Regulatory clause adaptation processing method and processing program

Abstract

The application discloses a kind of rule clause adaptive reference processing method and processing program, the processing method includes the following steps, S1 reads and identifies rules and regulations, S2 is parsed into tree structured clause data and is stored for rules and regulations, the structured clause data includes rules and regulations code, parent code, rules and regulations title and rules and regulations content, S3 sets review dataset, constructs the mapping relation of each review task in the review dataset with structured clause data, S4 analyzes keyword in review work form and matches review dataset, and the clause data corresponding to the review task in review dataset is associated with the corresponding matter in review work form.The application carries out tree structured analysis into corresponding clause for various rules, simultaneously, the content in review work form is analyzed and decomposed into review dataset, then each review task in review dataset is mapped with rule clause, so that the convenience of relevant rules and regulations reference is effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically relating to a method and procedure for handling adaptive citations of regulations and clauses. Background Technology

[0002] Currently, in traditional event compliance evaluation methods, the compliance evaluation of an event is often based on the company's internal system and past experience. If the event does not comply with the company's internal system, it is considered to have a compliance risk, and the compliance risk is classified into levels based on the impact of the historical occurrence of such compliance events.

[0003] However, traditional event compliance evaluation methods only consider the mismatch between internal company policies and actual events. When facing industry product audits, such as those for drones, which involve a wide variety of products and complex regulations, manual comparison is too inefficient. Specifically, personnel need to manually retrieve and visually match various regulations, significantly increasing time and labor costs. Moreover, there is no standard to measure the accuracy of manual comparisons, making them susceptible to subjective human factors, resulting in unstable matching results and generating much repetitive work. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a regulation processing system that automatically analyzes regulations and, by establishing mapping relationships, effectively improves the speed and accuracy of marking regulations in audit forms.

[0005] This invention is achieved through the following technical solution:

[0006] A method for handling adaptive citations of regulatory clauses includes the following steps:

[0007] S1 reads and identifies regulations and rules.

[0008] S2 parses the rules and regulations into tree-structured clause data and stores it. The structured clause data includes the rule code, parent code, rule title, and rule content.

[0009] S3 sets up the review dataset and constructs a mapping relationship between each review task and the structured clause data in the review dataset.

[0010] S4 analyzes keywords in the review worksheet and matches them with the review dataset, then associates the clause data corresponding to the review task in the review dataset with the corresponding items in the review worksheet.

[0011] As one preferred embodiment, the reading and identification of regulations includes the steps of periodically retrieving updated regulations from a regulations database or a pre-defined website, and specially marking data on any modifications or additions to the regulations.

[0012] As one preferred option, the method also includes a step of performing OCR recognition on non-text files.

[0013] As one preferred solution, the rules and regulations in S2 are parsed by identifying and analyzing the structure and hierarchical paragraphs of the rules based on the paragraph numbering information at the beginning of the paragraphs.

[0014] As one preferred approach, a mapping update step is also included.

[0015] Compare the information on the approved referenced clauses in the review worksheet with the related information in step S4.

[0016] Statistical analysis review tasks include updates, additions, deletions, or changes to corresponding mappings.

[0017] If the update rate of the approved reference clauses and related information for a review task exceeds a predetermined ratio, the review task will be redefined based on the work items in the review worksheet corresponding to the review task, and a new mapping relationship will be established.

[0018] If the number of additions, deletions, or changes to the terms for the same review task exceeds the predetermined number, the mapping relationship will be updated.

[0019] As one preferred option, it also includes the step of matching the review form with the review task in the review data or determining a new review task based on the semantic recognition of AI.

[0020] As one preferred option, the regulations in question are those applicable to unmanned aerial vehicles.

[0021] A processing procedure for implementing the adaptive reference processing method for the aforementioned regulatory clauses.

[0022] The advantages and beneficial effects of this invention are as follows:

[0023] The present invention provides a method for adaptively referencing regulations and clauses. This method analyzes various regulations into corresponding clauses using a tree structure, decomposes the content of the review work form into a review dataset, and then maps each review task in the review dataset to the regulations and clauses. This effectively improves the convenience of referencing relevant regulations and systems, and the relationship established based on the database has high robustness and reliability. Attached Figure Description

[0024] Figure 1 This is a screenshot of a portion of the regulatory document "Airworthiness Standards for Transport Category Aircraft [CCAR-25-R4]".

[0025] Figure 2 It is the parsed content obtained from processing regulatory documents.

[0026] Figure 3 It is formatted rule content.

[0027] Figure 4 This is a schematic diagram of the mapping relationship.

[0028] Figure 5 A schematic diagram illustrating the automatic matching process for creating a "Fuel System Inspection" review business set.

[0029] Figure 6 This is a schematic diagram showing the structure of some of the relevant structured clauses.

[0030] Figure 7 This is a diagram illustrating the correspondence between the final matching and review of the business set and the structured terms.

[0031] For those skilled in the art, other related figures can be obtained from the above figures without any creative effort. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0033] This invention provides a method for adaptively referencing regulatory clauses, primarily applicable to situations requiring numerous regulations during review. The review worksheet is customized with different review items based on actual product needs. Each review item has one or more corresponding regulatory clauses as review standards or institutional support. For example, the review of regulations applicable to unmanned aerial vehicles published by the Civil Aviation Administration of China requires reviewing each item in the review worksheet. However, the sheer number and frequent updates of regulations create significant inconvenience for the review process. To improve the efficiency of regulation processing and identification, and thus ensure rapid and accurate review, this invention's method for adaptively referencing regulatory clauses includes the following steps.

[0034] S1 reads and identifies rules and regulations, wherein the reading includes manual updating, automatic updating from the rules and regulations database, or automatic crawling and updating of rules and regulations by linking to the corresponding website, and also includes the step of OCR recognition of non-text files to ensure the timely updating of various rules and regulations;

[0035] S2 parses regulations and rules into tree-structured clause data and stores it. This structured clause data includes the regulation code, parent code, regulation title, and regulation content. By parsing and processing a regulation into a tree-structured data store and storing it in the database, the system can reference the clause content of specific regulatory documents. For example... Figure 1 As shown, taking the regulation document "Airworthiness Standards for Transport Category Aircraft [CCAR-25-R4]" as an example, after importing and analyzing it, the automatic identification of the regulation structure and hierarchical paragraphs is achieved based on the paragraph beginning number information. For example, Section A and Section B are Level 1; Article XX.X is Level 2; (a), (b)... are Level 3; (1) is Level 4; (i), (ii)... are Level 5, and so on. The identified content is stored in the service memory, including four columns: regulation code, parent code, regulation title, and regulation content. For example, Figure 2 As shown. Then, based on the parsed content, the system automatically generates batch SQL statements for `INSERT INTO` and inserts them into the database. After insertion, the formatted regulations can be viewed in the system; for example... Figure 3 As shown.

[0036] S3 sets up a review dataset and constructs a mapping relationship between review tasks and structured clause data in the review dataset. First, the review dataset is reviewed based on the review work resume. This review dataset is mainly a collection of review types. Taking UAV review as an example, the review dataset includes review tasks that reviewers need to pay attention to during actual review, such as "flight area restrictions", "altimeter adjustment procedure" and "operation within visual range". After constructing the review dataset, a mapping relationship is established between the review tasks and structured clause data. Here, one review task can correspond to multiple clause data, or multiple clause data can correspond to multiple review tasks.

[0037] S4 analyzes keywords in the review worksheets and matches them with the review dataset, associating the clause data corresponding to the review tasks in the review dataset with the corresponding items in the review worksheets. Because there are numerous regulations and rules, proactively matching them allows for efficient and rapid access to this information, speeding up the review process. The associated clause data can be restricted as needed, specifying the name of the regulation, the specific clauses involved, and their content.

[0038] Specifically, when the review team conducts a review of a project, they create action items for the inspection items, i.e., create a review dataset, such as inspections of the fuel system, engine, propeller, noise, etc. Figure 5-7 As shown, taking fuel system inspection as an example, after creating an action item, the system automatically associates the corresponding structured clauses based on the action item, i.e., the keywords of the inspection task in the inspection dataset, using semantic recognition and the results of large model training. Since there are many clauses, manual selection is inefficient and prone to errors. This system, however, can quickly and accurately associate all clauses related to this content from multiple regulations, greatly improving efficiency and accuracy.

[0039] Furthermore, to adapt to the update of the system, the read and recognition regulations include the steps of regularly obtaining updated regulations from the regulation library or a predetermined website and specially marking the data of modified or newly added clauses. The special marking can be in the form of bold, color, or underline, etc. After a new update or a predetermined time, such as 3 months or 1 year later, the special marking is automatically cancelled. The special reminder can effectively promote the implementation and understanding of the new system.

[0040] Because the content of each review work form may also be different or have different implementation intensities, the method for adaptively referencing regulation clauses of the present invention further includes a mapping relationship update step.

[0041] Compare the approved reference clause information in the review work form with the associated information in step S4.

[0042] Statistically analyze the update of the review task, the addition, deletion, or change of the corresponding mapping. Among them,

[0043] If the update rate of the approved reference clause of the review task and the associated information exceeds a predetermined ratio, such as 70%, redefine the review task based on the work items in the review work form corresponding to the review task and establish a new mapping relationship.

[0044] If the addition, deletion, or change of the clauses for the same review task exceeds a predetermined number of times, update the mapping relationship.

[0045] Among them, when conducting a review, the system automatically brings out the formatted regulation clause content according to the previously set mapping relationship. At the same time, the reviewer can delete or re - formulate new referenced regulation clauses on this basis. The approved reference clause in the review work form is the actual reference clause after the review is completed. The data in the review work form after actual use is used to update and modify the regulations recommended by the above - mentioned method. Among them, if the change is too large, the review task is identified as the content of a new review data set, and a new mapping relationship is established, realizing the supplementary iteration and improvement of the review data set. If the same review task has been done several times, such as deleting the same or several associated clause data more than three times, update the mapping relationship of the review task. This adjustment can be changed based on the actual review trend, etc. After statistical analysis, it is adjusted and iterated into the review data set, effectively improving the timeliness of this processing method and ensuring the best policy state.

[0046] In response to the above-mentioned update system, a neural network algorithm can also be used for updates. After completing the initial data sample library training and neural network construction, statistical analysis is performed to verify the review task update of the reference clause information and the related information in step S4, as well as the addition, deletion or modification of the corresponding mapping. The statistical analysis result is used as a correction value to expand the sample and then merged with the original sample part as new sample data for the neural network model to iterate, which can effectively improve the accuracy of the above mapping.

[0047] To improve response speed, the process also includes steps for matching review forms with review tasks or identifying new review tasks based on AI's semantic recognition. Since review forms are intended for reviewers and their wording may vary, semantic understanding and stemming can be performed on the paragraph text of the review work forms before mapping regulations and clauses. The steps are as follows:

[0048] This study combines Natural Language Processing (NLP) artificial intelligence techniques to analyze paragraph syntax. Specifically, it uses the dependency relationships between words to represent the syntactic structure of individual words and employs a tree-like structure to represent the overall sentence structure. It also combines subject-verb structure analysis and modifier removal methods for stem extraction. Syntactic structure information includes subject-verb, verb-object, and attributive-head relationships; the overall sentence structure includes subject-verb-object and attributive-adverbial-complement relationships. The paragraph classification model is trained using deep learning (existing mature deep learning technologies) on large datasets to automatically learn deep semantic and word order features, achieving 95% coverage of the censorship dataset for classification tasks. Paragraph parsing includes three main functions: word segmentation, dependency parsing, and named entity recognition. It identifies basic words in the text string (word segmentation), reorganizes and tags these words to obtain their parts of speech. Dependency parsing is used to identify named entities, allowing for the identification of entities with specific meanings in the review worksheet. Key information extraction: Based on named entities, the main body of the paragraph syntax is identified, and the corresponding structural information is identified using sentence dependency relationships, thereby extracting information from different types of regulations and clauses. Through systematic analysis of a large number of regulations, the relationship between the review worksheet and its clauses is ultimately summarized and established.

[0049] In summary, the processing method of this invention enables the rule formatting system to perform batch conversion of PDF files such as management procedures and rules issued by relevant national departments. Then, according to the numbering information in the rule content, it realizes the automatic identification of the rule structure, parses it into a tree structure, and stores the structured data in the database. When used for administrative approval review, it can quickly reference the clauses of specific rule documents.

[0050] Especially in the field of drones, with the rise of the low-altitude economy, various specifications and models of drones and aircraft will emerge, requiring a large amount of flight airworthiness review work. By structuring and analyzing regulations and labels, appropriate regulations can be automatically matched according to different aircraft models and application scenarios, such as flight tests, aircraft power tests, and aircraft safety tests. The structured data can cross different regulations and directly generate matching checklists, i.e., review work forms. At the same time, by combining big data processing capabilities and AI analysis capabilities, the recommendation and feedback optimization logic and label marking, as well as the automatic interpretation of application scenarios, can analyze the characteristics of different aircraft models and provide data analysis capabilities on the demand side. This dual technological approach effectively improves the overall user experience.

[0051] The present invention has been described above by way of example. It should be noted that any simple modifications, alterations or other equivalent substitutions that can be made by those skilled in the art without creative effort without departing from the core of the present invention fall within the protection scope of the present invention.

Claims

1. A method of regulatory clause adaptive referencing, characterized by, It comprises the following steps, S1 reading and identifying the regulatory system, S2 parsing the regulatory system into tree-structured clause data and storing, the structured clause data comprising regulatory code, parent code, regulatory title and regulatory content, S3 setting the review data set and constructing the mapping relationship between each review task in the review data set and the structured clause data, S4 analyzing the keywords in the review work form and matching the review data set, and associating the clause data corresponding to the review task in the review data set with the corresponding matter in the review work form; According to the parsed content in the regulatory content, automatically generate the SQL statement of Insert into in batches, and insert into the database; It also comprises: matching the review task in the review form and the review data or determining a new review task through semantic recognition; analyzing the syntax of the paragraph through natural language processing, that is, using the dependency relationship between words in the sentence to represent the syntactic structure information of the words and using a tree structure to represent the structure of the whole sentence; based on subject-predicate structure analysis and de-modification, extracting the stem; wherein the syntactic structure information comprises: subject-predicate, verb-object, and modifier-object structure relationship; the structure of the whole sentence includes: subject-predicate-object, modifier-state complement; the paragraph classification model is obtained through deep learning training based on big data, which is used to identify semantic features and sequence features; by identifying the basic words in the text string, the words are reorganized and labeled to obtain the part of speech of the words; the named entity is identified by using the dependency syntax, so as to identify the entities with specific meanings in the review work form; based on the named entity, the subject part in the paragraph syntax is identified, and the corresponding structure information is identified by using the dependency relationship of the sentence, so as to extract information of different types of regulatory clauses.

2. The regulatory provision adaptive referencing method of claim 1, wherein, The reading and identifying the regulatory system comprises the step of regularly obtaining updated regulations from the regulatory system library or the predetermined website, and specially marking the modified or added new clause data.

3. The regulatory provision adaptive referencing method of claim 1, wherein, It also comprises the step of ocr recognizing non-text files.

4. The regulatory provision adaptive referencing method of claim 1, wherein, In S2, the regulatory system is parsed according to the paragraph beginning number information in the regulation to identify and analyze the regulation structure and hierarchical paragraphs.

5. The regulatory provision adaptive referencing method of claim 1, wherein, It also comprises a mapping relationship updating step, Comparing the approved reference clause information in the review work form with the associated information in step S4, Statistical analysis of review task update, corresponding mapping addition, deletion or change, wherein, If the approved reference clause of the review task and the update rate of the associated information exceed the predetermined proportion, redefine the review task in the work item according to the review work form corresponding to the review task and establish a new mapping relationship; If the clause addition, deletion or change of the same review task exceeds the predetermined number of times, update the mapping relationship.

6. The regulatory provision adaptive referencing method of claim 5, wherein, The regulatory system is a regulation applicable to unmanned aerial vehicles.

7. A processing program for performing the regulatory clause adaptability reference processing method according to any one of claims 1-6.

Images