A trademark official opinion letter intelligent grading and bilingual response letter automatic generation method based on a rule engine

Abstract

The present application belongs to the technical field of computer-aided legal document processing, and particularly relates to a trademark official opinion letter intelligent grading and bilingual response letter automatic generation method based on a rule engine, comprising the following steps: S1: client local analysis, S2: rule engine dispute identification, S3: weighted aggregation complexity judgment, S4: channelization dynamic pricing, and S5: bilingual response letter automatic generation. The present application realizes the sinking of the whole process of document analysis, dispute identification, grading calculation and letter generation to the local execution of the client, and the original OA document does not need to be uploaded to any server, so that the channel of sensitive commercial information transmission is cut off from the technical path, the data leakage risk is fundamentally avoided, and the cross-border data compliance requirements and the privacy protection needs of high-end customers are fully met.

Description

Technical Field

[0001] This invention belongs to the field of computer-aided legal document processing technology, specifically relating to a method for intelligent classification of trademark official opinion letters and automatic generation of bilingual response letters based on a rule engine. Background Technology

[0002] In global trademark application and maintenance, trademark authorities in various countries issue official opinions on trademark applications, renewals, and oppositions, raising various dispute requirements such as rejection, evidence supplementation, and formal amendments. After receiving the OA (Office of Action), the agency needs to complete a series of tasks, including dispute identification, complexity assessment, service quotation, and drafting response letters. This is a core and high-frequency part of trademark agency services.

[0003] Currently, there are two main technical models for OA (Office Automation) processing in the industry: The first is the purely manual processing model, where qualified trademark agents manually read through the entire OA document, identify the type of dispute, assess the complexity of the case based on their professional experience, manually calculate the price according to the internal price list, and finally manually draft a bilingual (Chinese and English) response letter. This model has significant drawbacks: First, manual assessment is extremely inefficient; reading and classifying a single complex OA document can take tens of minutes, and the classification and pricing standards are highly dependent on the agent's experience, resulting in strong subjectivity and inconsistent standards. The same OA document processed by different personnel may yield completely different classification and pricing results. Second, cross-border trademark business requires the simultaneous production of bilingual (Chinese and English) materials. Manual translation and drafting can easily lead to inconsistencies in professional terminology and non-standard expressions, and repeated proofreading is time-consuming. Third, manual price quotations are prone to errors such as channel mismatch and incorrect classification, affecting the accuracy of the price and client trust.

[0004] The second type is the server-side parsing and processing mode: Some online systems support users uploading OA documents to a cloud server, where the server-side program performs document parsing, text extraction, and preliminary recognition before returning the results. While this mode improves processing efficiency to some extent, it poses serious privacy and compliance risks: Trademark OA documents contain a large amount of sensitive business information, such as applicant information, trademark layout strategies, and commercial logo planning. Uploading the original documents to a third-party server for storage and parsing poses a risk of data leakage and information misuse, making it difficult to meet cross-border data compliance requirements and the privacy protection needs of high-end clients.

[0005] Furthermore, existing technologies have not yet formed an end-to-end automated processing solution of "local resolution - intelligent classification - dynamic pricing - bilingual email generation", which cannot simultaneously meet the triple requirements of processing efficiency, standardization and privacy compliance, thus restricting the digital upgrade of the trademark agency industry. Summary of the Invention

[0006] The purpose of this invention is to provide a method for intelligent classification of official trademark opinion letters and automatic generation of bilingual response letters based on a rule engine. This method decentralizes the entire process to the client-side and achieves standardized dispute identification and quantitative complexity classification through a rule engine. It also enables channel-based dynamic pricing and automatic generation of bilingual (Chinese and English) letters. Under the premise of ensuring data privacy and compliance, this method significantly improves OA processing efficiency and result consistency.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for intelligent classification of official trademark opinion letters and automatic generation of bilingual response letters based on a rule engine, comprising the following steps: S1. Local parsing on the client: The official trademark opinion letter PDF document is extracted page by page on the client using a PDF parsing library. The original document is stored locally on the client and is not uploaded to the server. S2. Dispute Identification by the Rule Engine: Based on a pre-built dispute rule dictionary, the extracted text is scanned and matched for keywords to identify all disputed entries contained in the document; the dispute rule dictionary adopts a mapping structure of "keyword - dispute type - Chinese and English name - complexity label"; S3, Weighted Aggregate Complexity Rating: Preset a corresponding complexity weight for each type of dispute, sum the weights corresponding to all hit dispute items to obtain the aggregate weight value, and map the complexity level of the entire trademark official opinion letter based on the aggregate weight value; S4. Channel-based dynamic pricing: Using "customer source channel × complexity level" as the query index, the corresponding service price is obtained by matching from the price list; S5. Automatic generation of bilingual response letters: Based on the identified list of disputed items and their complexity level, the system calls a preset bilingual letter template and automatically generates a draft response letter in both Chinese and English.

[0008] As a method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on rule engine of the present invention, preferably, in step S1, the PDF text extraction supports limiting the maximum number of pages to be parsed, and only extracts the first N pages of text content of the trademark official opinion, where N is a preset positive integer; the client is a browser client, the PDF parsing library is a front-end JavaScript parsing library, and the text extraction process is completed entirely in the browser's local memory.

[0009] As a method for intelligent classification of trademark official opinion letters and automatic generation of bilingual response letters based on rule engine of the present invention, preferably, the construction method of the dispute rule dictionary in S2 is as follows: for each type of trademark official dispute, the corresponding Chinese and English official terms and common expression variations are included as matching keywords, and the standard Chinese and English names of the dispute type and the corresponding complexity tags are associated at the same time; the matching process adopts full-text keyword hit determination, and when the corresponding keywords appear in the text, it is determined that the dispute type exists.

[0010] As a method for intelligently classifying trademark official opinion letters and automatically generating bilingual response letters based on a rule engine according to the present invention, preferably, the dispute types include at least confusion rejection, descriptive rejection, sample / use evidence issues, abandonment of exclusive rights, translation declaration, modification of goods / services items, and general form issues.

[0011] As a method for intelligently classifying trademark official opinion letters and automatically generating bilingual response letters based on a rule engine according to the present invention, preferably, the complexity weight in S3 is divided into three levels: a weight value of 2 for complex disputes, a weight value of 1 for medium disputes, and a weight value of 0 for simple disputes; the complexity level is divided into three levels: simple, medium, and complex, and the corresponding level is obtained by comparing and mapping the preset weight threshold range with the aggregated weight value.

[0012] As a method for intelligently classifying trademark official opinion letters and automatically generating bilingual response letters based on a rule engine according to the present invention, preferably, in step S3, while obtaining the complexity level, a preset response success rate estimate value is matched according to the complexity level and output synchronously with the complexity level.

[0013] As a method for intelligently classifying trademark official opinion letters and automatically generating bilingual response letters based on a rule engine according to the present invention, preferably, the price list in S4 includes a built-in price list and a user-defined uploaded CSV format price list. The price list distinguishes different business types such as new applications, renewals, and oppositions and sets corresponding price columns. When querying, the corresponding price column is selected according to the current business type for quotation matching.

[0014] As a method for intelligently classifying trademark official opinion letters and automatically generating bilingual response letters based on a rule engine according to the present invention, preferably, the bilingual letter template in S5 pre-sets corresponding Chinese and English bilingual response key paragraphs for each type of dispute. When generating the letter, the corresponding response key points are automatically spliced ​​according to the disputed items, and the complexity level and general customer information fields are filled in to form a complete response letter draft; the generated letter supports one-click copying and local export.

[0015] As a method for intelligent classification of trademark official opinion letters and automatic generation of bilingual response letters based on rule engine of the present invention, preferably, in step S5, while generating the draft response letter, a sub-order creation template is automatically generated according to the dispute list, complexity level and corresponding quotation. The template includes the dispute type, complexity level, quotation amount and business type fields corresponding to the order.

[0016] As a method for intelligent grading of trademark official opinion letters and automatic generation of bilingual response letters based on rule engine of the present invention, preferably, the method is executed entirely on the browser client, and the original trademark official opinion letter document and the parsed complete text content are not uploaded to the server during all text parsing, rule matching, grade calculation, quotation query and letter generation processes.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This invention decentralizes the entire process of document parsing, dispute identification, hierarchical calculation, and letter generation to the client's local machine. The original OA documents do not need to be uploaded to any server, which technically cuts off the channels for the leakage of sensitive business information, fundamentally avoids the risk of data leakage, and fully meets the requirements of cross-border data compliance and the privacy protection needs of high-end customers.

[0018] This invention achieves automated identification of dispute types through a structured rule engine, and determines the complexity level by combining a weighted aggregation quantitative calculation model. It completely replaces subjective human judgment, and the classification results are reproducible and the standards are unified. It eliminates the judgment bias caused by the difference in experience of different agents, and ensures the fairness and consistency of case classification and pricing.

[0019] This invention achieves a fully automated closed loop from document parsing, dispute identification, complexity classification to pricing and letter generation. It can complete the processing work that originally required several hours of manual work with just one click, significantly shortening the response cycle of trademark OA and increasing the daily case processing capacity of a single agent by several times.

[0020] This invention achieves dynamic pricing by employing a two-dimensional indexing mechanism of "customer source channel × complexity level". It supports both the system's built-in price list and user-defined CSV price list, automatically matching corresponding quotations and completely avoiding mismatches and omissions caused by manual table lookup. This ensures that the pricing standards for cases from different channels and at different levels are consistent and the calculations are accurate. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1This is a flowchart illustrating the method for automatically generating official trademark opinion letters based on a rule engine, with intelligent classification and bilingual response letters.

[0022] Figure 2 This is a mapping table from a keyword dictionary to dispute type and complexity weight for a rule-engine-based method for intelligent grading of official trademark opinion letters and automatic generation of bilingual response letters. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Please see Figure 1-2 The present invention provides the following technical solution: a method for intelligent classification of trademark official opinion letters and automatic generation of bilingual response letters based on rule engine, comprising the following steps: S1: local parsing on the client, S2: dispute identification by rule engine, S3: weighted aggregation complexity classification, S4: channel-based dynamic pricing, S5: automatic generation of bilingual response letters.

[0025] Client-side local parsing principle: This stage is built upon native front-end PDF parsing technology. The core principle is to completely offload document parsing capabilities to the user's terminal, achieving a privacy-preserving architecture where "data remains unchanged, program operates." After the user selects a local OA PDF file in their browser, the system calls the front-end JavaScript PDF parsing library to directly read the file's binary stream from the browser's memory, decoding and extracting the text content page by page. Throughout the process, the original file, file fragments, or the complete parsed text are not uploaded to any server. Simultaneously, the system supports setting a maximum number of pages to parse. For regular OA documents, only the first few pages of core review content are extracted, skipping non-core pages such as attachments and receipts. This further improves parsing speed and reduces terminal performance consumption while ensuring recognition accuracy.

[0026] The principle of dispute identification by the rule engine: The dispute identification process is based on a pre-built structured dispute rule dictionary and adopts a four-layer mapping architecture: keyword layer → dispute type layer → Chinese and English name layer → complexity label layer.

[0027] Keyword layer: For each type of trademark dispute, it includes official standard expressions, commonly used official abbreviations, and industry-standard synonym variations from trademark offices of various countries. It also covers bilingual expressions in Chinese and English to ensure that OAs from different countries and with different expression styles can be accurately identified. Matching Logic: The system traverses and scans the extracted full-text. When any valid keyword of a certain type of dispute is matched in the text, the system determines that the dispute type exists and finally outputs a complete list of matched dispute entries. This rule engine adopts deterministic matching logic, ensuring that the matching results are not affected by human factors and that the recognition standards are consistent. It also supports rapid adaptation to new dispute types and countries / regions by expanding the keyword dictionary, demonstrating strong scalability.

[0028] The weighted aggregation complexity grading principle: Complexity grading adopts a quantitative calculation principle of "single-item weight assignment + overall aggregation summation + threshold mapping grading," transforming the originally vague "complexity" into a calculable numerical indicator. First, based on the difficulty, workload, and professional requirements of handling various disputes, a corresponding complexity weight is predefined for each dispute type, divided into three levels: complex (weight 2), medium (weight 1), and simple (weight 0). After the rule engine identifies all dispute items, the system automatically sums the weight values ​​of all items to obtain the total aggregate weight score for the entire OA (Office Automation) document. Finally, through a preset weight threshold range (e.g., 0 points for simple, 1-2 points for medium, and 3 points and above for complex), the total aggregate score is mapped to the three overall complexity levels: simple, medium, and complex. Simultaneously, based on historical case handling results data, the system presets a corresponding success rate reference value for each complexity level, outputting it synchronously with the level to provide decision-making reference for business personnel.

[0029] Channel-based dynamic pricing principle: The pricing process employs a two-dimensional index matching mechanism, using "customer source channel" and "complexity level" as the two query keys to construct a two-dimensional price list. The price list supports two modes: one is a system-built-in standard price list, adaptable to general business scenarios; the other is a user-defined CSV format price list, adaptable to the agency's own pricing system. The price list can be set with multiple columns of prices according to business type (such as new application, renewal, objection, cancellation, etc.), adapting to pricing differences in different business scenarios. During a query, the system reads the customer channel and the determined complexity level of the current case, using them as a two-dimensional index to locate the corresponding cell in the price list, directly retrieving the quoted price value. The entire process is automated, eliminating the need for manual table lookups and fundamentally avoiding errors caused by manual price checks.

[0030] The automatic bilingual letter generation principle: Letter generation adopts a combination of "general template filling + dispute module splicing". The system has a pre-set standardized general framework template for Chinese-English bilingual letters, including general paragraphs such as letter header, salutation, introduction, and closing; at the same time, for each type of dispute, corresponding professional response paragraphs are pre-set, and each paragraph is accompanied by standard Chinese-English bilingual text. When generating a letter, the system first fills in the general fields of the letter according to case information, complexity level, etc.; then, according to the list of dispute items, it retrieves the corresponding response paragraphs of the dispute in the standardized order and automatically splices them into the body of the letter; finally, it generates a complete and standardized Chinese-English bilingual response letter draft.

[0031] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligently classifying official trademark opinion letters and automatically generating bilingual response correspondence based on a rule engine, characterized in that, Includes the following steps: S1. Local parsing on the client: The official trademark opinion letter PDF document is extracted page by page on the client using a PDF parsing library. The original document is stored locally on the client and is not uploaded to the server. S2. Dispute Identification by the Rule Engine: Based on a pre-built dispute rule dictionary, the extracted text is scanned and matched for keywords to identify all disputed entries contained in the document; the dispute rule dictionary adopts a mapping structure of "keyword - dispute type - Chinese and English name - complexity label"; S3, Weighted Aggregate Complexity Rating: Preset a corresponding complexity weight for each type of dispute, sum the weights corresponding to all hit dispute items to obtain the aggregate weight value, and map the complexity level of the entire trademark official opinion letter based on the aggregate weight value; S4. Channel-based dynamic pricing: Using "customer source channel × complexity level" as the query index, the corresponding service price is obtained by matching from the price list; S5. Automatic generation of bilingual response letters: Based on the identified list of disputed items and their complexity level, the system calls a preset bilingual letter template and automatically generates a draft response letter in both Chinese and English.

2. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: In S1, the PDF text extraction supports limiting the maximum number of pages to be parsed, extracting only the first N pages of the trademark official opinion letter, where N is a preset positive integer; the client is a browser client, the PDF parsing library is a front-end JavaScript parsing library, and the text extraction process is completed entirely in the browser's local memory.

3. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: The dispute rule dictionary in S2 is constructed as follows: for each type of official trademark dispute, the corresponding official Chinese and English terms and common expression variations are included as matching keywords, and the standard Chinese and English names of the dispute type are associated with the corresponding complexity labels; the matching process adopts full-text keyword hit determination, and when the corresponding keywords appear in the text, it is determined that the dispute type exists.

4. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: The types of disputes include at least confusion rejection, descriptive rejection, sample / use evidence issues, waiver of exclusive rights, translation statement, modification of goods / services, and general form issues.

5. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: In S3, the complexity weight is divided into three levels: a weight value of 2 corresponds to a complex dispute, a weight value of 1 corresponds to a medium dispute, and a weight value of 0 corresponds to a simple dispute. The complexity level is divided into three levels: simple, medium, and complex. The corresponding level is obtained by comparing and mapping the preset weight threshold range with the aggregated weight value.

6. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: In step S3, while obtaining the complexity level, a preset estimated success rate value is matched according to the complexity level and output synchronously with the complexity level.

7. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: The price list in S4 includes a built-in price list and a user-defined CSV format price list. The price list distinguishes between different business types such as new applications, renewals, and objections, and sets corresponding price columns. When querying, the corresponding price column is selected according to the current business type for quotation matching.

8. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: The bilingual letter template in S5 pre-sets corresponding Chinese and English bilingual response points for each type of dispute. When generating a letter, it automatically splices the corresponding response points according to the disputed items and fills in the complexity level and customer information general fields to form a complete response letter draft. The generated letter supports one-click copying and local export.

9. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: In S5, while generating a draft response letter, a sub-order creation template is automatically generated based on the dispute list, complexity level, and corresponding quotation. The template includes fields for the dispute type, complexity level, quotation amount, and business type corresponding to the order.

10. The method for intelligent classification and automatic generation of bilingual response letters for trademark official opinions based on a rule engine, as described in claim 1, is characterized in that: The method is executed entirely on the browser client. All text parsing, rule matching, rating calculation, quotation query, and letter generation processes do not upload the original trademark official opinion document or the complete parsed text content to the server.

Images