A business travel automation monthly settlement method, device and equipment based on RPA

Abstract

The present invention relates to the field of artificial intelligence and discloses a method, device, and equipment for automated monthly settlement of business travel based on RPA. The method automatically acquires business travel data through Robotic Process Automation (RPA), parses the data to obtain a first expense item and corresponding amount information; performs OCR image recognition and logical reconstruction on archived invoices to obtain a second expense item and corresponding amount information; and determines the logical and arithmetic relationship between the first expense item and the corresponding second expense item to determine whether there are omissions or inconsistencies. The present invention not only reduces the workload of manual verification and significantly improves the efficiency of the reimbursement process, but also enhances the accuracy and completeness of reimbursement data and reduces the risk of human error.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, in particular to a business travel automation monthly settlement method, device and equipment based on RPA. BACKGROUND

[0002] Currently, travel expense reimbursement is one of the important links of financial management of various enterprises and organizations, which usually involves a large amount of manual operation and data processing work. In the traditional travel expense reimbursement process, the travel personnel need to manually submit the reimbursement application, and the financial personnel need to check, verify and approve the reimbursement documents one by one. This process usually includes multiple manual operations, such as logging into the business travel system, extracting relevant travel data, generating reimbursement documents, verifying relevant invoices and expenses, etc. This process not only consumes time and effort, but also is prone to human errors.

[0003] The manual processing method of travel reimbursement faces problems such as low efficiency, prone to errors and high cost. First, the collection and arrangement of data require the staff to perform complex operations such as data extraction, classification and calculation between various systems, which is prone to data loss or format errors. Second, the manual checking and verification process is complex, which is prone to missing some reimbursement documents or repeating the calculation of the travel expenses that have been processed. In addition, due to the long process time, the reimbursement application cannot be processed in real time, which may cause the financial reimbursement work to lag behind the actual business needs. SUMMARY

[0004] Therefore, the present application provides a business travel automation monthly settlement method based on RPA to solve the technical problems in the prior art.

[0005] According to a first aspect of the present application, a business travel automation monthly settlement method based on RPA is provided, comprising:

[0006] automatically obtaining business travel data by robot process automation (RPA), analyzing the business travel data to obtain first expense items and corresponding amount information;

[0007] performing OCR image recognition and logical reconstruction on the archived reimbursement invoices to obtain second expense items and corresponding amount information;

[0008] judging the logical relationship and arithmetic relationship between the first expense items and the corresponding second expense items to determine whether there is omission or inconsistency;

[0009] When it is determined that the second expense items are missing, the relevant personnel are notified to supplement; when it is determined that the first expense items and the corresponding second expense items are inconsistent, the relevant personnel are notified to modify.

[0010] According to a second aspect of the present application, a business travel automation monthly settlement device based on RPA is provided, comprising:

[0011] The first obtaining module is configured to automatically obtain business travel data by using RPA, analyze the business travel data, and obtain first expense items and corresponding amount information.

[0012] The second obtaining module is configured to perform OCR image recognition and logical reconstruction on the archived reimbursement receipts, and obtain second expense items and corresponding amount information.

[0013] The judging module is configured to judge the logical relationship and arithmetic relationship between the first expense items and the corresponding second expense items, and determine whether there is omission or inconsistency.

[0014] The notifying module is configured to notify relevant personnel to supplement when it is determined that the second expense items are omitted, and notify relevant personnel to modify when it is determined that the first expense items and the corresponding second expense items are inconsistent.

[0015] According to a third aspect of the present application, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the RPA-based business travel automation monthly settlement method when executing the computer program.

[0016] According to a fourth aspect of the present application, a computer readable storage medium is provided, which stores a computer program, and the computer program implements the steps of the RPA-based business travel automation monthly settlement method when executed by a processor.

[0017] According to the above technical solutions, the RPA-based business travel automation monthly settlement method, device, equipment and medium provided by the present application automatically obtain business travel data by using RPA, analyze the business travel data, and obtain first expense items and corresponding amount information; perform OCR image recognition and logical reconstruction on the archived reimbursement receipts, and obtain second expense items and corresponding amount information; judge the logical relationship and arithmetic relationship between the first expense items and the corresponding second expense items, and determine whether there is omission or inconsistency; notify relevant personnel to supplement when it is determined that the second expense items are omitted, and notify relevant personnel to modify when it is determined that the first expense items and the corresponding second expense items are inconsistent. The RPA-based business travel automation monthly settlement method provided by the present application not only reduces the workload of manual checking, significantly improves the efficiency of the reimbursement process, but also enhances the accuracy and completeness of the reimbursement data, reduces the risk of human error, and realizes the automation and intelligentization of accounting and monthly settlement reconciliation.

[0018] The above description is only a summary of the technical solutions of the present application. In order to enable a person skilled in the art to better understand the technical means of the present application and can be implemented according to the content of the description, and in order to enable the above and other purposes, characteristics and advantages of the present application to be more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0019] The drawings described herein are used to provide further understanding of the present application, and form a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 A schematic diagram of an application scenario of a business travel automation monthly settlement method based on RPA provided in an embodiment of the present application is shown;

[0021] Figure 2 A schematic diagram of a flow of a business travel automation monthly settlement method based on RPA provided in an embodiment of the present application is shown;

[0022] Figure 3 A schematic diagram of a flow of automatically obtaining business travel data by RPA provided in an embodiment of the present application is shown;

[0023] Figure 4 A schematic diagram of a method flow of performing OCR image recognition and logical reconstruction on the archived reimbursement receipts provided in an embodiment of the present application is shown;

[0024] Figure 5 A schematic diagram of a method flow of performing improved OCR image recognition on the archived reimbursement receipts provided in an embodiment of the present application is shown;

[0025] Figure 6 A schematic diagram of a structure of a business travel automation monthly settlement device based on RPA provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] In the following, the specific embodiments of the present application will be described in detail with reference to the drawings and in combination with the embodiments. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0027] The business travel automation monthly settlement method based on RPA provided in an embodiment of the present application can be applied to the reimbursement reconciliation scenario of the business trip documents as shown in Figure 1 The enterprise digital business travel automation monthly settlement (referred to as “monthly settlement” for short) process usually involves multiple steps from business trip data collection, reimbursement document generation to auditing, reconciliation and settlement. The complete process is divided into business trip application, business travel reservation, reimbursement document filling, auditing / approval, accounting / monthly settlement reconciliation and tax ticket input.

[0028] Business trip application: Before going on a business trip, the trip personnel need to submit a business trip application through the company's internal system, filling in the purpose, time, location, and other relevant information. The system will review the business trip application according to the company's travel policy.

[0029] Business travel reservation: After the business trip application is approved, the trip personnel can reserve business travel services through the system, such as booking tickets, hotels, etc. The business travel reservation data generated in this step, including transportation and accommodation cost information, will serve as an important basis for reimbursement verification.

[0030] Reimbursement form filling: After the trip is completed, the trip personnel need to fill out the reimbursement form and submit the reimbursement application. At this time, the business travel reservation information and actual expenses will be automatically retrieved by the system for preliminary data filling. RPA robots will automatically extract the corresponding expense data from the business travel reservation to generate a preliminary reimbursement form.

[0031] Review / Approval: After filling out the reimbursement form, the financial department or relevant approval personnel will review the business trip expense reimbursement application. Through intelligent verification technology, the system automatically compares the business travel reservation expenses with the actual expenses recorded in the reimbursement form to check for any inconsistencies or missing data. If there are problems, the system will automatically identify and notify the trip personnel to make modifications or supplements.

[0032] Monthly reconciliation: All approved reimbursement forms are summarized to generate the monthly settlement account. The system will classify expenses by department, project, or business unit to generate corresponding monthly reports. The summary report includes each employee's business trip expense details, total expenses for each category, trip locations, and trip duration. The system will automatically reconcile business travel expense data with billing data in the financial system to ensure that each expense is correctly recorded in the financial accounts. During the reconciliation process, the completeness, consistency, and accuracy of expenses need to be checked, and any non-compliant items need to be identified. If abnormal items (such as multiple entries or omissions) are found, the system will automatically generate a report and notify the financial personnel to handle them.

[0033] Tax ticket input: After the review, the trip personnel will input the actual expense-related invoices into the system. After scanning the invoices, the OCR technology will perform text recognition to automatically extract expense-related information and perform logical reconstruction through a finite state machine. The recognition results will be matched with the data in the reimbursement form.

[0034] Through this automated travel expense reimbursement verification process, enterprises can reduce the workload of manual processing, improve the efficiency and accuracy of the reimbursement process, and at the same time, monitor data problems in the reimbursement process in real time, ensuring the transparency and reliability of the financial reimbursement process. The present application provides an RPA-based business travel automation monthly settlement method, which optimizes the accounting / monthly settlement reconciliation process. The business travel data is automatically obtained by robotic process automation RPA (Robotic Process Automation), and the business travel data is analyzed to obtain the first expense item and the corresponding amount information. The archived reimbursement receipts are subjected to OCR (Optical Character Recognition) image recognition and logical reconstruction to obtain the second expense item and the corresponding amount information. The logical relationship and arithmetic relationship between the first expense item and the corresponding second expense item are determined to determine whether there is omission or inconsistency. When it is determined that the second expense item is missing, the relevant personnel are notified to supplement. When it is determined that the first expense item and the corresponding second expense item are inconsistent, the relevant personnel are notified to modify. The system automatically extracts the expense data related to travel (i.e. the first expense item) from the business travel system through RPA technology, including detailed information of air tickets, hotels, transportation and other expenses. This process no longer requires manual intervention, and RPA can efficiently obtain and analyze these data, and structure the corresponding expense information for subsequent reconciliation verification. For the archived reimbursement receipts, the system uses OCR technology for image recognition to automatically extract the expense items, amounts and other information in the receipts (i.e. the second expense item). At the same time, with the help of a finite state machine, the logical reconstruction of the recognition results is performed to ensure that the context and logical relationship of the content of the receipts can be accurately analyzed. This reconstruction ensures the correct identification of complex receipt content, such as the association between the trip location, expense type and amount. The system automatically determines the logical relationship and arithmetic relationship between the first expense item extracted from the business travel data and the second expense item identified in the reimbursement receipts through intelligent algorithms. This step includes checking for duplication, omission, or amount inconsistency. When the system determines that the second expense item is missing, an exception report is automatically generated, and the relevant travel personnel are notified to supplement the reimbursement materials or receipts. When the system finds that the first expense item and the second expense item are inconsistent in amount or other abnormalities, it automatically notifies the relevant personnel to modify the reimbursement application, ensuring the accuracy of the reimbursement amount and content. Through the RPA-based business travel automation monthly settlement method provided by the present application, not only the workload of manual checking is reduced, the efficiency of the reimbursement process is significantly improved, but also the accuracy and completeness of the reimbursement data are enhanced, the risk of human error is reduced, and the automation and intelligence of the accounting and monthly settlement reconciliation are realized.

[0035] The present application will be described in detail below through specific embodiments.

[0036] Example 1:

[0037] like Figure 2 As shown in FIG, an RPA-based automated monthly settlement method for business travel provided in an embodiment of the present invention includes:

[0038] Step 201: Automatically obtain business travel data through Robotic Process Automation (RPA), and parse the business travel data to obtain a first expense item and corresponding amount information;

[0039] Step 202: Perform OCR image recognition and logic reconstruction on the archived reimbursement invoice to obtain the second expense item and the corresponding amount information;

[0040] Step 203: Determine the logical and arithmetic relationship between the first expense item and the corresponding second expense item to determine whether there is any omission or inconsistency;

[0041] Step 204: When it is determined that the second expense item is missing, notify the relevant personnel to supplement it; when it is determined that the first expense item is inconsistent with the corresponding second expense item, notify the relevant personnel to modify it.

[0042] This paper provides an automated monthly settlement method for business travel based on RPA. This method combines Robotic Process Automation (RPA) and Optical Character Recognition (OCR) technology for the travel expense reimbursement verification process, achieving fully automated processing from business travel system data acquisition to invoice content recognition for the first time. RPA automatically acquires business travel expense data, while OCR recognizes and logically reconstructs the image content of reimbursement invoices. This technology combination effectively improves data processing efficiency, reduces manual intervention, and can automatically process complex reimbursement data.

[0043] Example 2:

[0044] Based on the first embodiment, a process for automatically obtaining business travel data through RPA is provided, such as Figure 3 As shown, the automation process is realized by RPA as follows:

[0045] Step 301: Automatically log in to the business travel system through Robotic Process Automation (RPA) and access a preset business travel data interface or database to obtain business travel data within a specified time range.

[0046] Step 302: Filter data based on business trip record parameters and extract expenses associated with expense items;

[0047] The business trip record parameters include employee name, business trip destination, departure time and return time, and expense items include airfare, hotel fees and transportation expenses;

[0048] Step 303, checking and verifying the expenses associated with the expense item according to the format requirements of the expense item amount, whether the amount is missing or not;

[0049] Step 304, generating a first expense item table with the expenses associated with the expense item.

[0050] Embodiment two of the present application provides a method for automatically obtaining business travel data through RPA. RPA (robotic process automation) can automatically log in to a business travel management system or database by simulating human operations and performing batch data acquisition operations. This technology can replace manual extraction of data from multiple systems or platforms, such as flight booking information, hotel reservation records, and transportation costs. Through automated operations, RPA can quickly and accurately obtain large amounts of business travel data, avoiding the tediousness and errors of manual operations. RPA not only enables batch acquisition of business travel data, but also intelligently parses web forms and Excel tables based on pre-set rules. For example, RPA can automatically classify data into different expense items (such as flight costs, hotel costs, and transportation costs) and extract the corresponding amount information for each expense item. This intelligent data processing capability significantly reduces the need for manual intervention.

[0051] Embodiment three:

[0052] Although traditional OCR technology can efficiently recognize text in images, the recognition results are often isolated characters or lines, lacking logical association in context. Especially when dealing with complex documents (such as reimbursement receipts and invoices), the business logic between lines and paragraphs is easily overlooked. Embodiment three of the present application provides a method for performing OCR image recognition on archived reimbursement receipts, as shown in Figure 4 , which includes:

[0053] Step 401, performing OCR image recognition on the archived reimbursement receipt to obtain a text recognition result;

[0054] Prior to step 401, the archived reimbursement receipt is also pre-processed, which includes binarization processing, image denoising, and tilt correction.

[0055] Step 402, logically reconstructing the text recognition result according to a finite state machine, obtaining text line states by analyzing document layout information, and constructing a line state machine;

[0056] Step 403, logically reconstructing the text recognition result through the state machine to form a text result with contextual logical semantics;

[0057] Step 404, identifying the logical relationship between lines through the finite state machine;

[0058] The logical relationship includes a continuation relationship, a different attribute relationship, a parallel relationship, and a containing relationship.

[0059] To better illustrate the method of logical reconstruction of the OCR recognition result based on the finite state machine, the actual business reimbursement ticket recognition is taken as an example in Embodiment Three of the present application for illustration as follows:

[0060] Step 1: OCR recognition of the ticket

[0061] In the OCR recognition of the business reimbursement ticket, the text content in the ticket is extracted. The OCR technology can recognize the expense item, date, amount and other contents in the ticket, but these contents are usually scattered and have no context logic. For example, line 1: "air ticket fee", line 2: "October 15, 2024", line 3: "800 yuan", although these text contents are recognized, the logical relationship between them is not clear.

[0062] Step 2: Explicit line state extraction

[0063] In this step, the "explicit line state" of each line of text in the OCR recognition result is extracted. The explicit line state is based on the format features (such as numbers, punctuation, date format, etc.) and layout information (such as left margin, alignment, etc.) to judge the basic classification of the text. In the business reimbursement ticket, the following several common states can be extracted by the explicit line state:

[0064] Expense type line: such as "air ticket fee", "hotel fee", etc.; date line: such as "October 15, 2023"; amount line: such as "800 yuan". Through these features, the text recognized by the OCR can be initially classified into different categories, providing a basis for subsequent logical reconstruction.

[0065] Step 3: Building a finite state machine

[0066] In this step, a finite state machine is built to describe the relationship between different elements (expense type, date, amount, etc.) in the reimbursement ticket. The finite state machine contains the following states: expense type state: identifying the type of expense, such as "air ticket fee", "accommodation fee", etc.; date state: identifying the date related to business, such as departure date, return date, etc.; amount state: identifying the expense amount. The state transition rules of the state machine are as follows:

[0067] When the "expense type line" is recognized, the state enters the expense type state. Then the system expects to recognize the related date of this expense type, so the state machine enters the date state. After recognizing the date, the state machine expects the amount information of the expense, so it transitions to the amount state. After completing the recognition of a complete expense item, the state machine returns to the initial state, ready to process the next expense information.

[0068] Step 4: State Change and Logic Reconstruction

[0069] When the finite state machine starts running, the system reads each line from the text line recognized by OCR and changes the state according to the rules. For example: when the system recognizes "Ticket fee", it enters the expense type state; then it recognizes "2023 October 15", and the state changes to the date state; when it recognizes "800 yuan", the state changes to the amount state. In this process, the system gradually reconstructs this information and generates structured reimbursement document information according to the logical relationship of "expense type-date-amount".

[0070] Suppose the content of the reimbursement document recognized by OCR is as follows:

[0071] Ticket fee

[0072] 2023 October 15

[0073] 800 yuan

[0074] Hotel fee

[0075] 2023 October 16

[0076] 500 yuan

[0077] The logic of processing by the finite state machine is as follows:

[0078] Line 1: "Ticket fee", enter the expense type state; Line 2: "2023 October 15", the state machine enters the date state; Line 3: "800 yuan", the state machine enters the amount state. Generate complete structured information: "Ticket fee, 2023 October 15, 800 yuan"; Line 4: "Hotel fee", the state machine returns to the expense type state; Line 5: "2023 October 16", the state machine enters the date state; Line 6: "500 yuan", the state machine enters the amount state. Generate complete structured information: "Hotel fee, 2023 October 16, 500 yuan".

[0079] Step 5: Logic Exception Handling

[0080] Among them, in the process of OCR recognition, if the situation does not meet the state transition rules, the system can automatically detect the exception and give a prompt. For example: if "Ticket fee" is recognized, followed by "500 yuan" without a date, the system can judge that the date information is missing and prompt the relevant personnel to supplement. If "Ticket fee" is followed by another expense item instead of a date or amount, the system can identify it as not logical and prompt for correction.

[0081] Step 6: Output Structured Data

[0082] The recognized text result is converted into structured data through a logic reconstruction process of a finite state machine. The data includes information such as expense type, date, and amount, and can be directly used in subsequent financial reimbursement verification and reconciliation processes.

[0083] The OCR recognition logic reconstruction method based on the finite state machine in the embodiment of the application can effectively process complex logical relationships and context dependencies in business trip reimbursement bills. Through explicit line state extraction and dynamic state conversion of the state machine, the system can generate structured information conforming to business logic from isolated text lines recognized by OCR, improve the accuracy and efficiency of reimbursement document processing, and reduce the possibility of human errors and omissions.

[0084] Embodiment Four

[0085] To further improve the accuracy of bill OCR text recognition, the application performs OCR image recognition on the archived reimbursement account to obtain text information. Embodiment Four of the application constructs a model in combination with an OCR confidence score and optimizes the training steps, as shown in Figure 5 , including

[0086] Step 501, recognized text, position information, and an OCR confidence score;

[0087] Step 502, based on each character box in the OCR output, calculate the confidence score of each character and associate it with the recognized character;

[0088] Step 503, combine the confidence score of each character with the word embedding in the BERT model to generate a confidence-aware word embedding;

[0089] Step 504, input the confidence-aware word embedding into a confidence-aware error detection model to classify the recognized text to determine whether there is an error in the text;

[0090] Step 505, for the detected error, semantic-based error correction can be used to select the best correction result through candidate generation, weight allocation, and scoring.

[0091] Before Step 504, the steps for training the confidence-aware error detection model are included, specifically including:

[0092] Step 504-1, obtain a data set containing OCR output with a confidence score;

[0093] The data set with a confidence score includes OCR-recognized original text, correct text corresponding to the OCR-recognized original text, and an OCR confidence score of each character.

[0094] Step 504-2: Align the OCR-recognized original text with the corresponding correct text character by character or word by word;

[0095] Step 504-3: Bind the OCR confidence score to each character;

[0096] Step 504-4: Label each character, label the incorrect character as 1 and the correct character as 0, to generate a training data set;

[0097] Step 504-5: Input the training dataset into the BERT model to generate a word embedding vector that integrates confidence perception, and pass it through the binary classification layer after the BERT model to classify each character as correct or incorrect;

[0098] Among them, the word embedding vector integrating confidence perception is e ci =(1-α)·Emb(t i )+α·(1-p ocr (t i )), where t is a character token list t={t1,t2,...,t n}, Emb is the standard embedding function, p ocr is the OCR confidence score, and α is a trainable parameter used to control the degree of noise used in the model;

[0099] Step 504-6: Perform back propagation parameter optimization based on the cross entropy loss function. When the change in the cross entropy loss function before and after iteration is less than a preset threshold or the number of iterations reaches a preset number, output a confidence-aware error detection model.

[0100] Among them, the cross entropy loss function cross entropy loss function Among them, y i is the label of the character as correct or wrong, y i is the probability of the model predicting correctly or incorrectly, N is the total number of samples, and the change in the cross entropy loss function before and after iteration is less than the preset threshold, indicating that the training optimization space is small and training can be stopped. Alternatively, when the number of iterations is greater than the preset number, it also means that training can be stopped.

[0101] The embodiment four of the present application fuses the OCR confidence and the BERT word embedding, and constructs a confidence-aware error detection model, which can realize accurate error detection in the text recognition process, and then automatically correct errors (the automatic correction method is prior art, for example, a variety of methods such as candidate word-based correction method, context-based language model correction, edit distance-based correction method, part-of-speech and grammar rule-based correction, etc. will not be repeated here), which significantly improves the accuracy, clarity and logical consistency of the recognition result. At the same time, by reducing the need for manual intervention, the automation level of text recognition is improved, which is suitable for high-precision demand OCR recognition scenarios, and provides reliable technical support for various document processing.

[0102] Further, as Figures 2 to 5 The specific implementation of the method, the embodiment of the present application provides a business travel automation monthly settlement device based on RPA, as shown in Figure 6 The device comprises:

[0103] The first acquisition module 610 is configured to automatically acquire business travel data by robot process automation (RPA), analyze the business travel data, and acquire first expense items and corresponding amount information;

[0104] The second acquisition module 620 is configured to perform OCR image recognition and logical reconstruction on the archived reimbursement receipts, and acquire second expense items and corresponding amount information;

[0105] The judgment module 630 is configured to judge the logical relationship and arithmetic relationship between the first expense items and the corresponding second expense items, and judge whether there is omission or inconsistency;

[0106] The notification module 640 is configured to notify relevant personnel to supplement when it is judged that the second expense items are omitted, and notify relevant personnel to modify when it is judged that the first expense items and the corresponding second expense items are inconsistent.

[0107] The embodiment of the present application provides a computer device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the business travel automation monthly settlement method based on RPA when executing the computer program, comprising:

[0108] The first acquisition module 610 is configured to automatically acquire business travel data by robot process automation (RPA), analyze the business travel data, and acquire first expense items and corresponding amount information;

[0109] The second acquisition module 620 is configured to perform OCR image recognition and logical reconstruction on the archived reimbursement receipts, and acquire second expense items and corresponding amount information;

[0110] Determine the logical and arithmetic relationship between the first expense item and the corresponding second expense item, and determine whether there are omissions or inconsistencies;

[0111] When it is determined that there is omission in the second expense item, the relevant personnel are notified to supplement it; when it is determined that the first expense item is inconsistent with the corresponding second expense item, the relevant personnel are notified to modify it.

[0112] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the following steps are implemented:

[0113] Automatically acquire business travel data through Robotic Process Automation (RPA) and analyze it to obtain the first expense item and the corresponding amount.

[0114] Perform OCR image recognition and logic reconstruction on archived reimbursement invoices to obtain the second expense item and the corresponding amount information;

[0115] Determine the logical and arithmetic relationship between the first expense item and the corresponding second expense item, and determine whether there are omissions or inconsistencies;

[0116] When it is determined that there is omission in the second expense item, the relevant personnel are notified to supplement it; when it is determined that the first expense item is inconsistent with the corresponding second expense item, the relevant personnel are notified to modify it.

[0117] It should be noted that the above embodiments only use travel reimbursement as an example to illustrate the principles and implementation steps of the embodiments of the present invention, and do not specifically limit the actual application scenarios, such as cost budgeting and funding application. The technical solution of the present invention can also be applied to extended scenarios of line loss analysis and evaluation in various substations. Regarding the functions or steps that can be implemented by computer-readable storage media or computer equipment, please refer to the aforementioned method embodiments. To avoid repetition, they will not be described one by one here.

[0118] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0119] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0120] The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A business travel automated monthly settlement method based on RPA, characterized by: include: Automatically obtain business travel data through robotic process automation (RPA), and parse the business travel data to obtain a first expense item and corresponding amount information; Performing OCR image recognition and logical reconstruction on the archived reimbursement invoices to obtain the second expense item and the corresponding amount information, specifically including preprocessing the archived reimbursement invoices, wherein the preprocessing includes binarization processing, image denoising and tilt correction; performing OCR image recognition on the archived reimbursement invoices to obtain text recognition results; logically reconstructing the text recognition results according to a finite state machine, obtaining text line states by analyzing document layout information, and constructing a line state machine; completing logical reconstruction of the text recognition results through the state machine to form a text result with contextual logical semantics; identifying logical relationships between lines through the finite state machine, wherein the logical relationships include continuation relationships, different attribute relationships, parallel relationships and inclusion relationships; Determining the logical and arithmetic relationships between the first expense item and the corresponding second expense item, and determining whether there are omissions or inconsistencies; When it is determined that there is omission in the second expense item, the relevant personnel are notified to supplement it; when it is determined that the first expense item is inconsistent with the corresponding second expense item, the relevant personnel are notified to modify it.

2. The RPA-based automated monthly settlement method for business travel according to claim 1 is characterized in that: The step of automatically acquiring business travel data through robotic process automation (RPA) and parsing the business travel data to obtain the first expense item and corresponding amount information includes: Automatically log into the business travel system through Robotic Process Automation (RPA) and access the pre-set business travel data interface or database to obtain business travel data within a specified time range; Filtering data based on business trip record parameters to extract expenses associated with expense items, wherein the business trip record parameters include employee name, business trip destination, departure time, and return time, and the expense items include airfare, hotel expenses, and transportation expenses; Check the expenses associated with the expense items based on the format requirements of the expense item amounts and whether the amounts are missing; Generate a first expense item table for the expenses associated with the expense items.

3. The RPA-based automated monthly settlement method for business travel according to claim 1 is characterized in that: The step of performing OCR image recognition on the archived reimbursement invoice to obtain text recognition results includes: Perform OCR image recognition on archived expense accounts to obtain text information, including recognized text, location information, and OCR confidence scores; Based on each character box output by OCR, calculate the confidence score of each character and associate it with the recognized character; Combine the confidence score of each character with the word embedding in the BERT model to generate confidence-aware word embeddings; The confidence-aware word embedding is input into the confidence-aware error detection model to classify the recognized text to determine whether there are errors in the text; For detected errors, semantic-based error correction is used to select the best correction result through candidate generation, weight assignment and scoring.

4. The RPA-based automated monthly settlement method for business travel according to claim 1 is characterized in that: Before the step of inputting the confidence-aware word embedding into the confidence-aware error detection model to classify the recognized text to determine whether the text contains errors, the step of training and generating the confidence-aware error detection model is included, specifically including: Obtaining a data set with confidence scores containing OCR output, wherein the data set with confidence scores includes the original text recognized by the OCR, the correct text corresponding to the original text recognized by the OCR, and the OCR confidence score of each character; Align the OCR-recognized original text with the corresponding correct text, character by character or word by word; Bind an OCR confidence score to each character; Label each character, mark the wrong character as 1, and the correct character as 0 to generate a training data set; The training data set is input into the BERT model to generate a word embedding vector that integrates confidence perception. Then, each character is classified as correct or incorrect through the binary classification layer after the BERT model. The word embedding vector that integrates confidence perception is ,in, is the token of the i-th character, is the standard embedding function, is the OCR confidence score, It is a trainable parameter used to control the degree of noise used in the model; Back-propagation parameter optimization is performed based on the cross-entropy loss function. When the change in the cross-entropy loss function before and after iteration is less than a preset threshold or the number of iterations reaches a preset number, a confidence-aware error detection model is output.

5. The RPA-based automated monthly settlement method for business travel according to claim 4 is characterized in that: The cross entropy loss function ,in, is the label for the character as true or false, is the probability that the model predicts correctly or incorrectly, and N is the total number of samples.

6. An RPA-based automated monthly settlement device for business travel, characterized in that: include: a first acquisition module, configured to automatically acquire business travel data through Robotic Process Automation (RPA), and parse the business travel data to obtain a first expense item and corresponding amount information; The second acquisition module is used to perform OCR image recognition and logical reconstruction on the archived reimbursement bills to obtain the second expense item and the corresponding amount information, specifically including preprocessing the archived reimbursement bills, wherein the preprocessing includes binarization processing, image denoising and tilt correction; performing OCR image recognition on the archived reimbursement bills to obtain text recognition results; logically reconstructing the text recognition results according to a finite state machine, obtaining text line states by analyzing document layout information, and constructing a line state machine; completing the logical reconstruction of the text recognition results through the state machine to form a text result with contextual logical semantics; identifying the logical relationship between lines through the finite state machine, wherein the logical relationship includes a continuation relationship, a different attribute relationship, a parallel relationship and an inclusion relationship; a judgment module, configured to judge the logical and arithmetic relationship between the first expense item and the corresponding second expense item, and to judge whether there is any omission or inconsistency; The notification module is used to notify relevant personnel to supplement when it is determined that there is omission in the second expense item; when it is determined that the first expense item is inconsistent with the corresponding second expense item, notify relevant personnel to modify it.

7. 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 computer program, the steps of the RPA-based business travel automated monthly settlement method as described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the RPA-based business travel automated monthly settlement method as described in any one of claims 1 to 5 are implemented.

Images