CN115050042A - Claims data entry method and device, computer equipment and storage medium - Google Patents

Abstract

The embodiment of the application belongs to the field of big data, and relates to a method and a device for inputting claim data, computer equipment and a storage medium, wherein the method comprises the following steps: acquiring claim data to be recorded, wherein the claim data comprises a plurality of images; processing the images through a preset OCR text detection model to obtain corresponding text contents, performing first preprocessing on the text contents to generate target feature vectors of the images so as to perform document classification on the images; and performing second preprocessing on the text content based on the document classification result, respectively inputting the second preprocessing result into a preset rule extractor and a preset neural network model, respectively obtaining corresponding output results to perform model integration processing so as to extract non-standard text elements, and inputting the extracted results into a claim settlement system. The application also relates to a block chain technology, and private information of users in claim settlement data can be stored in the block chain. According to the method and the device, efficient automatic and intelligent entry of claim settlement data can be realized, the data entry speed is increased, and the cost is saved.

Description

A claim settlement data entry method, device, computer equipment and storage medium

technical field

The present application relates to the technical field of big data, and in particular, to a method, device, computer equipment and storage medium for inputting claim settlement data.

Background technique

In the process of claim settlement, the customer is required to submit relevant case materials to prove the rationality and authenticity of the case. The case materials include customer certificate information, medical diagnosis and treatment information, and medical expense information. Medical diagnosis and treatment information includes outpatient medical records, inpatient medical records, pathology reports, etc. Expense data include medical invoices, medical bills, social security statements, etc., which need to be entered and archived.

The traditional data entry method is manual entry. The entire process of entry takes a long time, is inefficient, takes time and labor, and prolongs the claims settlement cycle, which affects the user experience.

SUMMARY OF THE INVENTION

The purpose of the embodiments of the present application is to provide a claim settlement data entry method, device, computer equipment and storage medium, so as to solve the problems of long entry time and low efficiency in the claim settlement data entry process in the prior art.

In order to solve the above-mentioned technical problems, the embodiment of the present application provides a method for inputting claim settlement data, which adopts the following technical solutions:

A claim settlement data entry method, comprising the following steps:

Obtaining claims data to be entered, the claim data including several images;

The image is processed through a preset OCR text detection model to obtain text content corresponding to the image, first preprocessing is performed on the text content, and target features of each image are generated according to the first preprocessing result vector, and then perform document classification on each of the images according to the target feature vector;

Based on the document classification result, a second preprocessing is performed on the text content, and the second preprocessing results are respectively input into the preset rule extractor and the preset neural network model, and corresponding output results are obtained respectively. The results are processed by model integration to extract non-standard text elements, and the extracted results are entered into the claim settlement system to complete the structured entry of claim settlement data.

Further, after inputting the second preprocessing result into the preset rule extractor, the method includes:

Element matching word location and neighborhood search are sequentially performed on the input second preprocessing result by the preset rule extractor, so as to perform named entity recognition on the text content, and obtain the value corresponding to each named entity and the corresponding value of each named entity. confidence in the value.

Further, when performing domain search on the input second preprocessing result by the preset rule extractor, the method includes:

Based on the position of the matched named entity in the text content, the operations of distance filtering, part-of-speech filtering and semantic filtering are sequentially performed on the second preprocessed text content to obtain a complete value corresponding to the named entity.

Further, after inputting the second preprocessing result into the preset neural network model, the method includes:

Multi-scale sliding window context splicing is performed on the second preprocessing result, and named entity recognition is performed according to the splicing content, so as to obtain the identified named entities and corresponding values and the confidence of each value.

Further, the step of performing model integration processing to extract non-standard text elements according to the output result includes:

The confidence level of each value output by the preset rule extractor and the preset neural network model is subjected to standardization processing, and then the preset rule extractor and the preset neural network model are identified. The normalized confidence of the same named entity is compared, and the value with higher confidence is selected as the final value of the named entity, and based on the identified named entity and its final value, unstandardized text element extraction is performed.

Further, after the step of entering the extracted result into the claim settlement system and completing the structured entry of the claim settlement data, the method further includes:

In response to the user's source tracing request, the extracted non-standard text elements are traced to the source, and the position of the extracted non-standard text elements in the original image is obtained, so that the user can check the entered text content.

Further, after the step of entering the extracted result into the claim settlement system and completing the structured entry of the claim settlement data, the method further includes:

Obtain new historical claims data, update the preset element positioning words and matching rules in the preset rule extractor according to all the historical claims data, and use all the historical claim data as model training data to update the preset neural network. The model is trained, and the parameters of the preset neural network model are updated.

In order to solve the above technical problems, the embodiment of the present application also provides a claim settlement data input device, which adopts the following technical solutions:

A claim settlement data entry device, comprising:

a data acquisition module, used for acquiring claim settlement data to be entered, the claim settlement data including several images;

The image classification module is used to process the image through a preset OCR text detection model to obtain text content corresponding to the image, perform first preprocessing on the text content, and generate each the target feature vector of the image, and then perform document classification on each of the images according to the target feature vector;

The data entry module is used to perform second preprocessing on the text content based on the document classification results, and input the second preprocessing results into the preset rule extractor and the preset neural network model respectively, and obtain corresponding outputs respectively As a result, model integration processing is performed according to the output results to extract non-standard text elements, and the extracted results are entered into the claim settlement system to complete the structured entry of the claim settlement data.

In order to solve the above-mentioned technical problems, the embodiment of the present application also provides a computer device, which adopts the following technical solutions:

A computer device includes a memory and a processor, wherein computer-readable instructions are stored in the memory, and when the processor executes the computer-readable instructions, the steps of the above-mentioned method for inputting claim settlement data are implemented.

In order to solve the above technical problems, the embodiments of the present application also provide a computer-readable storage medium, which adopts the following technical solutions:

A computer-readable storage medium, where computer-readable instructions are stored on the computer-readable storage medium, and when the computer-readable instructions are executed by a processor, implement the steps of the above-mentioned method for inputting claim settlement data.

Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects:

In this scheme, the claim settlement data is structured by document classification, and then extracting non-standard text elements from the text content based on the document classification result according to the preset rule extractor and the preset neural network model, thereby structuring the claim settlement data. Entry realizes efficient automation and intelligent entry of claims data, thereby improving the speed of data entry, saving costs, reducing the claim settlement cycle, and improving customer experience.

Description of drawings

In order to illustrate the solutions in the present application more clearly, the following will briefly introduce the accompanying drawings used in the description of the embodiments of the present application. For those of ordinary skill, other drawings can also be obtained from these drawings without any creative effort.

FIG. 1 is an exemplary system architecture diagram to which the present application can be applied;

2 is a flowchart of an embodiment of a claim settlement data entry method according to the present application;

3 is a schematic structural diagram of an embodiment of a claim settlement data input device according to the present application;

4 is a schematic structural diagram of another embodiment of the claim settlement data input device according to the present application;

FIG. 5 is a schematic structural diagram of an embodiment of a computer device according to the present application.

Detailed ways

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field of this application; the terms used herein in the specification of the application are for the purpose of describing specific embodiments only It is not intended to limit the application; the terms "comprising" and "having" and any variations thereof in the description and claims of this application and the above description of the drawings are intended to cover non-exclusive inclusion. The terms "first", "second" and the like in the description and claims of the present application or the above drawings are used to distinguish different objects, rather than to describe a specific order.

Reference herein to an "embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor a separate or alternative embodiment that is mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

In order to make those skilled in the art better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings.

As shown in FIG. 1 , the system architecture 100 may include terminal devices 101 , 102 , and 103 , a network 104 and a server 105 . The network 104 is a medium used to provide a communication link between the terminal devices 101 , 102 , 103 and the server 105 . The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages and the like. Various communication client applications may be installed on the terminal devices 101 , 102 and 103 , such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, and the like.

The terminal devices 101, 102, and 103 may be various electronic devices that have a display screen and support web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III, moving picture experts). Compression Standard Audio Layer 3), MP4 (Moving PictureExperts Group Audio Layer IV, Moving Picture Experts Compression Standard Audio Layer 4) Players, Laptops and Desktops, etc.

The server 105 may be a server that provides various services, such as a background server that provides support for the pages displayed on the terminal devices 101 , 102 , and 103 .

It should be noted that the claim settlement data entry method provided by the embodiment of the present application is generally performed by the server, and accordingly, the claim settlement data entry device is generally set in the server.

It should be understood that the numbers of terminal devices, networks and servers in FIG. 1 are merely illustrative. There can be any number of terminal devices, networks and servers according to implementation needs.

Continuing to refer to FIG. 2 , a flowchart of an embodiment of a claim settlement data entry method according to the present application is shown. The claim settlement data entry method includes the following steps:

Step S201: Acquire claim settlement data to be entered, where the claim settlement data includes several images.

When the user needs to make a claim, he/she sends a claim data upload request to the server that handles the collection and entry of claim data through the mobile phone or computer. Claims information includes customer certificate information, medical diagnosis and treatment information, and medical expense information required in the process of claim settlement. Medical diagnosis and treatment information includes outpatient medical records, inpatient medical records, pathology reports, etc. , the claims data of these cases need to be collected and stored in the claims system. These claims data are uploaded in the form of pictures, and the server side stores these pictures in the form of data sets after receiving the claim data.

In this embodiment, after obtaining the claim settlement data on the server side, the method further includes:

Each image is inspected to ensure that all images meet the image quality requirements. Specifically, image quality inspection can be performed through one or more of a sharpness detection model, an orientation correction model, and a remake detection model.

Among them, the sharpness detection model can use the existing image sharpness detection model for quality inspection, such as the gradient operator-based image sharpness detection model or the trained convolutional neural network model; the direction correction model can use the projection-based image sharpness detection model. method, method based on Hough transform, method based on linear fitting, method based on Fourier transform to frequency domain detection, etc. for direction correction; remake detection model can be detected by noise analysis, pixel value deviation detection, etc. For images that do not meet the detection standard, a prompt message for re-uploading is sent to the user's mobile phone or computer, and the detection is performed again after receiving a new uploaded image, and the subsequent steps are processed until all the detections are passed.

In some embodiments, the method further includes: performing a deduplication operation on the claim settlement data, so as to remove the same claim settlement data uploaded by the user in different upload portals due to improper operation, and prompt the user to confirm whether the claim settlement data is omitted.

Step S202, processing the image through a preset OCR text detection model to obtain text content corresponding to the image, performing a first preprocessing on the text content, and generating each of the images according to the first preprocessing result the target feature vector, and then perform document classification on each of the images according to the target feature vector.

In this step, when acquiring the text content according to the OCR text detection model, an existing OCR text detection model can be used to achieve it.

When the first preprocessing is performed on the text content, the purpose is to perform coarse-grained text content extraction, specifically, inputting the text content into the preprocessing module to perform operations such as morphological restoration, part-of-speech tagging, and number regularization. Shape restoration is for English words in the text. Part-of-speech tagging is to mark whether each word is a noun, verb, adjective or other part-of-speech. The main purpose of digital regularization is to regularize the numbers involved in expenses in images such as medical invoices in claims data. , in order to facilitate the document classification of images, where document classification refers to classifying the images of expense documents (such as diagnosis sheets, invoices, etc.) and identity certificates (such as ID cards, driver's licenses, etc.) Subsequent input of claim information in the claims system will be carried out in a targeted manner. When performing coarse-grained feature extraction, only the key text content that can be used for document classification is extracted as classification features.

When performing classification feature extraction, this embodiment can use a bag of words model, specifically, after word segmentation of the text content, a number of segmented words are extracted as feature fields, the feature fields are filtered to obtain a feature field set, and each feature field in the feature field set is determined in the feature field set. Based on the distribution of the histogram established by the bag-of-words model, the intermediate feature vector is obtained, and finally the intermediate feature vector is normalized and dimensionally reduced to obtain the target feature vector. The document classification result can be obtained by inputting the target feature vector into the document classification model. For claims materials, it can be divided into different document types such as certificates, medical records, diagnostic materials, invoices, lists, reimbursement certificates, bank cards, etc. Different types have different characteristic fields. For example, the characteristic fields of ID cards include name, number, etc. , validity period, etc. For the feature fields of medical records including hospital, admission and discharge diagnosis, admission and discharge date, name, etc., the dimension of the intermediate feature vector formed by the bag-of-words model’s understanding of the content of the text is large, so it is necessary to reduce the intermediate feature vector. dimensional processing of the target feature vector.

In some embodiments, a convolutional neural network model can also be used to extract classification features. Specifically, the text content of each image is spliced and input into the convolutional neural network model to obtain a target feature vector, and the obtained target feature vector Enter the document classification model to get the document classification result. Among them, when training the convolutional neural network model, the feature fields of different claims data types are manually marked. For example, the feature fields of ID cards include name, number, validity period, etc., and the feature fields of medical records include hospital, admission and discharge. Diagnosis, admission and discharge date, name, etc., the layout of different materials also has certain characteristics. For example, the title of the invoice will be in the middle and upper part of the document. These characteristics are artificially defined and interfered with. The training of the convolutional neural network model mainly includes the text In addition, it can also train the understanding of the document layout of the text, and establish the layout relationship of different material feature fields in the continuous learning process.

In other embodiments, the classification feature can also be extracted by combining the bag-of-words model and the convolutional neural network model. Specifically, according to the probability values of the target feature vectors obtained by the two, the target feature vector with a larger probability value is selected. as the final target feature vector.

Step S203, based on the document classification result, perform second preprocessing on the text content, input the second preprocessing results into the preset rule extractor and the preset neural network model respectively, and obtain corresponding output results respectively, according to The output results are subjected to model integration processing to extract non-standard text elements, and the extracted results are entered into the claim settlement system to complete the structured entry of the claim settlement data.

In this step, different from the first preprocessing, the purpose of performing the second preprocessing on the text content is to perform fine-grained text content extraction based on the document classification result. The claim settlement data presented in the form of images is divided into two image sets of documents, and the second preprocessing is performed on the two image sets respectively. The difference from the first preprocessing is that when the second preprocessing is performed on the two image sets, for There will be differences in the processing methods of the second preprocessing for the images of the document category and the images of the certification category, that is, the second preprocessing includes common processing methods (such as morphological restoration, part-of-speech tagging, etc.) and specific processing methods for image types. (such as digital regularization of document images, etc.).

In this embodiment, the output results of inputting the second preprocessing result into the preset rule extractor and the preset neural network model respectively are named entities obtained according to the text content after the second preprocessing, the values corresponding to the named entities, and Confidence for each value.

In this embodiment, after the second preprocessing result is input into a preset rule extractor, the method includes: performing element matching word positioning and Neighborhood search is used to perform named entity recognition on the text content, and obtain the value corresponding to each named entity and the confidence level of each value.

Specifically, the rule extractor extracts key-value pairs from the text content, the key is the named entity, and the value is the value of the named entity, such as name: Zhang San, height: 175cm, etc. This is obtained according to the rule extractor to the named entity and the corresponding value. The rule extractor locates named entities through element matching words to achieve the purpose of named entity recognition. Several element positioning words are pre-defined in your rule extractor, such as height, discharge diagnosis, etc. are defined element positioning words, through element positioning Words and their synonyms are retrieved and matched in the second preprocessed text content to identify named entities.

In this embodiment, when performing a domain search on the input second preprocessing result by using the preset rule extractor, the method includes: taking the position of the matched named entity in the text content as a benchmark , and sequentially perform distance filtering, part-of-speech filtering and semantic filtering operations on the second preprocessed text content to obtain a complete value corresponding to the named entity.

Specifically, after locating the named entity through the element matching word, the value corresponding to the named entity is further extracted through the domain search. In this embodiment, when the domain search is performed, the method includes using the matched named entity in the text content. Taking the position as the benchmark, the steps of distance filtering, part-of-speech filtering and semantic filtering are performed on the second preprocessed text content in sequence to obtain the complete value corresponding to the named entity. Among them, distance filtering is to spatially query multiple elements whose distance from the matched named entity meets the requirements. Generally speaking, it refers to several elements with a closer distance after the matched named entity. Part-of-speech filtering is for the query results. Perform part-of-speech recognition on multiple elements, and then perform element screening based on the matched named entities to obtain the first element. Semantic screening is to semantically identify and match other elements adjacent to the first element in the spatial position, and the matched elements As the second element, the first element and the second element are concatenated as the value corresponding to the matched named entity. For example, based on the element matching word "height", the named entity "height" is matched in the text content after the second preprocessing, and further, after "height", the nearest quantifier in space is matched, and then the nearest element is further matched based on the quantifier. , get the unit of the quantifier, and concatenate the quantifier and the unit to get the value of height. While outputting the named entity and the corresponding value, the rule extractor will also output the confidence of each value, which identifies the correct probability value for the named entity and the corresponding value.

In some embodiments, the value of the named entity does not necessarily have a unit, so semantic filtering in the domain search is not necessary, and can be adaptively performed according to the actual situation.

In some embodiments, if the same named entity is not matched according to the element locator, the named entity recognition may be further performed by calculating the edit distance between the word segmentation of the second preprocessed text content and the element locator to perform fuzzy matching. Because the naming of entities is often irregular, there may be various forms of deformation and spelling, which leads to a low recall rate of named entity recognition based on the exact matching of element positioning words. The use of edit distance to generalize from perfect matching to fuzzy matching effectively improves the recognition success. Rate.

In this embodiment, after the second preprocessing result is input into the preset neural network model, the method includes: performing multi-scale sliding window context splicing on the second preprocessing result, performing named entity recognition according to the splicing content, and obtaining The identified named entities and their corresponding values and the confidence level for each value. Among them, when the preset neural network model performs named entity recognition according to the spliced content, it performs structured analysis on the text content, and identifies all entities and corresponding values in the text.

In this embodiment, the step of performing model integration processing to extract non-standard text elements according to the output result includes: calculating the confidence level of each value output by the preset rule extractor and the preset neural network model Carry out benchmarking normalization processing, then compare the normalized confidence levels of the same named entity identified by the preset rule extractor and the preset neural network model, and select a value with a higher confidence level as the name The final value of the entity, and based on the identified named entity and its final value, non-standard text feature extraction is performed. After the extraction operation is completed, the extracted results are entered into the claim settlement system, and the structured entry of the claim settlement data is completed.

In this embodiment, after the step of entering the extracted results into the claim settlement system and completing the structured entry of the claim settlement data, the method further includes: tracing the extracted non-standard text elements in response to a user's traceability request, and obtaining The position of the extracted non-standard text elements in the original image is for the user to check the entered text content. Correspondingly, the method also includes the step of establishing the correspondence between the text and the image position, specifically determining the position of the text content in the image through target detection, and storing the mapping relationship between the text content and the image position in the form of coordinates. When the mouse stays on the entered text interface, in response to the stop action, the corresponding image is automatically obtained and displayed according to the text at the mouse stop position, which is convenient for manual verification of the automatically entered text content. Compared with the existing manual entry method, it can be greatly improved. Reduce entry time and improve overall timeliness.

In this embodiment, after the step of entering the extracted result into the claim settlement system and completing the structured entry of the claim settlement data, the method further includes: acquiring new historical claim settlement data, and adjusting the pre-determined claim settlement data according to all the historical claim settlement data. It is assumed that the preset element location words and matching rules in the rule extractor are updated, and the preset neural network model is trained according to all historical claim settlement data as model training data, and the parameters of the preset neural network model are updated. In this step, the preset rule extractor and the preset neural network model can be optimized. The optimization of the rule extractor is to optimize and update the preset element positioning words and matching rules according to the historical claim data. For the preset neural network The optimization of the model is based on the historical claim settlement data as model training data to perform optimization training and update the model parameters, so that the claim settlement service can be optimized through the optimized preset rule extractor and preset neural network model.

The above method of this application can be used for the data submitted by customers (including customer certificate data, medical diagnosis and treatment data, medical expense data, medical diagnosis and treatment data including outpatient medical records, inpatient medical records, pathology reports, etc., medical expenses data including medical invoices, medical bills, social security settlements, etc. Documents, etc.), are intelligently entered and archived, specifically by classifying the claims data by documents, and then extracting non-standard text elements from the text content based on the document classification results according to the preset rule extractor and the preset neural network model. , so as to carry out the structured input of claims data, realize the efficient automation and intelligent input of claim data, thereby improving the speed of data input, saving costs, reducing the claim cycle and improving customer experience.

It should be emphasized that, in order to further ensure the privacy and security of the above-mentioned claims data, the user privacy information in the above-mentioned claims data can also be stored in a node of a blockchain.

The blockchain referred to in this application is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. , when the program is executed, it may include the processes of the foregoing method embodiments. The aforementioned storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM).

It should be understood that although the various steps in the flowchart of the accompanying drawings are sequentially shown in the order indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless explicitly stated herein, the execution of these steps is not strictly limited to the order and may be performed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but may be executed at different times, and the execution sequence is also It does not have to be performed sequentially, but may be performed alternately or alternately with other steps or at least a portion of sub-steps or stages of other steps.

Further referring to FIG. 3 , as an implementation of the method shown in FIG. 2 above, the present application provides an embodiment of a claim settlement data input device, which corresponds to the method embodiment shown in FIG. 2 . Can be used in various electronic devices.

As shown in FIG. 3 , the claim settlement data entry device 300 in this embodiment includes: a data acquisition module 301 , an image classification module 302 and a data entry module 303 . in:

The data acquisition module 301 is used to acquire the claim settlement data to be entered, and the claim settlement data includes several images; the image classification module 302 is used to process the images through a preset OCR text detection model, and obtain the same data as the above. For the text content corresponding to the image, perform a first preprocessing on the text content, generate a target feature vector of each of the images according to the first preprocessing result, and then perform document classification on each of the images according to the target feature vector; The data entry module 303 is configured to perform second preprocessing on the text content based on the document classification results, and input the second preprocessing results into the preset rule extractor and the preset neural network model, respectively, to obtain corresponding According to the output results, model integration processing is performed to extract non-standard text elements, and the extracted results are entered into the claim settlement system to complete the structured entry of the claim settlement data.

When the user needs to make a claim, the data acquisition module 301 receives the request for uploading the claim data from the user through the mobile phone or the computer to the server that handles the collection and entry of the claim data, and receives the data uploaded by the user's mobile phone or computer after responding to the request for uploading the claim data. Claims data, these claims data include customer certificate data, medical diagnosis and treatment data, medical expense data required in the process of claim settlement, medical diagnosis and treatment data include outpatient medical records, inpatient medical records, pathology reports, etc. Medical expenses data include medical invoices, medical bills, Social security statements, etc., the claims data of these cases need to be collected and stored in the claims system. These claims data are uploaded in the form of pictures, and after receiving the claim data, the pictures are stored to the server in the form of data sets.

In this embodiment, the data acquisition module 301 is further configured to detect each image after obtaining the claim settlement data on the server side to ensure that all images meet the image quality requirements. Specifically, the definition detection model, the orientation correction model and the One or more of the remake detection models perform image quality inspection. Among them, the sharpness detection model can use the existing image sharpness detection model for quality inspection, such as the gradient operator-based image sharpness detection model or the trained convolutional neural network model; the direction correction model can use the projection-based image sharpness detection model. method, method based on Hough transform, method based on linear fitting, method based on Fourier transform to frequency domain detection, etc. for direction correction; remake detection model can be detected by noise analysis, pixel value deviation detection, etc. For images that fail to meet the detection standard, a prompt message for re-uploading is sent to the user's mobile terminal or computer terminal, and the detection is performed again after receiving a new uploaded image until all the detections pass.

In some embodiments, the data acquisition module 301 is further configured to perform a deduplication operation on the claim data, so as to remove the same claim data uploaded by the user in different upload portals due to improper operation, and prompt the user to confirm whether the claim data is missing.

In this step, when the image classification module 302 acquires the text content according to the OCR text detection model, the existing OCR text detection model can be used for implementation.

When the image classification module 302 performs the first preprocessing on the text content, the purpose is to perform coarse-grained text content extraction, specifically inputting the text content into the preprocessing module to perform morphological restoration, part-of-speech tagging, numbering Regularization and other operations, in which morphological restoration is for English words in the text, part-of-speech tagging is to tag each word as a noun, verb, adjective or other parts of speech, and digital regularization is mainly for medical invoices in claims data. The figures involved in expenses are organized to facilitate the document classification of images, where document classification refers to the expense documents (such as diagnosis sheets, invoices, etc.) and identity certificates (such as ID cards, driver's licenses, etc.) existing in the claims data. The images are classified, so that the claim information can be entered in the claims system in a targeted manner. When performing coarse-grained feature extraction, only the key text content that can be used for document classification is extracted as classification features.

When the image classification module 302 performs classification feature extraction, this embodiment may use a bag of words model, specifically, after word segmentation of the text content, a number of segmented words are extracted as feature fields, the feature fields are filtered to obtain a feature field set, and the feature fields are determined. The distribution of each feature field in the set in the histogram established based on the bag-of-words model is used to obtain the intermediate feature vector. Finally, the intermediate feature vector is normalized and dimensionally reduced to obtain the target feature vector. The document classification result can be obtained by inputting the target feature vector into the document classification model. For claims materials, it can be divided into different document types such as certificates, medical records, diagnostic materials, invoices, lists, reimbursement certificates, bank cards, etc. Different types have different characteristic fields. For example, the characteristic fields of ID cards include name, number, etc. , validity period, etc. For the feature fields of medical records including hospital, admission and discharge diagnosis, admission and discharge date, name, etc., the dimension of the intermediate feature vector formed by the bag-of-words model’s understanding of the content of the text is large, so it is necessary to reduce the intermediate feature vector. dimensional processing of the target feature vector.

In some embodiments, the image classification module 302 can also use a convolutional neural network model to extract classification features, specifically for splicing the text content of each image and then inputting it into the convolutional neural network model to obtain the target feature vector , and input the obtained target feature vector into the document classification model to obtain the document classification result. Among them, when training the convolutional neural network model, the feature fields of different claims data types are manually marked. For example, the feature fields of ID cards include name, number, validity period, etc., and the feature fields of medical records include hospital, admission and discharge. Diagnosis, admission and discharge date, name, etc., the layout of different materials also has certain characteristics. For example, the title of the invoice will be in the middle and upper part of the document. These characteristics are artificially defined and interfered with. The training of the convolutional neural network model mainly includes the text In addition, it can also train the understanding of the document layout of the text, and establish the layout relationship of different material feature fields in the continuous learning process.

In other embodiments, the image classification module 302 can also be used to extract classification features by combining the bag-of-words model and the convolutional neural network model. Specifically, according to the probability values of the target feature vectors obtained by the two, select The target feature vector with a larger probability value is used as the final target feature vector.

In this embodiment, different from the first preprocessing performed by the image classification module 302, the purpose of performing the second preprocessing on the text content by the data entry module 303 is to perform fine-grained text content extraction based on the document classification result , when the second preprocessing is performed, first, according to the document classification result, the claim settlement data presented in the form of images is divided into two image sets of documents, and the second preprocessing is carried out for the two image sets respectively, and the image classification module The difference between the first preprocessing performed by 302 is that when the data entry module 303 performs the second preprocessing on the two image sets, there will be differences in the processing methods of the second preprocessing for the images of the document type and the images of the certificate type. , that is, the second preprocessing includes common processing methods (such as lemmatization, part-of-speech tagging, etc.) and specific processing methods for image types (such as digital regularization of document-type images, etc.).

In this embodiment, the data entry module 303 inputs the second preprocessing results into the preset rule extractor and the preset neural network model respectively, and the output results are named entities obtained according to the second preprocessed text content, each The value corresponding to the named entity and the confidence level for each value.

In this embodiment, after the data entry module 303 inputs the second preprocessing result into the preset rule extractor, it is specifically configured to: sequentially perform the inputted second preprocessing result through the preset rule extractor. Element matching word location and neighborhood search are used to identify named entities on the text content, and obtain the values corresponding to each named entity and the confidence level of each value.

Specifically, the data entry module 303 is used to extract key-value pairs from the text content through a rule extractor, where the key is the named entity, and the value is the value of the named entity, such as name: Zhang San, height: 175cm, etc. This is the named entity and the corresponding value obtained according to the rule extractor. The rule extractor locates named entities through element matching words to achieve the purpose of named entity recognition. Several element positioning words are pre-defined in your rule extractor, such as height, discharge diagnosis, etc. are defined element positioning words, through element positioning Words and their synonyms are retrieved and matched in the second preprocessed text content to identify named entities.

In this embodiment, when the data entry module 303 performs a domain search on the input second preprocessing result through the preset rule extractor, it is specifically used for: using the matched named entity in the text content Based on the position in the second preprocessed text content, distance filtering, part-of-speech filtering, and semantic filtering operations are sequentially performed to obtain the complete value corresponding to the named entity.

Specifically, after the data entry module 303 locates the named entity through element matching words, it further extracts the value corresponding to the named entity through domain search. In this embodiment, when the data entry module 303 performs domain search, it is specifically used for Based on the position of the matched named entity in the text content, the steps of distance filtering, part-of-speech filtering and semantic filtering are sequentially performed on the second preprocessed text content to obtain a complete value corresponding to the named entity. Among them, distance filtering is to spatially query multiple elements whose distance from the matched named entity meets the requirements. Generally speaking, it refers to several elements with a closer distance after the matched named entity. Part-of-speech filtering is for the query results. Perform part-of-speech recognition on multiple elements, and then perform element screening based on the matched named entities to obtain the first element. Semantic screening is to semantically identify and match other elements adjacent to the first element in the spatial position, and the matched elements As the second element, the first element and the second element are concatenated as the value corresponding to the matched named entity. For example, based on the element matching word "height", the named entity "height" is matched in the text content after the second preprocessing, and further, after "height", the nearest quantifier in space is matched, and then the nearest element is further matched based on the quantifier. , get the unit of the quantifier, and concatenate the quantifier and the unit to get the value of height. While outputting the named entity and the corresponding value, the rule extractor will also output the confidence of each value, which identifies the correct probability value for the named entity and the corresponding value.

In some embodiments, the value of the named entity does not necessarily have a unit, so semantic filtering is not necessary when the data entry module 303 performs a domain search, and can be adaptively performed according to the actual situation.

In some embodiments, if the data entry module 303 does not match the same named entity according to the element location word, the data entry module 303 is further configured to calculate the word segmentation and element location word of the second preprocessed text content The edit distance is used for fuzzy matching to realize named entity recognition. Because the naming of entities is often irregular, there may be various forms of deformation and spelling, which leads to a low recall rate of named entity recognition based on the exact matching of element positioning words. The use of edit distance to generalize from perfect matching to fuzzy matching effectively improves the recognition success. Rate.

In this embodiment, after the data entry module 303 inputs the second preprocessing result into the preset neural network model, it is specifically used to perform multi-scale sliding window context splicing on the second preprocessing result, and name the splicing content according to the content. Entity identification, get the identified named entity and corresponding value and the confidence of each value. Among them, when the preset neural network model performs named entity recognition according to the spliced content, it performs structured analysis on the text content, and identifies all entities and corresponding values in the text.

In this embodiment, when the data input module 303 performs model integration processing according to the output result to extract non-standard text elements, it is specifically used to input the output of the preset rule extractor and the preset neural network model. The confidence level of the value is normalized, and then the normalized confidence level of the same named entity identified by the preset rule extractor and the preset neural network model is compared, and the confidence level is higher. The value of the named entity is used as the final value of the named entity, and based on the identified named entity and its final value, the non-standard text feature extraction is performed. After the extraction operation is completed, the extracted results are entered into the claim settlement system, and the structured entry of the claim settlement data is completed.

In this embodiment, as shown in FIG. 4 , the device further includes a traceability module 304, which is used for entering the extracted results into the claim settlement system in the data entry module 303, and after completing the structured entry of the claim settlement data, responds to the user's The source tracing request traces the extracted non-standard text elements to the source, and obtains the position of the extracted non-standard text elements in the original image, so that the user can check the entered text content. Correspondingly, the traceability module 34 is also used to establish the correspondence between the text and the image position, and is specifically used to determine the position of the text content in the image through target detection, and store the mapping relationship between the text content and the image position in the form of coordinates. , when the user's mouse stays on the entered text interface, in response to the stop action, the corresponding image is automatically obtained and displayed according to the text at the mouse stop position, which is convenient for manual verification of the automatically entered text content. Compared with the existing manual entry method in the whole process , which can greatly reduce the input time and improve the overall timeliness.

In this embodiment, as shown in FIG. 4 , the device further includes an update module 305 for entering the extracted results into the claim settlement system in the data entry module 303, and after completing the structured entry of the claim settlement data, obtains new Historical claim settlement data, update the preset element positioning words and matching rules in the preset rule extractor according to all historical claim settlement data, and train the preset neural network model according to all historical claim settlement data as model training data , and update the parameters of the preset neural network model. The update module optimizes the preset rule extractor and the preset neural network model, wherein the optimization of the rule extractor is to optimize and update the preset element locating words and matching rules according to historical claims data. The optimization of the network model is based on the historical claim settlement data as model training data to perform optimization training and update model parameters, so that the claim settlement service can be optimized through the optimized preset rule extractor and preset neural network model.

The above-mentioned device in this application can provide data submitted by customers (including customer certificate data, medical diagnosis and treatment data, medical expense data, medical diagnosis and treatment data include outpatient medical records, inpatient medical records, pathology reports, etc., and medical expenses data include medical invoices, medical bills, social security settlements, etc. Documents, etc.), are intelligently entered and archived, specifically by classifying the claims data by documents, and then extracting non-standard text elements from the text content based on the document classification results according to the preset rule extractor and the preset neural network model. , so as to carry out the structured input of claims data, realize the efficient automation and intelligent input of claim data, thereby improving the speed of data input, saving costs, reducing the claim cycle and improving customer experience.

To solve the above technical problems, the embodiments of the present application also provide computer equipment. Please refer to FIG. 5 for details. FIG. 5 is a block diagram of a basic structure of a computer device according to this embodiment.

The computer device 5 includes a memory 51 , a processor 52 , and a network interface 53 that communicate with each other through a system bus. The memory 51 stores computer-readable instructions, and when the processor 52 executes the computer-readable instructions, the steps of the claim settlement data entry method described in the foregoing embodiment are implemented.

It should be pointed out that only the computer device 5 with components 51-53 is shown in FIG. 5, but it should be understood that it is not required to implement all the shown components, and more or less components may be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculation and/or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, special-purpose Integrated circuit (ApplicationSpecific Integrated Circuit, ASIC), programmable gate array (Field-Programmable GateArray, FPGA), digital processor (Digital Signal Processor, DSP), embedded equipment, etc.

The computer equipment may be a desktop computer, a notebook computer, a palmtop computer, a cloud server and other computing equipment. The computer device can perform human-computer interaction with the user through a keyboard, a mouse, a remote control, a touch pad or a voice control device.

The memory 51 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static Random Access Memory (SRAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), Programmable Read Only Memory (PROM), Magnetic Memory, Magnetic Disk, Optical Disk, etc. In some embodiments, the memory 51 may be an internal storage unit of the computer device 5 , such as a hard disk or a memory of the computer device 5 . In other embodiments, the memory 51 may also be an external storage device of the computer device 5 , for example, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, flash memory card (FlashCard) and so on. Of course, the memory 51 may also include both the internal storage unit of the computer device 5 and its external storage device. In this embodiment, the memory 51 is generally used to store the operating system and various application software installed on the computer device 5 , such as computer-readable instructions of a method for inputting claim settlement data. In addition, the memory 51 can also be used to temporarily store various types of data that have been output or will be output.

In some embodiments, the processor 52 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. This processor 52 is typically used to control the overall operation of the computer device 5 . In this embodiment, the processor 52 is configured to execute computer-readable instructions stored in the memory 51 or process data, such as computer-readable instructions for executing the claim settlement data entry method.

The network interface 53 may include a wireless network interface or a wired network interface, and the network interface 53 is generally used to establish a communication connection between the computer device 5 and other electronic devices.

The above-mentioned computer equipment in this application can provide information about the client (including client certificate information, medical diagnosis and treatment information, medical expense information, medical diagnosis and treatment information including outpatient medical records, inpatient medical records, pathology reports, etc., and medical expenses information including medical invoices, medical bills, social security Settlement documents, etc.), intelligently enter and archive, specifically by classifying the claims data by documents, and then performing non-standard text elements on the text content based on the document classification results according to the preset rule extractor and the preset neural network model. Extraction, so as to carry out the structured entry of claims data, realize efficient automation and intelligent input of claim data, thereby improving the speed of data entry, saving costs, reducing the claims cycle, and improving customer experience.

The present application also provides another embodiment, that is, to provide a computer-readable storage medium, where the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to The at least one processor is caused to perform the steps of the claims data entry method as described above.

When the instructions stored in the above-mentioned computer-readable storage medium of this application are executed, the data submitted by the customer (including customer certificate data, medical diagnosis and treatment data, medical expense data, medical diagnosis and treatment data including outpatient medical records, inpatient medical records, pathology reports, etc., medical Expense data includes medical invoices, medical bills, social security statements, etc.), which are intelligently entered and archived. Specifically, the claim data is classified by documents, and then based on the results of the document classification according to the preset rule extractor and the preset neural network model. The non-standard text elements are extracted from the text content, so as to carry out the structured input of the claim data, and realize the efficient automation and intelligent input of the claim data, so as to improve the speed of data input, save the cost, reduce the claim cycle and improve the customer experience.

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

Obviously, the above-described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The accompanying drawings show the preferred embodiments of the present application, but do not limit the scope of the patent of the present application. This application may be embodied in many different forms, rather these embodiments are provided so that a thorough and complete understanding of the disclosure of this application is provided. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements for some of the technical features. . Any equivalent structure made by using the contents of the description and drawings of the present application, which is directly or indirectly used in other related technical fields, is also within the scope of protection of the patent of the present application.

Claims (10)

1. a claim settlement data entry method, is characterized in that, comprises the following steps: Obtaining claims data to be entered, the claim data including several images; The image is processed through a preset OCR text detection model to obtain text content corresponding to the image, first preprocessing is performed on the text content, and target features of each image are generated according to the first preprocessing result vector, and then perform document classification on each of the images according to the target feature vector; Based on the document classification result, a second preprocessing is performed on the text content, and the second preprocessing results are respectively input into the preset rule extractor and the preset neural network model, and corresponding output results are obtained respectively. The results are processed by model integration to extract non-standard text elements, and the extracted results are entered into the claim settlement system to complete the structured entry of claim settlement data. 2. The claim settlement data input method according to claim 1, wherein after inputting the second preprocessing result into the preset rule extractor, the method comprises: Element matching word location and neighborhood search are sequentially performed on the input second preprocessing result by the preset rule extractor, so as to perform named entity recognition on the text content, and obtain the value corresponding to each named entity and the corresponding value of each named entity. confidence in the value. 3 . The claim settlement data input method according to claim 2 , wherein, when performing a field search on the input second preprocessing result by the preset rule extractor, the method comprises: 3 . Based on the position of the matched named entity in the text content, the operations of distance filtering, part-of-speech filtering and semantic filtering are sequentially performed on the second preprocessed text content to obtain a complete value corresponding to the named entity. 4. The claim settlement data input method according to any one of claims 1 to 3, wherein after inputting the second preprocessing result into a preset neural network model, the method comprises: Multi-scale sliding window context splicing is performed on the second preprocessing result, and named entity recognition is performed according to the splicing content, so as to obtain the identified named entities and corresponding values and the confidence of each value. 5. The claim settlement data input method according to claim 4, wherein the step of performing model integration processing according to the output result to extract non-standard text elements comprises: The confidence level of each value output by the preset rule extractor and the preset neural network model is subjected to standardization processing, and then the preset rule extractor and the preset neural network model are identified. The normalized confidence of the same named entity is compared, and the value with higher confidence is selected as the final value of the named entity, and based on the identified named entity and its final value, unstandardized text element extraction is performed. 6. The claim settlement data entry method according to claim 5, characterized in that, after the described step of entering the extracted result into the claim settlement system and completing the structured entry of the claim settlement data, the method further comprises: In response to the user's source tracing request, the extracted non-standard text elements are traced to the source, and the position of the extracted non-standard text elements in the original image is obtained, so that the user can check the entered text content. 7. The claim settlement data entry method according to claim 6, characterized in that, after the described step of entering the extracted result into the claim settlement system and completing the structured entry of the claim settlement data, the method further comprises: Obtain new historical claims data, update the preset element positioning words and matching rules in the preset rule extractor according to all the historical claims data, and use all the historical claim data as model training data to update the preset neural network. The model is trained, and the parameters of the preset neural network model are updated. 8. A claim settlement data entry device, characterized in that it comprises: a data acquisition module, used for acquiring claim settlement data to be entered, the claim settlement data including several images; The image classification module is used to process the image through a preset OCR text detection model to obtain text content corresponding to the image, perform first preprocessing on the text content, and generate each the target feature vector of the image, and then perform document classification on each of the images according to the target feature vector; The data entry module is used to perform second preprocessing on the text content based on the document classification results, and input the second preprocessing results into the preset rule extractor and the preset neural network model respectively, and obtain corresponding outputs respectively As a result, model integration processing is performed according to the output results to extract non-standard text elements, and the extracted results are entered into the claim settlement system to complete the structured entry of the claim settlement data. 9. A computer device, comprising a memory and a processor, wherein computer-readable instructions are stored in the memory, and when the processor executes the computer-readable instructions, any one of claims 1 to 7 is implemented. Steps of the claim data entry method described in item . 10. A computer-readable storage medium, wherein computer-readable instructions are stored on the computer-readable storage medium, and when the computer-readable instructions are executed by a processor, any one of claims 1 to 7 is implemented. Steps of the claim data entry method described in item .

Images