Automatic audit method, device, electronic device and storage medium

Abstract

The present application discloses an automatic audit method, device, electronic device and storage medium, which belongs to the field of data processing technology. The method comprises: identifying multiple fields from the text information of the image to be audited; calculating the matching score and matching times of each field according to the business record corresponding to the text information; inputting the matching score of the field into at least one first decision tree to screen the target image to be audited associated with the business record; inputting the matching score and matching times of the field in the target image to be audited into the second decision tree, and when the output result of the second decision tree indicates that the text information meets the audit rules, the business record passes the audit.

Description

Technical Field

[0001] The present application belongs to the field of data processing technology, and specifically relates to an automatic review method, device, electronic equipment and storage medium. Background Art

[0002] With the development of informatization, a large amount of data is stored in computer systems. However, there are still many scenarios where paper documents are needed to verify the authenticity and validity of data in computer systems.

[0003] Most companies use manual auditing, which is costly. The related automatic auditing method uses optical character recognition (OCR) to recognize text in images of paper documents, stores the text in a database after structural analysis, and then verifies it by a computer system. However, this method has poor recognition effect on borderless text or blurred text, and cannot identify the types of different paper documents, and thus cannot determine whether the information in the paper documents is valid, resulting in low audit accuracy. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide an automatic audit method, device, electronic device and storage medium, which can improve the accuracy of automatic audit while reducing the audit cost.

[0005] In order to solve the above technical problems, this application is implemented as follows:

[0006] In a first aspect, an embodiment of the present application provides a method for automatic review, the method comprising: identifying multiple fields from the text information of an image to be reviewed; calculating the matching score and the number of matches of each of the fields according to the business record corresponding to the text information; inputting the matching score of the field into at least one first decision tree to screen a target image to be reviewed that is associated with the business record; inputting the matching score and the number of matches of the field in the target image to be reviewed into a second decision tree, and when the output result of the second decision tree indicates that the text information satisfies the review rules, the business record passes the review.

[0007] In the second aspect, an embodiment of the present application provides an automatic review device, which includes an identification module for identifying multiple fields from the text information of the image to be reviewed; a calculation module for calculating the matching score and the number of matches of each of the fields according to the business record corresponding to the text information; a screening module for inputting the matching score of the field into at least one first decision tree to screen the target image to be reviewed associated with the business record; and a review module for inputting the matching score and the number of matches of the field in the target image to be reviewed into a second decision tree, and when the output result of the second decision tree indicates that the text information meets the review rules, the business record passes the review.

[0008] In a third aspect, an embodiment of the present application provides an electronic device comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the automatic audit method described in the first aspect.

[0009] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the automatic audit method described in the first aspect are implemented.

[0010] In an embodiment of the present application, multiple fields are identified from the text information of the image to be reviewed; the matching score and the number of matches of each field are calculated according to the business record corresponding to the text information; the matching score of the field is input into at least one first decision tree to screen the target image to be reviewed associated with the business record; the matching score and the number of matches of the field in the target image to be reviewed are input into a second decision tree, and when the output result of the second decision tree indicates that the text information meets the review rules, the business record passes the review, which can reduce the review cost while improving the accuracy of automatic review. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0012] Figure 1 A schematic diagram showing a flow chart of the automatic audit method provided in an embodiment of the present application;

[0013] Figure 2 Another flow chart of the automatic audit method provided in the embodiment of the present application is shown;

[0014] Figure 3 A schematic diagram of a process for constructing a decision tree provided in an embodiment of the present application is shown;

[0015] Figure 4 A schematic diagram showing the structure of an automatic audit device provided in an embodiment of the present application is shown;

[0016] Figure 5 A schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0018] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0019] The automatic audit method provided in the embodiment of the present application is described in detail below through specific embodiments and their application scenarios in conjunction with the accompanying drawings.

[0020] Figure 1 1 is a flow chart of the automatic audit method provided by the embodiment of the present application, which can be executed by an electronic device, such as a terminal device or a server device. In other words, the method can be executed by software or hardware installed in the terminal device or the server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. As shown in the figure, the method may include the following steps.

[0021] S101: Identify multiple fields from text information of an image to be reviewed.

[0022] The image to be reviewed may be an image of a document such as an ID card or a bank card, or may be an image with text information such as a bill, a list or a report.

[0023] Specifically, the text information of the image to be reviewed can be extracted and identified using a general text interface, and then the key fields in the text information can be obtained through data mining. For example, for an ID card image, after data mining, the information corresponding to the attribute fields such as name, ethnicity, gender, date of birth, ID number, address, etc. in the ID card image can be extracted separately into a document. Optionally, special characters in the above text information, such as asterisks and colons, are filtered to avoid matching errors in subsequent steps caused by special characters.

[0024] S102: Calculate the matching score and matching times of each field according to the business record corresponding to the text information.

[0025] Specifically, the business record includes but is not limited to data records such as reimbursement applications, claim applications or vacation records. The business record has information of multiple fields, and according to the fields of the business record, zero, one or more corresponding fields can be matched in the text information of the image to be reviewed. Optionally, 1 indicates that the corresponding field is matched, and 0 indicates that the corresponding field is not matched. Each time the corresponding field is matched, the number of matches increases by one. For example, the name field in the business record can be matched in the text information of 5 images to be reviewed, and the number of matches and the matching score of the name field are (5,1).

[0026] S103: Input the matching scores of the fields into at least one first decision tree to screen target images to be reviewed that are associated with the business record.

[0027] Specifically, each of the first decision trees has a screening rule for a field of an image type. According to the screening rule, the images to be reviewed can be processed in batches, and the images to be reviewed that do not conform to the business record can be deleted. For example, the business record records the hospitalization data of user A, including the identity information, admission and discharge time, and medication record of user A. If the identity information and medication record of user A in the image to be reviewed are the same as those in the business record, but the admission and discharge time is different from that in the business record, then the above-mentioned image to be reviewed is not associated with the business record.

[0028] Optionally, the image type of the image to be reviewed can be determined according to the first decision tree. The image to be reviewed passes through at least one first decision tree, indicating that the image to be reviewed has at least one field that meets one image type.

[0029] S104: Input the matching score and matching times of the fields in the target image to be reviewed into a second decision tree. When the output result of the second decision tree indicates that the text information meets the review rules, the business record passes the review.

[0030] Specifically, the number of matches can reflect the number of target images to be reviewed, and it is more flexible to comprehensively judge the completeness of the images to be reviewed based on the number of target images to be reviewed and the matching score. For example, user A uploads 3 bill images as proof of business record 1. These 3 bill images can match all fields in business record 1, so business record 1 passes the review. User B loses one of the bills, but uploads multiple file images that can prove the information in the lost bill. The image uploaded by user B can match all fields in business record 2, that is, business record 2 passes the review.

[0031] Optionally, the audit rules include scoring weights for various fields in the business record, and different fields correspond to different score thresholds. For example, in the audit rules for ID card images, the score thresholds for ID card number and name are higher, while the score thresholds for fields such as gender, ethnicity, date of birth, etc. are lower.

[0032] The embodiment of the present application identifies multiple fields from the text information of the image to be reviewed; calculates the matching score and the number of matches of each field according to the business record corresponding to the text information; inputs the matching score of the field into at least one first decision tree to screen the target image to be reviewed associated with the business record; inputs the matching score and the number of matches of the field in the target image to be reviewed into a second decision tree, and when the output result of the second decision tree indicates that the text information meets the review rules, the business record passes the review, which can reduce the review cost while improving the accuracy of automatic review.

[0033] Figure 2 is another flow chart of the automatic audit method provided in the embodiment of the present application, such as Figure 2 As shown, the method may include the following steps.

[0034] S201: Identify multiple fields from text information of an image to be reviewed.

[0035] This step can be done using Figure 1 The description of step S101 of the embodiment will not be repeated here.

[0036] S202: searching the text information for a numerical field identical to the numerical field in the business record, and when an identical numerical field is found, setting the matching score to 1 and the matching times to 1.

[0037] Specifically, the numerical fields include but are not limited to age, ID number, time, date and other data. Optionally, multiple date representation formats are adapted, such as "January 1, 2022", "2022-01-01", "20220101", "2022 / 01 / 01", etc.

[0038] Optionally, when a numerical field in the image to be reviewed is partially blocked, resulting in an incomplete numerical field in the text information, a missing field query can be performed.

[0039] S203: searching the text information for a text field corresponding to the text field in the business record, and setting the matching score to 1 and the matching times to 1 when the same text field is found.

[0040] Specifically, the vector of each text field is obtained through the word vector model, and the similarity between the vector of the text field in the business record and the vector of the text field in the text information is calculated. When the similarity is 1, the text fields are determined to be the same, and the matching score is set to 1. Each time an identical text field is found, the number of matches is increased by 1.

[0041] Optionally, the similarity between vectors can be calculated using methods such as vector inner product or cosine similarity. For example, the angle between two vectors can be calculated using cosine similarity. The smaller the angle, the more similar the two vectors are. For example, the first vector is v(5,3,4,4) and the second vector is u(3,1,2,3). The cosine similarity between the first vector v and the second vector u is:

[0042]

[0043]

[0044] Among them, sim uv Represents the similarity between vectors u and v, and cos(u,v) represents the cosine angle between vectors u and v.

[0045] S204: When the same text field cannot be found, split the text field in the business record into at least one word, search the text information for a text field having similarity with the word, and when the similarity is greater than a preset threshold, set the matching score to the highest similarity and the number of matches to 1.

[0046] Specifically, the long text field in the business record is split into words, and the vector of the word is obtained through the word vector model. The word vector is used to perform a fuzzy search in the text field of the image to be reviewed, and the similarity between the word vector and the vector of the found text field is calculated. The text fields with similarity scores higher than the preset threshold are screened out, and the highest similarity score is used as the matching score of the word. The remaining words after the split also obtain corresponding matching scores according to the above steps. In the case of splitting multiple words, the highest matching score is selected as the matching score of the above long text field.

[0047] Optionally, when a text field in a business record matches text fields in multiple images to be reviewed, a median of the multiple matching scores is used as the matching score of the text field in the business record.

[0048] S205: Input the matching score of the field into at least one first decision tree to screen the target to-be-reviewed images associated with the business record.

[0049] This step can be done using Figure 1 The description of step S103 of the embodiment will not be repeated here.

[0050] S206: Calculate a comprehensive score of each of the fields according to the matching score and matching times of each of the fields.

[0051] Optionally, the product of the match score and the number of matches of the field is used as the comprehensive score of the field. For example, if the number of matches and the match score of the name field are (10, 1), the comprehensive score of the name field is 10. If the number of matches and the match score of the admission diagnosis disease name field are (3, 0.8), the comprehensive score of the admission diagnosis disease name field is 2.4. This method can indirectly verify whether the types of images to be reviewed are complete.

[0052] S207: Input the comprehensive score into a second decision tree. When the output result of the second decision tree indicates that the comprehensive score meets the audit rules, the business record passes the audit.

[0053] This step can be done using Figure 1 The description of step S104 of the embodiment will not be repeated here.

[0054] In one implementation, the automatic audit method further includes: before calculating the matching score and matching number of each field according to the business record corresponding to the text information, extracting any identity identification information in the text information; and matching the business record corresponding to the identity identification information.

[0055] Specifically, the text information includes at least one type of identity information, such as name, ID number, social security card number or unified social credit code, etc. According to the above identity information, at least one business record in the database can be matched. Optionally, the above identity information is composed of letters and / or numbers, and the general certificate number is set as a regular expression to improve the efficiency of matching business records.

[0056] The embodiment of the present application identifies multiple fields from the text information of the image to be reviewed; searches for the same numerical field as the numerical field in the business record in the text information, and sets the matching score to 1 and the matching number to 1 when the same numerical field is found; searches for the text field corresponding to the text field in the business record in the text information, and sets the matching score to 1 and the matching number to 1 when the same text field is found; if the same text field cannot be found, splits the text field in the business record into at least one word, searches for a text field with similarity to the word in the text information, and sets the matching score to the highest similarity and the matching number to 1 when the similarity is greater than a preset threshold; inputs the matching score of the field into at least one first decision tree to screen the target image to be reviewed associated with the business record; calculates the comprehensive score of each field according to the matching score and matching number of each field; inputs the comprehensive score into a second decision tree, and when the output result of the second decision tree indicates that the comprehensive score meets the review rule, the business record passes the review, which can improve the accuracy of automatic review and verify whether the types of images to be reviewed are complete.

[0057] The embodiment of the present application can improve the efficiency of matching business records by extracting any identity identification information in the text information before calculating the matching score and the number of matches of each field according to the business record corresponding to the text information; and matching the business record corresponding to the identity identification information.

[0058] Figure 3 It is a flow chart of constructing a decision tree provided in an embodiment of the present application. As shown in the figure, the method may include the following steps.

[0059] S301: constructing a first sample data set and a first verification data set according to fields and matching scores included in multiple image types.

[0060] Specifically, the fields contained in the various image types that need to be trained are obtained. For example, a hospital prescription contains at least fields such as hospital name, department name, date of consultation, name, gender, diagnosed disease name, doctor's name, total medical expenses, etc., and an ID card contains at least fields such as ID number, name, gender, ethnicity, date of birth, etc.

[0061] Optionally, the matching score is in the range of 0-1, 0 indicates that the sample image does not contain the field, 1 indicates that the sample image contains the same field, and 0.1-0.9 indicates the similarity score between the field in the sample image and the field contained in the above image type.

[0062] S302: Calculate the Gini index of each field in the first sample data set.

[0063] Specifically, the Gini index is used to determine the relevance of each field to the screening rule. The larger the Gini index, the smaller the impact of the corresponding field on the screening result, and the smaller the Gini index, the greater the impact of the corresponding field on the screening result. The calculation method of the Gini index is shown in Formula 1.

[0064]

[0065] S303: Select the field with the largest Gini index as the root node of the first decision tree.

[0066] Specifically, the first decision tree is a binary tree, that is, the screening results can be divided into two categories according to the root node field. For example, when screening ID card images, if the root node field is the ID card number, the images containing the ID card number pass the screening, and the images without the ID card number are deleted.

[0067] S304: Calculate the Gini index of each remaining field in at least one branch, and use the field with the largest Gini index as a child node until no field remains or the Gini index is less than a threshold.

[0068] Specifically, the Gini index of the remaining fields is calculated in each branch of the root node, and the field with the largest Gini index among the remaining fields is used as a branch node. Correspondingly, the above steps are repeated in each branch of the branch node until no field is left or the Gini index is less than a threshold, that is, the field corresponding to the Gini index cannot continue to divide the first sample data set.

[0069] S305: Evaluate and adjust nodes of the first decision tree according to the first validation data set to obtain a trained first decision tree.

[0070] Optionally, a post-pruning method is used to reduce the risk of overfitting of the first decision tree during the training process. Specifically, the first decision tree is examined from the bottom up to determine whether the screening effect of the first decision tree on the first validation data set is improved after the leaf node in the first decision tree is replaced by the subtree where the leaf node is located. After multiple judgments, the first decision tree with the best effect is selected.

[0071] In one implementation, the construction method of the second decision tree includes constructing a second sample data set and a second verification data set according to the number of matches, matching scores and audit results of multiple fields in the business record; calculating the Gini index of each field in the second sample data set; selecting the field with the largest Gini index as the root node of the second decision tree; calculating the Gini index of each remaining field in at least one branch, and using the field with the largest Gini index as a child node until no field remains or the Gini index is less than a threshold. The second decision tree is evaluated and nodes are adjusted according to the second verification data set to obtain a trained second decision tree. Specifically, the audit results of the business record include complete and incomplete. The construction method of the above-mentioned second decision tree is similar to the construction method of the first decision tree, and will not be repeated here.

[0072] The embodiment of the present application constructs a first sample data set and a first verification data set according to the fields and matching scores included in multiple image types; calculates the Gini index of each field in the first sample data set; selects the field with the largest Gini index as the root node of the first decision tree; calculates the Gini index of each remaining field in at least one branch, and uses the field with the largest Gini index as a child node until no fields remain or the Gini index is less than a threshold; evaluates and adjusts the nodes of the first decision tree according to the first verification data set to obtain a trained first decision tree, which can construct a first decision tree for screening target images to be reviewed associated with business records, thereby reducing the review cost.

[0073] The embodiment of the present application constructs a second sample data set and a second verification data set according to the number of matches, matching scores and audit results of multiple fields in the business record; calculates the Gini index of each field in the second sample data set; selects the field with the largest Gini index as the root node of the second decision tree; calculates the Gini index of each remaining field in at least one branch, and uses the field with the largest Gini index as a child node until no field remains or the Gini index is less than a threshold. The second decision tree is evaluated and nodes are adjusted according to the second verification data set to obtain a trained second decision tree, which can construct a second decision tree for auditing the completeness of the target image to be audited, thereby improving the accuracy of automatic auditing.

[0074] Figure 4 A schematic diagram of the structure of an automatic audit device provided in an embodiment of the present application is shown. The device 400 includes: an identification module 401, a calculation module 402, a screening module 403 and an audit module 404.

[0075] The identification module 401 is used to identify multiple fields from the text information of the image to be reviewed; the calculation module 402 is used to calculate the matching score and the number of matches of each field according to the business record corresponding to the text information; the screening module 403 is used to input the matching score of the field into at least one first decision tree to screen the target image to be reviewed associated with the business record; the review module 404 is used to input the matching score and the number of matches of the field in the target image to be reviewed into the second decision tree, and when the output result of the second decision tree indicates that the text information meets the review rules, the business record passes the review.

[0076] In one implementation, the calculation module 402 is also used to search for a numerical field in the text information that is identical to a numerical field in the business record, and when the identical numerical field is found, set the matching score to 1 and the number of matches to 1; search for a text field in the text information that corresponds to a text field in the business record, and when the identical text field is found, set the matching score to 1 and the number of matches to 1.

[0077] In one implementation, the calculation module 402 is also used to split the text field in the business record into at least one word when the same text field cannot be found, search for a text field with similarity to the word in the text information, and when the similarity is greater than a preset threshold, set the matching score to the highest similarity and the number of matches to 1.

[0078] In one implementation, the calculation module 402 is further configured to extract any identity information in the text information; and match the business record corresponding to the identity information.

[0079] In one implementation, the audit module 404 is also used to calculate a comprehensive score for each of the fields based on the matching scores and matching times of each of the fields; the comprehensive score is input into a second decision tree, and when the output result of the second decision tree indicates that the comprehensive score meets the audit rules, the business record passes the audit.

[0080] In one implementation, the screening module 403 is also used to construct a first sample data set and a first verification data set based on the fields and matching scores included in multiple image types; calculate the Gini index of each field in the first sample data set; select the field with the largest Gini index as the root node of the first decision tree; calculate the Gini index of each remaining field in at least one branch, and use the field with the largest Gini index as a child node until no fields remain or the Gini index is less than a threshold; evaluate and adjust the nodes of the first decision tree based on the first verification data set to obtain a trained first decision tree.

[0081] In one implementation, the audit module 404 is also used to construct a second sample data set and a second verification data set based on the number of matches, matching scores and audit results of multiple fields in the business record; calculate the Gini index of each field in the second sample data set; select the field with the largest Gini index as the root node of the second decision tree; calculate the Gini index of each remaining field in at least one branch, and use the field with the largest Gini index as a child node until no fields remain or the Gini index is less than a threshold; evaluate and adjust the nodes of the second decision tree according to the second verification data set to obtain a trained second decision tree.

[0082] Figure 5 The hardware structure diagram of the electronic device 500 provided by the embodiment of the present application is shown. Referring to the figure, at the hardware level, the electronic device 500 includes a processor 504, and optionally, includes an internal bus 503, a network interface 501, and a memory. Among them, the memory may include a memory 502, such as a high-speed random access memory (Random-Access Memory, RAM), and may also include a non-volatile memory 505 (non-volatile memory), such as at least one disk storage, etc. Of course, the electronic device 500 may also include hardware required for other services.

[0083] The processor 504, the network interface 501 and the memory can be connected to each other through the internal bus 503, and the internal bus 503 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one bidirectional arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0084] The memory is used to store programs. Specifically, the program may include program codes, and the program codes include computer operation instructions. The memory may include a memory 502 and a non-volatile memory 505, and provide instructions and data to the processor 504.

[0085] The processor 504 reads the corresponding computer program from the non-volatile memory 505 into the memory 502 and then runs it, forming a device for locating the target user at the logical level. The processor 504 executes the program stored in the memory and is specifically used to execute the method in the above embodiment. Figure 1-3 The methods described above are implemented in the above method embodiments Figure 1-3 The functions and beneficial effects of the methods described above will not be described in detail here.

[0086] The above application Figure 1-3 The method disclosed in the illustrated embodiment can be applied to the processor 504, or implemented by the processor 504. The processor 504 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 504. The above processor 504 can be a general-purpose processor, including a central processing unit (CPU), a network (NP), etc.; it can also be a digital signal (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module may be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor 504 reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0087] The electronic device 500 can also execute the method in the above embodiment Figure 1-3 The methods described above are implemented in the above method embodiments Figure 1-3 The functions and beneficial effects of the methods described above will not be described in detail here.

[0088] Of course, in addition to software implementation, the electronic device 500 of the present application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0089] The present application also provides a computer-readable storage medium, wherein the computer-readable medium stores one or more programs, and when the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes the method in the above method embodiment. Figure 1-3 The methods described above are implemented in the above method embodiments Figure 1-3 The functions and beneficial effects of the methods described above will not be described in detail here.

[0090] The computer-readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0091] Furthermore, the embodiment of the present application also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, and when the program instructions are executed by a computer, the method in the above embodiment is implemented. Figure 1-3 The functions and beneficial effects of the methods described above will not be described in detail here.

[0092] In short, the above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0093] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0094] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0095] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0096] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0097] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. An automatic audit method, characterized in that: include: identifying a plurality of fields from text information of the image to be reviewed; Calculate the matching score and matching times of each field according to the business record corresponding to the text information; Inputting the matching score of the field into at least one first decision tree to screen the target to-be-reviewed image associated with the business record; Inputting the matching score and matching number of the field in the target image to be reviewed into a second decision tree, and when the output result of the second decision tree indicates that the text information meets the review rule, the business record passes the review; Each of the first decision trees has a screening rule for a field of an image type, and the images to be reviewed that do not conform to the business record are deleted according to the screening rule. The first decision tree is also used to determine the image type of the image to be reviewed.

2. The automatic audit method according to claim 1, characterized in that: Calculating the matching score and the matching number of each field according to the business record corresponding to the text information includes: Searching for a numerical field in the text information that is identical to a numerical field in the business record, and if an identical numerical field is found, setting a matching score to 1 and a matching count to 1; A text field corresponding to the text field in the business record is searched in the text information, and when the same text field is found, the matching score is set to 1 and the matching number is set to 1.

3. The automatic audit method according to claim 2, characterized in that: Calculating the matching score and the matching number of each field according to the business record corresponding to the text information also includes: If the same text field cannot be found, the text field in the business record is split into at least one word, and a text field having similarity with the word is searched in the text information. If the similarity is greater than a preset threshold, the matching score is set to the highest similarity, and the number of matches is set to 1.

4. The automatic audit method according to claim 1, characterized in that: Before calculating the matching score and the matching number of each field according to the business record corresponding to the text information, the method further includes: Extracting any identity information from the text information; Match the business record corresponding to the identity information.

5. The automatic audit method according to claim 1, characterized in that: The step of inputting the matching score and the matching number of the field in the target image to be reviewed into the second decision tree, and when the output result of the second decision tree indicates that the text information satisfies the review rule, the business record passes the review, comprises: Calculating a comprehensive score of each of the fields according to the matching score and matching times of each of the fields; The comprehensive score is input into a second decision tree, and when the output result of the second decision tree indicates that the comprehensive score meets the audit rules, the business record passes the audit.

6. The automatic audit method according to claim 1, wherein: The method for constructing the first decision tree includes: Constructing a first sample data set and a first validation data set according to the fields and matching scores included in the multiple image types; Calculating the Gini index of each field in the first sample data set; Select the field with the largest Gini index as the root node of the first decision tree; Calculate the Gini index of each remaining field in at least one branch, and take the field with the largest Gini index as a child node until no field remains or the Gini index is less than a threshold; The first decision tree is evaluated and nodes are adjusted according to the first validation data set to obtain a trained first decision tree.

7. The automatic audit method according to claim 1, wherein: The construction method of the second decision tree includes: constructing a second sample data set and a second verification data set according to the matching times, matching scores and audit results of multiple fields in the business records; Calculating the Gini index of each field in the second sample data set; Select the field with the largest Gini index as the root node of the second decision tree; Calculate the Gini index of each remaining field in at least one branch, and take the field with the largest Gini index as a child node until no field remains or the Gini index is less than a threshold; The second decision tree is evaluated and nodes are adjusted according to the second validation data set to obtain a trained second decision tree.

8. An automatic audit device, characterized in that: include: A recognition module, for recognizing a plurality of fields from text information of the image to be reviewed; A calculation module, used to calculate the matching score and matching times of each field according to the business record corresponding to the text information; A screening module, used for inputting the matching scores of the fields into at least one first decision tree to screen the target to-be-reviewed images associated with the business record; An audit module, used for inputting the matching score and matching number of the fields in the target image to be audited into a second decision tree, and when the output result of the second decision tree indicates that the text information satisfies the audit rule, the business record passes the audit; Each of the first decision trees has a screening rule for a field of an image type, and the images to be reviewed that do not conform to the business record are deleted according to the screening rule. The first decision tree is also used to determine the image type of the image to be reviewed.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the automatic audit method as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the steps of the automatic audit method according to any one of claims 1 to 7 are implemented.

Images