Data resource management method and device, equipment and storage medium

Abstract

The invention belongs to the technical field of bill management, and relates to a data resource management method and device, equipment and a storage medium. Inputting into a bill category identification model; obtaining a category output node of each paperless bill; according to the category output nodes and the issuing numbers of the paperless bills, marking the paperless bills differently; and in combination with the differentiated labeling result and a preset data resource management strategy, differentiated management is carried out on the paperless bills of the multiple sources and channels. According to the method, automatic and intelligent category identification can be carried out on the electronic service bills generated in the business travel journey, manual arrangement consumption is reduced, the paperless bill classification efficiency is improved, and meanwhile, the classification efficiency of the electronic service bills can be improved in post-processing functions such as a business travel verification function and a financial staff statistics function. Paperless bill obtaining and searching can be carried out in combination with a differentiated management result, and the processing efficiency of a post-processing function is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bill management, and is applied to the scene of paperless management of electronic service bills generated in business trip, and relates to a data resource management method, device, equipment and storage medium. BACKGROUND

[0002] In the current digital era, paper bills are gradually replaced by paperless bills. For example, in the past business trip, a large number of paper bills need to be issued when traveling, lodging or eating. At present, whether traveling by car, high-speed rail or airplane, electronic tickets have gradually replaced paper tickets, and hotels have gradually transformed into issuing electronic lodging bills.

[0003] At present, different paperless bill issuing servers have different types and specifications of paperless bills, and different service APPs on the user's mobile terminal receive different types of paperless bills. For example, some service APPs can cache paperless travel tickets, and some service APPs can cache paperless lodging tickets. Therefore, how to efficiently manage paperless bills from different sources has become a technical problem to be solved. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a data resource management method, device, equipment and storage medium to efficiently manage paperless bills from different sources for subsequent business search.

[0005] In a first aspect, the embodiments of the present application provide a data resource management method, which adopts the technical scheme as follows: A data resource management method includes the following steps: Obtain paperless bills from multiple sources; Input all paperless bills into a pre-trained bill category recognition model; Obtain the category output node of the bill category recognition model when classifying each paperless bill; According to the category output node and the issuing number of the paperless bill, each paperless bill is distinguished and labeled; Combine the differentiated labeling result and the preset data resource management strategy to manage the paperless bills from multiple sources.

[0006] In a second aspect, the embodiments of the present application also provide a data resource management device, which adopts the technical scheme as follows: A data resource management device includes: The paperless bill acquisition module is configured to acquire paperless bills from multiple sources; The identification model input module is configured to input all the paperless bills into a pre-trained bill category identification model; The category output node acquisition module is configured to acquire category output nodes of the bill category identification model when classifying and outputting each paperless bill; The differentiation labeling module is configured to differentiate and label each paperless bill according to the category output nodes and the issuing number of the paperless bill; The differentiation management module is configured to combine the differentiation labeling result and a preset data resource management strategy to differentiate and manage the paperless bills from the multiple sources.

[0007] In a third aspect, the embodiments of the present application further provide a computer device, which adopts the technical scheme as follows: A computer device includes a memory and a processor, the memory stores computer readable instructions, and the processor executes the computer readable instructions to realize the steps of the data resource management method.

[0008] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium, which adopts the technical scheme as follows: A computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed by a processor to realize the steps of the data resource management method.

[0009] Compared with the prior art, the embodiments of the present application have the following beneficial effects: The data resource management method described in the application can be widely applied to the scene of paperless management of electronic service tickets generated in business trip, by obtaining paperless tickets from multiple sources; inputting all paperless tickets into a pre-trained ticket category identification model; obtaining the category output node of the ticket category identification model when classifying each paperless ticket; distinguishing and labeling each paperless ticket according to the category output node and the issuing number of the paperless ticket; combining the distinguishing and labeling result and the preset data resource management strategy to distinguish and manage the paperless tickets from multiple sources. The method can manage the electronic service tickets generated in business trip in a paperless manner, so as to combine the ticket category identification model to automatically and intelligently identify the paperless tickets generated in business trip, reduce the consumption of manual sorting, improve the classification efficiency of paperless tickets, and also enable the paperless tickets to be obtained and searched according to the distinguishing and management result in post-processing functions such as business trip audit and financial staff statistics, thereby improving the processing efficiency of the post-processing functions. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the scheme in the application, the following will briefly introduce the drawings needed in the description of the embodiments of the application. Obviously, the drawings in the following description are some embodiments of the application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0011] Figure 1 is an exemplary system architecture diagram to which the application can be applied; Figure 2 is a flowchart of an embodiment of a data resource management method according to the application; Figure 3 is a flowchart of a specific embodiment of the pre-training of the ticket category identification model in the data resource management method described in the application; Figure 4 is Figure 3 is a flowchart of a specific embodiment of step 303 shown in the figure; Figure 5 is Figure 4 is a flowchart of a specific embodiment of step 402 shown in the figure; Figure 6 is Figure 5 is a flowchart of a specific embodiment of step 503 shown in the figure; Figure 7 is Figure 2 is a flowchart of a specific embodiment of step 203 shown in the figure; Figure 8This is a schematic diagram of one embodiment of a data resource management device according to this application; Figure 9 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0012] like Figure 1 As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables.

[0013] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0014] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0015] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0016] It should be noted that the data resource management method provided in this application embodiment is generally executed by a server, and correspondingly, a data resource management device is generally installed in the server.

[0017] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0018] Continue to refer to Figure 2The diagram illustrates a flowchart of an embodiment of a data resource management method according to this application. The data resource management method includes the following steps: Step 201: Obtain paperless invoices from multiple sources.

[0019] In this embodiment, the "multiple source channels" refers to multiple service apps that provide paperless receipts. For example, various electronic service receipts generated during business trips, online ticketing receipts generated by hotel booking apps, online taxi receipts generated during taxi rides, paperless receipts generated during high-speed rail and air travel, and paperless receipts generated during meals during the trip. These online paperless receipts are often booked and cached by different service apps.

[0020] This embodiment aims to provide a business travel service APP that can interface with the invoice acquisition interfaces of the aforementioned different service APPs to obtain corresponding paperless invoices from these different service APPs. Of course, it is also possible that the business travel service APP itself includes some service invoice issuance functionality. This business travel service APP is a user-end APP that integrates paperless invoices from its own sources and other multiple sources to summarize paperless invoices generated during business travel, thereby improving the efficiency of subsequent verification and processing for personnel.

[0021] Step 202: Input all paperless tickets into the pre-trained ticket category recognition model.

[0022] In this embodiment, the pre-trained invoice category recognition model is pre-trained by the server corresponding to the business travel service APP for paperless invoice category recognition, and it can be enabled as an invoice category recognition function. When the business travel service APP obtains paperless invoices from multiple sources, the user terminal that has installed the business travel service APP can enable the invoice category recognition function provided by the corresponding server, thereby enabling the pre-trained invoice category recognition model.

[0023] Specifically, when sending an instruction to enable the invoice category recognition function to the server via the business travel service APP, all paperless invoices can be packaged and sent to the pre-trained invoice category recognition model simultaneously. Alternatively, the invoice category recognition function provided by the server can be pre-deployed as a built-in feature of the business travel service APP on the user's end, allowing the packaged sending of all paperless invoices to the pre-trained invoice category recognition model to be implemented directly within the business travel service APP on the user's end.

[0024] By inputting all paperless receipts into a pre-trained receipt category recognition model, the paperless receipts generated during business travel can be automatically and intelligently categorized, reducing manual processing costs and improving the efficiency of paperless receipt classification.

[0025] Step 203: Obtain the category output node when the ticket category recognition model classifies each paperless ticket.

[0026] In this embodiment, there is a one-to-one correspondence between the category output nodes in the ticket category recognition model and the categories of paperless tickets. That is, each output node of the ticket category recognition model corresponds to only one category of paperless tickets. For example, the category output node corresponding to the high-speed rail paperless ticket in the ticket category recognition model is node number 6; the category output node corresponding to the electronic ticket in the ticket category recognition model is node number 8.

[0027] By obtaining the category output node of the ticket category recognition model when classifying each paperless ticket, it is possible to identify paperless tickets of different categories based on the category output node.

[0028] Step 204: Based on the category output node and the issuance number of the paperless invoice, each paperless invoice is marked with a unique identifier.

[0029] In this embodiment, when differentiating each paperless invoice, the invoice's own issuance number is incorporated. Since paperless invoices of the same category are output from the same category output node, simply using the category output node for invoice labeling is insufficient for individual breakdown of each paperless invoice. Therefore, the category output node and the invoice's issuance number are introduced, and these two fields are concatenated to differentiate each paperless invoice, enabling individual breakdown of each paperless invoice and facilitating more refined management and retrieval of paperless invoices in the future.

[0030] Step 205: Based on the differentiated labeling results and the preset data resource management strategy, differentiated management is carried out on the paperless invoices from the multiple sources.

[0031] In this embodiment, the preset data resource management strategy can be pre-set according to different management needs. For example, it can differentiate the management of paperless receipts for accommodation, food, and transportation; or it can differentiate the management based on whether the paperless receipts are reimbursable; or it can differentiate the management based on the source channel of the paperless receipts. Here, the data resource management strategy refers to the pre-set differentiated management rules for paperless receipts.

[0032] By combining differentiated labeling results with preset data resource management strategies, paperless invoices from multiple sources are managed in a differentiated manner. This allows post-processing functions to leverage the differentiated management results for acquiring and searching paperless invoices, thereby improving the efficiency of post-processing. Specifically, these post-processing functions include, for example, business travel reimbursement functions and financial personnel statistical functions.

[0033] In this embodiment, the data resource management method can be widely applied to scenarios involving paperless management of electronic service receipts generated during business travel. It involves acquiring paperless receipts from multiple sources; inputting all paperless receipts into a pre-trained receipt category recognition model; obtaining the category output node of the receipt category recognition model when classifying each paperless receipt; and then, based on the category output node and the receipt's issue number, differentiated labeling of each paperless receipt; finally, combining the differentiated labeling results with a preset data resource management strategy to differentiate and manage paperless receipts from multiple sources. This method enables paperless management of electronic service receipts generated during business travel, facilitating automated and intelligent category recognition of these receipts using the receipt category recognition model. This reduces manual processing costs and improves the efficiency of paperless receipt classification. Furthermore, in post-processing functions, such as business travel reimbursement and financial personnel statistics, the differentiated management results can be used to acquire and search for paperless receipts, improving processing efficiency in post-processing functions.

[0034] In this embodiment, the step of obtaining paperless receipts from multiple sources specifically includes: obtaining paperless receipts received by multiple different service apps within a preset time period according to the third-party service interface connected to the target service app, as paperless receipts from channels other than its own source. The target service app includes a target business travel management app pre-installed and used on the user's mobile terminal. Specifically, the preset time period refers to the target user's business travel itinerary, and the multiple different services refer to third-party services connected to the target service app. Here, the third-party services refer to the mobile server installed by the target business travel management app and third-party service clients other than those providing the service client of the target business travel management app. The paperless receipts received by the target service app within the preset time period are obtained through the target service app's own paperless receipt receiving interface, as paperless receipts from its own source. The paperless receipts from channels other than its own source and the paperless receipts from its own source are then combined to form the paperless receipts from multiple sources.

[0035] By pre-integrating paperless invoices from its own source and other sources, the system can summarize the paperless invoices generated during business trips, thereby improving the efficiency of subsequent verification and processing by the verification personnel, eliminating the need to obtain paperless invoices from multiple sources during the verification stage.

[0036] Continue to refer to Figure 3 In some specific implementations, a pre-training step for the ticket category recognition model is included before step 202. Figure 3 This is a flowchart of a specific embodiment of the data resource management method described in this application for pre-training a bill category recognition model, including: Step 301: Obtain paperless invoices from multiple historical sources and generate a training set for invoice category recognition; In this embodiment, obtaining paperless tickets from multiple historical sources specifically refers to the server side of the target business travel management APP obtaining historically compiled paperless tickets from multiple sources in advance. Here, "multiple sources" can be, for example, from the high-speed rail paperless ticket database, the electronic ticket database, the hotel paperless ticket management database, etc.

[0037] By acquiring paperless invoices from multiple historical sources, a training set for invoice category recognition is generated to facilitate pre-training of the invoice category recognition model.

[0038] Step 302: Input the training set for bill category recognition into the pre-built bill category recognition model; Step 303: Using a discriminative training strategy, the first and second recognition layers in the pre-constructed ticket category recognition model are pre-trained sequentially. In this embodiment, the invoice category recognition model includes at least two recognition layers, namely a first recognition layer and a second recognition layer. Different training strategies are used to train the first recognition layer and the second recognition layer, so that the pre-trained first recognition layer and the second recognition layer can recognize the category of the input paperless invoice according to different recognition mechanisms.

[0039] Step 304: Obtain the pre-trained ticket category recognition model with both the first and second recognition layers trained.

[0040] By pre-training the invoice category recognition model and employing a differential training strategy for different recognition layers during pre-training, the pre-trained invoice category recognition model can identify invoice types based on different recognition mechanisms, thereby improving the accuracy of paperless invoice category recognition.

[0041] Continue to refer to Figure 4 ,Figure 4 yes Figure 3 A flowchart of a specific embodiment of step 303 shown includes: Step 401: In the first recognition layer, the aspect ratio of each paperless ticket in the ticket category recognition training set is measured using a preset ticket ratio measurement tool. Specifically, a pre-defined ticket ratio measurement tool is deployed in the first recognition layer. This ticket ratio measurement tool can measure the aspect ratio of the input paperless ticket. During the pre-training stage of the first recognition layer, this ticket ratio measurement tool can be used to measure the aspect ratio of each paperless ticket in the ticket category recognition training set.

[0042] In this embodiment, in the first recognition layer, a preset ticket ratio measurement tool is used to measure the length and width ratio of each paperless ticket in the ticket category recognition training set. Different ticket issuance service terminals are used, such as high-speed rail paperless tickets, hotel electronic service tickets, and catering paperless ticket issuance service terminals. Due to the differences in service providers, the generated paperless tickets have different length and width ratios, so that when the actual ticket category classification is carried out later, the paperless tickets can be classified first according to their length and width ratios.

[0043] Step 402: Based on the aspect ratio measurement results and the preset ticket theme extraction component, identify whether the category of the paperless ticket corresponding to each aspect ratio is unique. In this embodiment, the preset document theme extraction component can extract the substantive content of paperless documents. By extracting the substantive content of paperless documents through the preset document theme extraction component, and then identifying whether the category of paperless documents with the same aspect ratio is unique based on the aspect ratio measurement results, paperless documents that can be categorized by aspect ratio are filtered out.

[0044] Step 403: If the category of the paperless ticket corresponding to the current aspect ratio is unique, then deploy the node output mapping relationship between the current aspect ratio and the corresponding unique ticket category between the preset model output layer and the first recognition layer. In this embodiment, the invoice category recognition model further includes a model output layer. In step 402, when it is found that the category of the paperless invoice corresponding to the current aspect ratio is unique, the node output mapping relationship between the current aspect ratio and the corresponding unique invoice category is directly deployed between the preset model output layer and the first recognition layer. This is so that when performing category recognition on paperless invoices in the future, for this type of invoice, the category output node can be directly assigned to it when its aspect ratio is measured without extracting the invoice content.

[0045] Step 404: If the category of paperless tickets corresponding to the current aspect ratio is not unique, then in the second recognition layer, the ticket themes corresponding to different categories of paperless tickets under the current aspect ratio are classified and sorted, and the node output mapping relationship between ticket themes and ticket categories is deployed between the model output layer and the second recognition layer. Specifically, if it is found that the category of paperless tickets corresponding to the current aspect ratio is not unique, it means that there are at least two categories of paperless tickets corresponding to this aspect ratio. In this case, according to the ticket theme content proposed by the ticket theme extraction component preset in step 402, in the second recognition layer, the ticket themes corresponding to different categories of paperless tickets under the current aspect ratio are classified and organized. The node output mapping relationship between ticket theme and ticket category is deployed between the model output layer and the second recognition layer. This is so that when classifying paperless tickets in the future, for tickets that cannot be directly classified by aspect ratio, the corresponding paperless ticket classification category output node is realized in the second recognition layer based on the ticket theme content.

[0046] Step 405: The pre-training of the first recognition layer and the second recognition layer is completed until each type of paperless ticket has a corresponding category output node deployed in the model output layer.

[0047] By employing a differential training strategy for the first and second recognition layers in the invoice category recognition model during pre-training, the pre-trained invoice category recognition model can first identify the invoice type based on the aspect ratio recognition mechanism. For invoices that cannot be identified on the first attempt, the model can then identify the invoice type based on the invoice's subject content. This avoids blindly extracting data content directly from all paperless invoices, thereby improving the efficiency of paperless invoice category recognition to some extent.

[0048] Continue to refer to Figure 5 , Figure 5 yes Figure 4 A flowchart of a specific embodiment of step 402 shown includes: Step 501: Cluster the paperless tickets with the same aspect ratio to obtain the clustering results; Specifically, in step 401, after measuring the aspect ratio of each paperless ticket in the ticket category recognition training set using a preset ticket proportion measurement tool, the different aspect ratio measurement results can be directly used as the basis for clustering to perform clustering processing on the elements in the ticket category recognition training set, thereby obtaining paperless ticket subsets corresponding to different aspect ratio measurement results.

[0049] Step 502: Perform cyclic random sampling N times on the paperless tickets in each cluster to obtain the random sample ticket set corresponding to each cluster. In this embodiment, to avoid having too many or too few paper tickets in the paperless ticket subsets corresponding to different aspect ratios, the paperless tickets in each cluster are randomly sampled N times in a cyclic manner. This means randomly sampling the paperless tickets in each paperless ticket subset with replacement N times in a cyclic manner to construct the randomly sampled ticket sets corresponding to different aspect ratios, where N is a positive integer greater than 1.

[0050] Step 503: Use the bill theme extraction component to extract the bill theme content for the elements in the random sample bill set to obtain the bill theme content corresponding to each random sample bill set. In this embodiment, during the model pre-training stage, to avoid extracting the subject content of all paperless tickets, aspect ratio measurement, clustering, and cyclic random sampling are performed first. Then, the ticket subject extraction component is used to extract the subject content of the randomly sampled ticket set. This significantly reduces the workload of the ticket subject extraction component and improves the pre-training efficiency of the ticket category recognition model.

[0051] Step 504: Identify whether the subject content of the bills corresponding to the current randomly sampled bill set is consistent; Step 505: If the subject content of the paperless tickets corresponding to the current random sample ticket set is consistent, then the category of the paperless ticket corresponding to the current aspect ratio is unique. Specifically, if the theme content of the invoice is extracted by the invoice theme extraction component and the theme content of the invoices corresponding to the current random sample invoice set is consistent, it means that the category of the paperless invoice corresponding to the current aspect ratio is unique.

[0052] Step 506: If the subject content of the paperless tickets corresponding to the current random sample ticket set is inconsistent, then the category of the paperless ticket corresponding to the current aspect ratio is not unique.

[0053] In this embodiment, by combining the aspect ratio measurement results and the preset document theme extraction component, it is identified whether the category of the paperless document corresponding to each aspect ratio is unique, so as to filter out paperless documents that can be categorized by aspect ratio and paperless documents that cannot be categorized by aspect ratio.

[0054] Continue to refer to Figure 6 , Figure 6 yes Figure 5 A flowchart of a specific embodiment of step 503 shown includes: Step 601: According to the method of extracting the pattern mark of the bill at the target fixed position, extract the pattern mark contained in the bill from each randomly sampled bill set, and use the extracted pattern mark as the subject content of the bill. Specifically, for example, high-speed rail e-tickets, e-bus tickets, and e-air tickets all use different patterns to represent high-speed rail, bus, or airfare tickets. These patterns are located in specific positions within the ticket according to regulations. Therefore, by directly extracting the pattern symbols from each randomly sampled ticket set according to the method for extracting the pattern symbols at the target fixed position, the pattern symbols contained in the ticket can be extracted. The extracted pattern symbols can then be used as the subject content of the ticket, thus achieving the extraction of the corresponding subject content from the pattern symbols in the paperless ticket.

[0055] Step 602: According to the method of extracting the subject area of ​​the bill, extract the key text information contained in the target division area of ​​each randomly sampled bill set, and use the key text information as the subject content of the bill.

[0056] Specifically, this applies to paperless tickets without graphic symbols. The text content of the ticket can be directly extracted. For example, if the main content of the current paperless ticket is set as the first line of text on the ticket, and the first line of text is located in the middle of the ticket, then the first line of text can be directly extracted as the main content of the ticket.

[0057] In this embodiment, two methods for extracting the subject matter of paper tickets are provided: one for paperless tickets with graphic markings and the other for paperless tickets with key text information. This ensures that the subject matter of paperless tickets can be fully extracted, thereby improving the accuracy of the subject matter extraction of the paper ticket category recognition model.

[0058] Continue to refer to Figure 7 , Figure 7 yes Figure 2 A flowchart of a specific embodiment of step 203 shown includes: Step 701: In the first identification layer, the aspect ratio of the current paperless document is measured using the document proportion measurement tool as the aspect ratio to be classified. Step 702: Identify whether the aspect ratio to be classified has a corresponding category output node in the model output layer; Specifically, in the actual category recognition stage for paperless tickets, firstly, in the first recognition layer, the length-to-width ratio of the current paperless ticket is measured using the ticket ratio measurement tool as the length-to-width ratio to be classified. Since the category output nodes corresponding to the length-to-width ratio have been deployed in the pre-training stage, it is possible to directly identify whether the length-to-width ratio to be classified has a corresponding category output node in the model output layer.

[0059] Step 703: If the aspect ratio to be classified has a corresponding category output node in the model output layer, then mark the category output node as the category output node when the current paperless ticket is classified and output. Specifically, if the aspect ratio to be classified has a corresponding category output node in the model output layer, it means that the category of the current paperless ticket is clear, and there is no need to perform subsequent identification processing steps. The category output node is directly marked as the category output node when the current paperless ticket is classified and output.

[0060] Step 704: If the aspect ratio to be classified does not have a corresponding category output node in the model output layer, then the invoice theme extraction component is used to extract the invoice theme content of the current paperless invoice, obtain the invoice theme content corresponding to the current paperless invoice, identify the category output node corresponding to the invoice theme in the model output layer, and mark it as the category output node when the current paperless invoice is classified and output.

[0061] Specifically, when the aspect ratio to be classified cannot be matched to the corresponding category output node, it means that the invoice theme extraction component is also needed to extract the invoice theme content of the current paperless invoice. Only based on the invoice theme content can the category of the current paperless invoice be determined.

[0062] By obtaining the category output node of the invoice category recognition model when classifying each paperless invoice, the invoice category of each paperless invoice to be identified can be determined, which facilitates subsequent invoice post-processing based on the invoice category.

[0063] In this embodiment, the target business travel management APP includes a paperless invoice resource management component. Before executing the step of differentiated management of paperless invoices from multiple sources by combining differentiated labeling results and preset data resource management strategies, the method further includes: pre-selecting all paperless invoices that are allowed to be reimbursed according to preset invoice reimbursement regulations; obtaining the category output nodes when the invoice category recognition model classifies and outputs all paperless invoices that are allowed to be reimbursed, and constructing a first type of output node set; summarizing the remaining output nodes in the model output layer other than the first type of output node set, and constructing a second type of output node set; allocating differentiated cache space for the paperless invoices corresponding to the first type of output node set and the second type of output node set in the paperless invoice resource management component, wherein the paperless invoices corresponding to the first type of output node set are cached in the first cache space, and the paperless invoices corresponding to the second type of output node set are cached in the second cache space.

[0064] Specifically, based on the reimbursement regulations, all eligible paperless invoices are pre-selected, and the corresponding category output nodes for each eligible paperless invoice are identified, constructing a first-class output node set. This ensures that if the current paperless invoice's corresponding category output node belongs to the first-class output node set, then the current paperless invoice is eligible for reimbursement; otherwise, it is not eligible. Using this method, the reimbursement and non-reimbursement classification of electronic service invoices generated during business travel is quickly achieved, improving the efficiency of paperless invoice verification.

[0065] In this embodiment, the step of performing differentiated management of paperless invoices from multiple sources by combining differentiated labeling results and preset data resource management strategies specifically includes: identifying the category output node corresponding to each invoice from multiple sources based on the differentiated labeling results; if the category output node corresponding to the current paperless invoice is an element of the first type of output node set, adding the current paperless invoice to the first cache space in the paperless invoice resource management component for caching; if the category output node corresponding to the current paperless invoice is an element of the second type of output node set, adding the current paperless invoice to the second cache space in the paperless invoice resource management component for caching.

[0066] In this embodiment, the data resource management method can be widely applied to scenarios involving paperless management of electronic service receipts generated during business travel. It involves acquiring paperless receipts from multiple sources; inputting all paperless receipts into a pre-trained receipt category recognition model; obtaining the category output node of the receipt category recognition model when classifying each paperless receipt; and then, based on the category output node and the receipt's issue number, differentiated labeling of each paperless receipt; finally, combining the differentiated labeling results with a preset data resource management strategy to differentiate and manage paperless receipts from multiple sources. This method enables paperless management of electronic service receipts generated during business travel, facilitating automated and intelligent category recognition of these receipts using the receipt category recognition model. This reduces manual processing costs and improves the efficiency of paperless receipt classification. Furthermore, in post-processing functions, such as business travel reimbursement and financial personnel statistics, the differentiated management results can be used to acquire and search for paperless receipts, improving processing efficiency in post-processing functions.

[0067] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0068] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0069] Further reference Figure 8 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of a data resource management device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0070] like Figure 8 As shown, the data resource management device 800 described in this embodiment includes: a paperless ticket acquisition module 801, a recognition model input module 802, a category output node acquisition module 803, a differentiation labeling module 804, and a differentiation management module 805. Wherein: The paperless invoice acquisition module 801 is used to acquire paperless invoices from multiple sources. The recognition model input module 802 is used to input all paperless tickets into the pre-trained ticket category recognition model; The category output node acquisition module 803 is used to acquire the category output node when the ticket category recognition model classifies and outputs each paperless ticket; The differentiation labeling module 804 is used to differentiate each paperless invoice based on the category output node and the issuance number of the paperless invoice; The differentiation management module 805 is used to differentiate and manage the paperless invoices from multiple sources by combining the differentiation labeling results and the preset data resource management strategy.

[0071] This application involves acquiring paperless invoices from multiple sources; inputting all paperless invoices into a pre-trained invoice category recognition model; obtaining the category output nodes of the invoice category recognition model when classifying each paperless invoice; and performing differentiated labeling on each paperless invoice based on the category output nodes and the invoice's issuance number. Combining the differentiated labeling results with a pre-set data resource management strategy, the paperless invoices from multiple sources are managed differentiatedly. This method enables paperless management of electronic service invoices generated during business travel, facilitating automated and intelligent category recognition of these invoices using the invoice category recognition model. This reduces manual processing costs and improves the efficiency of paperless invoice classification. Furthermore, in post-processing functions, such as business travel reimbursement and financial personnel statistics, the differentiated management results can be used to acquire and search for paperless invoices, improving post-processing efficiency.

[0072] In this embodiment, the paperless document acquisition module 801 includes a first paperless document acquisition unit, a second paperless document acquisition unit, and a paperless document aggregation unit. Wherein: The first paperless invoice acquisition unit is used to acquire paperless invoices received by multiple different service apps within a preset time period based on the third-party service interface connected to the target service app, as paperless invoices from channels other than its own source. The second paperless invoice acquisition unit is used to acquire paperless invoices received by the target service APP within the preset time period through the paperless invoice receiving interface of the target service APP itself, and use them as paperless invoices from its own source channel. The paperless invoice aggregation unit is used to aggregate paperless invoices from non-originating channels and paperless invoices from its own originating channels as the multi-originating channel paperless invoices.

[0073] In this embodiment, the data resource management device 800 further includes a bill category recognition training set generation module, a bill category recognition training set input module, a recognition layer distinction training module, and a bill category recognition model acquisition module. Wherein: The bill category recognition training set generation module is used to acquire paperless bills obtained from multiple historical sources and generate a bill category recognition training set. The bill category recognition training set input module is used to input the bill category recognition training set into the pre-built bill category recognition model; The recognition layer discrimination training module is used to pre-train the first recognition layer and the second recognition layer in the pre-constructed ticket category recognition model in sequence using a discrimination training strategy; The bill category recognition model acquisition module is used to obtain a bill category recognition model that has been trained on both the first recognition layer and the second recognition layer as the pre-trained bill category recognition model.

[0074] In this embodiment, the recognition layer differentiation training module includes a ticket aspect ratio measurement unit, a category uniqueness recognition unit, a first deployment unit for node output mapping relationships, a second deployment unit for node output mapping relationships, and a recognition layer pre-training completion unit. Wherein: The document aspect ratio measurement unit is used to measure the aspect ratio of each paperless document in the document category recognition training set in the first recognition layer using a preset document aspect ratio measurement tool. The category uniqueness identification unit is used to identify whether the category of the paperless ticket corresponding to each aspect ratio is unique based on the aspect ratio measurement results and the preset ticket theme extraction component. The first deployment unit for node output mapping relationship is used to deploy the node output mapping relationship between the current aspect ratio and the corresponding unique paperless ticket category between the preset model output layer and the first recognition layer if the category of the paperless ticket corresponding to the current aspect ratio is unique. The second deployment unit for node output mapping relationship is used to classify and organize the paper ticket themes corresponding to different categories of paperless tickets under the current aspect ratio in the second recognition layer if the categories of paperless tickets under the current aspect ratio are not unique, and to deploy the node output mapping relationship between the paper ticket theme and the paper ticket category between the model output layer and the second recognition layer. The recognition layer pre-training completion unit is used to complete the pre-training of the first recognition layer and the second recognition layer until each type of paperless ticket has a corresponding category output node deployed in the model output layer.

[0075] In this embodiment, the data resource management device 800 further includes a first invoice subject content extraction module and a second invoice subject content extraction module. Wherein: The first module for extracting the subject content of the bill is used to extract the pattern marks contained in the bill from each randomly sampled bill set according to the extraction method of the bill pattern marks at the target fixed position, and use the extracted pattern marks as the subject content of the bill. The second module for extracting the subject content of invoices is used to extract key text information contained in the target division area of ​​each randomly sampled invoice set according to the subject area division extraction method, and to use the key text information as the subject content of the invoice.

[0076] In this embodiment, the category output node acquisition module 803 includes a dimension-to-width ratio measurement unit, a category output node identification unit, a first category output node marking unit, and a second category output node marking unit. Wherein: The aspect ratio measurement unit to be classified is used to measure the aspect ratio of the current paperless ticket as the aspect ratio to be classified in the first recognition layer using the ticket aspect ratio measurement tool. The category output node identification unit is used to identify whether the aspect ratio to be classified has a corresponding category output node in the model output layer. The first marking unit for category output nodes is used to mark the category output node as the category output node when the current paperless ticket is classified and output if the aspect ratio to be classified exists in the model output layer. The second labeling unit for category output nodes is used to extract the subject content of the current paperless ticket by the ticket subject extraction component if the aspect ratio to be classified does not have a corresponding category output node in the model output layer, thereby obtaining the subject content of the current paperless ticket, identifying the category output node corresponding to the subject in the model output layer, and labeling it as the category output node when the current paperless ticket is classified and output.

[0077] In this embodiment, the data resource management device 800 further includes a paperless ticket selection module, a first type of output node set construction module, a second type of output node set construction module, and a differential cache space allocation module. Wherein: The paperless invoice selection module is used to pre-select all paperless invoices that are eligible for reimbursement according to preset invoice reimbursement regulations. The first type of output node set construction module is used to obtain the category output nodes when the invoice category recognition model classifies and outputs all paperless invoices that are allowed to be reimbursed, and construct the first type of output node set; The second type of output node set construction module is used to summarize the remaining output nodes in the model output layer other than the first type of output node set and construct the second type of output node set; The differential cache space allocation module is used to allocate differential cache space for the paperless tickets corresponding to the first type of output node set and the second type of output node set in the paperless ticket resource management component. The paperless tickets corresponding to the first type of output node set are cached in the first cache space, and the paperless tickets corresponding to the second type of output node set are cached in the second cache space.

[0078] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0079] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0080] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 9 , Figure 9 This is a basic structural block diagram of the computer device in this embodiment.

[0081] The computer device 9 includes a memory 9a, a processor 9b, and a network interface 9c that are interconnected via a system bus. It should be noted that... Figure 9 Only a computer device 9 with component memory 9a, processor 9b, and network interface 9c is shown. However, it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead. Those skilled in the art will understand that the computer device described herein is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions. Its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0082] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0083] The memory 9a includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 9a may be an internal storage unit of the computer device 9, such as the hard disk or memory of the computer device 9. In other embodiments, the memory 9a may also be an external storage device of the computer device 9, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 9. Of course, the memory 9a may also include both the internal storage unit and its external storage device of the computer device 9. In this embodiment, the memory 9a is typically used to store the operating system and various application software installed on the computer device 9, such as computer-readable instructions for a data resource management method. In addition, the memory 9a can also be used to temporarily store various types of data that have been output or will be output.

[0084] In some embodiments, the processor 9b may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 9b is typically used to control the overall operation of the computer device 9. In this embodiment, the processor 9b is used to execute computer-readable instructions stored in the memory 9a or to process data, for example, to execute computer-readable instructions of the data resource management method described above.

[0085] The network interface 9c may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 9 and other electronic devices.

[0086] The computer equipment proposed in this embodiment belongs to the field of invoice management technology and is applied to the scenario of paperless management of electronic service invoices generated during business travel. This application acquires paperless invoices from multiple sources; inputs all paperless invoices into a pre-trained invoice category recognition model; obtains the category output node of the invoice category recognition model when classifying each paperless invoice; differentiates each paperless invoice based on the category output node and the invoice's issuance number; and performs differentiated management of paperless invoices from multiple sources by combining the differentiated labeling results with a preset data resource management strategy. This method enables paperless management of electronic service invoices generated during business travel, facilitating automated and intelligent category recognition of these invoices using the invoice category recognition model. This reduces manual processing costs and improves the efficiency of paperless invoice classification. Furthermore, in post-processing functions, such as business travel reimbursement and financial personnel statistics, the differentiated management results can be used to acquire and search for paperless invoices, improving the processing efficiency of post-processing functions.

[0087] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by a processor to cause the processor to perform the steps of the data resource management method described above.

[0088] The computer-readable storage medium proposed in this embodiment belongs to the field of invoice management technology and is applied to the scenario of paperless management of electronic service invoices generated during business travel. This application obtains paperless invoices from multiple sources; inputs all paperless invoices into a pre-trained invoice category recognition model; obtains the category output node of the invoice category recognition model when classifying each paperless invoice; differentiates each paperless invoice based on the category output node and the invoice's issuance number; and performs differentiated management of paperless invoices from multiple sources by combining the differentiated labeling results with a preset data resource management strategy. This method enables paperless management of electronic service invoices generated during business travel, facilitating automated and intelligent category recognition of these invoices using the invoice category recognition model. This reduces manual processing costs and improves the efficiency of paperless invoice classification. Furthermore, in post-processing functions, such as business travel reimbursement and financial personnel statistics, the differentiated management results can be used to acquire and search for paperless invoices, improving the processing efficiency of post-processing functions.

[0089] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0090] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to make the disclosure of this application more thorough and comprehensive. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application. Software tools or components not belonging to this company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

Claims

1. A data resource management method, characterized in that, The method comprises the following steps: obtaining paperless bills from multiple sources; inputting all paperless bills into a pre-trained bill classification model; obtaining the classification output nodes of the bill classification model for each paperless bill; distinguishing and labeling each paperless bill according to the classification output nodes and the issuing number of the paperless bill; combining the distinguishing and labeling results with the preset data resource management strategy to manage the paperless bills from multiple sources.

2. The data resource management method of claim 1, wherein, Before the step of inputting all paperless bills into a pre-trained bill classification model, the method further comprises: obtaining paperless bills from multiple sources to generate a bill classification training set; inputting the bill classification training set into a pre-constructed bill classification model; pre-training the first recognition layer and the second recognition layer of the pre-constructed bill classification model in sequence using a distinguishing training strategy; obtaining a bill classification model with the first recognition layer and the second recognition layer both trained as the pre-trained bill classification model.

3. The data resource management method of claim 2, wherein, The step of pre-training the first recognition layer and the second recognition layer of the pre-constructed bill classification model in sequence using a distinguishing training strategy comprises: in the first recognition layer, measuring the length-width ratio of each paperless bill in the bill classification training set using a preset bill proportion measurement tool; according to the length-width ratio measurement result and a preset bill theme extraction component, identifying whether the category of the corresponding paperless bill under each length-width ratio is unique; if the category of the corresponding paperless bill under the current length-width ratio is unique, deploying a node output mapping relationship between the preset model output layer and the first recognition layer for the current length-width ratio and the corresponding unique bill category; if the category of the corresponding paperless bill under the current length-width ratio is not unique, classifying and arranging the bill themes corresponding to the paperless bills of different categories under the current length-width ratio in the second recognition layer, and deploying a node output mapping relationship between the bill theme and the bill category in the model output layer and the second recognition layer; until each paperless bill corresponds to a corresponding category output node in the model output layer, the first recognition layer and the second recognition layer are pre-trained.

4. The data resource management method of claim 3, wherein, The step of identifying whether the category of the corresponding paperless bill under each length-width ratio is unique according to the length-width ratio measurement result and the preset bill theme extraction component comprises: clustering the paperless bills with the same length-width ratio to obtain clustering results; performing loop random sampling on the paperless bills in each clustering cluster N times to obtain a random sampling bill set corresponding to each clustering cluster; extracting bill theme content from the elements in the random sampling bill set using the bill theme extraction component to obtain bill theme content corresponding to each random sampling bill set; identifying whether the bill theme content corresponding to the current random sampling bill set is consistent; If the current random sampling ticket set corresponds to the same ticket theme content, the corresponding paperless ticket under the current aspect ratio is unique in class; If the current random sampling ticket set corresponds to different ticket theme content, the corresponding paperless ticket under the current aspect ratio is not unique in class.

5. The data resource management method of claim 3, wherein, The step of obtaining the class output node of the ticket class identification model when each paperless ticket is classified and output, specifically includes: In the first identification layer, the aspect ratio of the current paperless ticket is measured as the aspect ratio to be classified by using the ticket ratio measuring tool; Identify whether there is a corresponding class output node in the model output layer for the aspect ratio to be classified; If the aspect ratio to be classified has a corresponding class output node in the model output layer, mark the class output node as the class output node when the current paperless ticket is classified and output; If the aspect ratio to be classified does not have a corresponding class output node in the model output layer, use the ticket theme extraction component to extract the ticket theme content of the current paperless ticket, obtain the corresponding ticket theme content of the current paperless ticket, and identify the corresponding class output node of the ticket theme in the model output layer, and mark it as the class output node when the current paperless ticket is classified and output.

6. The data resource management method of claim 1, wherein, The step of obtaining paperless tickets from multiple sources includes: According to the third party service interface connected to the target service APP, a plurality of different service APPs respectively received paperless tickets in a preset time period are obtained as paperless tickets from non-self source channels, and the target service APP includes a target business travel management APP pre-installed and used by a user's mobile terminal; Through the paperless ticket receiving interface of the target service APP itself, the paperless tickets received by the target service APP in the preset time period are obtained as paperless tickets from the self source channel; The paperless tickets from non-self source channels and the paperless tickets from self source channels are summarized as paperless tickets from multiple sources.

7. The data resource management method of claim 6, wherein, The target business travel management APP includes a paperless ticket resource management component, and the step of combining the differentiated annotation result and the preset data resource management strategy to manage the paperless tickets from multiple sources includes: According to the differentiated annotation result, identify the corresponding class output node of each ticket in the paperless tickets from multiple sources; If the corresponding class output node of the current paperless ticket is an element in the first class output node set, the current paperless ticket is added to the first cache space in the paperless ticket resource management component for caching; If the corresponding class output node of the current paperless ticket is an element in the second class output node set, the current paperless ticket is added to the second cache space in the paperless ticket resource management component for caching.

8. A data resource management apparatus, characterized by comprising: It includes: A paperless ticket acquisition module for acquiring paperless tickets from multiple sources; An identification model input module for inputting all paperless tickets into a pre-trained ticket class identification model; An identification model input module for inputting all paperless tickets into a pre-trained ticket class identification model; A category output node acquisition module is configured to acquire a category output node when the bill category identification model classifies each paperless bill; A differentiation labeling module is configured to differentiate and label each paperless bill according to the category output node and the issuing number of the paperless bill; A differentiation management module is configured to combine the differentiation labeling result and a preset data resource management strategy to differentiate and manage the paperless bills from the multiple sources.

9. A computer device, comprising: A computer readable storage medium stores computer readable instructions, and a processor executes the computer readable instructions to implement the steps of the data resource management method according to any one of claims 1 to 7.

10. A computer readable storage medium characterized by The computer readable storage medium stores computer readable instructions, and a processor executes the computer readable instructions to implement the steps of the data resource management method according to any one of claims 1 to 7.

Images