An aviation enterprise standardization management system

Abstract

The application relates to the technical field of enterprise management, and specifically discloses an aviation enterprise standardized management system, which, when receiving an approval request, uses natural language processing technology to analyze the approval request to extract key information such as initiator information, request type, and approval content details, then carries out semantic embedding coding based on word granularity and local context semantic association optimization on the approval request analysis result to realize deep understanding of the approval request, and intelligently matches a suitable first-level approver type based on the context semantic information of the approval request analysis result. In this way, the approval path can be dynamically adjusted according to the specific content and background of the approval request, and the most suitable approver can be selected, so that various approval requests can be flexibly handled, and the approval efficiency and decision quality are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of enterprise management, and more specifically, to an aviation enterprise standardized management system. BACKGROUND

[0002] Aviation enterprise management refers to the management work carried out for the operation and development of an airline, involving flight scheduling, aircraft maintenance, route development, and other aspects, to ensure the realization of enterprise goals and the improvement of customer satisfaction. In the daily operation of an aviation enterprise, standardized enterprise management is crucial for ensuring efficient, safe, and compliant operations. As the complexity and scale of business continue to expand, traditional management methods have become inadequate to meet the needs of modern aviation enterprises in rapidly changing market environments. To address this challenge, many enterprises have begun to adopt digital means to optimize their internal processes, improve decision-making efficiency, and ensure coordination and consistency between departments.

[0003] For example, the invention patent with publication number CN115271294A proposes an enterprise standardized management system that creates standardized departments and positions by loading a standardized organizational structure tree and using parent ID to associate upper and lower levels. At the same time, a similar architecture tree is established in the original organizational structure module, and by corresponding the nodes in the standardized organizational structure tree with the nodes in the original organizational structure, changes in the standardized template or new projects are automatically updated to the original organizational structure module, maintaining consistency in departments and positions.

[0004] In the prior art, in terms of approval processes, standardized positions are directly and uniformly configured, and the project manager under the project where the approval request initiator is located is automatically captured as the first-level approver to meet the requirements of standardized management. This approval process management mainly relies on pre-set fixed approval paths. For example, "flight dispatcher submits -> maintenance engineer reviews -> aviation safety director approves".

[0005] However, in practical applications, this fixed approval path has some limitations. First, the fixed approval path lacks flexibility and is difficult to adapt to the specific circumstances and urgency of different approval requests. For example, some urgent maintenance requests may need to skip the regular review steps and be directly approved by higher-level decision-makers to minimize the impact on flight operations. However, under the existing fixed approval path, such flexible adjustment is often difficult to achieve, resulting in low efficiency of approval. Second, the fixed approval path is difficult to fully consider the differences between approval requests. Different approval requests may involve different professional fields, risk levels, or urgency, requiring approval persons with different backgrounds and professional skills to evaluate. However, under the fixed approval path, the selection of approval persons is often based on job level rather than the specific content of the approval request, which may lead to uneven quality of approval decisions, and even in some cases, approval errors.

[0006] Therefore, an optimized aviation enterprise standardized management system is expected. SUMMARY

[0007] To solve the above technical problems, the present application is proposed. The embodiments of the present application provide an aviation enterprise standardized management system, which, when receiving an approval request, uses natural language processing technology to analyze the approval request to extract key information such as initiator information, request type, and approval content details. Then, through semantic embedding coding based on word granularity and local context semantic association optimization of the approval request analysis result, the depth understanding of the approval request is realized, so as to intelligently match the appropriate first-level approver type based on the context semantic information of the approval request analysis result. In this way, the approval path can be dynamically adjusted according to the specific content and background of the approval request, and the most suitable approver can be selected, so as to flexibly cope with various approval requests and improve the efficiency and quality of decision-making.

[0008] Accordingly, according to one aspect of the present application, an aviation enterprise standardized management system is provided, which comprises: a standardized module for loading a standardized organizational structure tree, the organizational structures in the standardized organizational structure tree being associated in upper and lower levels by parent ID; an original organizational structure module for loading a company organizational structure tree, the organizational structures in the company organizational structure tree being associated in upper and lower levels by the parent ID, and when a new organizational structure node is created at the company level or the branch level, adding lower department and post data according to the preset node in the standardized module; an association module for corresponding the nodes in the standardized module to the nodes in the original organizational structure module, realizing the association operation in the authorization, approval, modification, and review processes in each level through the association relationship between the standardized module and the original organizational structure module.

[0009] In the aviation enterprise standardized management system, the correlation module comprises:

[0010] The approval request analysis unit is configured to analyze the received approval request to obtain an approval request analysis result.

[0011] The semantic embedding coding unit is configured to perform word granularity semantic embedding coding on the approval request analysis result to obtain a sequence of approval request word granularity semantic embedding coding vectors.

[0012] The medium granularity semantic correlation optimization unit is configured to perform medium granularity semantic correlation constraint optimization on the sequence of approval request word granularity semantic embedding coding vectors to obtain a sequence of medium granularity semantic reinforced approval request word granularity semantic embedding coding vectors.

[0013] The approver matching unit is configured to determine the type of the first-level approver based on the context semantic features of the sequence of medium granularity semantic reinforced approval request word granularity semantic embedding coding vectors.

[0014] Compared with the prior art, the aviation enterprise standardized management system provided by the application can use natural language processing technology to analyze the approval request when receiving the approval request, extract key information such as initiator information, request type, and approval content details, then perform word granularity-based semantic embedding coding and local context semantic correlation optimization on the approval request analysis result to realize deep understanding of the approval request, and thus intelligently match the appropriate first-level approver type based on the context semantic information of the approval request analysis result. In this way, the approval path can be dynamically adjusted according to the specific content and background of the approval request, and the most suitable approver can be selected, so as to flexibly cope with various approval requests and improve the approval efficiency and decision quality. BRIEF DESCRIPTION OF DRAWINGS

[0015] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description of embodiments of the present application, taken in conjunction with the accompanying drawings. The drawings provided in the specification and the embodiments of the present application together with the detailed description serve to provide a thorough understanding of the application. However, the application should not be construed as being limited to the embodiments set forth herein. In the drawings, like reference numerals refer to like elements or steps throughout.

[0016] Figure 1 FIG. 1 is a block diagram of an aviation enterprise standardized management system according to an embodiment of the present application.

[0017] Figure 2 FIG. 1 is a data flow diagram of an aviation enterprise standardized management system according to an embodiment of the present application.

[0018] Figure 3A block diagram of a medium-granularity semantic correlation optimization unit in the aviation enterprise standardized management system according to an embodiment of the present application.

[0019] Figure 4 A block diagram of an approval person matching unit in the aviation enterprise standardized management system according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] As shown in the present application and claims, unless the context clearly indicates otherwise, the words "one", "an", "a", and / or "the" do not mean "only one", "single" or "just one", but can include a plurality or "one or more" unless the context clearly indicates otherwise. Generally, the terms "comprise", "comprising", "include", "including" and the like indicate the inclusion of a stated step or element but not to the exclusion of any other steps or elements that are not specifically stated.

[0021] Although the present application makes various references to certain modules in the system according to embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server. The modules are merely illustrative, and different aspects of the system and method can use different modules.

[0022] Flowcharts are used in the present application to illustrate the operations performed by the system according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in the exact order. Instead, various steps can be processed in reverse order or at the same time, as desired. Other operations can also be added to or removed from these processes, or one or more steps can be removed from these processes.

[0023] In the following, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and are not all the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein.

[0024] It is worth noting that in the present application, all actions of obtaining data are carried out in compliance with the corresponding data protection regulations and policies of the country where the device is located, and with the authorization given by the owner of the corresponding device.

[0025] As mentioned in the background section, patent CN115271294A proposes an enterprise standardization management system, which includes: a standardization module for loading a standardized organizational structure tree, the organizational structure in the standardized organizational structure tree is associated by parent ID up and down level; an original organizational structure module for loading a company organizational structure tree, the organizational structure in the company organizational structure tree is associated by the parent ID up and down level, and when a new organizational structure node is created by selecting a company level or a branch level, lower-level department and post data are added according to the preset node in the standardization module; an association module for corresponding the node in the standardization module to the node in the original organizational structure module, through the association relationship between the standardization module and the original organizational structure module, the association operation in the authorization, approval, modification, and review process in each level is realized.

[0026] Firstly, loading the standardized organizational structure tree can provide a general and reusable organizational structure template for enterprises. This template is associated by parent ID up and down level, which ensures that the relationship between each department and post is clear and easy to maintain. Specifically, the system defines the highest level department or functional area of the enterprise as the top node, such as "head office", "finance department", "operation department", etc. These nodes represent the main business areas within the enterprise. Then, the relationship between levels is established through the parent ID, and each child node points to its superior node, forming a complete organizational structure tree. This hierarchical association not only simplifies the representation of organizational structure, but also facilitates subsequent queries and operations. For example, when looking for the department where a certain employee works, the parent ID can be traced up layer by layer to quickly locate it. In addition, the standardized organizational structure tree also includes detailed post information, such as job description, responsibility range, required skills, etc., which helps ensure consistency and accuracy when creating new nodes.

[0027] At the same time, loading the company organizational structure tree is to map the actual enterprise organizational structure into the system. Similarly, this architecture tree is also associated by parent ID up and down level, but it reflects the specific department setup and post distribution situation within the current enterprise. When creating a new organizational structure node, whether selecting a company level or a branch level, the system will add the corresponding lower-level department and post data according to the preset node in the standardization module. This means that whenever a new department or post is added, the system will automatically reference the corresponding node in the standardized organizational structure tree and fill in the relevant information according to the preset rules, such as department name, job description, required skills, etc. The benefit of this is to ensure the consistency of new nodes with the existing architecture, while also reducing the possibility of manual input errors.

[0028] In the process of creating new organizational structure nodes, if there are special requirements or specific industry demands, appropriate adjustments can be made based on standardized templates. For example, some airlines may need to establish additional departments specifically responsible for security checks and technical support. In this case, these specific nodes can be added outside the standardized templates. In this way, the advantages of standardized modules are maintained while being flexible to different business scenarios. At the same time, the system also considers the size and development stage of the enterprise. Large enterprises may require more detailed hierarchical structures, while start-ups can adopt more concise architectural designs. This flexibility ensures that enterprises at any stage of development can have an organizational structure that meets their needs.

[0029] In addition, to ensure the real-time updating and effectiveness of the organizational structure, the system needs to have a powerful change management function. Whenever there are significant changes within the enterprise, such as merging departments or adjusting job responsibilities, the system should promptly reflect these changes to ensure that all relevant parties can access the latest information. Change management not only involves modifying the organizational structure tree itself, but also includes synchronously adjusting related approval processes, permission settings, etc. For example, if a department is dissolved, the approval path related to it also needs to be re-evaluated and optimized; if an employee is promoted to a higher position, their access permissions should be adjusted accordingly. Through such mechanisms, the system can always maintain high consistency with the actual situation of the enterprise, avoiding decision-making errors or inefficiencies due to outdated information.

[0030] In the above-mentioned enterprise standardized management system approval process, by configuring standardized positions, the first-level approver is automatically captured from the project responsible person under the project according to the approval request initiator, to meet the requirements of standardized management. However, this approval process management mainly relies on pre-set fixed approval paths, which has some limitations. First, the fixed approval process lacks flexibility and is difficult to adapt to the specific circumstances and urgency of different approval requests. For example, some urgent maintenance requests may need to skip the regular review steps and be quickly approved by higher-level decision-makers to reduce the impact on flight operations. However, under the fixed approval process, such flexible adjustments are often difficult to achieve, resulting in reduced approval efficiency. Second, the fixed approval process is difficult to fully consider the differences between approval requests. Different approval requests may involve different professional fields, risk levels, or urgency, requiring different backgrounds and professional skills of the approvers for evaluation. However, under the fixed approval process, the selection of approvers is often based on job titles rather than the specific content of the approval request, which may lead to inconsistent quality of approval decisions and even approval errors in some cases.

[0031] To solve the above technical problems, the application provides an optimized aviation enterprise standardized management system. When receiving an approval request, the system uses natural language processing technology to analyze the approval request to extract key information such as initiator information, request type, and approval content details. Then, the system performs semantic embedding coding and local context semantic association optimization based on the analysis result of the approval request to achieve deep understanding of the approval request, and intelligently matches a suitable first-level approver type based on the context semantic information of the analysis result of the approval request. In this way, the approval path can be dynamically adjusted according to the specific content and background of the approval request, and the most suitable approver can be selected, so that various approval requests can be flexibly handled, and the approval efficiency and decision quality can be improved.

[0032] Figure 1 A block diagram of an aviation enterprise standardized management system according to an embodiment of the application. Figure 2 A data flow diagram of an aviation enterprise standardized management system according to an embodiment of the application. As shown in Figure 1 and Figure 2 The aviation enterprise standardized management system 100 includes an approval request analysis unit 110 configured to analyze a received approval request to obtain an approval request analysis result; a semantic embedding coding unit 120 configured to perform word granularity semantic embedding coding on the approval request analysis result to obtain a sequence of approval request word granularity semantic embedding coding vectors; a medium granularity semantic association optimization unit 130 configured to perform medium granularity semantic association constraint optimization on the sequence of approval request word granularity semantic embedding coding vectors to obtain a sequence of medium granularity semantic reinforced approval request word granularity semantic embedding coding vectors; and an approver matching unit 140 configured to determine a first-level approver type based on context semantic features of the sequence of medium granularity semantic reinforced approval request word granularity semantic embedding coding vectors.

[0033] In the above-mentioned aviation enterprise standardized management system, the approval request analysis unit 110 is configured to analyze the received approval request to obtain an approval request analysis result. It should be understood that, considering that the approval request usually contains information in various formats such as forms and text descriptions, in order to accurately extract the key information in the approval request and guide the specific approval process, the present application first uses natural language processing (NLP) technology to structurally analyze the text information in the approval request to extract the initiator information, request type, approval content details and other key information, thereby obtaining the approval request analysis result. In specific implementation, a rule-based method can be used, and a series of syntax rules related to the format of the approval request are defined in advance, for example, it is specified that the initiator information is located at the beginning of the request text and follows the format of “initiator: [name / department name]”; the request type is matched through a keyword list, such as “purchase approval” and “project approval”; the approval content details are extracted by scanning the text paragraph or sentence describing the specific matter, and the corresponding information is recognized and extracted according to the rules, and is arranged into a structured data format such as JSON or XML to form the approval request analysis result.

[0034] Specifically, in order to accurately extract the key information in the approval request and guide the specific approval process, considering that the approval request usually contains information in various formats such as forms and text descriptions, the present application first uses natural language processing (NLP) technology to structurally analyze the received approval request. The application of NLP technology aims to ensure that regardless of the source and form of information, the initiator information, request type, approval content details and other key information can be effectively analyzed and extracted, thereby obtaining a comprehensive and accurate approval request analysis result.

[0035] Firstly, the system needs to preprocess the approval request. The approval request can come from various business systems within the aviation enterprise, such as flight scheduling system, maintenance management system or financial management system, and the users of these systems include but are not limited to flight dispatchers, maintenance engineers, aviation safety directors, project managers, etc. Each approval request carries information about the decision needed, which is crucial for determining the approval path. The preprocessing stage includes removing irrelevant characters, punctuation marks, and converting the text into a unified format, such as all lowercase or standardized abbreviations. In addition, for requests in a multilingual environment, language detection and translation are also needed to ensure that all requests can be correctly analyzed.

[0036] The next stage is text segmentation and tokenization. Typically, an NLP tool is used to split the text into words or phrases, known as "word granularity." Then, part-of-speech tagging (POS tagging) is applied to add grammatical tags to each word, such as noun, verb, etc. This helps understand the sentence structure and lays the foundation for the next step of information extraction. Named entity recognition (NER) techniques can automatically identify specific types of entities from the text, such as names of people, places, dates, amounts, and other key information. This is crucial for determining the relevant parties, timeframes, and other important details of the approval request. For example, if the approval request mentions "Zhang San" and "January 5, 2025," it can be identified that this is the initiator's name and the estimated completion date.

[0037] The information extraction (IE) stage utilizes relationship extraction and event extraction algorithms to further mine the implicit information from the text. For example, finding out who initiated the request, which departments or projects are involved, what are the specific requirements, and so on. These pieces of information are critical factors in deciding the approval path. In practical operations, rule-based methods are widely used for approval request parsing. This method relies on a series of syntax rules related to the format of the approval request defined in advance. For example, it is stipulated that the initiator information is located in the beginning part of the request text and follows the format of "Initiator: [Name / Department Name]"; the request type is matched through a keyword list, such as "purchase approval," "project approval," etc. The extraction of approval content details involves in-depth analysis of the paragraphs or sentences in the text that describe specific matters. By scanning the text word by word and sentence by sentence, the system identifies and extracts the corresponding information according to the pre-set rules. For example, if the text contains a sentence such as "Need to purchase the following equipment," the system may consider it as the content of the purchase approval and further look for the specific equipment list and other related details.

[0038] After the identification and extraction of information are completed, the next task is to organize these information into a structured data format, such as JSON or XML. This step not only helps maintain data consistency and readability but also facilitates subsequent processing and storage. Using NLP techniques is not limited to simple keyword matching but also includes more advanced functions, such as named entity recognition (NER), relationship extraction, event extraction, etc. These functions can help the system better understand the overall semantics of the text rather than just focusing on the meaning of individual words. For example, through NER, it can identify the names of people, places, dates, and other specific types of entities mentioned in the text; while relationship extraction and event extraction can capture the connections between different entities and the events that occur, which are particularly important for understanding complex approval requests.

[0039] In the aviation enterprise standardized management system, the semantic embedding coding unit 120 is configured to perform word granularity semantic embedding coding on the approval request analysis result to obtain a sequence of approval request word granularity semantic embedding coding vectors. In one specific example of the present application, the semantic embedding coding unit 120 is configured to input the approval request analysis result after word segmentation processing into a word semantic embedding encoder based on a Bert model to obtain the sequence of approval request word granularity semantic embedding coding vectors. It should be understood that, since the text information of the approval request often contains a large number of professional terms and complex expressions, in order to accurately capture the semantic features of the approval request, the present application further adopts the Bert model which performs well in the field of text semantic understanding to perform semantic coding processing on the approval request analysis result, so as to convert each word in the approval request analysis result into a high-dimensional semantic vector representation. Specifically, first, the existing Chinese word segmentation tool, such as Jieba word segmentation, is used to perform word segmentation operation on the approval request analysis result, so as to split the continuous text into independent words. Then, each word after word segmentation processing is input into the pre-trained Bert model. The Bert model is trained on a large amount of corpus, and for each input word, it can output a corresponding high-dimensional semantic vector based on the rich language knowledge and semantic representation learned in the pre-training process, so as to represent the semantic features of the word. In this way, the text information of the approval request is converted into the sequence of approval request word granularity semantic embedding coding vectors, thereby providing accurate semantic basis for subsequent approval personnel matching.

[0040] In the aviation enterprise standardized management system, the medium granularity semantic association optimization unit 130 is configured to perform medium granularity semantic association constraint optimization on the sequence of approval request word granularity semantic embedding coding vectors to obtain a sequence of medium granularity semantic reinforced approval request word granularity semantic embedding coding vectors. It should be understood that, considering that each word in the text information of the approval request analysis result may have different semantic meanings in different context environments, in order to more accurately understand the complete semantics of the approval request, the present application proposes a medium granularity semantic association constraint optimization method, which identifies the local context window associated with each approval request word granularity semantic embedding coding vector by further performing syntax analysis on the sequence of approval request word granularity semantic embedding coding vectors, and performs semantic reinforcement processing on each approval request word granularity semantic embedding coding vector based on the context in the local window, so as to enhance the accuracy and richness of the semantic representation of the approval request. Among them, Figure 3 The block diagram of the medium granularity semantic association optimization unit in the aviation enterprise standardized management system according to the embodiment of the present application. As shown in FIG. 6, the medium granularity semantic association optimization unit 130 is configured to include a local context window identification unit 131 and a semantic reinforcement unit 132. Figure 3As shown, the medium-granularity semantic association optimization unit 130 includes: a grammatical depth analysis subunit 131, which is used to perform grammatical depth analysis on each approval request word granularity semantic embedding coding vector in the sequence of the approval request word granularity semantic embedding coding vector to obtain a sequence distribution of approval request word granularity grammatical depth implicit representation values; a semantic association window search subunit 132, which is used to determine the medium-granularity semantic association window of each approval request word granularity semantic embedding coding vector in the sequence of the approval request word granularity semantic embedding coding vector based on the sequence distribution of the approval request word granularity grammatical depth implicit representation values; a contextual semantic association enhancement subunit 133, which is used to perform medium-granularity contextual semantic association enhancement on each approval request word granularity semantic embedding coding vector based on the medium-granularity semantic association window of each approval request word granularity semantic embedding coding vector to obtain a sequence of the medium-granularity semantically enhanced approval request word granularity semantic embedding coding vectors.

[0041] Specifically, in a specific example of the present application, the syntax depth analysis subunit 131 is used to: first, map each approval request word granularity semantic embedding coding vector in the sequence of approval request word granularity semantic embedding coding vectors to a hyperbolic space to obtain a sequence of hyperbolic space approval request word granularity semantic embedding coding vectors, which is expressed as follows:

[0042] T={v1,v2,...,v n}

[0043] h i =f hyp (v i )=W1v i W2

[0044] Where T represents the sequence of the granular semantic embedding encoding vectors of the approval request word, v1, v2, v i 、v n Respectively represent the first, second, i-th, and n-th approval request word granularity semantic embedding coding vectors in the sequence of the approval request word granularity semantic embedding coding vectors, f hyp (·) is the hyperbolic space mapping function, W1 and Q2 represent the first mapping weight matrix and the second mapping weight matrix respectively, h i Indicates the v i The corresponding hyperbolic space approval request word granularity semantic embedding encoding vector.

[0045] That is, when processing text data, the present application takes into account the complex hierarchical relationships between words, including but not limited to hyponymy, synonymy and antonymy, etc. Therefore, in order to better represent the hierarchical structure of the approval request parsing result, the present application adopts an advanced mapping strategy to further map each approval request word granularity semantic embedding code vector into hyperbolic space. This method makes use of the unique properties of hyperbolic geometry and can more effectively express the hierarchical relationships between data, thus better revealing the semantic hierarchy and relevance between words in the approval request parsing result.

[0046] Then, the syntactic depth implicit representation value of each hyperbolic space approval request word granularity semantic embedding code vector in the sequence of hyperbolic space approval request word granularity semantic embedding code vectors is calculated to obtain the sequence distribution of the approval request word granularity syntactic depth implicit representation value, which is represented by the formula:

[0047]

[0048] where ‖·‖ 2 represents the square of the norm of the vector, log2(·) represents the logarithmic function with base 2, f(·) represents the syntactic depth metric function, e i represents the v i corresponding approval request word granularity syntactic depth implicit representation value.

[0049] That is, by calculating the syntactic depth implicit representation value of each approval request word granularity semantic embedding code vector mapped into hyperbolic space, the position of these words in the syntactic tree structure is evaluated, thus analyzing their grammatical roles in specific context. This not only reveals the importance of each word in sentence construction, but also demonstrates the relationship between each word and other language elements. For example, certain words or phrases may occupy the core position of the sentence, conveying the main semantic content, while other words may play a modifying or supplementary role, having relatively less impact on the overall semantics. In this way, the actual meaning of the words can be better understood according to their grammatical characteristics, strengthening the understanding of the text content and helping to improve the accuracy of information interpretation in the approval process. Using this grammar depth-based analysis method, the approval request processing process can be further optimized to ensure that key information is given proper attention.

[0050] Specifically, in one specific example of the present application, the semantic association window finding subunit 132 is configured to: first, extract the syntactic depth implicit representation value of the first approval request word granularity semantic embedding encoding vector in the sequence of approval request word granularity semantic embedding encoding vectors as the starting position of the medium-granularity semantic association window; then, along the sequence distribution of the syntactic depth implicit representation values of the approval request word granularity, find the approval request word granularity syntactic depth implicit representation value closest to the syntactic depth implicit representation value of the first approval request word granularity semantic embedding encoding vector as the termination position of the medium-granularity semantic association window; finally, based on the starting position of the medium-granularity semantic association window and the termination position of the medium-granularity semantic association window, determine the medium-granularity semantic association window of the first approval request word granularity semantic embedding encoding vector from the sequence of approval request word granularity semantic embedding encoding vectors, which can be expressed by the formula:

[0051] w i～j = arg min j {|e i -e k |}

[0052] wherein e k is the syntactic depth implicit representation value of the kth approval request word granularity semantic embedding encoding vector in the sequence of approval request word granularity semantic embedding encoding vectors, |·| represents taking the absolute value, arg min represents taking the index corresponding to the minimum value, w i～j represents the medium-granularity semantic association window of the v i , i is the position index of the v i , and j is the position index of the approval request word granularity syntactic depth implicit representation value closest to the syntactic depth implicit representation value of the v i .

[0053] It can be understood that by reasonably setting the semantic association window, the semantic features of the words can be strengthened while ensuring that the key context information is not lost and the interference caused by the introduction of too much irrelevant information is avoided. This helps to accurately capture the appropriate scale of semantic relationship between the words and the surrounding vocabulary in a specific context, thereby enhancing the understanding ability of the model for complex semantic structures, while simplifying the calculation process.

[0054] Specifically, in one specific example of the present application, the contextual semantic association reinforcement subunit 133 is configured to: first, calculate the semantic association scores of each of the first approval request word granularity semantic embedding encoding vectors in the medium granularity semantic association window of the first approval request word granularity semantic embedding encoding vector to obtain a sequence of first approval request word granularity semantic association scores; then, normalize the sequence of first approval request word granularity semantic association scores to obtain a sequence of normalized first approval request word granularity semantic association scores; and finally, take the sequence of normalized first approval request word granularity semantic association scores as a sequence of weights, and calculate the position-weighted sum between each of the first approval request word granularity semantic embedding encoding vectors in the medium granularity semantic association window to obtain a medium granularity semantic reinforcement approval request word granularity semantic embedding encoding vector corresponding to the first approval request word granularity semantic embedding encoding vector, which can be expressed by the formula:

[0055]

[0056] wherein, represents a matrix multiplication operation, v k is the kth approval request word granularity semantic embedding encoding vector in the sequence of approval request word granularity semantic embedding encoding vectors, v t is a score reference vector, w t is a score reference weight matrix, b t is a score bias vector, exp(·) is a natural exponential function, v i ’ represents the v i corresponding medium granularity semantic reinforcement approval request word granularity semantic embedding encoding vector.

[0057] Here, after determining the medium granularity semantic association window, the contextual information within the window is further utilized to optimize the semantic representation of each word, thereby realizing medium granularity contextual semantic reinforcement of each approval request word granularity semantic embedding encoding vector. That is, based on the semantic and grammatical relationship between each word and other words within a specific window in the approval request analysis result, the feature representation of these words is adjusted to more accurately reflect the actual meaning carried in the current context. In this way, the understanding ability of long-distance semantic dependency relationships in the approval request analysis result can be significantly improved, thereby improving the accuracy of the semantic expression of each word.

[0058] In the above-mentioned aviation enterprise standardized management system, the approver matching unit 140 is configured to determine the type of the first-level approver based on the contextual semantic features of the sequence of medium granularity semantic reinforcement approval request word granularity semantic embedding encoding vectors. Wherein, Figure 4A block diagram of an approver matching unit in an aviation enterprise standardized management system according to an embodiment of the present application. As shown in Figure 4 The approver matching unit 140 includes a global context semantic coding subunit 141 configured to globally and contextually code the sequence of mid-granularity semantic reinforced approval request word granularity semantic embedding coding vectors to obtain an approval request context semantic coding vector; and an approval personnel type determination subunit 142 configured to input the approval request context semantic coding vector into an intelligent approval module based on a classifier to obtain an approval result, which is used to represent a type label of the first-level approver.

[0059] Specifically, in one specific example of the present application, the global context semantic coding subunit 141 is configured to input the sequence of mid-granularity semantic reinforced approval request word granularity semantic embedding coding vectors into a context encoder based on a transformer module to obtain the approval request context semantic coding vector. It should be understood that the present application takes into account that relying only on the semantic information representation of word granularity cannot sufficiently capture the more complex semantic structure and context relationship in the approval request resolution result. Therefore, in order to further understand the complete semantics of the approval request from a global perspective, the present application introduces a context encoder based on a transformer module to globally and contextually code and integrate information of the sequence of mid-granularity semantic reinforced approval request word granularity semantic embedding coding vectors. It should be known by those skilled in the art that the transformer module can simultaneously focus on different positions in the input sequence and perform parallel calculation on them through the self-attention mechanism and position encoding, so as to capture the long-distance semantic dependency relationship and context information in the text sequence, comprehensively interact the semantics and integrate the information of each word in the entire approval request resolution result, thereby achieving comprehensive understanding and representation of the context information of the approval request, and generating the approval request context semantic coding vector.

[0060] Specifically, in one specific example of the present application, the approval personnel type determination subunit 142 is configured to: perform full connection coding on the approval request context semantic coding vector using the full connection layer of the intelligent approval module to obtain an approval request context semantic full connection coding vector; input the approval request context semantic full connection coding vector into a Softmax classification function of the intelligent approval module to obtain probability values of the approval request context semantic coding vector belonging to various classification labels, wherein the classification labels are used to represent the types of first-level approval personnel; and determine the classification label corresponding to the maximum probability value as the approval result. Specifically, the classifier establishes a mapping relationship model between the semantic features of the approval request and the types of the first-level approval personnel through learning a large amount of labeled approval request data. In actual application, when the approval request context semantic coding vector is input into the classifier, the classifier can accurately infer the type label of the first-level approval personnel that best matches the current approval request, such as “purchase department manager”, “technical director”, “financial supervisor”, and the like, by using the mapping relationship model learned in the training process to perform classification prediction on the approval request context semantic coding vector. In this way, the system can automatically route the approval request to the corresponding approval personnel according to the obtained approval result, realize automatic processing of the approval process, and thus greatly improve the approval efficiency and accuracy.

[0061] In one preferred example of the present application, inputting the approval request context semantic coding vector into the intelligent approval module based on the classifier to obtain an approval result comprises:

[0062] First, the distance, such as the L2 distance, between each pair of feature values of the approval request context semantic coding vector is calculated, and the square root of the distance is taken to obtain an approval request context semantic coding distance representation matrix, which is represented by the formula:

[0063]

[0064] wherein D i,j represents the element value at the (i, j) position in the approval request context semantic coding distance representation matrix, x i , and x j represent the feature values at the i-th position and the j-th position in the approval request context semantic coding vector, respectively, d(·, ·) represents a distance measurement function, and X represents the approval request context semantic coding vector.

[0065] Second, an approval request context semantic coding self-association matrix of the approval request context semantic coding vector as a row vector is obtained, which is represented by the formula:

[0066]

[0067] wherein D' represents the self-association matrix of the approval request context semantic coding, and X is a row vector, (·) T denotes the transpose of a vector, denotes matrix multiplication;

[0068] Then, the approval request context semantic coding vector is multiplied by the approval request context semantic coding distance representation matrix to obtain an approval request context semantic coding first-level mapping vector, which is expressed by the formula:

[0069]

[0070] wherein X' represents the approval request context semantic coding first-level mapping vector, and D represents the approval request context semantic coding distance representation matrix;

[0071] Next, the approval request context semantic coding first-level mapping vector is multiplied by the matrix product of the approval request context semantic coding distance representation matrix and the approval request context semantic coding self-association matrix to obtain an approval request context semantic coding multi-level mapping vector, which is expressed by the formula:

[0072]

[0073] wherein X" represents the approval request context semantic coding multi-level mapping vector;

[0074] Then, the approval request context semantic coding multi-level mapping vector is point-added with an approval request context semantic coding association eigen vector composed of eigenvalues of the approval request context semantic coding self-association matrix to obtain an optimized approval request context semantic coding vector, wherein the eigenvalues are interpolated or zero-padded in the case of deficiency;

[0075] Finally, the optimized approval request context semantic coding vector is input into the classifier-based intelligent approval module to obtain an approval result.

[0076] Specifically, the present application considers that each of the medium-granularity semantic reinforced approval request word granularity semantic embedding coding vectors in the sequence of the medium-granularity semantic reinforced approval request word granularity semantic embedding coding vectors respectively represents a local attention reinforced-based approval request word granularity semantic embedding coding feature. When performing context coding based on the transformer module, the multi-head self-attention mechanism is essentially equivalent to a secondary attention mechanism, which will cause attention redundancy coding and thus cause the context fine-granularity semantic feature instance of the approval request context semantic coding vector to be missing in the judgment, thereby affecting the accuracy of the approval result obtained by the classifier-based intelligent approval module.

[0077] Therefore, the application realizes the multi-level distribution hierarchy-based secondary target mapping representation of the complete similarity instantiation of the self-association of the approval request context semantic coding vector by linear target mapping representation of the similarity distance representation matrix based on the approval request context semantic coding vector, and compensates for the associated mismatch negative influence factor by associated fusion kernel bias, so as to improve the eigenvalue instantiation degree of the approval request context semantic coding vector under the similarity limitation, that is, the significance degree of the eigenvalue as an instance for classification regression judgment, and improve the accuracy of the approval result of the approval request context semantic coding vector obtained by the intelligent approval module based on the classifier.

[0078] After obtaining the approval result, that is, determining the type label of the first-level approver, the system needs to filter out the qualified candidates from the enterprise internal database according to this label. This not only depends on the role matching-based method, which uses the organizational structure tree and the post information in the standardization module in the enterprise to find the roles or positions matched with the label, for example, if the label is “maintenance engineer”, the system will find all employees with this title; but also considers the professional skills and past experience of the candidates. By analyzing historical data, the system can identify which approvers perform well when handling similar requests and preferentially recommend these people. In addition, in order to avoid the overload of individual approvers, the system also checks the current workload of the candidates and selects those who have enough time and energy to handle new requests.

[0079] After determining the approver, an effective information transmission mechanism must be established to ensure that the approval request can be delivered to the relevant personnel in a timely manner. Notification pushing is an important part of it, using email, instant messaging tools or a dedicated approval management platform to send notifications to selected approvers, informing the approvers of the pending requests. The notification content should include necessary background information and operation guidelines to enable the approver to quickly understand the situation and take action. At the same time, task assignment is also crucial, formally assigning the approval request to the designated approver and marking it as a to-do item in the enterprise's task management system, so that the approver will not miss any important request. Permission settings cannot be ignored, ensuring that the approver has all the permissions needed to view and handle the request, including access to relevant documents, data modification, etc., while setting appropriate restrictions to prevent unauthorized operations. The design of the entire information transmission mechanism is to ensure that the approval request can quickly and accurately reach the correct approver and that the approver has enough resources and support to complete the task.

[0080] In summary, the aviation enterprise standardized management system based on the embodiments of the present application is illustrated, when receiving an approval request, the approval request is parsed using natural language processing technology to extract key information such as initiator information, request type, and approval content details, then, the approval request parsing result is encoded based on the semantic embedding of the word granularity and the local context semantic association optimization, to realize the deep understanding of the approval request, and then intelligently match the appropriate first-level approver type based on the context semantic information of the approval request parsing result. In this way, according to the specific content and background of the approval request, the approval path can be dynamically adjusted, and the most suitable approver can be selected, so as to flexibly cope with various approval requests, and improve the approval efficiency and decision quality.

[0081] The basic principles of the present application are described above in conjunction with specific embodiments, but it should be noted that the advantages, advantages, effects, etc. mentioned in the present application are only examples and not limitations, and these advantages, advantages, effects, etc. cannot be considered as the must-have of each embodiment of the present application. In addition, the specific details of the above embodiments are only for the purpose of example and for the purpose of understanding, and the above details do not limit the present application to the above specific details.

[0082] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in a certain embodiment can be referred to the related description of other embodiments. In the several embodiments provided by the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the unit division is only a logical function division, and the actual implementation can have another division way. The units described as separated components can or can not be physically separated, and the components displayed as units can be or can not be physical units, that is, they can be located in one place or distributed on a network. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0083] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be considered as limiting the claims to which they relate.

[0084] Furthermore, the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural and vice-versa, unless the context clearly requires these exclusions. The conjunction "or" is used to link items in a list or a set of alternatives, and is not disjunctive, unless the context clearly requires it to be disjunctive. The conjunction "and" is used to link items in a list or a set of alternatives, and is not conjunctive, unless the context clearly requires it to be conjunctive. The prefix "re-" when used in the context of the term to which it relates, means "not" or "opposite of", unless the context clearly requires it to have another meaning.

[0085] Finally, it should be noted that the description has been given for illustrative and descriptive purposes only and is not intended to limit the technical solutions of the present application. Furthermore, the above examples are merely used to illustrate the technical solutions of the present application but not to limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalent replaced without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. An aviation enterprise standardization management system, comprising: The standardized module is used for loading a standardized organization architecture tree, and organization architectures in the standardized organization architecture tree are associated in up and down levels through a parent ID; The original organization architecture module is used for loading a company organization architecture tree, and organization architectures in the company organization architecture tree are associated in up and down levels through the parent ID, and when a new organization architecture node is created by selecting a company level or a branch company level, lower-level department and post data are added according to a preset node in the standardized module; and the association module is used for associating nodes in the standardized module with nodes in the original organization architecture module, and through the association relationship between the standardized module and the original organization architecture module, associated operations in authorization, approval, modification and review processes in each level are realized, and the association module comprises: An approval request analysis unit is configured to analyze the received approval request to obtain an approval request analysis result; A semantic embedding coding unit is configured to perform word granularity semantic embedding coding on the approval request analysis result to obtain a sequence of approval request word granularity semantic embedding coding vectors; A medium granularity semantic association optimization unit is configured to perform medium granularity semantic association constraint optimization on the sequence of approval request word granularity semantic embedding coding vectors to obtain a sequence of medium granularity semantic reinforced approval request word granularity semantic embedding coding vectors; An approver matching unit is configured to determine the type of a first-level approver based on the context semantic features of the sequence of medium granularity semantic reinforced approval request word granularity semantic embedding coding vectors. The medium granularity semantic association optimization unit comprises: A syntax depth analysis subunit is configured to map each approval request word granularity semantic embedding coding vector in the sequence of approval request word granularity semantic embedding coding vectors to a hyperbolic space respectively to obtain a sequence of hyperbolic space approval request word granularity semantic embedding coding vectors; The syntax depth implicit representation value of each hyperbolic space approval request word granularity semantic embedding coding vector in the sequence of hyperbolic space approval request word granularity semantic embedding coding vectors is calculated to obtain a sequence distribution of the approval request word granularity syntax depth implicit representation values; A semantic association window searching subunit is configured to extract the syntax depth implicit representation value of a first approval request word granularity semantic embedding coding vector in the sequence of approval request word granularity semantic embedding coding vectors as a starting bit of the medium granularity semantic association window; Along the sequence distribution of the approval request word granularity syntax depth implicit representation values, an approval request word granularity syntax depth implicit representation value that is closest to the syntax depth implicit representation value of the first approval request word granularity semantic embedding coding vector is searched as a terminal bit of the medium granularity semantic association window; Based on the starting bit of the medium granularity semantic association window and the terminal bit of the medium granularity semantic association window, a medium granularity semantic association window of the first approval request word granularity semantic embedding coding vector is determined from the sequence of approval request word granularity semantic embedding coding vectors. The contextual semantic association reinforcement subunit is configured to calculate semantic association scores of each of the first approval request word granularity semantic embedding encoding vectors in the medium granularity semantic association window of the first approval request word granularity semantic embedding encoding vector relative to the first approval request word granularity semantic embedding encoding vector to obtain a sequence of first approval request word granularity semantic association scores. The sequence of first approval request word granularity semantic association scores is normalized to obtain a sequence of normalized first approval request word granularity semantic association scores. The sequence of normalized first approval request word granularity semantic association scores is used as a sequence of weights to calculate a position-weighted sum between each of the first approval request word granularity semantic embedding encoding vectors in the medium granularity semantic association window of the first approval request word granularity semantic embedding encoding vector to obtain a medium granularity semantic reinforcement approval request word granularity semantic embedding encoding vector corresponding to the first approval request word granularity semantic embedding encoding vector.

2. The aviation business standardization management system according to claim 1, characterized by, The semantic embedding encoding unit is configured to: input the segmented approval request into a word semantic embedding encoder based on a Bert model to obtain a sequence of approval request word granularity semantic embedding encoding vectors.

3. The aviation enterprise standardization management system of claim 2, wherein, The approver matching unit includes: a global contextual semantic encoding subunit configured to perform global contextual semantic association encoding on the sequence of medium granularity semantic reinforcement approval request word granularity semantic embedding encoding vectors to obtain an approval request contextual semantic encoding vector; an approver type determination subunit configured to input the approval request contextual semantic encoding vector into an intelligent approval module based on a classifier to obtain an approval result, which is used to represent a type label of the first-level approver.

4. The aviation enterprise standardization management system of claim 3, wherein, The global contextual semantic encoding subunit is configured to: input the sequence of medium granularity semantic reinforcement approval request word granularity semantic embedding encoding vectors into a contextual encoder based on a transformer module to obtain the approval request contextual semantic encoding vector.

5. The aviation enterprise standardization management system of claim 4, wherein, The approver type determination subunit is configured to: perform fully connected encoding on the approval request contextual semantic encoding vector using a fully connected layer of the intelligent approval module to obtain an approval request contextual semantic fully connected encoding vector; input the approval request contextual semantic fully connected encoding vector into a Softmax classification function of the intelligent approval module to obtain probability values of the approval request contextual semantic encoding vector belonging to each classification label, wherein the classification label is used to represent the type of the first-level approver; determine a classification label corresponding to the largest probability value in the probability values as the approval result.

Images