hierarchical intelligent collaborative patent full-process assisted drafting method, applied to the aforementioned hierarchical intelligent collaborative patent full-process assisted drafting system, includes the following steps: S1, receiving manually input patent core technology disclosure and inventiveness information, and transmitting it to the text-assisted writing intelligent module; S2, the text-assisted writing intelligent module intelligently generates a complete patent text draft and outputs structured technical data with semantic tags, while simultaneously sending a drawing generation request to the dedicated line drawing auxiliary module; S3, the dedicated line drawing auxiliary module parses the drawing request tags, automatically generates two-dimensional black and white line drawing drawings according to patent specifications, and automatically completes format verification; S4, prior art auxiliary inspection... The search module extracts core technical features from structured technical data, performs prior art semantic retrieval, and assists in outputting novelty risk assessment results with risk level indicators; S5, the compliance auxiliary review and verification module, in conjunction with the search results, adopts a search risk-driven targeted auxiliary review method to conduct linked auxiliary compliance review of the patent text and figures, and outputs text optimization reference schemes; S6, if the review is passed, the format standardization auxiliary processing module corrects the style and format of the document and outputs a patent application preparation document for manual finalization and submission; S7, if the review is failed, the dynamic feedback unit automatically triggers the corresponding module to make corrections, and after the corrections are completed, the review process is restarted.

[0008] Furthermore, step S5 focuses on verifying high-risk technical features and providing targeted optimization suggestions. Moreover, the entire process features automatic data flow and seamless workflow integration across all modules, reducing repetitive workloads in manual document editing, drawing, searching, and formatting, and supporting intelligent parallel processing of multiple batches of patent application documents.

[0009] Beneficial Effects: Layered Collaborative Closed-Loop System Significantly Improves Efficiency. This invention employs a layered, collaborative, and closed-loop architecture with five types of dedicated intelligent modules, featuring pre-defined time-sequenced workflows. This breaks down data barriers throughout the patent drafting process, ensuring seamless integration of each stage. Industry testing shows that the average time for manually drafting an invention patent is 5 days, while using this system reduces the average time to 1.5 days, resulting in an efficiency improvement of over 230%.

[0010] With strong compliance, the risk of application rejection is significantly reduced. Relying on a professional search and verification and compliance review tiered mechanism, it provides optimized references from multiple dimensions such as novelty screening, text logic verification, and consistency check of text and image content, and unifies review standards. Among them, the rejection rate for patent formal defects has been reduced from 12% in the traditional model to below 0.5%, effectively reducing the probability of subsequent patent examination rejection.

[0011] It fully complies with the patent law examination rules, which clearly stipulate that the core inventive concept must be provided by humans, and the intelligent agent only undertakes auxiliary processing tasks. There are no defects in the subject matter created autonomously by AI, the authorization is highly stable, and there is no risk of examination red lines.

[0012] With strong architectural versatility and long-term stable value, this invention only protects the top-level layered collaborative architecture and process logic, without limiting specific intelligent models, manufacturers, or deployment carriers. It is unaffected by AI technology iterations, and the patent protection barrier remains effective for a long period. Attached Figure Description

[0013] Figure 1 This is a module architecture diagram of the system described in this invention;

[0014] Figure 2 This is an overall flowchart of the method described in this invention. Detailed Implementation

[0015] The present invention will be further described in detail below with reference to specific embodiments. Example:

[0016] This embodiment provides a specific data flow method for a layered intelligent collaborative patent end-to-end assisted compilation system:

[0017] Manual Input Technology Disclosure: This invention relates to a collaborative scheduling method for intelligent agents in cross-border e-commerce, including an order receiving module, an intelligent allocation module, and a logistics tracking module. A system architecture diagram and flowchart are required.

[0018] The text-assisted writing of the intelligent module parsing disclosure document outputs the following structured JSON data (example): {"technical_features": ["Order Receiving", "Intelligent Allocation", "Logistics Tracking", "Multi-Agent Collaboration"],"drawing_requirements": [{"type": "system_architecture", "modules": ["Order Receiving Module", "Intelligent Allocation Module", "Logistics Tracking Module"]},{"type": "flow_chart", "steps": ["Order Receiving", "Intelligent Allocation", "Logistics Tracking", "Order Completion"]}],

[0019] "risk_level": 2,

[0020] "audit_priority": "medium"}

[0021] The dedicated line drawing auxiliary module parses the drawing_requirements field in the above JSON, and automatically generates two two-dimensional black and white line drawings that conform to the specifications of the State Intellectual Property Office based on the vector graphics automatic drawing algorithm, and automatically verifies the size, scale and annotation format.

[0022] The existing technology-assisted retrieval module extracts the technical_features field, compares its semantic similarity with the global patent database, and outputs a risk assessment report based on the similarity results and a preset threshold: 3 related patents were found, none of which have completely identical technical solutions, and the novelty risk level is 2.

[0023] The compliance-assisted review and verification module, based on risk level 2, focuses on reviewing the technical features of multi-agent collaboration in the claims and outputs optimization suggestions: it is recommended to supplement the specific logic of intelligent allocation in claim 2. For low-risk patents with a risk level below the preset threshold (such as risk level 1), the dynamic feedback unit can skip the manual intervention step and directly enter the format standardization process.

[0024] The dynamic feedback unit will push optimization suggestions to the intelligent text-assisted writing module, automatically correct the corresponding content, and then resubmit for review.

[0025] After the review is approved, the format standardization auxiliary processing module converts the file into the State Intellectual Property Office standard XML format and outputs the final application preparation document.

Claims

1. A hierarchical intelligent collaborative patent full-process assisted compilation system, characterized in that, The system includes a text-assisted writing intelligent module, a dedicated line drawing assistance module, a prior art retrieval assistance module, a compliance assistance review and verification module, a format standardization assistance processing module, and a dynamic feedback unit. The text-assisted writing intelligent module receives externally inputted technical disclosure and inventiveness information, intelligently assists in generating initial drafts of the patent text, and outputs structured technical data with semantic tags. The dedicated line drawing assistance module, communicatively connected to the text-assisted writing intelligent module, parses the figure requirement tags in the structured technical data and assists in generating two-dimensional black-and-white line drawings that conform to patent application specifications. The prior art retrieval assistance module connects to both the text-assisted writing intelligent module and the compliance assistance review and verification module, extracting core technical features from the structured technical data, conducting prior art comparison searches, and assisting in outputting novelty risk assessment results with risk level indicators. The compliance assistance review and verification module, based on the search risk assessment results, assists in compliance verification of the patent text and corresponding figures, focusing on high-risk technical features, and outputs optimization and modification suggestions. The format standardization assistance processing module, communicatively connected to the compliance assistance review and verification module, assists in converting the optimized patent documents into the official standard application format. The dynamic feedback unit connects to all modules and automatically triggers the correction process of the corresponding module based on the review results, forming a closed-loop collaborative mechanism. Each module flows sequentially according to the preset business time sequence and achieves interconnection through standardized data interaction interfaces, forming an intelligent auxiliary closed loop for the entire patent application document process.

2. The system according to claim 1, characterized in that, The standardized data interaction interface transmits data based on JSON format.

3. The system according to claim 1, characterized in that, The intelligent text-assisted writing module has the capabilities of continuous processing of long contexts, automatic output of structured content, and low-latency, high-frequency response. It can intelligently assist in the drafting of complete manuscripts of patent specifications, claims, and embodiments.

4. The system according to claim 1, characterized in that, The dedicated line drawing auxiliary drawing module is based on a vector graphics automatic drawing algorithm and incorporates the patent drawing format verification rules of the State Intellectual Property Office. It generates only pure two-dimensional black and white engineering line drawings with no fill, no shadow, regular lines, and conforming to standard size proportions. It automatically filters out artistic decoration elements and outputs drawings that directly meet the requirements of patent application drawings, reducing the workload of manual drawing.

5. The system according to claim 1, characterized in that, The existing technology-assisted retrieval module uses semantic similarity retrieval to compare the technical solutions with existing technologies across all dimensions, automatically identify duplicate technical features and generate a quantitative risk reference report to assist manual prediction of application risks.

6. The system according to claim 1, characterized in that, The compliance assistance review and verification module includes at least a claim logic verification unit, a technical feature integrity verification unit, and a graphic content consistency verification unit. It can assist in identifying potential rejection risks of patent applications from multiple dimensions and provide optimization reference directions.

7. The system according to claim 1, characterized in that, The dynamic feedback unit supports dynamic adjustment of business timing and can automatically adjust the execution order and priority of each module according to the risk level.

8. A hierarchical intelligent collaborative patent full-process assisted compilation method, characterized in that, Includes the following steps: S1. Receive manually inputted patent core technology disclosure and inventiveness information, and transmit it to the text-assisted writing intelligent module; S2. The text-assisted writing intelligent module intelligently generates a complete draft of the patent text and outputs structured technical data with semantic tags, while simultaneously sending a drawing generation request to the dedicated line drawing auxiliary module; S3. The dedicated line drawing auxiliary module parses the drawing request tags, automatically generates two-dimensional black and white line drawings according to patent specifications, and automatically completes format verification; S4. The prior art auxiliary retrieval module extracts the core technical features from the structured technical data, performs prior art semantic retrieval, and assists in outputting novelty risk assessment results with risk level indicators; S5. The compliance auxiliary review and verification module, combined with the search results, adopts a search risk-driven targeted auxiliary review method to conduct a linked auxiliary compliance review of the patent text and drawings, and outputs a text optimization reference scheme; S6. If the review is passed, the format standardization auxiliary processing module corrects the style and format of the document, and outputs a patent application preparatory document for manual finalization and submission; S7. If the review is failed, the dynamic feedback unit automatically triggers the corresponding module to make corrections, and after corrections are completed, the review process is restarted.

9. The method according to claim 8, characterized in that, In step S5, key verification and targeted optimization suggestions are output for high-risk technical characteristics.

10. The method according to claim 8, characterized in that, The entire process features automatic data flow and seamless connection between modules, reducing repetitive workloads such as manual document editing, drawing, searching, and formatting. It also supports intelligent parallel processing of multiple batches of patent application documents.

Images