Artificial intelligence-based data processing method and device, computer device and medium

Abstract

This invention relates to the field of artificial intelligence technology, and particularly to a data processing method, apparatus, computer equipment, and medium based on artificial intelligence. This invention performs character recognition on a target image to obtain indicator representation quantities and their values, as well as conclusion descriptive quantities. Based on the indicator representation quantities and conclusion descriptive quantities, it determines the indicator threshold range and descriptive information. When the representation value exceeds the indicator threshold range, the indicator representation quantity is identified as an anomalous indicator, and the corresponding descriptive information is determined. Based on the anomalous indicator's representation value, indicator threshold range, corresponding descriptive information, and descriptive information corresponding to the conclusion descriptive quantity, the anomalous level is determined, and the descriptive category to which the anomalous indicator belongs is extracted. The descriptive category and anomalous level are input into a trained classification model, and the classification result corresponding to the target image is output. By determining the anomalous level of the target image through anomaly analysis of the indicator representation quantities and conclusion descriptive quantities, the accuracy of data processing is improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a data processing method, apparatus, computer equipment, and medium based on artificial intelligence. Background Technology

[0002] Currently, in image data processing scenarios, it mainly relies on business personnel to manually process target images submitted by customers, determine the abnormality level of indicators in the target image and the classification results, and then provide targeted suggestions to customers. This requires business personnel to have certain relevant knowledge to make reasonable judgments and processes of various indicators in the image. Moreover, different business personnel have different judgment criteria. Therefore, the requirements for business personnel to manually process target images are relatively high, resulting in low accuracy of data processing for target images.

[0003] Therefore, in the field of artificial intelligence technology, how to improve the accuracy of data processing has become an urgent problem to be solved. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a data processing method, apparatus, computer equipment, and medium based on artificial intelligence to solve the problem of low accuracy in existing data processing methods that rely on manual review.

[0005] In a first aspect, embodiments of the present invention provide a data processing method based on artificial intelligence, the data processing method comprising:

[0006] The target image sent by the user is acquired, and character recognition is performed on the target image to obtain the indicator representation quantity and its representation value, as well as the conclusion description quantity;

[0007] Based on the indicator representation quantity and the conclusion description quantity, the indicator threshold range corresponding to the indicator representation quantity and the description information corresponding to the conclusion description quantity are respectively matched from the first database;

[0008] When the value of the indicator exceeds the threshold range, the indicator is determined to be an abnormal indicator, and the descriptive information corresponding to the abnormal indicator is determined from the second database.

[0009] The anomaly level is determined based on the characterization value of the anomaly indicator, the threshold range of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity.

[0010] Extract the description category to which the anomaly index belongs from the third database, input the description category and the anomaly level into the trained classification model, and output the classification result corresponding to the target image.

[0011] Secondly, embodiments of the present invention provide a data processing device based on artificial intelligence, the data processing device comprising:

[0012] The image recognition module is used to acquire the target image sent by the user, perform character recognition on the target image, and obtain the indicator representation quantity and its representation value, as well as the conclusion description quantity;

[0013] The information matching module is used to match the indicator threshold range corresponding to the indicator characterization quantity and the descriptive information corresponding to the conclusion description quantity from the first database, respectively, based on the indicator characterization quantity and the conclusion description quantity.

[0014] An abnormal indicator judgment module is used to determine that the indicator is an abnormal indicator when the indicator value exceeds the indicator threshold range, and to determine the description information corresponding to the abnormal indicator from the second database.

[0015] An anomaly level determination module is used to determine the anomaly level based on the characterization value of the anomaly indicator, the threshold range of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity.

[0016] The image classification module is used to extract the description category to which the anomaly index belongs from the third database, input the description category and the anomaly level into the trained classification model, and output the classification result corresponding to the target image.

[0017] Thirdly, embodiments of the present invention provide a computer device, the computer device including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the data processing method as described in the first aspect.

[0018] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the data processing method as described in the first aspect.

[0019] The beneficial effects of this invention compared to existing technologies are as follows: By performing character recognition on the target image, indicator representation quantities and their representation values, as well as conclusion descriptive quantities, are obtained. Based on the indicator representation quantities and conclusion descriptive quantities, the indicator threshold range corresponding to the indicator representation quantity and the descriptive information corresponding to the conclusion descriptive quantity are matched from a first database. When the representation value of the indicator representation quantity is detected to exceed the indicator threshold range, the indicator representation quantity is determined to be an abnormal indicator. The descriptive information corresponding to the abnormal indicator is determined from a second database. Based on the representation value of the abnormal indicator, the indicator threshold range, the descriptive information corresponding to the abnormal indicator, and the descriptive information corresponding to the conclusion descriptive quantity, the anomaly level is determined. The descriptive category to which the abnormal indicator belongs is extracted from a third database. The descriptive category and anomaly level are input into a trained classification model, and the classification result corresponding to the target image is output. By analyzing the anomalies of the indicator representation quantities and conclusion descriptive quantities, the anomaly level of the target image is determined, improving the accuracy of data processing. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of an application environment for a data processing method provided in Embodiment 1 of the present invention;

[0022] Figure 2 This is a flowchart illustrating a data processing method provided in Embodiment 1 of the present invention;

[0023] Figure 3 This is a schematic diagram of the structure of a data processing device provided in Embodiment 2 of the present invention;

[0024] Figure 4 This is a schematic diagram of the structure of a computer device provided in Embodiment 3 of the present invention. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0026] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0027] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0029] Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0030] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0031] The embodiments of this invention can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that utilize digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0032] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0033] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0034] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0035] The data processing method provided in Embodiment 1 of this invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server. Clients include, but are not limited to, handheld computers, desktop computers, laptops, ultra-mobile personal computers (UMPCs), netbooks, cloud computing devices, and personal digital assistants (PDAs). The server can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0036] See Figure 2 This is a flowchart illustrating a data processing method provided in Embodiment 1 of the present invention. The above data processing method can be applied to... Figure 1 For clients in the network, the data processing method may include the following steps:

[0037] Step S201: Obtain the target image sent by the user, perform character recognition on the target image, and obtain the indicator representation quantity and its representation value, as well as the conclusion description quantity.

[0038] In processing image data, the business side needs to review the target images provided by users to identify any anomalies in the user's various metrics in order to determine the classification result of the target image. Therefore, in order to obtain the various metrics contained in the target image, it is necessary to perform character recognition on the target image, extract the metric representation quantities and their corresponding representation values ​​contained in the target image, as well as the conclusion descriptive quantities contained in the target image.

[0039] Character recognition refers to the process of examining printed characters on a target image using electronic devices, determining the shape of the characters by detecting bright and dark patterns, and then translating the shape into computer text using character recognition methods. For example, OCR (Optical Character Recognition) technology can recognize optical characters on a target image through image processing and pattern recognition techniques to obtain the corresponding characters on the target image, including index representation quantities and their representation values, as well as conclusion description quantities.

[0040] For example, in an insurance underwriting scenario, the target image sent by the user can be a physical examination image sent by the user to be underwritten. The indicator representation quantity can be the name of the physical examination indicator in the image, such as white blood cells, red blood cells, glucose, and platelets. The representation value can be the corresponding indicator value of each physical examination indicator. The conclusion description quantity can be the description of the physical examination result in the image. Correspondingly, after character recognition of the physical examination image sent by the user to be underwritten, the corresponding indicator representation quantity and its representation value, as well as the conclusion description quantity, are obtained. These are used to make anomaly judgments on the various physical examination indicators of the user to be underwritten to determine the health status of the user, thereby providing targeted product recommendations for the user to be underwritten.

[0041] This embodiment obtains indicator representation quantities and their representation values, as well as conclusion description quantities, by performing character recognition on the target image. The characters in the target image are converted into computer text, which serves as the basis for subsequent judgment of various abnormal user indicators, thereby improving the accuracy of judging abnormal user situations.

[0042] Optionally, character recognition is performed on the target image to obtain the indicator representation quantity and its representation value, as well as the conclusion descriptive quantity, including:

[0043] Image segmentation is performed on the target image to obtain the indicator region and conclusion region in the target image;

[0044] Character recognition is performed on the indicator area to obtain the indicator representation quantity and its representation value;

[0045] Character recognition is performed on the conclusion region to obtain the conclusion description.

[0046] The target image includes not only information valid for identifying abnormalities in various user indicators, but also information invalid for identifying abnormalities in various user indicators. In this embodiment, in order to improve the accuracy of judging abnormalities in users, the target image is first segmented into different regions, and valid indicator regions and conclusion regions are extracted, while interference from other invalid regions is discarded.

[0047] Specifically, the indicator area includes the corresponding indicator content, and the conclusion area includes the corresponding conclusion content. Therefore, character recognition is performed on the extracted indicator area to obtain the corresponding indicator representation quantity and its representation value within that area, and character recognition is performed on the extracted conclusion area to obtain the corresponding conclusion description quantity within that area.

[0048] This embodiment performs image segmentation on the target image to obtain the indicator region and the conclusion region in the target image. Then, it performs character recognition on the indicator region and the conclusion region respectively to obtain the corresponding indicator representation quantity and its representation value, as well as the conclusion description quantity. By extracting the effective indicator region and the conclusion region and discarding the interference of other invalid regions, the accuracy of judging abnormal situations of users is improved.

[0049] Optionally, image segmentation is performed on the target image to obtain the indicator region and conclusion region in the target image, including:

[0050] The target image is input into a trained image segmentation model to obtain the index region and conclusion region in the target image.

[0051] The image segmentation model includes a segmentation encoder and a segmentation decoder. The target image is used as the training sample, and the actual index region and actual conclusion region of the training sample are used as training labels to train the image segmentation model.

[0052] Image segmentation is the technique and process of dividing an image into several specific regions with unique properties and extracting targets of interest. In this embodiment, the purpose of image segmentation of the target image is to extract effective indicator regions and conclusion regions. In data processing activities, in order to process user target images in batches, a pre-trained image segmentation model is used to segment the target images to obtain indicator regions and conclusion regions in the target images.

[0053] The image segmentation model includes a segmentation encoder and a segmentation decoder. The segmentation encoder extracts features from the input target image to obtain the region features of the target image, and the segmentation decoder decodes these region features to output a segmented image. In this embodiment, the image segmentation model uses the target image as a training sample and the actual indicator region and actual conclusion region of the training sample as training labels to evaluate the segmented image output by the image segmentation model. This allows the image segmentation model to be trained based on the evaluation results, resulting in a trained image segmentation model used to segment the user's target image to obtain the indicator region and conclusion region in the target image.

[0054] This embodiment inputs the target image into a trained image segmentation model to obtain the indicator region and conclusion region in the target image. This ensures that the target image indicator region and conclusion region can be extracted in batches, quickly and effectively during data processing, thus effectively improving the efficiency and accuracy of image segmentation.

[0055] Optionally, the training process of the image segmentation model includes:

[0056] The target image of the training sample is input into the segmentation encoder for feature extraction to obtain the sample region features;

[0057] The sample region features are input into the segmentation decoder to obtain a two-channel segmentation image, where the first channel segmentation image is the sample index region and the second channel segmentation image is the sample conclusion region.

[0058] The segmentation category of each pixel in the target image is determined based on the sample index region and the sample conclusion region. The actual category of each pixel in the target image is determined based on the actual index region and the actual conclusion region. The first loss function is calculated based on the segmentation category and the actual category of each pixel. The parameters of the segmentation encoder and the segmentation decoder are corrected in reverse using the gradient descent method until the first loss function converges, thus obtaining the trained image segmentation model.

[0059] Since the purpose of image segmentation of the target image is to extract effective indicator regions and conclusion regions, the segmentation decoder of the image segmentation model is set to two channels during the training process. The first channel is used to output the indicator region of the target image, and the second channel is used to output the conclusion region of the target image.

[0060] Specifically, in the training process of the image segmentation model, the training samples are a large number of target images. The segmentation encoder extracts features from the target images used as training samples to obtain sample region features, and inputs the sample region features into the segmentation decoder to obtain two-channel segmentation images. The first channel segmentation image is the sample index region, and the second channel segmentation image is the sample conclusion region.

[0061] Then, the segmentation category of each pixel in the target image is determined based on the sample index region and the sample conclusion region, and the actual category of each pixel in the target image is determined based on the actual index region and the actual conclusion region. The first loss function is calculated using the segmentation category and the actual category of each pixel. When the first loss function is small, it indicates that the difference between the sample index region and the sample conclusion region obtained by the image segmentation model and the actual index region and the actual conclusion region is small, indicating that the accuracy of the image segmentation model is high. When the first loss function is large, it indicates that the difference between the sample index region and the sample conclusion region obtained by the image segmentation model and the actual index region and the actual conclusion region is large, indicating that the accuracy of the image segmentation model is low. It is necessary to back-correct the parameters of the segmentation encoder and the segmentation decoder using the gradient descent method until the first loss function converges, thus obtaining the trained image segmentation model.

[0062] For example, when determining the segmentation category of each pixel in the target image based on the sample index region and the sample conclusion region, the segmentation category of the pixel in the sample index region is recorded as 1, and the segmentation category of the pixel in the sample conclusion region is recorded as 2. When determining the actual category of each pixel in the target image based on the actual index region and the actual conclusion region, the actual category of the pixel in the actual index region is recorded as 1, and the actual category of the pixel in the actual conclusion region is recorded as 2.

[0063] For each pixel in the target image, the segmentation category and the actual category can be determined. For each pixel, if its segmentation category and the actual category are consistent, it means that the segmentation result of the pixel is correct. Conversely, if its segmentation category and the actual category are inconsistent, it means that the segmentation result of the pixel is incorrect. Therefore, the first loss function can be calculated based on the difference between the segmentation category and the actual category of each pixel.

[0064] In this embodiment, the number of pixels is denoted as N, where N is a positive integer, and the segmentation category of the i-th (i = 1, 2, ..., N) pixel is denoted as F. i Then F i =1 or F i =2, and let S be the actual category of the i-th pixel. i Then S i =1 or S i =2, let the first loss function be denoted as Loss1, then the first loss function Loss1 is:

[0065]

[0066] In the formula, N is the number of pixels, and F i S represents the segmentation category of the i-th pixel. i 1. The actual category of the i-th pixel.

[0067] This embodiment trains an image segmentation model and determines the segmentation category and actual category of each pixel based on the segmentation results and training labels. A first loss function is then calculated to evaluate the accuracy of the image segmentation model. When the first loss function converges, a well-trained image segmentation model is obtained, thus ensuring the segmentation accuracy of the image segmentation model.

[0068] Step S202: Based on the indicator representation quantity and the conclusion description quantity, the indicator threshold range corresponding to the indicator representation quantity and the description information corresponding to the conclusion description quantity are matched from the first database respectively.

[0069] The first database stores indicator features and their corresponding indicator threshold ranges, as well as conclusion descriptions and their corresponding description information. The indicator threshold range is used to represent the normal range of the representation values ​​of the corresponding indicator features and is used to make anomaly judgments on the indicator representations and their representation values ​​extracted from the target image. The description information is used to make anomaly judgments on the conclusion descriptions extracted from the target image, so as to make anomaly judgments on the target image sent by the user.

[0070] Specifically, character matching is performed on the first database based on the indicator representation quantity to find the indicator representation quantity in the first database. Then, the indicator threshold range corresponding to the indicator representation quantity is obtained based on the matching result. Keywords of the conclusion description quantity are extracted according to the keyword extraction algorithm. Character matching is performed on the first database based on the keywords to find the conclusion description quantity in the first database. Then, the description information corresponding to the conclusion description quantity is obtained based on the matching result.

[0071] For example, in an insurance underwriting scenario, the indicator representation quantity can be the name of a physical examination indicator in a physical examination image, such as white blood cells. Then, character matching is performed from the first database to find the white blood cells and the corresponding indicator threshold range in the first database, such as [4,10]. Then, the indicator threshold range corresponding to the indicator representation quantity can be obtained from the first database.

[0072] This embodiment uses the steps of matching the index threshold range corresponding to the index characterization quantity and the descriptive information corresponding to the conclusion description quantity from the first database, respectively, to determine the index threshold range indicating that the index characterization quantity is normal, as well as the descriptive information of the conclusion description quantity, thereby improving the accuracy of anomaly judgment for target images.

[0073] Step S203: When the value of the indicator is detected to exceed the threshold range, the indicator is determined to be an abnormal indicator, and the descriptive information corresponding to the abnormal indicator is determined from the second database.

[0074] The second database stores indicator representation quantities and their descriptive information, which are used to describe the abnormal situations of each indicator. For the indicator representation quantities in the target image, based on the identification of abnormal indicators, the descriptive information corresponding to the abnormal indicators can be determined from the second database to describe the abnormal situation of the abnormal indicators in the target image, which serves as the basis for subsequent anomaly judgment of the target image.

[0075] Specifically, the representation value of the indicator is compared with the corresponding indicator threshold range. When the representation value exceeds the indicator threshold range, the corresponding indicator is determined to be an abnormal indicator. Based on the indicator representation value of the abnormal indicator, character matching is performed in the second database to find the corresponding indicator representation value and corresponding description information in the second database, and the description information corresponding to the abnormal indicator is determined.

[0076] In this embodiment, when the value of an indicator exceeds the threshold range, the indicator is identified as an abnormal indicator. The step of determining the descriptive information corresponding to the abnormal indicator from the second database improves the accuracy of anomaly judgment in the target image by identifying the abnormal indicator and its descriptive information.

[0077] Step S204: Determine the anomaly level based on the characterization value of the abnormal indicator, the threshold range of the indicator, the descriptive information corresponding to the abnormal indicator, and the descriptive information corresponding to the conclusion descriptive quantity.

[0078] Among them, the characterization value of the abnormal indicator is not within the threshold range of the indicator, and the greater the difference between the characterization value and the threshold range of the indicator, the higher the degree of abnormality of the abnormal indicator. The degree of abnormality of the abnormal indicator can be characterized by the characterization value and the threshold range of the indicator, combined with the descriptive information corresponding to the abnormal indicator, and the degree of abnormality of the conclusion descriptive quantity can be characterized by the descriptive information corresponding to the conclusion descriptive quantity. Finally, the abnormality level of the target image is determined, which serves as the basis for determining the classification result of the target image.

[0079] Optionally, the anomaly level is determined based on the characteristic value of the anomaly indicator, the threshold range of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity, including:

[0080] Calculate the difference between the characteristic value and the threshold range of the abnormal indicator, and determine the degree of abnormality of each abnormal indicator based on the difference.

[0081] The abnormality level of the indicator, the descriptive information corresponding to the abnormal indicator, and the descriptive information corresponding to the conclusion descriptive quantity are input into the trained abnormality evaluation model to obtain the abnormality level of the abnormal indicator.

[0082] The greater the difference between the characterization value and the threshold of the indicator range, the higher the degree of abnormality of the abnormal indicator. Therefore, in this embodiment, the difference between the characterization value of the abnormal indicator and the threshold range of the indicator is calculated to determine the degree of abnormality of each abnormal indicator based on the difference. Then, the degree of abnormality of the indicator, the descriptive information corresponding to the abnormal indicator and the descriptive information corresponding to the conclusion descriptive quantity are input into the trained abnormality evaluation model to obtain the abnormality level of the abnormal indicator.

[0083] In this embodiment, the characteristic value of the abnormal indicator is denoted as Z, and the indicator threshold range is denoted as W = [w1, w2], where w1 is the lower limit of the indicator threshold range and w2 is the upper limit of the indicator threshold range. For characteristic values ​​that are not within the indicator threshold range, when the characteristic value Z is greater than the upper limit w2 of the indicator threshold range, the difference between the characteristic value Z and the upper limit w2 of the indicator threshold range is calculated as the indicator difference between the characteristic value and the indicator threshold range; when the characteristic value Z is less than the lower limit w1 of the indicator threshold range, the difference between the characteristic value Z and the lower limit w1 of the indicator threshold range is calculated as the indicator difference between the characteristic value and the indicator threshold range. The larger the indicator difference, the larger the difference between the characteristic value and the indicator range threshold.

[0084] Then, the differences in indicators are normalized, and the normalization result is used as the degree of abnormality of the abnormal indicators to eliminate the influence of the order of magnitude of the characteristic values ​​on the judgment of the degree of abnormality and improve the accuracy of the judgment of the degree of abnormality.

[0085] This embodiment calculates the difference between the characterization value of the abnormal indicator and the threshold range of the indicator, determines the degree of abnormality of each abnormal indicator based on the difference, and inputs the degree of abnormality, the descriptive information corresponding to the abnormal indicator, and the descriptive information corresponding to the conclusion descriptive quantity into the trained anomaly evaluation model to obtain the anomaly level of the abnormal indicator. This effectively characterizes the degree of abnormality of the abnormal indicators and conclusion descriptive quantities in the target image, obtains the anomaly level of the target image, improves the accuracy of anomaly degree judgment, and thus improves the accuracy of data processing.

[0086] Optionally, the anomaly evaluation model includes an evaluation encoder and an evaluation fully connected layer, with the degree of anomaly of the indicators, the descriptive information corresponding to the anomaly indicators, and the descriptive information corresponding to the conclusion descriptive quantity as training samples, and the actual anomaly level of the training samples as training labels.

[0087] The training process of the anomaly assessment model includes:

[0088] The abnormality level of the training sample indicators, the descriptive information corresponding to the abnormal indicators, and the descriptive information corresponding to the conclusion descriptive quantity are input into the evaluation encoder for feature extraction to obtain the abnormal features of the sample.

[0089] The abnormal features of the samples are input into the evaluation fully connected layer to obtain the abnormality level of the samples;

[0090] The second loss function is calculated based on the sample anomaly level and the corresponding actual anomaly level. The parameters of the evaluation encoder and the evaluation fully connected layer are then corrected in reverse using the gradient descent method until the second loss function converges, thus obtaining the trained anomaly evaluation model.

[0091] The training samples consist of a large number of corresponding indicators of abnormality, descriptive information corresponding to abnormal indicators, and descriptive information corresponding to conclusion descriptions. The training labels are the actual abnormality levels of the training samples determined by the instructor. These labels are used to evaluate the abnormality levels obtained by the abnormality evaluation model, so that the abnormality evaluation model can be trained based on the evaluation results to obtain a well-trained abnormality evaluation model.

[0092] In the training process of the anomaly assessment model, the anomaly degree of the training samples, the descriptive information corresponding to the anomaly indicators, and the descriptive information corresponding to the conclusion descriptive quantity are first input into the assessment encoder for feature extraction to obtain sample anomaly features. Then, the sample anomaly features are input into the assessment fully connected layer to obtain the sample anomaly level. Next, a second loss function is calculated based on the sample anomaly level and the actual anomaly level. When the second loss function is small, it indicates that the difference between the sample anomaly level obtained by the anomaly assessment model and the actual anomaly level is small, indicating that the accuracy of the anomaly assessment model is high. When the second loss function is large, it indicates that the difference between the sample anomaly level obtained by the anomaly assessment model and the actual anomaly level is large, indicating that the accuracy of the anomaly assessment model is low. It is necessary to back-correct the parameters of the assessment encoder and the assessment fully connected layer using the gradient descent method until the second loss function converges, thus obtaining the trained anomaly assessment model.

[0093] This embodiment trains the anomaly evaluation model based on the actual anomaly level determined by humans, thereby improving the accuracy of the anomaly evaluation model and obtaining a well-trained anomaly evaluation model to improve the accuracy of subsequent data processing.

[0094] Step S205: Extract the description category of the anomaly index from the third database, input the description category and anomaly level into the trained classification model, and output the classification result corresponding to the target image.

[0095] The third database stores indicator representations and their descriptive categories, used to distinguish between various indicators. For indicator representations in a target image, after identifying anomalous indicators, the corresponding descriptive categories can be determined from the third database. These descriptive categories and anomaly levels are then input into a trained classification model, outputting the classification result for the target image, thereby improving the accuracy of data processing.

[0096] For example, in an insurance underwriting scenario, the target image could be a user's medical examination report. The classification result corresponding to the target image could be no impact on underwriting, slight impact on underwriting, or severe impact on underwriting, indicating the impact of the medical examination report provided by the user on underwriting. By determining the classification result corresponding to the target image provided by the user, targeted insurance recommendations can be provided to the user. Furthermore, by improving the accuracy of data processing, the accuracy of the target image classification result can be improved, thereby improving the accuracy of insurance product recommendations to the user.

[0097] Optionally, the classification model includes a classification encoder and a classification fully connected layer, using descriptions of categories and anomaly levels as training samples, and the actual classification results of the training samples as training labels;

[0098] The training process for a classification model includes:

[0099] The description category and anomaly level are input into the classification encoder for feature extraction to obtain the sample description features;

[0100] The sample description features are input into the classification fully connected layer to obtain the sample classification result;

[0101] The third loss function is calculated based on the sample classification results and the corresponding actual classification results. The parameters of the classification encoder and the classification fully connected layer are then corrected in reverse using the gradient descent method until the third loss function converges, thus obtaining the trained classification model.

[0102] The training samples consist of a large number of corresponding descriptive categories and anomaly levels, while the training labels are the actual classification results of the training samples determined by the user. These labels are used to evaluate the classification results obtained by the classification model, so that the classification model can be trained based on the evaluation results to obtain a well-trained classification model.

[0103] During the training of the classification model, the description category and anomaly level of the training samples are first input into the evaluation encoder for feature extraction to obtain sample description features. Then, the sample description features are input into the classification fully connected layer to obtain the sample classification result. Next, a third loss function is calculated based on the sample classification result and the actual classification result. When the third loss function is small, it indicates that the difference between the sample classification result obtained by the classification model and the actual classification result is small, indicating that the accuracy of the classification model is high. When the third loss function is large, it indicates that the difference between the sample classification result obtained by the classification model and the actual classification result is large, indicating that the accuracy of the classification model is low. It is necessary to backtrack and correct the parameters of the classification encoder and the classification fully connected layer using the gradient descent method until the third loss function converges, thus obtaining the trained classification model.

[0104] This embodiment trains the classification model based on the actual classification results determined by humans, thereby improving the accuracy of the classification model and obtaining a well-trained classification model to improve the accuracy of data processing.

[0105] This invention, through character recognition of a target image, obtains indicator representation quantities and their values, as well as conclusion descriptive quantities. Based on the indicator representation quantities and conclusion descriptive quantities, it matches the indicator threshold range corresponding to the indicator representation quantity and the descriptive information corresponding to the conclusion descriptive quantity from a first database. When the indicator representation quantity's value exceeds the indicator threshold range, the indicator representation quantity is determined to be an anomalous indicator. The descriptive information corresponding to the anomalous indicator is then determined from a second database. Based on the anomalous indicator's value, threshold range, and descriptive information, the anomalous level is determined. The descriptive category to which the anomalous indicator belongs is extracted from a third database. The descriptive category and anomalous level are input into a trained classification model, outputting the classification result corresponding to the target image. By analyzing the anomalies in the indicator representation quantities and conclusion descriptive quantities, the anomalous level of the target image is determined, improving the accuracy of data processing.

[0106] Corresponding to the data processing method in the above embodiments, Figure 3 A structural block diagram of the data processing device provided in Embodiment 2 of the present invention is given. For ease of explanation, only the parts related to the embodiments of the present invention are shown.

[0107] See Figure 3 The data processing device includes:

[0108] Image recognition module 31 is used to acquire the target image sent by the user, perform character recognition on the target image, and obtain the indicator representation quantity and its representation value, as well as the conclusion description quantity;

[0109] Information matching module 32 is used to determine that the indicator is an abnormal indicator when the indicator value exceeds the indicator threshold range, and to determine the description information corresponding to the abnormal indicator from the second database.

[0110] The abnormal indicator judgment module 33 is used to determine that the indicator is an abnormal indicator when the indicator value exceeds the indicator threshold range, and to determine the description information corresponding to the abnormal indicator from the second database.

[0111] The anomaly level determination module 34 is used to determine the anomaly level based on the characterization value of the anomaly indicator, the threshold range of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity.

[0112] The image classification module 35 is used to extract the description category to which the anomaly index belongs from the third database, input the description category and anomaly level into the trained classification model, and output the classification result corresponding to the target image.

[0113] Optionally, the image recognition module 31 mentioned above includes:

[0114] The image segmentation submodule is used to segment the target image to obtain the indicator region and conclusion region in the target image.

[0115] The first character recognition submodule is used to perform character recognition on the indicator area to obtain the indicator representation quantity and its representation value.

[0116] The second character recognition submodule is used to perform character recognition on the conclusion region to obtain the conclusion description.

[0117] Optionally, the above image segmentation submodule includes:

[0118] The image segmentation unit is used to input the target image into the trained image segmentation model to obtain the index region and conclusion region in the target image;

[0119] The first model training unit, which is used for image segmentation model including segmentation encoder and segmentation decoder, uses the target image as training sample and the actual index region and actual conclusion region of the training sample as training labels to train the image segmentation model.

[0120] Optionally, the first model training unit mentioned above includes:

[0121] The first feature extraction subunit is used to input the target image of the training sample into the segmentation encoder for feature extraction to obtain the sample region features;

[0122] The segmentation image determination subunit is used to input the sample region features into the segmentation decoder to obtain a two-channel segmentation image, wherein the first channel segmentation image is the sample index region and the second channel segmentation image is the sample conclusion region.

[0123] The first parameter correction subunit is used to determine the segmentation category of each pixel in the target image based on the sample index region and the sample conclusion region, determine the actual category of each pixel in the target image based on the actual index region and the actual conclusion region, calculate the first loss function based on the segmentation category and the actual category of each pixel, and correct the parameters of the segmentation encoder and segmentation decoder in reverse according to the gradient descent method until the first loss function converges, thus obtaining the trained image segmentation model.

[0124] Optionally, the above-mentioned anomaly level determination module 34 includes:

[0125] The anomaly degree calculation submodule is used to calculate the difference between the characteristic value of the anomaly indicator and the threshold range of the indicator, and to determine the degree of anomaly of each anomaly indicator based on the difference.

[0126] The anomaly level determination submodule is used to input the degree of anomaly of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity into the trained anomaly evaluation model to obtain the anomaly level of the anomaly indicator.

[0127] Optionally, the above-mentioned anomaly level determination submodule includes:

[0128] An anomaly evaluation model determination unit is used to determine the anomaly evaluation model, which includes an evaluation encoder and an evaluation fully connected layer. The training samples are the degree of anomaly of the indicators, the descriptive information corresponding to the anomaly indicators, and the descriptive information corresponding to the conclusion descriptive quantity. The actual anomaly level of the training samples is used as the training label.

[0129] The second feature extraction unit is used to input the abnormality degree of the indicators of the training samples, the descriptive information corresponding to the abnormal indicators, and the descriptive information corresponding to the conclusion descriptive quantity into the evaluation encoder for feature extraction, so as to obtain the abnormal features of the samples.

[0130] The grade determination unit is used to input the sample anomaly features into the evaluation fully connected layer to obtain the sample anomaly grade;

[0131] The second parameter correction unit is used to calculate the second loss function based on the sample anomaly level and the corresponding actual anomaly level, and to correct the parameters of the evaluation encoder and the evaluation fully connected layer in reverse according to the gradient descent method until the second loss function converges, thus obtaining the trained anomaly evaluation model.

[0132] Optionally, the image classification module 35 mentioned above includes:

[0133] The classification model determination submodule is used to determine the classification model, which includes a classification encoder and a classification fully connected layer. The training samples describe the category and the anomaly level, and the actual recommended category of the training samples are used as the training labels.

[0134] The third feature extraction submodule is used to input the description category and anomaly level into the classification encoder for feature extraction to obtain sample description features;

[0135] The recommendation category determination submodule is used to input sample description features into the classification fully connected layer to obtain the sample classification result;

[0136] The third parameter correction submodule is used to calculate the third loss function based on the sample classification results and the corresponding actual classification results. It then uses the gradient descent method to back-correct the parameters of the classification encoder and the classification fully connected layer until the third loss function converges, thus obtaining the trained classification model.

[0137] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0138] Figure 4 This is a schematic diagram of the structure of a computer device provided in Embodiment 3 of the present invention. Figure 4 As shown, the computer device of this embodiment includes: at least one processor ( Figure 4 Only one is shown in the diagram), a memory, and a computer program stored in the memory and executable on at least one processor, wherein the processor executes the computer program to implement the steps in any of the above-described data processing method embodiments.

[0139] This computer device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that... Figure 4 The examples of computer devices are merely examples and do not constitute a limitation on computer devices. Computer devices may include more or fewer components than shown, or combinations of certain components, or different components, such as network interfaces, displays, and input devices.

[0140] The processor referred to can be a CPU, but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0141] Memory includes readable storage media, internal memory, etc., wherein internal memory can be the RAM of a computer device, providing an environment for the operation of the operating system and computer-readable instructions stored in the readable storage media. The readable storage media can be the hard drive of a computer device, or in other embodiments, it can be an external storage device of the computer device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, memory can include both internal storage units and external storage devices of a computer device. Memory is used to store the operating system, applications, bootloader, data, and other programs, such as program code for computer programs. Memory can also be used to temporarily store data that has been output or will be output.

[0142] Those skilled in the art will understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the functions described above can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above device can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the methods of the above embodiments by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. A computer-readable medium can include at least: any entity or device capable of carrying computer program code, a recording medium, a computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0143] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be accomplished by a computer program product. When the computer program product is run on a computer device, the computer device executes the steps in the above method embodiments.

[0144] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0145] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0146] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / computer devices and methods can be implemented in other ways. For example, the apparatus / computer device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0147] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0148] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A data processing method based on artificial intelligence, characterized in that, The data processing method includes: The target image sent by the user is acquired, and character recognition is performed on the target image to obtain the indicator representation quantity and its representation value, as well as the conclusion description quantity; Based on the indicator representation quantity and the conclusion description quantity, the indicator threshold range corresponding to the indicator representation quantity and the description information corresponding to the conclusion description quantity are respectively matched from the first database; When the value of the indicator exceeds the threshold range, the indicator is determined to be an abnormal indicator, and the descriptive information corresponding to the abnormal indicator is determined from the second database. The anomaly level is determined based on the characterization value of the anomaly indicator, the threshold range of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity. Extract the description category to which the anomaly index belongs from the third database, input the description category and the anomaly level into the trained classification model, and output the classification result corresponding to the target image; The step of determining the anomaly level based on the characterization value of the anomaly indicator, the threshold range of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity includes: Calculate the difference between the characterization value of the abnormal indicator and the threshold range of the indicator, and determine the degree of abnormality of each of the abnormal indicators based on the difference. The abnormality level of the indicator, the descriptive information corresponding to the abnormal indicator, and the descriptive information corresponding to the conclusion descriptive quantity are input into the trained abnormality evaluation model to obtain the abnormality level of the abnormal indicator. The anomaly evaluation model includes an evaluation encoder and an evaluation fully connected layer. The anomaly degree of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity are used as training samples, and the actual anomaly level of the training samples is used as the training label. The training process of the anomaly evaluation model includes: The abnormality degree of the indicator of the training sample, the descriptive information corresponding to the abnormal indicator, and the descriptive information corresponding to the conclusion descriptive quantity are input into the evaluation encoder for feature extraction to obtain the abnormal features of the sample. The sample anomaly features are input into the evaluation fully connected layer to obtain the sample anomaly level; The second loss function is calculated based on the sample anomaly level and the corresponding actual anomaly level. The parameters of the evaluation encoder and the evaluation fully connected layer are then corrected in reverse using the gradient descent method until the second loss function converges, thus obtaining the trained anomaly evaluation model.

2. The data processing method according to claim 1, characterized in that, The process of performing character recognition on the target image to obtain indicator representation quantities and their representation values, as well as conclusion description quantities, includes: The target image is segmented to obtain the indicator region and the conclusion region in the target image; Character recognition is performed on the indicator region to obtain the indicator representation quantity and its representation value; Character recognition is performed on the conclusion region to obtain the conclusion description quantity.

3. The data processing method according to claim 2, characterized in that, The step of segmenting the target image to obtain the indicator region and conclusion region in the target image includes: The target image is input into a trained image segmentation model to obtain the indicator region and conclusion region in the target image; The image segmentation model includes a segmentation encoder and a segmentation decoder. The target image is used as a training sample, and the actual index region and actual conclusion region of the training sample are used as training labels to train the image segmentation model.

4. The data processing method according to claim 3, characterized in that, The training process of the image segmentation model includes: The target image of the training sample is input into the segmentation encoder for feature extraction to obtain the sample region features; The sample region features are input into the segmentation decoder to obtain a two-channel segmentation image, wherein the segmentation image of the first channel is the sample index region and the segmentation image of the second channel is the sample conclusion region. The segmentation category of each pixel in the target image is determined based on the sample index region and the sample conclusion region. The actual category of each pixel in the target image is determined based on the actual index region and the actual conclusion region. A first loss function is calculated based on the segmentation category and the actual category of each pixel. The parameters of the segmentation encoder and the segmentation decoder are corrected in reverse using the gradient descent method until the first loss function converges, thus obtaining the trained image segmentation model.

5. The data processing method according to claim 1, characterized in that, The classification model includes a classification encoder and a classification fully connected layer, using the described category and anomaly level as training samples and the actual classification results of the training samples as training labels. The training process of the classification model includes: The description category and anomaly level are input into the classification encoder for feature extraction to obtain sample description features; The sample description features are input into the classification fully connected layer to obtain the sample classification result; The third loss function is calculated based on the sample classification result and the corresponding actual classification result. The parameters of the classification encoder and the classification fully connected layer are then corrected in reverse using the gradient descent method until the third loss function converges, thus obtaining the trained classification model.

6. A data processing device based on artificial intelligence, characterized in that, The data processing device includes: The image recognition module is used to acquire the target image sent by the user, perform character recognition on the target image, and obtain the indicator representation quantity and its representation value, as well as the conclusion description quantity; The information matching module is used to match the indicator threshold range corresponding to the indicator characterization quantity and the descriptive information corresponding to the conclusion description quantity from the first database, respectively, based on the indicator characterization quantity and the conclusion description quantity. An abnormal indicator judgment module is used to determine that the indicator is an abnormal indicator when the value of the indicator exceeds the threshold range, and to determine the description information corresponding to the abnormal indicator from the second database. An anomaly level determination module is used to determine the anomaly level based on the characterization value of the anomaly indicator, the threshold range of the indicator, the descriptive information corresponding to the anomaly indicator, and the descriptive information corresponding to the conclusion descriptive quantity. The image classification module is used to extract the description category to which the anomaly index belongs from the third database, input the description category and the anomaly level into the trained classification model, and output the classification result corresponding to the target image. The anomaly level determination module includes: The anomaly degree calculation submodule is used to calculate the difference between the characteristic value of the anomaly indicator and the threshold range of the indicator, and to determine the degree of anomaly of each anomaly indicator based on the difference. The anomaly level determination submodule is used to input the degree of anomaly of the indicator, the descriptive information corresponding to the abnormal indicator, and the descriptive information corresponding to the conclusion descriptive quantity into the trained anomaly evaluation model to obtain the anomaly level of the abnormal indicator. The anomaly level determination submodule includes: An anomaly evaluation model determination unit is used to determine the anomaly evaluation model, which includes an evaluation encoder and an evaluation fully connected layer. The training samples are the degree of anomaly of the indicators, the descriptive information corresponding to the anomaly indicators, and the descriptive information corresponding to the conclusion descriptive quantity. The actual anomaly level of the training samples is used as the training label. The second feature extraction unit is used to input the abnormality degree of the indicators of the training samples, the descriptive information corresponding to the abnormal indicators, and the descriptive information corresponding to the conclusion descriptive quantity into the evaluation encoder for feature extraction, so as to obtain the abnormal features of the samples. The grade determination unit is used to input the sample anomaly features into the evaluation fully connected layer to obtain the sample anomaly grade; The second parameter correction unit is used to calculate the second loss function based on the sample anomaly level and the corresponding actual anomaly level, and to correct the parameters of the evaluation encoder and the evaluation fully connected layer in reverse according to the gradient descent method until the second loss function converges, thus obtaining the trained anomaly evaluation model.

7. A computer device, characterized in that, The computer device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the data processing method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the data processing method as described in any one of claims 1 to 5.

Images