Expressway business form extraction method and system based on neural network

Abstract

The invention provides an expressway business form extraction method and system based on a neural network, and the method comprises the steps: obtaining to-be-processed sub-business forms, carrying out the data cleaning, and generating rule data; performing compliance verification on the rule data based on the agreed rule, and if the verification is qualified, storing the rule data as a standard business form according to the category; traversing each standard service form in sequence, identifying coincident entries, taking the coincident entries as association keys, and extracting a high-speed service form set belonging to the same user based on the association keys; and inputting the high-speed business form set into a pre-trained statement extraction model to generate a passing behavior statement. According to the invention, by integrating the expressway business forms, standardized processing and efficient management of data are realized. According to the method, the data compliance is ensured, the business forms of the same user are identified and associated, the traffic behavior statements are generated by using the statement extraction model, and convenient and accurate information support is provided for expressway business management.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a method and system for extracting expressway business forms based on neural networks. Background Art

[0002] Highway business data is characterized by massiveness, diversity, real-time, dynamics, and high value. These characteristics make highway business data have broad application prospects and important value in the fields of traffic management, urban planning, emergency response, etc. Based on this, high requirements are placed on the performance of business data cleaning and view display platforms. However, most current platforms rely on management platforms for data management and display pages, and use multiple platforms and multiple methods to clean and integrate heterogeneous data, which cannot form a unified visual display and effectively display and utilize the inherent connections between the data. At the same time, there are a large number of similar or even identical businesses on highways, and multiple platforms process data separately, resulting in unclear business logic. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes a highway business form extraction method and system based on neural network, which realizes the standardized processing and efficient management of data by integrating highway business forms; it not only ensures the compliance of data, but also intelligently identifies and associates the business forms of the same user, and then uses the sentence extraction model to generate the passage behavior statement, integrates and effectively utilizes the correlation between different sub-business data, and provides convenient and accurate information support for highway business management.

[0004] In order to achieve the above object, the present invention adopts the following technical solution:

[0005] In a first aspect, the present invention provides a method for extracting highway business forms based on a neural network, comprising:

[0006] Obtain each sub-business form to be processed to clean up the data and generate rule data;

[0007] Compliance check is performed on rule data based on agreed rules. If the check is qualified, it is saved as a standard business form according to the category.

[0008] Traverse each standard business form in turn, identify overlapping entries, use the overlapping entries as association keys, and extract a high-speed business form set belonging to the same user based on the association keys;

[0009] Input high-speed business form sets into the pre-trained sentence extraction model to generate traffic behavior sentences.

[0010] Preferably, the obtaining of each sub-business form to be processed to clean up data and generate rule data specifically includes:

[0011] Unify each sub-business form into JSON Schema format, delete missing and duplicate information, and obtain rule data;

[0012] The sub-business form includes a vehicle information form, a fee information form and a toll station information form, and the toll station information form contains business flow data when a vehicle passes through a toll station.

[0013] Preferably, the agreed rules include cleaning rules, correction rules and desensitization rules;

[0014] The cleaning rules are used to check and clean transaction failures, duplications, invalid transaction flow data, missing data in required fields, and expiration time extraction data at the entry and provincial boundary entry, exit and provincial boundary exit in the agreed rules;

[0015] The correction rule is used to query the associated inlet and outlet flow data according to the outlet flow data to correct the inlet and vehicle information corresponding to the outlet flow data;

[0016] The desensitization rules are used to verify sensitive data and perform desensitization processing.

[0017] Preferably, if the verification is qualified, the form is saved as a standard business form according to the category, which specifically includes:

[0018] According to the qualified data, data is stored through the AI ​​data analysis module according to its original sub-business type; specifically, a distributed storage architecture is adopted to automatically create partition tables according to the three dimensions of year, month, and day to realize automatic data storage. At the same time, the original data source flow generation time is analyzed for each new data, and then the partition table is automatically matched according to the algorithm to form a standard business form.

[0019] Preferably, the step of sequentially traversing each standard business form, identifying overlapping entries, using the overlapping entries as association keys, and extracting a high-speed business form set belonging to the same user based on the association keys specifically includes:

[0020] Traverse each standard business form in turn, and take the same terms appearing in different standard business forms as overlapping terms, thereby obtaining multiple overlapping terms;

[0021] Use overlapping entries as association keys to extract multiple sets of standard business forms to be integrated;

[0022] The unique identity information in the standard business form set to be integrated is extracted as the first item to obtain a high-speed business form set belonging to the same user; the unique identity information includes a pass identification ID, a license plate number and a mobile phone number.

[0023] Preferably, the high-speed business form set includes a form and a table.

[0024] Preferably, the training process of the sentence extraction model is:

[0025] Obtaining a historical high-speed business form and labeling it, wherein the label is a given pass behavior statement;

[0026] Train the sentence extraction model, input the historical high-speed business forms into the sentence extraction model, and obtain the predicted traffic behavior sentences for each form;

[0027] The loss function of the model is calculated based on the predicted passage behavior sentence of each form and the given passage behavior sentence. When the loss is minimized, a trained sentence extraction model is obtained.

[0028] In a second aspect, the present invention provides a highway business form extraction system based on a neural network, comprising:

[0029] The pre-processing module is used to obtain the sub-business forms to be processed, clean the data, and generate rule data;

[0030] The data center is used to check the compliance of rule data. If the check is qualified, it will be saved as a standard business form according to the category; each standard business form is traversed in turn, overlapping entries are identified, and the overlapping entries are used as association keys to extract the high-speed business form set belonging to the same user based on the association keys;

[0031] The sentence extraction module is used to input the high-speed business form set into the pre-trained sentence extraction model to generate the pass behavior sentence.

[0032] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a highway business form extraction method based on a neural network as described in the first aspect.

[0033] In a fourth aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for extracting highway business forms based on a neural network described in the first aspect are implemented.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. The present invention realizes the standardized processing of highway business forms through a unified data cleaning and rule data generation process, breaking the heterogeneous data barriers caused by multiple platforms and multiple methods. Furthermore, by using overlapping entries as association keys, the high-speed business form sets belonging to the same user are successfully extracted and integrated, thereby revealing the inherent connection between the data and providing the possibility for unified visual display. In addition, this solution clearly defines the business logic through an intelligent data processing flow, avoiding the problem of unclear business logic caused by multiple platforms processing data separately. The present invention not only solves the problems of data integration and visual display in the prior art, but also improves the efficiency and accuracy of business processing, providing strong support for highway business management.

[0036] 2. The present invention uses a view-driven method based on business data cleaning, which greatly saves the time of module development, maintains the consistency of layout and visualization effects among various business modules, reduces common vulnerabilities and defects caused by repeated development, improves R&D efficiency, improves platform operation performance, and reduces operation and maintenance costs.

[0037] 3. The present invention constructs a "double verification" mechanism by first unifying the format and then performing detailed data processing. First, each sub-business form is unified into the JSON Schema format, and a standard specification is established for subsequent processing to avoid data processing confusion caused by format differences. On this basis, missing information and duplicate information are deleted, which can effectively reduce data redundancy and errors and preliminarily ensure data quality. Then, compliance verification is performed based on the agreed rules. Cleaning rules check and clean up various problem data, such as transaction failures, duplications, and invalid transactions, to prevent bad data from affecting subsequent analysis; correction rules can ensure that the information related to the export transaction is accurate and ensure the relevance and accuracy of the data; desensitization rules protect the security of sensitive data. This double verification checks both the source and content of the data, comprehensively improves data quality, and makes subsequent data-based business form integration, statement generation and other operations more reliable and efficient, providing strong support for highway business management.

[0038] 4. In terms of data storage, the present invention uses a neural network to automatically learn and adjust data classification strategies without manual intervention. Its high-precision data classification capabilities can handle complex data relationships, accurately divide data tables, and accurately classify data types with fuzzy boundaries. Moreover, through continuous optimization, data storage can be made more efficient, storage resources can be reasonably allocated, and redundant storage can be avoided. When extracting traffic behavior statements from the form set, the neural network can well understand the form semantic information and accurately extract key semantic features such as vehicle driving direction and toll type. At the same time, it can also process the complex logical relationships implicit in the form data and generate logical statements. In addition, in the face of form data in different formats, the generalization ability of the neural network enables it to effectively extract key information and generate correct statements.

[0039] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention but do not constitute a limitation of the present invention.

[0041] Figure 1 A main flow chart of a method for extracting highway business forms based on a neural network provided by an embodiment of the present invention;

[0042] Figure 2 A data cleaning and verification flow chart provided for an embodiment of the present invention;

[0043] Figure 3 A flow chart for obtaining a standard business form provided by an embodiment of the present invention;

[0044] Figure 4 A schematic diagram of generating statements based on a business form provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0046] Embodiment 1

[0047] like Figure 1 As shown, this embodiment discloses a method for extracting highway business forms based on a neural network, comprising the following steps:

[0048] S1: Obtain each sub-business form to be processed, perform data cleaning, and generate rule data.

[0049] Specifically, the sub-business forms include a vehicle information form, a fee information form and a toll station information form.

[0050] Among them, the vehicle information form includes the owner’s name, owner’s mobile phone number, access identification ID, etc.

[0051] The cost information form includes export fees, etc.

[0052] The toll station information form contains the business flow data of the vehicle when it passes through the toll station, including the entrance toll station code, entrance toll station personnel code, entrance lane, entrance time, entrance license plate, entrance license plate color, entrance vehicle axle number, entrance vehicle model, entrance weight, exit toll station code, exit toll station personnel code, exit lane, exit time, exit license plate, exit license plate color, exit vehicle axle number, exit vehicle model, exit weight, overload axle number, overload weight, etc.

[0053] Perform data cleaning according to standard rules to generate data that can be parsed by the data center. Specifically, obtain the data API interface documents of each business system; determine the data required by the main business process based on the data API interface documents; transmit the data of the business subsystem to the data cleaning program in real time through the Kafka message queue; the data cleaning program uses the data collector at the receiving end to clean the sub-business system data into JSON Schema format data that can be stored in the business center; filter out garbage (incomplete data) and duplicate information during the data cleaning stage.

[0054] S2: Perform compliance check on rule data based on agreed rules. If the check is qualified, save it as a standard business form according to the category.

[0055] like Figure 2 As shown, when the data center receives the rule data after processing by the data cleaning program, it checks whether the data format is correct, whether the data is complete, and whether data desensitization is required through agreed rules; after the data center completes the inspection, if it passes, it will be stored in the database; if it fails, the data cleaning program will be notified.

[0056] Specifically, the agreed rules include cleaning rules, correction rules and desensitization rules.

[0057] The cleaning rules are used to verify and clean transaction failures, duplications, invalid transaction flow data at the entrances and provincial boundary entrances, exits and provincial boundary exits that exist in the agreed rules, missing data in business required fields (missing key data, such as pass identification), and invalid time extracted data (the time when the extracted data was generated is not within the valid time range).

[0058] The correction rule is used to query the associated inlet and outlet flow data according to the outlet flow data to correct the inlet and vehicle information corresponding to the outlet flow data.

[0059] The desensitization rules are used to verify sensitive data and perform desensitization processing. According to relevant regulations, in order to protect personal privacy and maintain data security, the user's sensitive data and private data are desensitized.

[0060] The general data desensitization process is divided into the data layer and the view layer. The data layer desensitization method uses MD5 encryption to desensitize the toll station staff information including user login passwords, contact information, etc. The view layer desensitization method uses intermediate digital desensitization for data that needs to be desensitized, such as vehicle identification information (license plate, color, etc.), personal identity information (mobile phone number of vehicle driving license personnel, ID card number, contact number, home address, work unit, etc.), transaction information (partial details of ETC toll records, timestamps, etc.).

[0061] As an implementation method, the data may be reprocessed into a unified unit. The cost may be RMB or 10,000 yuan, the time format may be YYYY-MM-DD or YYYYMM-DDhh:mm:ss, and the type may be dictionary table association or automatic generation. Those skilled in the art may select the unit format according to actual conditions, and this embodiment does not specifically limit it.

[0062] like Figure 3 As shown, after the data center receives the verified data, it stores it in different sub-business standard business forms according to different business types.

[0063] After receiving the data processed and verified by the data cleaning program, the data center must complete a data analysis based on the processed JSON data; the purpose of data analysis is to create different data tables based on the data of each business subsystem to store different types of data; data analysis is a continuous self-learning process. As the amount of data in each business subsystem increases and the data types become more extensive, the AI ​​data analysis module of the data center will continue to self-learn and update, and the automatic creation of data table classification and data storage will become more and more accurate and efficient. Specifically, according to business needs, a distributed storage architecture is adopted to automatically create partition tables in three dimensions: year, month, and day for automatic data storage. Each new data will parse the time when the original data source flow is generated, and automatically match the partition table according to the algorithm for data storage.

[0064] Among them, the AI ​​data analysis module is built on neural networks, such as recursive neural networks (RNN) or long short-term memory networks (LSTM). When faced with new data, the trained neural network can automatically infer the appropriate data table classification and data storage method based on the patterns it has learned. Moreover, by continuously inputting new data for learning, it can continuously update and optimize its own model, thereby achieving the goal of automatically creating data table classification and data storage with increasing accuracy and efficiency.

[0065] This embodiment adopts a distributed storage architecture and creates partition tables to store data by year, month, and day. In terms of data management, partition storage can quickly locate and retrieve data based on the time dimension, greatly improving query efficiency, and facilitating data cleaning and archiving, reducing storage costs; from a business analysis perspective, it helps to analyze the changing trends of data over time, providing a strong basis for decision-making, and meeting regulatory requirements for data storage periods, ensuring data compliance and security.

[0066] S3: Traverse each standard business form in turn, identify overlapping entries, use the overlapping entries as association keys, and extract the high-speed business form set belonging to the same user based on the association keys. Among them:

[0067] S301: traverse each standard business form in turn, and take the same entries appearing in different standard business forms as overlapping entries, thereby obtaining multiple overlapping entries;

[0068] S302: Using the overlapping terms as association keys, extracting multiple sets of standard business forms to be integrated;

[0069] S303: Extracting the unique identity information in the standard business form set to be integrated as the first item to obtain a high-speed business form set belonging to the same user; the unique identity information includes a traffic identification ID, a license plate number and a mobile phone number.

[0070] Specifically, after completing the storage operation, the data center needs to identify the primary key of the data collected by each business subsystem, and the data of each business subsystem needs to be linked to the data of each business subsystem through the primary key of the main table.

[0071] like Figure 4 As shown, the "Vehicle Information Table" containing the owner's name information is used as the main table, and the owner's mobile phone number in the vehicle information table is used as the keyword to query the vehicle information; the license plate number in the vehicle information is used as the association key to query the fee information in the "Fee Information Table", and then the toll station name is used as the association key to query the toll station information through the "Toll Station Information Table", and so on. The unique data primary key of each business sub-table can be used to automatically associate the information at a certain time, a certain stage, and a certain rule.

[0072] The data analysis results generated through automatic association can also be associated with the data source of the low-code platform components to automatically generate valuable data tables or reports; according to different sub-business tables, they can be associated with the form module of the low-code platform to automatically match enumeration items; the data source can be associated with the visualization part of the low-code platform to automatically generate visual analysis; and finally a fast, efficient, and valuable automatic data collection low-code generation platform can be completed.

[0073] The highway business form set includes forms and tables. The highway business form set must use a unique field (such as access identification ID, license plate number, mobile phone number, etc.) as the first item and describe the field to be stored in the database. The remaining data is spliced ​​with a unique identifier; after parsing by the data center, vehicle access information can be automatically generated based on the storage information.

[0074] This embodiment can efficiently integrate scattered related data and clearly sort out the complete business situation of a specific user by extracting overlapping entries and using them as association keys to collect high-speed business forms of the same user, facilitating subsequent centralized analysis and processing. At the same time, using the unique identity information as the first item is like setting a clear "header" or "index", which can quickly and accurately locate the form set of a specific user, greatly improving the efficiency of data query and retrieval. When processing a large amount of high-speed business data, it can effectively improve the convenience and accuracy of data management, and provide strong support for related business decisions.

[0075] S4: Input the high-speed business form set into the pre-trained sentence extraction model to generate the pass behavior sentence.

[0076] like Figure 4 As shown, after data analysis of the high-speed business form set, the traffic behavior statements belonging to the car owner were obtained.

[0077] Among them, the training process of the sentence extraction model is:

[0078] S401: Obtain a historical high-speed business form and mark it with a label, where the label is a given traffic behavior statement;

[0079] S402: training a sentence extraction model, inputting historical high-speed business forms into the sentence extraction model, and obtaining a predicted traffic behavior sentence for each form;

[0080] S403: Calculate the loss function of the model according to the predicted traffic behavior sentence of each form and the given traffic behavior sentence, and obtain a trained sentence extraction model when the loss is minimum.

[0081] It should be understood that the sentence extraction model is a natural language model and can be built based on a neural network. As a specific implementation, the sentence extraction model of this embodiment is preferably a Chat2DB model. The use process of the sentence extraction model specifically includes: first, using Chat2DB natural language to generate SQL, the specific operation is as follows: configure Chat2DB to the database, then open the Chat2DB interface, enter the natural language description you want to query in the input box, and then Chat2DB will generate the corresponding SQL query statement based on the input natural language. Secondly, verify the extracted SQL statement, mainly including checking the generated SQL query statement to confirm its correctness, manually adjusting the SQL statement if necessary, and finally executing the generated SQL query and viewing the results.

[0082] This embodiment extracts a high-speed business form set belonging to a certain user from multiple business forms, and can bring together relevant information scattered in different forms. In the original multi-business form state, the data is scattered, for example, vehicle information is in one form, fee information is in another form, and toll station information is in a third form. If a user wants to query the complete information of a certain pass, it is necessary to search and piece together relevant data in multiple forms, which is cumbersome and easy to miss information. By extracting the business form set belonging to the same user, all the pass information related to the user can be integrated together, which is convenient for subsequent processing and query, and greatly improves the efficiency of data query and management.

[0083] Subsequently, the business form set is further converted into user pass behavior statements, making the data more intuitive and easy to understand. User pass behavior statements present complete pass events in the form of natural language. This natural language representation allows users to intuitively capture key information without having to interpret the various fields and data relationships in the form, which enhances the readability and interpretability of the data.

[0084] This specific embodiment achieves standardized processing of original form data through data cleaning and rule data generation, effectively improving the quality and availability of data. Secondly, compliance verification and classification are saved as standard business forms to ensure the accuracy and compliance of the data, providing a reliable basis for subsequent business processing. Furthermore, by identifying overlapping entries and using them as association keys, the high-speed business form set belonging to the same user is successfully extracted, and data association and integration are achieved, facilitating comprehensive analysis of user behavior. Finally, the integrated high-speed business form set is input into the pre-trained sentence extraction model to generate a traffic behavior statement, which provides intuitive and accurate information support for highway management and decision-making, and helps to improve management efficiency and decision-making level.

[0085] Embodiment 2

[0086] This embodiment provides a highway business form extraction system based on a neural network, including:

[0087] The pre-processing module is used to obtain the sub-business forms to be processed, clean the data, and generate rule data;

[0088] The data center is used to perform compliance verification on rule data based on agreed rules. If the verification is qualified, it will be saved as a standard business form according to the category; each standard business form is traversed in turn, overlapping entries are identified, and the overlapping entries are used as association keys to extract the high-speed business form set belonging to the same user based on the association keys;

[0089] The sentence extraction module is used to input the high-speed business form set into the pre-trained sentence extraction model to generate the pass behavior sentence.

[0090] Embodiment 3

[0091] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the method for extracting highway business forms based on a neural network as described in the first embodiment above are implemented.

[0092] Embodiment 4

[0093] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for extracting highway business forms based on neural networks as described in the first embodiment above are implemented.

[0094] The steps or modules involved in the above embodiments 2 to 4 correspond to those in embodiment 1. For the specific implementation, please refer to the relevant description of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0095] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for extracting highway business forms based on neural networks, characterized in that: include: Obtain each sub-business form to be processed to clean up the data and generate rule data; Compliance check is performed on rule data based on agreed rules. If the check is qualified, it is saved as a standard business form according to the category. Traverse each standard business form in turn, identify overlapping entries, use the overlapping entries as association keys, and extract a high-speed business form set belonging to the same user based on the association keys; Input high-speed business form sets into the pre-trained sentence extraction model to generate traffic behavior sentences.

2. A method for extracting highway business forms based on neural networks as claimed in claim 1, characterized in that: The obtaining of each sub-business form to be processed to clean up the data and generate rule data specifically includes: Unify each sub-business form into JSON Schema format, delete missing and duplicate information, and obtain rule data; The sub-business form includes a vehicle information form, a fee information form and a toll station information form, and the toll station information form contains business flow data when a vehicle passes through a toll station.

3. A method for extracting highway business forms based on neural networks as claimed in claim 2, characterized in that: The agreed rules include cleaning rules, correction rules and desensitization rules; The cleaning rules are used to check and clean transaction failures, duplications, invalid transaction flow data, missing data in required fields, and expiration time extraction data at the entry and provincial boundary entry, exit and provincial boundary exit in the agreed rules; The correction rule is used to query the associated inlet and outlet flow data according to the outlet flow data to correct the inlet and vehicle information corresponding to the outlet flow data; The desensitization rules are used to verify sensitive data and perform desensitization processing.

4. A method for extracting highway business forms based on neural network as claimed in claim 1, characterized in that: If the verification is qualified, it is saved as a standard business form according to the category, specifically including: According to the qualified data, data is stored through the AI ​​data analysis module according to its original sub-business type; specifically, a distributed storage architecture is adopted to automatically create partition tables according to the three dimensions of year, month, and day to realize automatic data storage. At the same time, the original data source flow generation time is analyzed for each new data, and then the partition table is automatically matched according to the algorithm to form a standard business form.

5. A method for extracting highway business forms based on neural network as claimed in claim 1, characterized in that: The method of sequentially traversing each standard business form, identifying overlapping entries, using the overlapping entries as association keys, and extracting a high-speed business form set belonging to the same user based on the association keys specifically includes: Traverse each standard business form in turn, and take the same terms appearing in different standard business forms as overlapping terms, thereby obtaining multiple overlapping terms; Use overlapping entries as association keys to extract multiple sets of standard business forms to be integrated; The unique identity information in the standard business form set to be integrated is extracted as the first item to obtain a high-speed business form set belonging to the same user; the unique identity information includes a pass identification ID, a license plate number and a mobile phone number.

6. A method for extracting highway business forms based on neural network as claimed in claim 1, characterized in that: The high-speed business form set includes forms and tables.

7. A method for extracting highway business forms based on neural network as claimed in claim 1, characterized in that: The training process of the sentence extraction model is as follows: Obtaining a historical high-speed business form and labeling it, wherein the label is a given pass behavior statement; Train the sentence extraction model, input the historical high-speed business forms into the sentence extraction model, and obtain the predicted traffic behavior sentences for each form; The loss function of the model is calculated based on the predicted passage behavior sentence of each form and the given passage behavior sentence. When the loss is minimized, a trained sentence extraction model is obtained.

8. A highway business form extraction system based on neural network, characterized in that: include: The pre-processing module is used to obtain the sub-business forms to be processed, clean the data, and generate rule data; The data center is used to perform compliance verification on rule data based on agreed rules. If the verification is qualified, it will be saved as a standard business form according to the category; each standard business form is traversed in turn, overlapping entries are identified, and the overlapping entries are used as association keys to extract the high-speed business form set belonging to the same user based on the association keys; The sentence extraction module is used to input the high-speed business form set into the pre-trained sentence extraction model to generate the pass behavior sentence.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, the steps in the highway business form extraction method based on neural network as described in any one of claims 1-7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the highway business form extraction method based on neural network as described in any one of claims 1-7 are implemented.

Images