A method and device for generating an annual fee plan and risk reminding based on multi-source identification and structured management of patent files, an electronic device, and a storage medium

Abstract

The application discloses a kind of based on patent file multi-source identification and structured management's annual fee plan generation and risk reminding method and device.The method is executed PDF text extraction, image conversion and direction correction to original patent file such as patent certificate, handling registration procedures notice;Based on at least two identification sources, generate candidate identification text, according to field hit quantity, integrity, legality and confidence determine main identification result;According to file type template, extract patent basic field and filter noise;After patent number normalization, match existing records according to four-level rules, generate records to be confirmed when multiple record hits or field conflicts;With application date, generate patent annual interval, calculate registration payment deadline according to different caliber before and after policy switching, determine authorized year should be paid annual and generate annual fee plan.The application improves the accuracy of field in patent file structured processing, matching consistency and authorized year determination reliability.

Description

Technical Field

[0001] This invention relates to the fields of document recognition, patent data structuring, patent annuity plan generation, and risk alerts, and particularly to a method, apparatus, electronic device, and storage medium for annuity plan generation and risk alerts based on multi-source patent document recognition and structuring governance. Background Technology

[0002] Patent certificates, patent certificate renewal pages, registration notices, patent termination notices, and archived patent documents are important data sources for enterprises or agencies to establish patent annuity ledgers and generate payment plans. These documents may exist in the form of raw PDF text, scanned copies, image-based pages, or archived documents, and often contain non-target fields such as rotated pages, watermarks, footers, page numbers, legal clauses, and tabulation instructions, leading to inconsistent single OCR recognition results.

[0003] The same patent may appear in different patent documents with different numbering formats. For example, the patent number may contain prefixes such as ZL and CN, spaces, separators, English punctuation marks, Chinese punctuation marks, or differences in capitalization. If the system directly uses the original recognized text for matching, it may easily identify the same patent as different records, resulting in duplicate filing or the inability to associate subsequent notifications and termination notices with the existing patent ledger.

[0004] When documents from multiple sources are archived successively, inconsistencies may arise in fields such as patent name, application date, patentee, patent type, application number, or patent number for the same patent. If the system directly uses the new identification fields to overwrite existing manually confirmed fields, it may contaminate critical fields, thereby affecting the generation of annual fee plans, risk alerts, and subsequent manual verification.

[0005] The patent year due for authorization is not simply determined by the authorization announcement date, calendar year, or document generation date. Instead, it requires consideration of the patent year range formed by the application date and the deadline for registration and payment. Furthermore, the calculation methods for the registration and payment deadline differ before and after policy changes. Relying on manual judgment or inference from a single date field can easily lead to misjudgments of the patent year due for authorization.

[0006] Therefore, a structured governance method is needed after the identification of official patent documents. This method can reduce the risk of incorrect filing, incorrect coverage, and misjudgment of the year of authorization by using multi-source candidate identification and scoring, template constraint field extraction, number standardization matching, field conflict confirmation, and determination of the patent year range where the registration and payment deadline falls. Summary of the Invention

[0007] To address the aforementioned issues, this invention provides a method for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents. This method resolves problems such as noise in the identification of official patent documents, inconsistent numbering formats, automatic overwriting of conflicting fields, and misjudgment of the payment year in the year of authorization, which can lead to incorrect filing and payment years.

[0008] The method of this invention includes: acquiring the original patent document and performing PDF native text extraction, page image conversion, and page orientation correction; generating candidate recognition text based on at least two recognition sources, and determining the main recognition result according to the number of field hits, field completeness, field legality, and recognition confidence; extracting basic patent fields and filtering noisy text according to the patent document type template and field constraint rules; standardizing the patent number and / or application number to generate a simplified patent number, and matching existing patent records according to a four-level matching rule; generating a record to be confirmed without automatically overwriting existing records when multiple records are hit or field conflicts occur; generating a patent year interval based on the application date, calculating the deadline for registration and payment, and determining the annual fee payable for the current year of authorization; and generating a patent annuity plan based on the standardized patent records, the annual fee payable for the current year of authorization, and the patent year interval.

[0009] Compared with the prior art, the beneficial effects of the present invention are:

[0010] 1. By selecting the main result through multi-source candidate recognition scoring, the original PDF text, local image OCR recognition results, backup OCR recognition results and image text recognition results can be comprehensively compared based on field hit, field completeness, field legality and recognition confidence, reducing the probability of misidentification of key fields due to quality fluctuations of a single recognition source.

[0011] 2. Field extraction is performed using patent document type templates and field constraint rules. Combined with field start keywords, field end markers, negative keywords, and cross-page duplicate text exclusion rules, the risk of watermarks, page numbers, legal provisions, explanatory text, and duplicate footers entering the basic patent fields is reduced.

[0012] 3. By removing the ZL prefix, CN prefix, spaces, separators, Chinese punctuation, and English period, and unifying the capitalization of letters, a simplified patent number is generated. Then, the record is matched using a four-level matching rule to reduce the risk of duplicate filing or incorrect association of the same patent due to differences in number format.

[0013] 4. By generating a record to be confirmed when multiple records are hit or when the identified field conflicts with an existing field, instead of automatically overwriting existing patent records, the probability of existing manually confirmed fields being contaminated by incorrect identification results is reduced.

[0014] 5. By mapping the deadline for registration and payment to the patent year range based on the application date, the year for payment due in the year of authorization is determined, thereby improving the consistency and traceability of the determination of the year for payment due in the year of authorization.

[0015] 6. By retaining the source field association in the annual fee plan, and recording which file, page, identification text, and field extraction rules the structured fields come from, the traceability of subsequent verification, error correction, and manual confirmation is improved. Attached Figure Description

[0016] Figure 1 The overall flowchart of the annual fee plan generation method based on multi-source identification and structured governance of patent documents provided in the embodiments of the present invention includes step labels S101 to S108; Figure 2 This is a flowchart of multi-source candidate identification and template field extraction provided in an embodiment of the present invention, which includes step labels S201 to S207; Figure 3 The flowchart for number standardization, four-level matching and pending confirmation processing provided in the embodiments of the present invention includes step markers S301 to S309. Figure 4 The flowchart for determining the patent year range and the annual tax payable in the year of authorization is provided for the embodiments of the present invention, and the flowchart includes step markers S401 to S408; Figure 5 The flowchart for annual fee plan generation, fee calculation and risk status refresh provided in the embodiments of the present invention includes step labels S501 to S512.

[0017] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application; other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are all within the scope of protection of this application.

[0019] like Figure 1 As shown, a method for generating annual fee plans and risk alerts based on multi-source identification and structured governance of patent documents includes steps S1 to S6.

[0020] For ease of explanation with reference to the accompanying drawings, steps S1 to S6 in the claims are labeled as the main process flow of the method. Figures 1 to 5S101 to S512 in the figures are step labels in the specific embodiments. These step labels are only used to illustrate the processing order and module relationships in the embodiments and are not used to limit the scope of protection of the claims.

[0021] Please see Figure 1 , Figure 1 The overall processing flow of an embodiment of the present invention is shown, wherein S101 represents obtaining a patent certificate, certificate renewal page, registration procedure notification, termination notice, or archived patent document; S102 represents document type identification and preprocessing, including PDF native text extraction, image conversion, and page orientation correction; S103 represents multi-source text recognition, including PDF text, local OCR, backup OCR, and image text recognition results; S104 represents extracting fields such as patent name, type, patent number, application number, and application date based on the patent document template; S105 represents standardizing the patent number and application number and generating abbreviated patent number; S106 represents multi-level matching of existing patent records, supplementing fields when there is a unique match, and creating a new record when there is no match; S107 represents generating a record to be confirmed when there is a conflict, suspected duplication, or missing key fields; S108 represents generating the patent annual range, annual fee plan, and refreshing risk reminders.

[0022] Please see Figure 2 , Figure 2 The process of multi-source candidate identification and template field extraction is shown. S201 represents obtaining the original text of the PDF; S202 represents obtaining the page image or rotated candidate image; S203 represents reading the archived structured information; S204 represents performing multi-source identification and candidate text set, and recording the identification source, page number, confidence level and identification time; S205 represents distinguishing patent certificates, certificate renewal pages, registration procedure notices and termination notices according to document type templates; S206 represents performing field constraints and noise elimination, excluding headers, footers, watermarks, legal provisions and irrelevant prompts; S207 represents outputting field candidate values, and entering the pending confirmation stage when the confidence level is low or the key field is empty.

[0023] Please see Figure 3 , Figure 3The diagram illustrates the numbering standardization, four-level matching, and pending confirmation process. Specifically, S301 represents the first matching based on identical abbreviated patent numbers; S302 indicates a high-confidence match when a unique record is found, and missing fields are supplemented; S303 represents the second matching based on identical application numbers; S304 generates a suspected duplicate anomaly when multiple records are found; S305 represents the third matching based on patent name, application date, and patentee; S306 indicates that when a unique match is found, the source file is associated and the source identification is saved; S307 represents the fourth matching based on name similarity, customer, type, and application date proximity; S308 indicates that records with medium to low confidence are not automatically merged and enter the pending confirmation stage; and S309 represents the pending confirmation process, including manual supplementation, merging existing records, ignoring, marking non-patent records, re-identifying, and saving modified records.

[0024] Please see Figure 4 , Figure 4 The document illustrates the process for determining the patent year range and the annual fee due in the year of authorization. Specifically, S401 reads the application date and patent type; S402 generates the patent year range based on the application date; S403 identifies the date of issuance of the registration procedure notification; S404 calculates the registration and fee payment deadline based on the policy switch date; S405 uses the last day of the target month if no corresponding date exists; S406 determines the patent year range into which the registration and fee payment deadline falls; S407 determines the patent fee payment year due in the year of authorization; and S408 generates the normal payment deadline and annual fee plan for subsequent years.

[0025] Please see Figure 5 , Figure 5 The flowchart illustrates the process of generating an annual fee plan, calculating fees, and updating risk status. Specifically, S501 retrieves the patent type and patent payment year; S502 queries the fee rules table and obtains the standard annual fee; S503 retrieves the payment status; S504 retrieves the customer's annual fee reduction status; S505 determines if there are co-owners; S506 isolates manual adjustments to the amount; S507 calculates the reduced amount; S508 calculates late payment fees based on the standard annual fee if overdue; S509 calculates the total amount due; S510 determines the grace period, recovery period, and invalidation status; S511 updates the annual fee status and risk level; and S512 generates reminders, to-do items, and operation logs.

[0026] Step S1: Obtain the original patent documents, which include at least one of the following: patent certificate, patent certificate renewal page, notification of registration procedures, notification of patent termination, or archived patent documents. For original patent documents in PDF format, the system first performs native PDF text extraction; when the native PDF text is empty, the quality is below the threshold, or key fields are not fully matched, the page is converted into a page image; for image-based pages, the system performs page orientation correction to reduce the impact of page rotation on subsequent field extraction.

[0027] In step S2, the system generates candidate recognition text based on at least two recognition sources: the original PDF text, the local image OCR recognition result, the backup OCR recognition result, and the image text recognition result. Each candidate recognition text can be associated with information such as recognition source, recognition time, page number, confidence level, and whether it comes from the original PDF text. The system scores the candidate recognition text according to the number of field hits, field completeness, field validity, and recognition confidence level, and selects the candidate recognition text with the highest comprehensive score and that meets the field validity condition as the main recognition result.

[0028] In a multi-source recognition scoring example, the original PDF text only matches the patent name and patentee, but not the application number, application date, and date of issuance of the registration notification. The local image OCR text matches the patent name, patentee, application number, application date, and date of issuance of the registration notification, and the number and date formats are valid. The backup OCR text matches fewer fields or has an invalid date format. The system calculates a comprehensive score based on field match weights, date validity, number format validity, template context matching, and OCR confidence, and determines the local image OCR text as the primary recognition result. If the local image OCR text and the original PDF text conflict on key fields and the difference in comprehensive scores is less than a preset threshold, a set of fields to be confirmed is generated.

[0029] Step S3: The system extracts basic patent fields from the main recognition result and / or candidate recognition text based on the patent document type template and field constraint rules. The patent document type template includes at least one of the following: patent certificate template, patent certificate renewal page template, registration procedure notification template, and patent termination notice template. Different templates correspond to different field start keywords, field end markers, field format constraints, negative keywords, and cross-page duplicate text exclusion rules.

[0030] For example, in the patent certificate template, the system can extract fields based on contextual keywords such as certificate number, patent name, inventor or designer, patent number, application date, patentee, address, and authorization announcement date; in the registration procedure notification template, the system can prioritize extracting the application number, patent name, date of issuance of the registration procedure notification, and deadline for payment of registration fees; in the patent right termination notification template, the system can prioritize extracting the patent number, application number, date of issuance of the termination notification, reason for termination, and restoration period.

[0031] In the noise filtering implementation, when the same legal prompts or page numbers appear at the bottom of consecutive pages, and are not located between the field's start keyword and end marker, the system identifies them as duplicate footers or explanatory text spanning multiple pages and excludes them. For watermarks, legal provision prompts, tabular instructions, and text that does not match the field context, the system filters them using negative keywords and field format constraints to prevent such text from entering basic patent fields such as patent name, address, patentee, and application date.

[0032] In step S4, the system normalizes the patent number and / or application number in the basic patent fields to generate abbreviated patent numbers for deduplication and matching. Normalization includes removing the ZL prefix, CN prefix, spaces, separators, Chinese punctuation, and English periods, and converting all letters to uppercase. For example, if the original number contains ZL, spaces, and periods, the system retains the original number while generating abbreviated patent numbers for matching.

[0033] The system matches existing patent records according to a four-level matching rule. The first match is an exact match of the abbreviated patent number; the second match is an exact match of the application number; the third match is an exact match of the patent name, application date, and patentee; and the fourth match is an exact match of the similarity of the patent name, client name, patent type, and application date meeting preset conditions. The preset similarity threshold and preset number of days can be set according to the actual patent ledger size and field quality, for example, the patent name similarity threshold can be set to 0.8, and the application date difference can not exceed 30 days.

[0034] In the example where a fourth-level match triggers confirmation, the system normalizes the identified patent number to generate a simplified patent number. When the first match queries existing patent records using the simplified patent number, it finds two records. The system does not automatically merge or overwrite either record. Instead, it generates a suspected duplicate record to be confirmed, retaining the candidate record, source file, candidate fields, conflict fields, and matching reasons. The system then performs merging, ignoring, re-identification, or saving the changes only after manual confirmation.

[0035] When a unique patent record is matched, the system fills in the missing fields in the unique patent record and associates it with the original patent document; when no existing patent record is matched, the system generates a new patent record; when multiple existing patent records meet the matching conditions, or when the identification field conflicts with the existing field, the system generates a record to be confirmed instead of automatically overwriting the existing patent record.

[0036] Step S5: The system generates consecutive patent year intervals based on the application date. Specifically, the first year is defined as the same month and day of the following year from the application date, and subsequent patent year intervals are generated sequentially. For example, if the application date is November 15, 2099, the first year interval is from November 15, 2099 to November 15, 2100; the second year interval is from November 15, 2100 to November 15, 2101; and so on for subsequent years. The dates mentioned above are for illustrative purposes only and do not correspond to actual cases.

[0037] The calculation of the registration and payment deadline varies depending on the pre-set policy switch date. When the date of the registration procedure notification is earlier than the pre-set policy switch date, the deadline is calculated as fifteen days plus two months from the date of the notification. When the date of the registration procedure notification is not earlier than the pre-set policy switch date, the deadline is calculated as two months from the date of the notification. When adding a month to the date, if the target month does not have a corresponding date (e.g., adding one month from the 31st of a certain month when the target month does not have a 31st), the last day of the target month is used as the calculation result.

[0038] The system compares the registration and payment deadline with the continuous patent year range and determines the patent year into which the deadline falls as the year of authorization for which fees are due. This method avoids misusing the authorization announcement date, calendar year, or document generation date as the basis for determining the year of authorization for fee payment.

[0039] Step S6: The system generates a patent annuity plan based on the standardized patent records, the annual payment due date for the current year of authorization, and the patent year interval. The patent annuity plan includes at least the patent payment year, the year start date, the year end date, the normal payment deadline, the annual payment due date identifier for the current year of authorization, and the source field association relationships. The source field association relationships are used to record which file, page, identified text, and field extraction rule each structured field originates from.

[0040] In a subordinate embodiment, the system can further calculate the standard fee amount, the reduced fee amount, the late payment fee amount, and the total amount payable based on the customer's annual fee reduction status, patent type, patent year, and co-owner information. When the customer's annual fee reduction status is unconfirmed or temporarily unconfirmed, the system displays the unreduced amount and generates a fee reduction status confirmation prompt. This fee calculation relies on the aforementioned structured fields and the annual payment determination result for the current year of authorization, rather than a separate business registration process.

[0041] When there are two or more patent holders, the system applies the fee reduction rules corresponding to each co-owner individually, without changing the fee reduction calculation results for other patents under the same client's name. If there are manual adjustments to the annual fee plan, the system only updates the calculated amount and does not automatically overwrite the manually confirmed final amount due. Late payment penalties are calculated based on the standard annual fee, not the reduced amount.

[0042] In the risk refresh implementation, the system refreshes the annual fee status and risk level based on the normal payment deadline, grace period deadline, recovery period deadline, payment status, termination notice, abandonment status, rejection status, or expiration status. The annual fee status can include categories such as pending confirmation, normal, nearing deadline, overdue, recovery period, termination, or abandonment; the risk level can be graded based on the deadline and the rights status. The above status refresh relies on the generated patent annual fee plan, source documents, and structured fields.

[0043] The system records a complete workflow log, including file identification, field extraction, numbering standardization, record matching, pending confirmation processing, annual fee calculation, risk updates, and manual modifications. The logs can include processing time, processing source, pre-processing value, post-processing value, candidate identification source, matching rules, conflicting fields, and manual processing results. Through these log entries, the system can trace the basis for generating a specific field, annual plan, or risk alert during subsequent verification.

[0044] The pending confirmation records support manual addition of fields, merging of existing records, ignoring, marking as non-patented, re-identifying, or saving modified records. After key fields such as application date, patent type, patent number, or date of issuance of the registration procedure notification are manually added, the system can re-execute the steps of number standardization, four-level matching, determination of the annual fee payable for the current year of authorization, and generation of the annual fee plan.

[0045] This invention can be deployed on electronic devices or servers capable of performing document recognition, field extraction, record matching, and annual fee plan generation. The above deployment method does not affect the technical processing chain of this invention, namely, multi-source candidate identification and scoring, template-constrained field extraction, number standardization matching, conflict confirmation, and annual fee determination for the current year.

[0046] Each module in this embodiment of the invention can be implemented by software programs, hardware circuits, or a combination of software and hardware. When the processor executes the computer program stored in the memory, it can implement any of the above method steps.

[0047] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents, characterized in that, Includes the following steps: Step S1: Obtain the original patent document, which includes at least one of the following: patent certificate, patent certificate renewal page, registration procedure notification, patent termination notice, or archived patent document; perform PDF native text extraction, page image conversion, and page orientation correction on the original patent document; Step S2: Generate candidate recognition text based on at least two recognition sources among the PDF native text, local image OCR recognition results, backup OCR recognition results, and image text recognition results; The candidate texts are scored based on the number of field hits, field completeness, field validity, and recognition confidence. Based on the comprehensive score, the candidate text with the highest comprehensive score that meets the field validity condition is selected as the main recognition result. Step S3: According to the patent document type template and field constraint rules, basic patent fields are extracted from the main recognition result and / or the candidate texts, and watermarks, page numbers, explanatory text, legal clause prompts, duplicate footers, and noisy text that does not match the field context are filtered out. The patent document type template includes patent certificate templates and patent certificate continuation pages. At least one of the following: a template for registration procedures, a template for a notice of patent termination; Step S4: Standardize the patent number and / or application number in the basic patent fields to generate abbreviated patent numbers for deduplication and matching, and match existing patent records according to a four-level matching rule, which includes: first matching is complete consistency of abbreviated patent numbers, second matching is complete consistency of application numbers, third matching is consistency of patent name, application date, and patentee, and fourth matching is that the similarity of patent name, customer name, patent type, and application date meets preset conditions; when a unique match is found... When recording a patent, the missing fields of the unique patent record are supplemented and associated with the original patent document; when no existing patent record is matched, a new patent record is generated; when multiple existing patent records meet the matching conditions, or when the identified fields conflict with existing fields, a record to be confirmed is generated without automatically overwriting the existing patent record; in step S5, a continuous patent year interval is generated based on the application date; when the date of issuance of the registration procedure notification is earlier than the preset policy switch date, the deadline for registration and payment is calculated as fifteen days plus two months from the date of issuance of the registration procedure notification; when the date of issuance of the registration procedure notification is earlier than the preset policy switch date, the deadline for registration and payment is calculated as fifteen days plus two months from the date of issuance of the registration procedure notification. No earlier than the preset policy switch date, the deadline for registration and payment is calculated by adding two months to the date of issuance of the registration procedure notification; the deadline for registration and payment is compared with the patent year range, and the patent year in which it falls is determined as the year of payment due for the authorized patent; in step S6, a patent annuity plan is generated based on the standardized patent records, the year of payment due for the authorized patent, and the patent year range. The patent annuity plan includes at least the patent payment year, the normal payment deadline, the identifier of the year of payment due for the authorized patent, and the relationship between the source field. The normal payment deadline is determined according to the expiration date of the corresponding patent year range.

2. The method for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents according to claim 1, characterized in that, In step S2, scoring the candidate text includes calculating a comprehensive score, which includes at least field hit weight, date validity score, number format validity score, template context matching score, and OCR confidence score. When the difference in comprehensive scores among multiple candidate texts is less than a preset threshold and there are conflicts in key fields, a set of fields to be confirmed is generated instead of directly determining a unique primary recognition result. The page orientation correction includes generating candidate images at multiple rotation angles for the same page, performing text recognition on each candidate image, and using the orientation of the candidate image with the highest number of field hits, field validity, or recognition confidence as the effective recognition orientation.

3. The method for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents according to claim 1, characterized in that, Each patent document type template is configured with field start keywords, field end markers, field format constraints, negative keywords, and cross-page duplicate text exclusion rules. Text that is in the same position, has the same or similar content, and is not located between the field start keyword and the field end marker in consecutive pages is identified as a cross-page duplicate footer and excluded. The negative keywords are used to exclude legal provisions, explanatory text, page numbers, watermarks, tabulation instructions, and text unrelated to the basic patent fields.

4. The method for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents according to claim 1, characterized in that, The standardization process for patent numbers and / or application numbers includes removing the ZL prefix, CN prefix, spaces, separators, Chinese punctuation, and English period marks, and converting all letters to uppercase to generate the abbreviated patent number. In the four-level matching results, if the first or second match matches multiple records, a suspected duplicate record to be confirmed is directly generated, and existing patent records are not automatically merged. The preset conditions include that the patent name similarity is not lower than a preset similarity threshold, the difference in application date does not exceed a preset number of days, and the customer name and patent type simultaneously meet the matching requirements.

5. The method for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents according to claim 1, characterized in that, The patent year is calculated based on the application date. The first year is defined as the same month and day from the application date to the following year, and subsequent patent year intervals are generated accordingly. When adding a month to the date, if there is no corresponding date in the target month, the last day of the target month is taken. The patent year into which the registration and payment deadline falls is the year for which fees should be paid in the year of authorization.

6. The method for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents according to claim 1, characterized in that, The patent annuity plan calculates the standard fee amount, the reduced fee amount, the late payment fee amount, and the total amount due based on the customer's annual fee reduction status, patent type, patent year, and co-owner information. When the customer's annual fee reduction status is unconfirmed or temporarily unconfirmed, the unreduced amount is displayed, and a "fee reduction status pending confirmation" prompt is generated. When there are two or more patent holders, the fee reduction rules corresponding to the co-owners are applied separately for that patent, without changing the fee reduction calculation results for other patents under the same customer's name. When there are manual adjustments to the annual fee plan, only the system-calculated amount is updated, without automatically overwriting the final amount due after manual confirmation. Late payment fees are calculated based on the standard annuity, not the reduced fee amount.

7. The method for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents according to claim 1, characterized in that, The annual fee status and risk level are updated based on the normal payment deadline, grace period deadline, recovery period deadline, payment status, termination notice, abandonment status, rejection status, or expiration status. The system records the entire operation log, including file identification, field extraction, numbering standardization, record matching, pending confirmation processing, annual fee calculation, risk refresh, and manual modification. The pending confirmation records support manual field addition, merging of existing records, ignoring, marking as non-patented, re-identifying, or saving of modified records.

8. A device for generating annual fee plans and providing risk alerts based on multi-source identification and structured governance of patent documents, characterized in that, include: The document preprocessing module is used to obtain the original patent documents and perform PDF native text extraction, page image conversion, and page orientation correction. The multi-source recognition module is used to generate candidate recognition text based on at least two recognition sources, and determine the main recognition result according to the number of field hits, field completeness, field legality and recognition confidence; the field extraction and noise filtering module is used to extract the basic patent fields and filter noisy text according to the patent document type template and field constraint rules. The standardization matching module is used to standardize patent numbers and / or application numbers and match existing patent records according to the four-level matching rules; the pending confirmation processing module is used to save candidate fields, conflicting fields, source files, and matching reasons when multiple existing patent records meet the matching conditions or when there is a conflict between the identification field and existing fields, and to prevent automatic overwriting of existing patent records; the authorization year determination module is used to generate a patent year range based on the application date, calculate the deadline for registration and payment, and determine the annual fee payable in the authorization year; the annual fee plan generation module is used to generate a patent annual fee plan based on the standardized patent records, the annual fee payable in the authorization year, and the patent year range.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1 to 7.

Images