Object detection device, method, electronic device, storage medium and program product

Abstract

The present disclosure provides an object detection device, method, electronic device, storage medium and program product, which relate to the field of computer technology, especially to the field of artificial intelligence and deep learning technology. According to the object detection device of the present disclosure, it includes a classification module, an identification module, a document value detection result determination module, a document validity determination module and a document detection result determination module, wherein the classification module is used to classify multiple document objects to obtain document categories; the identification module is used to respectively identify the numerical information of multiple document objects to obtain document values; the document value detection result determination module is used to obtain document value detection results according to the document value association relationship of multiple document objects; the document validity determination module is used to perform validity detection on multiple document objects to obtain document validity; the document detection result determination module is used to determine the document detection result according to the document value detection result and document validity.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, in particular to the field of artificial intelligence and deep learning technology, and specifically to an object detection device, method, electronic device, storage medium, and program product. Background Art

[0002] There are many types of document objects, so detecting document objects is a tedious task. How to improve the efficiency of document object detection is an urgent problem to be solved. Summary of the invention

[0003] The present disclosure provides an object detection device, method, electronic device, storage medium, and program product.

[0004] According to one aspect of the present disclosure, there is provided an object detection device, comprising: a classification module, an identification module, a document value detection result determination module, a document validity determination module and a document detection result determination module, wherein the classification module is used to classify a plurality of document objects to obtain document categories; the identification module is used to respectively identify the value information of the plurality of document objects to obtain document values; the document value detection result determination module is used to obtain a document value detection result according to a document value association relationship of the plurality of document objects; the document validity determination module is used to perform validity detection on the plurality of document objects to obtain document validity; the document detection result determination module is used to determine a document detection result according to the document value detection result and the document validity.

[0005] According to another aspect of the present disclosure, there is provided an object detection method, comprising: classifying a plurality of document objects to obtain document categories; respectively identifying numerical information of the plurality of document objects to obtain document numerical values; obtaining a document numerical detection result based on a document numerical association relationship of the plurality of document objects; performing validity detection on the plurality of document objects to obtain document validity; and determining a document detection result based on the document numerical detection result and the document validity.

[0006] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in an embodiment of the present disclosure.

[0007] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method according to the embodiment of the present disclosure.

[0008] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the method according to the embodiment of the present disclosure is implemented.

[0009] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0011] Figure 1 The following schematically shows a system architecture diagram of an object detection method and device according to an embodiment of the present disclosure;

[0012] Figure 2 A flowchart of an object detection method according to an embodiment of the present disclosure is schematically shown;

[0013] Figure 3 A schematic diagram schematically shows an object detection method according to an embodiment of the present disclosure;

[0014] Figure 4 A schematic diagram schematically shows an object detection method according to another embodiment of the present disclosure;

[0015] Figure 5 A schematic diagram schematically shows an object detection method according to another embodiment of the present disclosure;

[0016] Figure 6 A schematic diagram schematically shows an object detection method according to another embodiment of the present disclosure;

[0017] Figure 7 A schematic diagram schematically shows an object detection method according to another embodiment of the present disclosure;

[0018] Figure 8 A schematic diagram schematically shows the application of the object detection method according to an embodiment of the present disclosure to a disease document object claim settlement scenario;

[0019] Fig. 9 A schematic diagram schematically shows an object detection device according to an embodiment of the present disclosure;

[0020] Fig.10 The block diagram schematically shows an electronic device that can implement the object detection method according to the embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0022] Figure 1 The system architecture of the object detection method and device according to the embodiment of the present disclosure is schematically shown. It should be noted that: Figure 1 What is shown is merely an example of a system architecture to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical content of the present disclosure, but it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0023] like Figure 1 As shown, the system architecture 100 according to this embodiment may include clients 101A, 101B, 101C, a network 102, and a server 103. The network 102 is used to provide a medium for communication links between the clients 101A, 101B, 101C and the server 103. The network 102 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0024] Users can use clients 101A, 101B, 101C to interact with server 103 through network 102 to receive or send messages, etc. Various communication client applications can be installed on clients 101A, 101B, 101C, such as navigation applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0025] The clients 101A, 101B, and 101C may be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers, etc. The clients 101A, 101B, and 101C of the disclosed embodiment may, for example, run applications.

[0026] The server 103 may be a server that provides various services, such as a backend management server that provides support for websites browsed by users using the clients 101A, 101B, and 101C (for example only). The backend management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the client. In addition, the server 103 may also be a cloud server, that is, the server 103 has a cloud computing function.

[0027] It should be noted that the object detection method provided in the embodiment of the present disclosure can be executed by the server 103. Accordingly, the object detection device provided in the embodiment of the present disclosure can be arranged in the server 103. The object detection method provided in the embodiment of the present disclosure can also be executed by a server or a server cluster that is different from the server 103 and can communicate with the client 101A, 101B, 101C and / or the server 103. Accordingly, the object detection device provided in the embodiment of the present disclosure can also be arranged in a server or a server cluster that is different from the server 103 and can communicate with the client 101A, 101B, 101C and / or the server 103.

[0028] In an example, the server 103 may obtain the document object from the clients 101A, 101B, and 101C through the network 102 .

[0029] It should be understood that Figure 1 The number of clients, networks and servers in the embodiment is only for illustration. Any number of clients, networks and servers may be provided as required.

[0030] It should be noted that in the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0031] Figure 2 The flowchart of the object detection method according to an embodiment of the present disclosure is exemplified.

[0032] like Figure 2 As shown, the object detection method 200 according to an embodiment of the present disclosure includes operations S210 to S250.

[0033] In operation S210, multiple document objects are classified to obtain document categories.

[0034] The object detection method of the disclosed embodiment will be described below by taking disease insurance claims as an example. According to the requirements of insurance claims, the document objects required for insurance claims include: medical records, prescriptions, payment slips, examination slips, invoices, etc., which can be used as different document categories.

[0035] For example, the image information of the document object can be obtained by scanning, and the image information of the document object can be input into the document classification model to obtain the document category. The document classification model can include a VGG model, a ResNet model, and a DenseNet model. Among them, VGG is Visual Geometry Group. ResNet is a deep residual network. DenseNet is a densely connected network.

[0036] In operation S220, numerical information of a plurality of document objects is respectively identified to obtain document values.

[0037] During the insurance claim process, it is necessary to detect the numerical information of the document object, which can be understood as information indicating the amount. For example, the payment amount of the document object, such as the payment slip, is the document value.

[0038] It should be understood that there are some document objects of the document category from which the document value cannot be obtained. For example, the document object of the document category of medical records does not include amount information, so the document value representing the amount information cannot be obtained from the medical records.

[0039] For example, the numerical information of the document object can be identified by optical character recognition technology to obtain the document value. Optical character recognition is Optical Character Recognition, referred to as OCR. Specifically, the staff can mark the location of the numerical information of the document object in the image information of the document object, and then identify the text information at the location by optical character recognition technology to obtain the document value.

[0040] In operation S230, a document value detection result is obtained according to the document value association relationship of the plurality of document objects.

[0041] During the insurance claims process, there is a mutual verification relationship between various document objects. For example, multiple document objects have a document value association relationship. The document value of each document object can be verified by the document value association relationship to obtain a document value detection result.

[0042] In operation S240, validity check is performed on multiple document objects to obtain document validity.

[0043] Document validity can be understood as the validity of the document object for the current insurance claim. For example, for the insurance claim of hypertension, the document object related to the car accident surgery does not have document validity.

[0044] In operation S250, a document detection result is determined according to the document value detection result and the document validity.

[0045] The object detection method of the disclosed embodiment can classify document objects and determine the document category. For example, document objects of different document categories can be used to verify whether the document objects for insurance claims are complete. The document values ​​are also identified and detected, and the document value detection result obtained can indicate whether the document value of the document object is correct. In addition, the validity detection is also performed on multiple document objects, and the object detection result is accurately determined by comprehensively considering the three aspects of whether the document objects are complete, whether the values ​​of the document objects are correct, and whether the document objects are valid. This can realize automatic detection of document objects, save human resources, and improve object detection efficiency.

[0046] It should be noted that the above operation numbers S210 to S250 only represent different operations, and do not represent the order of operations. The document detection results of the embodiment of the present disclosure can also be obtained by adjusting the above operation order.

[0047] Figure 3 A schematic diagram of an object detection method according to an embodiment of the present disclosure is schematically shown.

[0048] Operations S310 to S350 of the object detection method 300 according to the embodiment of the present disclosure correspond to the above-mentioned operations S210 to S250, respectively. That is, in operation S310, multiple document objects 301 are classified to obtain document categories 302. In operation S320, the numerical information of multiple document objects 301 is respectively identified to obtain document numerical values ​​303. In operation S330, based on the document numerical association relationship of multiple document objects, a document numerical detection result 304 is obtained. In operation S340, validity detection is performed on multiple document objects 301 to obtain document validity 305. In operation S350, based on the document numerical detection result 304 and the document validity 305, a document detection result 306 is determined.

[0049] Figure 4 A schematic diagram of an object detection method according to another embodiment of the present disclosure is schematically shown.

[0050] like Figure 4 As shown, according to an object detection method 400 of another embodiment of the present disclosure, the document object may include a total document object and a sub-document object, the numerical information of the total document object includes the total number of documents, and the object detection method 400 may further include operations S460 to S470.

[0051] In operation S460 , the total document value 403 of the total document object 401 is verified to obtain a total document value verification result 407 .

[0052] In operation S470, the total document target detection result 408 is determined according to the document detection result 406 and the document total value verification result 407 of the total document object 401.

[0053] The document detection result 406 of the total document object 401 is obtained through operations S420 to S450. Operations S420 to S450 correspond to operations S220 to S250 of the above embodiment, respectively. It should be understood that the document value detection result 404 obtained through operation S430 is the detection result of the total document value 403, the document validity 405 obtained through operation S440 is the validity of the total document object 401, and the document detection result 406 obtained through operation S450 is the document detection result of the total document object 401.

[0054] The total document object includes the total document value, and the document value of the sub-document object can be included in the total document value. For example, when making a claim for disease insurance, the total amount invoice includes the amount information of each inspection form, payment form, etc. The total amount invoice can be understood as the total document object, the total amount of the total amount invoice can be understood as the total document value, and each inspection form and payment form can be understood as a sub-document object. The total document value includes the document value of the sub-document object.

[0055] The total number of documents of the total document object can be used as reference data for the insurance claim amount. The object detection method of the embodiment of the present disclosure can improve the accuracy of object detection by verifying the total number of documents of the total document object, thereby avoiding object detection errors caused by unreasonable total number of documents of the total document object or forged document objects.

[0056] Figure 5 The figure schematically shows a schematic diagram of performing validity detection on multiple document objects in an object detection method according to yet another embodiment of the present disclosure.

[0057] According to another embodiment of the present disclosure, the following embodiments may be used to implement validity detection of multiple document objects in an object detection method to obtain specific examples of document validity.

[0058] like Figure 5 As shown, in operation S541, medical information of multiple document objects 501 is identified to obtain document object knowledge graph elements 502 of the multiple document objects 501.

[0059] For example, when making a claim for disease insurance, the document object includes medical information related to the disease. Medical information may include: disease characteristics, disease symptoms, treatment methods, treatment drugs, precautions, etc. The knowledge graph elements of the document object may include disease characteristics, disease symptoms, treatment methods, treatment drugs, precautions, etc.

[0060] In operation S542 , the document object knowledge graph element coverage 504 is determined based on the document object knowledge graph element 502 and the medical knowledge graph 503 .

[0061] The medical knowledge graph can be understood as a continuously updated, medically relevant, and complete knowledge graph. The document object knowledge graph element coverage rate can be expressed, for example, as the ratio of the number of document object knowledge graph elements that appear in the medical knowledge graph to the number of document object knowledge graph elements. The document object knowledge graph element coverage rate can indicate the degree of relevance between the document object and the medical knowledge graph.

[0062] In operation S543, the document object knowledge graph element coverage 504 is compared with the validity threshold 505 to obtain the document validity 506.

[0063] Exemplarily, the effectiveness threshold may be a threshold range greater than 70%.

[0064] The object detection method of the disclosed embodiment introduces a knowledge graph, and evaluates the validity of the document by comparing the coverage of the document object knowledge graph with the validity threshold. Accurate document validity can still be obtained without manual intervention in evaluating the document validity.

[0065] Figure 6 A schematic diagram schematically shows a method for verifying the total number of documents of a total document object in an object detection method according to another embodiment of the present disclosure.

[0066] According to another embodiment of the present disclosure, the following embodiments can be used to implement the object detection method to verify the total number of documents of the total document object and obtain a specific example of the verification result of the total number of documents. Among them, the document object knowledge graph element can include a document object entity.

[0067] like Figure 6 As shown, in operation S661, the sub-document object entity of the sub-document object is determined.

[0068] Still taking the above-mentioned general document object including the total amount invoice, and the sub-document objects including the inspection form and the payment form as an example, the sub-document object entities may include: disease characteristics, therapeutic drugs, and treatment methods.

[0069] In operation S662, an initial word vector of the sub-document object entity is determined.

[0070] Word vectors can be understood as the vector representation of words in natural language processing tasks. The name of the document entity represented by word vectors can reflect the similarity between different sub-document object entities and can also contain more information. Each dimension of the initial word vector has a specific meaning. Therefore, the initial word vector can better represent the sub-document object entity.

[0071] In operation S663, pre-training is performed independently on each sub-document object to obtain a target word vector for each sub-document object.

[0072] Pre-training can be understood as the process of using as much training data as possible to extract as many common features as possible from it, thereby reducing the learning burden of the model for specific tasks.

[0073] The target word vector can be understood as the vector representation of the sub-document object entity that integrates the text semantics of the current sub-document object.

[0074] In operation S664, the document standard total value 601 of the total document object is determined according to the target word vector.

[0075] In operation S665, a document total value verification result 607 is determined based on the document standard total value 601 and the document total value 603.

[0076] Exemplarily, the document total value verification result can be determined based on the ratio of the absolute value of the difference between the document standard total value and the document total value to the document total value. For example, when the ratio of the absolute value of the difference between the document standard total value and the document total value to the document total value is less than 20%, the document total value verification result is determined to be passed.

[0077] The object detection method of the disclosed embodiment combines the document object with the knowledge graph, and the sub-document object entity can represent the medical information of the sub-document object. Since medical information has a specific meaning in the medical field, the specific meaning needs to be determined in combination with the text information of the current sub-document object. Therefore, the object detection method of the disclosed embodiment, the target word vector obtained by pre-training can more accurately represent the name of the sub-document object entity that integrates the specific meaning in the medical field. At the same time, the semantics expressed by the text information of different sub-document objects are also different. Therefore, the object detection method of the disclosed embodiment, by pre-training each sub-document object independently, will not mix the semantics expressed by the text information of different sub-document objects, so that the target word vectors determined by different sub-document objects are more accurate, thereby improving the accuracy of the document total value verification result and the object detection efficiency.

[0078] Figure 6 The schematic diagram of determining the verification result of the total number of documents according to multiple sub-document objects is exemplified. Among them, sub-document object entity 11 and sub-document object entity 1n are determined according to sub-document object 1, and sub-document object entity N1 and sub-document object entity Nn are determined according to sub-document object N. The corresponding initial word vector 111, initial word vector 1nn, initial word vector N11 and initial word vector Nnn are also determined according to each sub-document object entity. Each sub-document object is pre-trained independently to obtain the corresponding target word vector 1v and target word vector Nv.

[0079] Figure 7A schematic diagram of obtaining a target word vector for each sub-document object in an object detection method according to yet another embodiment of the present disclosure is schematically shown.

[0080] According to another embodiment of the present disclosure, the following embodiment can be used to implement the object detection method to independently pre-train each sub-document object to obtain a specific example of the target word vector of each sub-document object.

[0081] like Figure 7 As shown, in operation S731, the disease characteristic 702 of the document object 701 is determined.

[0082] In operation S732 , a target pre-trained model 704 is determined according to a mapping relationship between the disease feature 702 and the pre-trained model 703 .

[0083] In operation S733 , the initial word vector 705 of the sub-document object is pre-trained using the target pre-training model 704 to obtain the target word vector 706 .

[0084] Disease characteristics can be understood as the name of the disease, and the initial word vector is a vector representation of the sub-document object entity. The sub-document object entity reflects a certain aspect of the disease. For example, a certain sub-document object entity is a disease symptom. It may happen that different disease characteristics have the same disease symptoms, but the treatment methods and therapeutic drugs for different disease characteristics are not the same, and the text information of the corresponding sub-document objects will be very different. Therefore, the object detection method of the embodiment of the present disclosure, when determining the corresponding target word vector based on the initial sub-vector, determines the target pre-training model based on the mapping relationship between the disease characteristics and the pre-training model, and uses the target pre-training model to pre-train the initial word vector of the sub-document object, so as to obtain a more accurate target word vector and improve the efficiency of object detection.

[0085] According to the object detection method of the embodiment of the present disclosure, in the above embodiment, determining the total number of document standards for the total document objects based on the target word vector may include: inputting the target word vector into a document numerical prediction model to obtain the total number of document standards.

[0086] The document value prediction model can be understood as a prediction model that has been pre-trained with a large number of samples, whose input is the target word vector and whose output is the document standard total value. Exemplarily, the document value prediction model can include a logistic regression model.

[0087] The object detection method of the embodiment of the present disclosure, when including multiple sub-document objects, can splice the target word vectors of each sub-document object, and input the word vector after splicing the multiple target word vectors into the document numerical prediction model to obtain the total value of the document standard.

[0088] The object detection method of the disclosed embodiment can automatically and accurately obtain the standard total value of the document according to the document value prediction model.

[0089] Exemplarily, according to the object detection method of the embodiment of the present disclosure, in the above embodiment, obtaining the document value detection result according to the document value association relationship of multiple document objects may include: determining the sum of the document values ​​of at least one sub-document object according to the association relationship between the total document object and the sub-document object to obtain the document sub-value sum; comparing the total document value with the document sub-value sum to obtain the document value comparison result; and determining the document value detection result according to the document value comparison result.

[0090] For example, when the total document object is a total amount invoice and the sub-document object is a check order, the association relationship between the total document object and the sub-document object includes: the association relationship between the sum of the document values ​​of each sub-document object and the total document value of the total document object. Exemplarily, when the document value comparison result is that the total document value is less than or equal to the sum of the document sub-values, the document value detection result is determined to be passed.

[0091] When making an insurance claim, the claim conditions include: the total number of documents of the total document object is less than or equal to the sum of the document sub-values. The object detection method of the embodiment of the present disclosure can detect the document value in combination with the specific scenario of insurance claims to determine the document value detection result.

[0092] It should be understood that when there is only one sub-document object, the above-mentioned "sum of the document values ​​of at least one sub-document object" should be understood as the document value of the only sub-document object.

[0093] According to the object detection method of the embodiment of the present disclosure, the document total value verification result may include pass and fail, and the object detection method may further include: in response to the document total value verification result being fail, sending a manual review request.

[0094] In some abnormal cases, the verification result of the total number of documents may be wrong. To ensure the accuracy of object detection, the verification result of the total number of documents can be manually reviewed again.

[0095] The object detection method according to the embodiment of the present disclosure may further include: generating a document object detection report.

[0096] Exemplarily, the document object detection report may include: document category, document validity, document value detection result and document total value verification result.

[0097] Figure 8 The following is a schematic diagram showing the application of the object detection method according to an embodiment of the present disclosure in a disease document object claim settlement scenario.

[0098] like Figure 8 As shown, the user uses the claim paper document object, and can scan the claim paper document object into image information through a scanner, and input the document object in image format into the document classification model to obtain the document category, which may include: invoices, medical records, prescriptions, inspection applications, inspection reports and other document objects. Identify the document values ​​of the document objects of each document category, and detect the document values ​​to obtain the document value detection result. For example, when the document value detection result is not passed, it indicates that the user's claim document object has a document value problem. After passing the document value detection, the document validity test can be performed, that is, first extract the document object knowledge graph elements related to medicine, and obtain the document validity according to the document object knowledge graph element coverage rate relative to the medical knowledge graph document object knowledge graph element. For example, when the document validity is not passed, it indicates that the user's claim document object has a document validity problem. After passing the document validity test, the total value of the document can be verified, that is, according to the sub-document object entity of the sub-document object, the corresponding initial word vector is determined. After pre-training the initial word vector, the target word vector can be obtained. The sub-document object entity here is the document object knowledge graph element of the sub-document object. The target word vector is used as the input of the document value prediction model. After the document value prediction model predicts, the standard total value of the document of the total document object can be obtained. The standard total value of the document is compared with the total value of the document to obtain the verification result of the total value of the document. For example, when the verification result of the total value of the document fails, it means that the user's claim document object has a problem with the total value of the document. When at least one of the document value problem, document validity problem and document total value problem occurs, you can send an email to the relevant personnel for feedback. When none of the above problems occur, you can enter the claim system to make a claim.

[0099] Figure 8 The illustrated operations S810 to S840 and S860 correspond to the operations S210 to S240 and S460 of the above-mentioned embodiment, respectively, and will not be described in detail here. It should be understood that, Figure 8 In the illustrated embodiment, each operation is performed in sequence, and after operation S840 is performed on the basis of operation S830, a document detection result can be obtained according to the document value detection result obtained by operation S830 and the document validity obtained by operation S840. Similarly, after operation S860 is performed, a total document target detection result can be obtained according to the document detection result of the total document object and the document total value verification result.

[0100] According to an embodiment of the present disclosure, the present disclosure also provides an object detection device.

[0101] like Fig. 9As shown, the object detection device 900 according to the embodiment of the present disclosure includes: a classification module 910, a recognition module 920, a document value detection result determination module 930, a document validity determination module 940 and a document detection result determination module 950.

[0102] The classification module 910 is used to classify multiple document objects to obtain document categories. In one embodiment, the classification module 910 can be used to perform the above operation S210, which will not be described in detail here.

[0103] The identification module 920 is used to identify the numerical information of multiple document objects respectively to obtain the document value. In one embodiment, the identification module 920 can be used to perform the above operation S220, which will not be described in detail here.

[0104] The document value detection result determination module 930 is used to obtain the document value detection result according to the document value association relationship of multiple document objects. In one embodiment, the document value detection result determination module 930 can be used to perform the above operation S230, which will not be described in detail here.

[0105] The document validity determination module 940 is used to perform validity detection on multiple document objects to obtain document validity. In one embodiment, the document validity determination module 940 can be used to perform the above operation S240, which will not be described in detail here.

[0106] The document detection result determination module 950 is used to determine the document detection result according to the document value detection result and the document validity. In one embodiment, the document detection result determination module 950 can be used to perform the above operation S250, which will not be described in detail here.

[0107] According to the object detection device of the embodiment of the present disclosure, wherein the document object includes a total document object and a sub-document object, and the numerical information of the total document object includes a total document value, the object detection device may further include: a total document value verification module and a total document target detection result determination module.

[0108] The document total value verification module is used to verify the document total value of the total document object and obtain the document total value verification result.

[0109] The total document target detection result determination module is used to determine the total document target detection result based on the document detection result of the total document object and the document total value verification result.

[0110] According to the object detection device of an embodiment of the present disclosure, the document validity determination module may include: a document object knowledge graph element determination submodule, a document object knowledge graph element coverage determination submodule and a document validity determination submodule.

[0111] The document object knowledge graph element determination submodule is used to identify the medical information of multiple document objects and obtain the document object knowledge graph elements of multiple document objects.

[0112] The document object knowledge graph element coverage determination submodule is used to determine the document object knowledge graph element coverage based on the document object knowledge graph elements and the medical knowledge graph.

[0113] The document validity determination submodule is used to compare the document object knowledge graph element coverage with the validity threshold to obtain the document validity.

[0114] According to the object detection device of the embodiment of the present disclosure, the document object knowledge graph element may include a document object entity, and the document total value verification module may include: a sub-document object entity determination sub-module, an initial word vector determination sub-module, a target word vector determination sub-module, a document standard total value determination sub-module and a document total value verification result determination sub-module.

[0115] The sub-document object entity determination sub-module is used to determine the sub-document object entity of the sub-document object.

[0116] The initial word vector determination submodule is used to determine the initial word vector of the sub-document object entity.

[0117] The target word vector determination submodule is used to pre-train each sub-document object independently to obtain the target word vector of each sub-document object.

[0118] The document standard total value determination submodule is used to determine the document standard total value of the total document object according to the target word vector.

[0119] The document total value verification result determination submodule is used to determine the document total value verification result based on the document standard total value and the document total value.

[0120] According to the object detection device of the embodiment of the present disclosure, the target word vector determination submodule may include: a disease feature determination unit, a target pre-trained model determination unit and a target word vector determination unit.

[0121] The disease characteristic determination unit is used to determine the disease characteristics of the document object.

[0122] The target pre-training model determination unit is used to determine the target pre-training model according to the mapping relationship between the disease characteristics and the pre-training model.

[0123] The target word vector determination unit is used to pre-train the initial word vector of the sub-document object using the target pre-training model to obtain the target word vector.

[0124] According to the object detection device of the embodiment of the present disclosure, the document standard total value determination submodule may include: a document standard total value determination unit, which is used to input the target word vector into the document numerical prediction model to obtain the document standard total value.

[0125] According to the object detection device of the embodiment of the present disclosure, the document value detection result determination module may include: a document sub-value sum determination submodule, a document value comparison result determination submodule and a document value detection result determination submodule.

[0126] The document sub-value sum determination submodule is used to determine the sum of the document values ​​of at least one sub-document object according to the association relationship between the total document object and the sub-document object, so as to obtain the document sub-value sum.

[0127] The document value comparison result determination submodule is used to compare the document total value with the sum of the document sub-values ​​to obtain the document value comparison result.

[0128] The document value detection result determination submodule is used to determine the document value detection result based on the document value comparison result.

[0129] According to the object detection device of the embodiment of the present disclosure, the document value verification result may include pass and fail, and the object detection device may further include: a manual review request sending module for sending a manual review request in response to the document total value verification result being fail.

[0130] According to the object detection device of the embodiment of the present disclosure, the object detection report generation module may further include: an object detection report generation module, which is used to generate a document object detection report.

[0131] It should be understood that the embodiments of the equipment part of the present disclosure are the same or similar to the embodiments of the method part of the present disclosure, and the technical problems solved and the technical effects achieved are also the same or similar, and the present disclosure will not go into details here.

[0132] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0133] Fig.10A schematic block diagram of an example electronic device 1000 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0134] like Fig.10 As shown, the device 1000 includes a computing unit 1001, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1002 or a computer program loaded from a storage unit 1008 into a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for the operation of the device 1000 can also be stored. The computing unit 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0135] A number of components in the device 1000 are connected to the I / O interface 1005, including: an input unit 1006, such as a keyboard, a mouse, etc.; an output unit 1007, such as various types of displays, speakers, etc.; a storage unit 1008, such as a disk, an optical disk, etc.; and a communication unit 1009, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1009 allows the device 1000 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0136] The computing unit 1001 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1001 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1001 performs the various methods and processes described above, such as the object detection method. For example, in some embodiments, the object detection method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 1008. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1000 via the ROM 1002 and / or the communication unit 1009. When the computer program is loaded into the RAM 1003 and executed by the computing unit 1001, one or more steps of the object detection method described above may be performed. Alternatively, in other embodiments, the computing unit 1001 may be configured to perform the object detection method in any other appropriate manner (e.g., by means of firmware).

[0137] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0138] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0139] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0140] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0141] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0142] A computer system may include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship to each other.

[0143] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0144] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An object detection device, comprising: The classification module is used to classify multiple document objects to obtain document categories; An identification module, used to identify the numerical information of the plurality of document objects respectively and obtain the document values; A document value detection result determination module, used for obtaining a document value detection result according to a document value association relationship of the plurality of document objects; A document validity determination module, used to perform validity detection on the multiple document objects to obtain document validity; as well as A document detection result determination module, used to determine the document detection result according to the document value detection result and the document validity; The document object includes a total document object and a sub-document object, the numerical information of the total document object includes a total number of documents, and the object detection device further includes: A document total value verification module, used to verify the document total value of the total document object to obtain a document total value verification result; and A total document target detection result determination module, used to determine the total document target detection result according to the document detection result of the total document object and the document total value verification result; Wherein, the document validity determination module includes: A document object knowledge graph element determination submodule, used to identify the medical information of the multiple document objects and obtain the document object knowledge graph elements of the multiple document objects; A document object knowledge graph element coverage determination submodule, used to determine the document object knowledge graph element coverage based on the document object knowledge graph elements and the medical knowledge graph; and The document validity determination submodule is used to compare the document object knowledge graph element coverage with the validity threshold to obtain the document validity.

2. The device according to claim 1, wherein: The document object knowledge graph element includes a document object entity, and the document total value verification module includes: A sub-document object entity determination sub-module, used to determine the sub-document object entity of the sub-document object; An initial word vector determination submodule, used to determine the initial word vector of the sub-document object entity; A target word vector determination submodule is used to independently pre-train each of the sub-document objects to obtain a target word vector for each of the sub-document objects; A document standard total value determination submodule, used to determine the document standard total value of the total document object according to the target word vector; and The document total value verification result determination submodule is used to determine the document total value verification result according to the document standard total value and the document total value.

3. The device according to claim 2, wherein: The target word vector determination submodule includes: a disease characteristic determination unit, used to determine the disease characteristic of the document object; a target pre-training model determining unit, configured to determine a target pre-training model according to a mapping relationship between the disease characteristics and the pre-training model; and A target word vector determination unit is used to pre-train the initial word vector of the sub-document object using the target pre-training model to obtain the target word vector.

4. The device according to claim 2, wherein: The document standard total value determination submodule includes: The document standard total value determination unit is used to input the target word vector into the document value prediction model to obtain the document standard total value.

5. The device according to any one of claims 2 to 4, wherein: The document value detection result determination module includes: A document sub-value sum determination submodule, configured to determine the sum of the document values ​​of at least one of the sub-document objects according to the association relationship between the total document object and the sub-document objects, to obtain the document sub-value sum; A document value comparison result determination submodule, used for comparing the document total value with the sum of the document sub-values ​​to obtain a document value comparison result; and The document value detection result determination submodule is used to determine the document value detection result according to the document value comparison result.

6. The device according to any one of claims 2 to 4, wherein the document value verification result includes pass and fail, and the object detection device further comprises: The manual review request sending module is used to send a manual review request in response to the document total value verification result being failed.

7. The device according to any one of claims 1 to 4, further comprising: The object detection report generation module is used to generate document object detection reports.

8. An object detection method, comprising: Classify multiple document objects to obtain document categories; Respectively identifying the numerical information of the plurality of document objects to obtain the document values; Obtaining a document value detection result according to the document value association relationship of the multiple document objects; Performing validity checks on the multiple document objects to obtain document validity; as well as Determine the document detection result according to the document value detection result and the document validity; The document object includes a total document object and a sub-document object, the numerical information of the total document object includes a total number of documents, and the object detection method further includes: Verifying the total number of documents of the total document object to obtain a verification result of the total number of documents; and Determine the total document target detection result according to the document detection result of the total document object and the document total value verification result; The performing validity check on the multiple document objects to obtain the document validity includes: Identify medical information of the multiple document objects and obtain document object knowledge graph elements of the multiple document objects; Determining the document object knowledge graph element coverage rate according to the document object knowledge graph element and the medical knowledge graph; and Compare the document object knowledge graph element coverage rate with the validity threshold to obtain the document validity.

9. The method according to claim 8, wherein: The document object knowledge graph element includes a document object entity, and the verification of the total document value of the total document object to obtain the verification result of the total document value includes: Determine a sub-document object entity of the sub-document object; Determining an initial word vector of the sub-document object entity; Pre-training each of the sub-document objects independently to obtain a target word vector for each of the sub-document objects; Determining a standard total value of the total document object according to the target word vector; and The document total value verification result is determined based on the document standard total value and the document total value.

10. The method according to claim 9, wherein: The pre-training of each sub-document object independently to obtain the target word vector of each sub-document object includes: Determine the disease characteristics of the subject of the document; Determining a target pre-training model according to the mapping relationship between the disease characteristics and the pre-training model; and The initial word vector of the sub-document object is pre-trained using the target pre-training model to obtain the target word vector.

11. The method according to claim 9, wherein: Determining the total number of standard bills of the total bill object according to the target word vector includes: The target word vector is input into the document value prediction model to obtain the standard total value of the document.

12. The method according to any one of claims 8 to 11, wherein: The obtaining of the document value detection result according to the document value association relationship of the plurality of document objects comprises: Determine the sum of the document values ​​of at least one of the sub-document objects according to the association relationship between the general document object and the sub-document objects, and obtain the document sub-value sum; Compare the document total value with the sum of the document sub-values ​​to obtain a document value comparison result; and The document value detection result is determined according to the document value comparison result.

13. According to the method of any one of claims 8 to 11, the document value verification result includes pass and fail, and the object detection method further comprises: In response to the document total value verification result being failed, a manual review request is sent.

14. The method according to any one of claims 8 to 11, further comprising: Generate a document object detection report.

15. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 8 to 14.

16. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 8-14.

17. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 8 to 14.

Images