Claim settlement intention identification method and device, computer equipment and storage medium

Abstract

The invention belongs to the technical field of artificial intelligence and management, and relates to a claim settlement intention recognition method and device, computer equipment and a storage medium, and the method comprises the steps: carrying out the intention recognition of a to-be-recognized insurance claim settlement text through employing a trained claim settlement intention recognition model, and obtaining a claim settlement intention type; determining a claim settlement processing flow corresponding to the to-be-identified insurance claim settlement text according to the claim settlement intention category; wherein the claim settlement intention recognition model performs error correction processing on insurance claim settlement text data for training based on a pointer network model guided by a confusion set; performing claim settlement intention data expansion according to the insurance claim settlement text data after error correction by adopting a hybrid generation model combining a generative adversarial network and a variational auto-encoder; and performing model training based on the insurance claim settlement text data after error correction and the expanded claim settlement intention data. The claim settlement intention can be efficiently identified under the small sample condition.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence and management technology, and in particular to a method, apparatus, computer equipment and storage medium for identifying claim intention. Background Art

[0002] In the daily operation of the insurance industry, claims processing, as one of its core business processes, is directly related to the insurance company's operational efficiency, customer satisfaction and risk control capabilities. As a key link in the claims processing process, accurate identification and prediction of claims intentions are of immeasurable value in accelerating the claims process, improving service quality and effectively preventing fraud. However, the acquisition and application of insurance claims intention data currently faces severe challenges.

[0003] Due to the particularity of the insurance industry, claims data are often subject to strict compliance restrictions and privacy protection requirements, which greatly limits the widespread collection and sharing of high-quality claims intention data. Especially when dealing with rare or complex claims cases, the small number of such cases makes the problem of small samples that can be used for algorithm model training particularly prominent. Small sample data is not only difficult to support the training needs of complex models, but may also lead to limited model generalization capabilities and inability to effectively respond to diverse claims scenarios, thus seriously restricting the further development and application of claims intention recognition technology.

[0004] Therefore, how to effectively identify claim intention under small sample conditions becomes an urgent problem that needs to be solved. Summary of the invention

[0005] The purpose of the embodiments of the present application is to propose a method, apparatus, computer equipment and storage medium for identifying claim intention, so as to achieve efficient identification of claim intention under small sample conditions.

[0006] In order to solve the above technical problems, the present application embodiment provides a method for identifying claim intention, which adopts the following technical solution:

[0007] Obtain the insurance claim text to be identified;

[0008] Using the trained claim intention recognition model to perform intention recognition on the insurance claim text to be recognized, and obtaining a claim intention category corresponding to the insurance claim text to be recognized;

[0009] Determining the claims processing flow corresponding to the insurance claims text to be identified according to the claims intention category;

[0010] Among them, the claim intention recognition model is obtained by pre-correcting the insurance claim text data used for training using a pointer network model guided by a confusion set; using a hybrid generation model combining a generative adversarial network and a variational autoencoder to expand the claim intention data based on the corrected insurance claim text data; and training the model based on the corrected insurance claim text data and the expanded claim intention data.

[0011] In order to solve the above technical problems, the embodiment of the present application further provides a claim intention recognition device, which adopts the following technical solution:

[0012] A text acquisition module, used to acquire the insurance claim text to be identified;

[0013] An intention recognition module, used to use the trained claim intention recognition model to perform intention recognition on the insurance claim text to be recognized, and obtain a claim intention category corresponding to the insurance claim text to be recognized;

[0014] A process determination module, used to determine the claim processing process corresponding to the insurance claim text to be identified according to the claim intention category;

[0015] Among them, the claim intention recognition model is obtained by pre-correcting the insurance claim text data used for training using a pointer network model guided by a confusion set; expanding the claim intention data based on the corrected insurance claim text data using a hybrid generation model combining a generative adversarial network and a variational autoencoder; and training the model based on the corrected insurance claim text data and the expanded claim intention data.

[0016] In order to solve the above-mentioned technical problems, an embodiment of the present application also provides a computer device, which adopts the technical solution described below: the computer device includes a memory and a processor, the memory stores computer-readable instructions, and the processor implements the steps of the claim intention identification method as described in any one of the above items when executing the computer-readable instructions.

[0017] In order to solve the above technical problems, an embodiment of the present application also provides a computer-readable storage medium, which adopts the following technical solution: the computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the claim intention identification method as described in any one of the above items are implemented.

[0018] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0019] The embodiment of the present application adopts a pointer network model guided by confusion sets to perform error correction on the insurance claim text data for training, which can accurately identify and correct the erroneous text fragments in the claim text, thereby improving the accuracy and reliability of the data; by combining the hybrid generation model of the generative adversarial network and the variational autoencoder, the claim intention data is expanded according to the insurance claim text data after error correction, which increases the number and diversity of sample data, and helps the model to learn broader and deeper features during the training process, thereby improving the generalization ability of the model; based on the high-quality data after error correction and the diversified data after expansion, the claim intention recognition model is trained to more accurately identify the intent category of the insurance claim text to be identified. By using the trained claim intention recognition model to identify the intent category of the insurance claim text to be identified, the corresponding claim processing process is automatically determined, which not only improves the efficiency of claim processing, but also reduces the need for manual intervention and reduces the risk of human error. Based on the present application scheme, the problem of low accuracy of claim intention recognition caused by data scarcity in the insurance industry is solved, the automation processing capability is improved, and the burden of manual review is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0022] Figure 2 It is a flowchart of an embodiment of the claim intention identification method of the present application;

[0023] Figure 3 yes Figure 2 A flowchart of a specific implementation method before step S201;

[0024] Figure 4 It is a structural schematic diagram of an embodiment of the claim intention identification device of the present application;

[0025] Figure 5 It is a structural diagram of an embodiment of a computer device of the present application. DETAILED DESCRIPTION

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0027] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with 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.

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0029] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0030] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0031] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to a laptop computer 1011, a tablet computer 1012 or a mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0032] The server 103 may be a server that provides various services, such as a background server that provides support for a web page displayed on the terminal device 101 .

[0033] It should be noted that the claim intention identification method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the claim intention identification device is generally set in the server / terminal device.

[0034] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0035] Continue to refer Figure 2 , shows a flow chart of an embodiment of a method for identifying claim intention according to the present application. The method for identifying claim intention comprises the following steps:

[0036] Step S201, obtaining the insurance claim text to be identified;

[0037] In this embodiment, the electronic device (eg Figure 1 The server / terminal device shown in the figure) can obtain the insurance claim text to be identified through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 2G / 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0038] In this embodiment, the insurance claim text refers to the text content submitted by the customer when applying for insurance claims, including information such as claim request, accident description, loss situation, etc. The source of the insurance claim text may include the claim application form filled out by the customer, the written report submitted, or the claim application information submitted through channels such as telephone and online platforms.

[0039] Specifically, obtaining the insurance claim text to be identified may, for example, be collecting and extracting text information from a claim application submitted by a customer as the insurance claim text to be identified.

[0040] Step S202, using the trained claim intention recognition model to perform intention recognition on the insurance claim text to be recognized, and obtaining a claim intention category corresponding to the insurance claim text to be recognized;

[0041] Among them, the claim intention recognition model is obtained by pre-correcting the insurance claim text data used for training using a pointer network model guided by a confusion set; using a hybrid generation model combining a generative adversarial network and a variational autoencoder to expand the claim intention data based on the corrected insurance claim text data; and training the model based on the corrected insurance claim text data and the expanded claim intention data.

[0042] In this embodiment, the claim intention category refers to the claim type or demand divided according to the content of the insurance claim text. The claim intention category may include but is not limited to vehicle accident claims, medical compensation, property damage claims, etc. The claim intention recognition model is an algorithm model based on machine learning or deep learning technology. Its function is to automatically analyze and process the insurance claim text to identify and classify the claim intention expressed in the text. The claim intention recognition model learns and extracts key features related to the claim intention by training a large amount of text data containing claim requests, accident descriptions, loss conditions and other information, and calculates the probability of the category to which the key features belong.

[0043] Specifically, a trained claim intention recognition model is obtained, and the insurance claim text to be recognized is input into the trained claim intention recognition model. The insurance claim text to be recognized is analyzed and processed by the trained claim intention recognition model to recognize and classify the claim intention expressed in the text, and obtain the claim intention category corresponding to the insurance claim text to be recognized.

[0044] More specifically, before inputting the insurance claim text to be identified into the trained claim intention recognition model, a process of pre-training the claim intention recognition model may also be included. The claim intention recognition model is constructed by a confusion set-guided pointer network model and a hybrid generation model combining a generative adversarial network (GANs) and a variational autoencoder (VAEs). The training process of the claim intention recognition model may include: constructing a confusion set in advance based on common spelling errors, grammatical errors or misuse of professional terms in insurance claim texts, wherein the confusion set includes incorrect words and their corresponding correct words; obtaining pre-collected insurance claim text data for training, using a pointer network model guided by the confusion set to perform error correction on the insurance claim text data for training, identifying and correcting incorrect words in the insurance claim text data for training, replacing the identified incorrect words with the corresponding correct words in the confusion set, and obtaining corrected insurance claim text data; using a hybrid generation model combining a generative adversarial network and a variational autoencoder to expand claim intention data based on the corrected insurance claim text data, so that the expanded claim intention data covers more claim scenarios and intent categories; merging the corrected insurance claim text data and the expanded claim intention data to obtain a training data set, and performing model training on the selected model architecture based on the training data set to obtain a trained claim intention recognition model.

[0045] In some optional implementations, reference Figure 3 Before obtaining the insurance claim text to be identified, the following steps may also be included:

[0046] Step S2001, obtaining insurance claim text data for training, wherein the insurance claim text data for training includes at least one claim record;

[0047] Specifically, insurance claim text data for training is collected from a database or a related data source, and the insurance claim text data for training includes at least one claim record. Each claim record may include but is not limited to basic information of the customer, accident description, loss details, etc.

[0048] Step S2002, using a pre-built pointer network model based on confusion set guidance to perform error correction processing on the insurance claim text data for training, to obtain the error-corrected insurance claim text data;

[0049] Specifically, the insurance claim text data for training is input into a pre-built pointer network model guided by a confusion set, and the pointer network model is used to perform error correction on the insurance claim text data for training, identify and correct erroneous words in the insurance claim text data for training, and replace the identified erroneous words with the corresponding correct words in the confusion set to obtain the corrected insurance claim text data.

[0050] Step S2003, inputting the error-corrected insurance claim text data into a pre-trained hybrid generative model combining a generative adversarial network and a variational autoencoder, and performing data expansion based on the error-corrected insurance claim text data through the hybrid generative model to obtain expanded claim intention data;

[0051] Specifically, the corrected insurance claim text data is input into a pre-trained hybrid generative model that combines a generative adversarial network and a variational autoencoder. The hybrid generative model is used to generate similar claim intention data based on the corrected insurance claim text data, so that the newly generated claim intention data covers more claim scenarios and intention categories, thereby obtaining expanded claim intention data.

[0052] Step S2004, combining the error-corrected insurance claim text data with the expanded claim intention data to form an insurance claim intention data set for training;

[0053] Specifically, according to the data format requirements of model training, data preprocessing is performed on the error-corrected insurance claim text data and the expanded claim intention data, and the preprocessed error-corrected insurance claim text data and the expanded claim intention data are combined to obtain an insurance claim intention data set for training.

[0054] Step S2005, training an initial claim intention recognition model based on the insurance claim intention data set for training to obtain a trained claim intention recognition model.

[0055] Specifically, based on the insurance claim intention dataset for training, the initial claim intention recognition model is trained, and the training process may include model parameter adjustment, feature learning, classifier optimization, etc. After multiple iterations and tuning, a trained claim intention recognition model is obtained. The trained claim intention recognition model can accurately identify insurance claim texts of various types and complexities, and classify the insurance claim texts into corresponding claim intention categories.

[0056] The embodiment of the present application uses a pointer network model guided by a confusion set for error correction processing, effectively solves the problems of spelling errors, grammatical errors or confusion of professional terms that may exist in the insurance claim text, and improves the accuracy and consistency of the data. By inputting the corrected insurance claim text data into a hybrid generation model combining a generative adversarial network and a variational autoencoder, more diverse data samples can be generated based on the existing data, thereby effectively expanding the number of claim intention data. Data expansion not only increases the diversity of training data, but also helps to improve the generalization ability of the model to unseen data, so that the trained claim intention recognition model can more accurately identify various claim intentions. The corrected insurance claim text data is combined with the expanded claim intention data to form a training insurance claim intention data set, so that the data set contains both high-quality original data and diversified data generated by the model. The claim intention recognition model is obtained by training based on the training insurance claim intention data set. Since the quality and quantity of the training data set have been significantly improved, the trained model not only has a higher recognition performance in identifying claim intentions, but also has a lower false alarm rate in practical applications, achieving higher accuracy and robustness.

[0057] In some optional implementations, the pre-built pointer network model based on confusion set guidance is used to perform error correction processing on the insurance claim text data for training to obtain the error-corrected insurance claim text data, which may include the following steps:

[0058] The insurance claim text data for training is input into the pointer network model guided by the confusion set, and a pre-integrated confusion set is obtained through the pointer network model guided by the confusion set. The erroneous text fragments existing in the insurance claim text data for training are identified according to the confusion set, and the identified erroneous text fragments are corrected to obtain the corrected insurance claim text data.

[0059] Specifically, the insurance claim text data for training is input into the pointer network model guided by the confusion set, wherein the insurance claim text data for training may include basic information of the customer, accident description, loss details and other information, and the information may contain spelling errors, grammatical errors or misuse of professional terms. The pointer network model guided by the confusion set internally integrates a pre-built confusion set, which contains several incorrect words and their corresponding correct words. Through the pointer network model guided by the confusion set, the pre-integrated confusion set is obtained, and the input insurance claim text data for training is scanned according to the confusion set to identify the incorrect text fragments in the insurance claim text data for training. When an incorrect text fragment is identified, the incorrect text fragment is corrected according to the correct words in the confusion set, such as replacing the incorrect words, adjusting the grammatical structure or correcting the use of professional terms, etc., to obtain the corrected insurance claim text data.

[0060] The pointer network model guided by confusion sets in the embodiments of the present application can effectively identify and correct spelling errors, grammatical errors, and misuse of professional terms in training data through pre-integrated confusion sets, significantly improving the accuracy and consistency of data. The automated error correction process reduces the need for manual participation and reduces the cost of manual intervention.

[0061] In some optional implementations, the above-mentioned inputting the error-corrected insurance claim text data into a pre-trained hybrid generative model combining a generative adversarial network and a variational autoencoder, and performing data expansion according to the error-corrected insurance claim text data through the hybrid generative model to obtain the expanded claim intention data may include the following steps:

[0062] Converting the corrected insurance claim text data into corrected text features, and inputting the corrected text features into the hybrid generative model, wherein the hybrid generative model is composed of a generative adversarial network and a variational autoencoder, and the generative adversarial network includes a generator;

[0063] The generator is used to expand the claim intention data of the error-corrected insurance claim text data, and the expanded claim intention data is reconstructed by the variational autoencoder to obtain the expanded claim intention data.

[0064] Specifically, the corrected insurance claim text data is subjected to text preprocessing, such as word segmentation, stop word removal, word embedding, etc., so as to convert the corrected insurance claim text data into text features in the form of vectors or matrices to obtain corrected text features. The converted corrected text features are input into a pre-trained hybrid generative model combining a generative adversarial network and a variational autoencoder. The model consists of a generative adversarial network (GAN) and a variational autoencoder (VAE), wherein the generative adversarial network includes a generator. In the hybrid generative model, the input corrected insurance claim text features are expanded with claim intention data by the generator. The generator expands new claim intention data with similar intentions but different expressions based on the corrected insurance claim text features. Subsequently, the expanded claim intention data is reconstructed by the variational autoencoder, for example, the generated claim intention data is mapped back to the original feature space, and optimized and corrected by removing noise and redundancy, so as to obtain more accurate and stable expanded claim intention data.

[0065] The embodiment of the present application uses the generator in the hybrid generation model to generate new claim intention data based on the text features after error correction. It can expand more new claim intention data with similar intentions but different expressions on the basis of maintaining the original intention data, greatly increasing the size and diversity of the training data set, and effectively solving the problem of low accuracy of claim intention recognition caused by data scarcity in the insurance industry. By reconstructing the generated claim intention data using a variational autoencoder, it is possible to remove noise and redundant information, retain key features, and thus improve the quality of the expanded data.

[0066] In some optional implementations, the training of the initial claim intention recognition model based on the insurance claim intention dataset for training to obtain the trained claim intention recognition model may include the following steps:

[0067] Dividing the insurance claim intention dataset for training into a training set and a validation set;

[0068] Obtaining a preset neural network model, inputting the training set into the neural network model for training, and obtaining an initial claim intention recognition model;

[0069] The initial claim intention recognition model is verified based on the verification set, and the parameters of the initial claim intention recognition model are adjusted according to the verification result to obtain the trained claim intention recognition model.

[0070] Specifically, the insurance claim intention data set for training is divided into a training set and a validation set, and both the training set and the validation set may include several insurance claim intention data for training. A preset neural network model is obtained as an initial claim intention recognition model, and the neural network model may be built based on a deep learning framework (such as TensorFlow, PyTorch, etc.), and has the ability to process text data and recognize claim intention. The insurance claim intention data in the training set are sequentially input into the preset neural network model for training. During the training process, the neural network model will learn how to map the input insurance claim intention data to the corresponding claim intention label to obtain the initial claim intention recognition model. The initial claim intention recognition model is verified based on the validation set. During the verification process, the initial claim intention recognition model predicts the data in the validation set and compares it with the actual label to obtain a verification result for evaluating the performance of the model. According to the verification results, the parameters of the initial claim intention recognition model are gradually adjusted, and the model is retrained. After multiple iterations of training and parameter adjustment, the trained claim intention recognition model is finally obtained.

[0071] In some optional implementations, the above-mentioned use of the trained claim intention recognition model to perform intent recognition on the insurance claim text to be recognized to obtain the claim intention category corresponding to the insurance claim text to be recognized may include the following steps:

[0072] Extracting text features of the insurance claim text to be identified to obtain text features to be identified;

[0073] The text features to be identified are input into the trained claim intention recognition model, and the probability values ​​of the text features to be identified belonging to each preset claim intention category are calculated through the trained claim intention recognition model, and the claim intention category with the largest probability value is determined as the claim intention category corresponding to the insurance claim text to be identified.

[0074] Specifically, the insurance claim text to be identified is subjected to text preprocessing, such as word segmentation, stop word removal, word embedding, word vector, etc., so as to convert the insurance claim text to be identified into text features in the form of vectors or matrices to obtain text features to be identified. The text features to be identified are input into the trained claim intention recognition model, and the probability values ​​of the text features to be identified belonging to each preset claim intention category are calculated through the trained claim intention recognition model. The probability values ​​reflect the possibility that the text features belong to each category. According to the calculated probability values, the claim intention category with the largest probability value is selected to determine the claim intention category corresponding to the insurance claim text to be identified.

[0075] Step S203, determining the claim processing flow corresponding to the insurance claim text to be identified according to the claim intention category.

[0076] Specifically, the claims processing process refers to a series of operations to process the customer's claim application according to the identified claim intention category and preset rules and processes. These operations may include verifying claim information, assessing the extent of loss, calculating the amount of compensation, reviewing claim materials, etc. Different claim intention categories may correspond to different claims processing processes.

[0077] Specifically, for the identified claim intention category, according to the preset matching rules, the claim processing flow corresponding to the claim intention category is matched as the claim processing flow corresponding to the insurance claim text to be identified.

[0078] In some optional implementations, the step of determining the claim processing flow corresponding to the insurance claim text to be identified according to the claim intention category may include the following steps:

[0079] Obtain the correspondence between the preset claim intention categories and the claim processing procedures;

[0080] According to the corresponding relationship, a claim processing flow that matches the claim intention category corresponding to the insurance claim text to be identified is determined, and the insurance claim text to be identified is sent to the claim processing flow for corresponding claim processing.

[0081] Specifically, a preset correspondence table or rule base is obtained, which contains the correspondence between each claim intention category and the claim processing process. According to the correspondence between each claim intention category and the claim processing process, a claim processing process matching the claim intention category corresponding to the insurance claim text to be identified is searched, and the insurance claim text to be identified is sent to the matched claim processing process for corresponding claim processing.

[0082] The embodiment of the present application can automatically match the corresponding claim processing process according to the claim intention category through the pre-set correspondence between the claim intention category and the claim processing process, without the need for human intervention. This not only improves the processing efficiency, but also reduces the possibility of human errors, thereby realizing the automation and intelligence of claim processing.

[0083] The embodiment of the present application adopts the above scheme, specifically by adopting the pointer network model guided by the confusion set to perform error correction processing on the insurance claim text data for training, it can accurately identify and correct the erroneous text fragments in the claim text, thereby improving the accuracy and reliability of the data; by combining the hybrid generation model of the generative adversarial network and the variational autoencoder, the claim intention data is expanded according to the insurance claim text data after error correction, which increases the number and diversity of sample data, and helps the model to learn broader and deeper features during the training process, thereby improving the generalization ability of the model; based on the high-quality data after error correction and the diversified data after expansion, the claim intention recognition model is trained to more accurately identify the intent category of the insurance claim text to be identified. By adopting the trained claim intention recognition model to identify the intent category of the insurance claim text to be identified, the corresponding claim processing process is automatically determined, which not only improves the efficiency of claim processing, but also reduces the need for manual intervention and reduces the risk of human error. Based on the scheme of this application, the problem of low accuracy of claim intention recognition caused by data scarcity in the insurance industry is solved, the automation processing capability is improved, and the burden of manual review is reduced.

[0084] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned insurance claim text to be identified, the trained claim intention recognition model and claim intention category and other information as well as other relevant data, the above-mentioned insurance claim text to be identified, the trained claim intention recognition model and claim intention category and other information as well as other relevant data can also be stored in a node of a blockchain.

[0085] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0086] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the 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 can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0087] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed 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 can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0088] Further references Figure 4 , as a response to the above Figure 2 In order to realize the method shown in the figure, the present application provides an embodiment of a claim intention recognition device, and the device embodiment is Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0089] like Figure 4 As shown, the claim intention recognition device 400 described in this embodiment includes: a text acquisition module 401, an intention recognition module 402 and a process determination module 403. Among them:

[0090] The text acquisition module 401 is used to acquire the insurance claim text to be identified;

[0091] The intention recognition module 402 is used to use the trained claim intention recognition model to perform intention recognition on the insurance claim text to be recognized, and obtain the claim intention category corresponding to the insurance claim text to be recognized;

[0092] A process determination module 403 is used to determine the claim processing process corresponding to the insurance claim text to be identified according to the claim intention category;

[0093] Among them, the claim intention recognition model is obtained by pre-correcting the insurance claim text data used for training using a pointer network model guided by a confusion set; expanding the claim intention data based on the corrected insurance claim text data using a hybrid generation model combining a generative adversarial network and a variational autoencoder; and training the model based on the corrected insurance claim text data and the expanded claim intention data.

[0094] In this embodiment, the insurance claim text refers to the text content submitted by the customer when applying for insurance claims, including information such as claim request, accident description, loss situation, etc. The source of the insurance claim text may include the claim application form filled out by the customer, the written report submitted, or the claim application information submitted through channels such as telephone and online platforms.

[0095] Specifically, the text acquisition module 401 is used to acquire the insurance claim text to be identified, for example, it can be to collect and extract text information from the claim application submitted by the customer as the insurance claim text to be identified.

[0096] Specifically, the claim intention category refers to the claim type or demand divided according to the content of the insurance claim text. The claim intention category may include but is not limited to vehicle accident claims, medical compensation, property damage claims, etc. The claim intention recognition model is an algorithm model based on machine learning or deep learning technology. Its function is to automatically analyze and process insurance claim texts to identify and classify the claim intentions expressed in the text. The claim intention recognition model learns and extracts key features related to the claim intention by training a large amount of text data containing claim requests, accident descriptions, loss conditions, etc., and calculates the probability of the key features belonging to the category.

[0097] Specifically, the intention recognition module 402 is used to obtain a trained claim intention recognition model, input the insurance claim text to be recognized into the trained claim intention recognition model, and analyze and process the insurance claim text to be recognized through the trained claim intention recognition model to recognize and classify the claim intention expressed in the text, and obtain the claim intention category corresponding to the insurance claim text to be recognized.

[0098] More specifically, before inputting the insurance claim text to be identified into the trained claim intention recognition model, a process of pre-training the claim intention recognition model may also be included. The claim intention recognition model is constructed by a confusion set-guided pointer network model and a hybrid generation model combining a generative adversarial network (GANs) and a variational autoencoder (VAEs). The training process of the claim intention recognition model may include: constructing a confusion set in advance based on common spelling errors, grammatical errors or misuse of professional terms in insurance claim texts, wherein the confusion set includes incorrect words and their corresponding correct words; obtaining pre-collected insurance claim text data for training, using a pointer network model guided by the confusion set to perform error correction on the insurance claim text data for training, identifying and correcting incorrect words in the insurance claim text data for training, replacing the identified incorrect words with the corresponding correct words in the confusion set, and obtaining corrected insurance claim text data; using a hybrid generation model combining a generative adversarial network and a variational autoencoder to expand claim intention data based on the corrected insurance claim text data, so that the expanded claim intention data covers more claim scenarios and intent categories; merging the corrected insurance claim text data and the expanded claim intention data to obtain a training data set, and performing model training on the selected model architecture based on the training data set to obtain a trained claim intention recognition model.

[0099] Specifically, the claims processing process refers to a series of operations to process the customer's claim application according to the identified claim intention category and preset rules and processes. These operations may include verifying claim information, assessing the extent of loss, calculating the amount of compensation, reviewing claim materials, etc. Different claim intention categories may correspond to different claims processing processes.

[0100] Specifically, the process determination module 403 is used to match the claim processing process corresponding to the identified claim intention category according to a preset matching rule as the claim processing process corresponding to the insurance claim text to be identified.

[0101] The claim intention recognition device 400 of the embodiment of the present application, by using a pointer network model guided by a confusion set to perform error correction processing on the insurance claim text data for training, can accurately identify and correct the erroneous text fragments in the claim text, thereby improving the accuracy and reliability of the data; by combining the hybrid generation model of the generative adversarial network and the variational autoencoder, the claim intention data is expanded according to the insurance claim text data after error correction, which increases the number and diversity of sample data, and helps the model to learn broader and deeper features during the training process, thereby improving the generalization ability of the model; based on the high-quality data after error correction and the diversified data after expansion, the claim intention recognition model is trained to more accurately identify the intent category of the insurance claim text to be identified. By using the trained claim intention recognition model to identify the intent category of the insurance claim text to be identified, the corresponding claim processing process is automatically determined, which not only improves the efficiency of claim processing, but also reduces the need for manual intervention and reduces the risk of human error. Based on the present application scheme, the problem of low accuracy of claim intention recognition caused by data scarcity in the insurance industry is solved, the automation processing capability is improved, and the burden of manual review is reduced.

[0102] In some optional implementations of this embodiment, the claim intention recognition device 400 may further include a data acquisition module, a text error correction module, a data expansion module, a data combination module, and a model training module. Among them:

[0103] A data acquisition module, used to acquire insurance claim text data for training, wherein the insurance claim text data for training includes at least one claim record;

[0104] A text error correction module, used to use a pre-built pointer network model based on confusion set guidance to perform error correction processing on the insurance claim text data for training, so as to obtain the insurance claim text data after error correction;

[0105] A data expansion module, used for inputting the error-corrected insurance claim text data into a pre-trained hybrid generative model combining a generative adversarial network and a variational autoencoder, and performing data expansion based on the error-corrected insurance claim text data through the hybrid generative model to obtain expanded claim intention data;

[0106] A data combination module, used to combine the error-corrected insurance claim text data with the expanded claim intention data to form an insurance claim intention data set for training;

[0107] The model training module is used to train an initial claim intention recognition model based on the insurance claim intention data set used for training, so as to obtain a trained claim intention recognition model.

[0108] Specifically, the data acquisition module is used to collect insurance claim text data for training from a database or related data source, and the insurance claim text data for training includes at least one claim record. Each claim record may include but is not limited to the basic information of the customer, accident description, loss details, etc.

[0109] Specifically, the text error correction module is used to input the insurance claim text data for training into a pre-built pointer network model guided by a confusion set, and perform error correction processing on the insurance claim text data for training through the pointer network model, identify and correct erroneous words in the insurance claim text data for training, and replace the identified erroneous words with the corresponding correct words in the confusion set to obtain the corrected insurance claim text data.

[0110] Specifically, the data expansion module is used to input the corrected insurance claim text data into a pre-trained hybrid generative model that combines a generative adversarial network and a variational autoencoder, and generate similar claim intention data based on the corrected insurance claim text data through the hybrid generative model, so that the newly generated claim intention data covers more claim scenarios and intention categories, thereby obtaining expanded claim intention data.

[0111] Specifically, the data combination module is used to perform data preprocessing on the error-corrected insurance claim text data and the expanded claim intention data according to the data format requirements of the model training, and to combine the preprocessed error-corrected insurance claim text data with the expanded claim intention data to obtain an insurance claim intention data set for training.

[0112] Specifically, the model training module is used to train the initial claim intention recognition model based on the insurance claim intention data set for training. The training process may include model parameter adjustment, feature learning, classifier optimization, etc. After multiple iterations and tuning, a trained claim intention recognition model is obtained. The trained claim intention recognition model can accurately identify insurance claim texts of various types and complexities, and classify the insurance claim texts into corresponding claim intention categories.

[0113] The claim intention recognition device 400 of the embodiment of the present application uses a pointer network model guided by a confusion set for error correction processing, which effectively solves the problems of spelling errors, grammatical errors or confusion of professional terms that may exist in the insurance claim text, and improves the accuracy and consistency of the data. By inputting the corrected insurance claim text data into a hybrid generation model that combines a generative adversarial network and a variational autoencoder, more diverse data samples can be generated based on existing data, thereby effectively expanding the amount of claim intention data. Data expansion not only increases the diversity of training data, but also helps to improve the generalization ability of the model to unseen data, so that the trained claim intention recognition model can more accurately identify various claim intentions. The corrected insurance claim text data is combined with the expanded claim intention data to form a training insurance claim intention data set, so that the data set contains both high-quality original data and diversified data generated by the model. The claim intention recognition model is obtained by training based on the insurance claim intention dataset used for training. Since the quality and quantity of the training dataset have been significantly improved, the trained model not only has high recognition performance in identifying claim intention, but also has a lower false alarm rate in practical applications, achieving higher accuracy and robustness.

[0114] In some optional implementations of this embodiment, the text error correction module may include a text error correction submodule.

[0115] The text error correction submodule is used to input the insurance claim text data for training into the pointer network model guided by the confusion set, obtain a pre-integrated confusion set through the pointer network model guided by the confusion set, identify erroneous text fragments in the insurance claim text data for training according to the confusion set, and correct the identified erroneous text fragments to obtain the corrected insurance claim text data.

[0116] Specifically, the text error correction submodule is used to input the insurance claim text data for training into the pointer network model guided by the confusion set, wherein the insurance claim text data for training may include the basic information of the customer, accident description, loss details and other information, which may contain spelling errors, grammatical errors or misuse of professional terms. The pointer network model guided by the confusion set internally integrates a pre-built confusion set, which contains several incorrect words and their corresponding correct words. Through the pointer network model guided by the confusion set, the pre-integrated confusion set is obtained, and the input insurance claim text data for training is scanned according to the confusion set to identify the incorrect text fragments in the insurance claim text data for training. When an incorrect text fragment is identified, the incorrect text fragment is corrected according to the correct words in the confusion set, such as replacing the incorrect words, adjusting the grammatical structure or correcting the use of professional terms, etc., to obtain the corrected insurance claim text data.

[0117] The text error correction submodule of the embodiment of the present application is based on a pointer network model guided by a confusion set, and can effectively identify and correct spelling errors, grammatical errors, and misuse of professional terms in training data through a pre-integrated confusion set, significantly improving the accuracy and consistency of the data. The automated error correction process reduces the need for manual participation and reduces the cost of manual intervention.

[0118] In some optional implementations of this embodiment, the data expansion module may include a text input submodule and a data expansion submodule.

[0119] A text input submodule, used for converting the corrected insurance claim text data into corrected text features, and inputting the corrected text features into the hybrid generative model, wherein the hybrid generative model is composed of a generative adversarial network and a variational autoencoder, and the generative adversarial network includes a generator;

[0120] The data expansion submodule is used to expand the claim intention data of the error-corrected insurance claim text data through the generator, and reconstruct the expanded claim intention data through the variational autoencoder to obtain the expanded claim intention data.

[0121] Specifically, the text input submodule is used to perform text preprocessing on the corrected insurance claim text data, such as word segmentation, stop word removal, word embedding, etc., so as to convert the corrected insurance claim text data into text features in the form of vectors or matrices to obtain corrected text features. The converted corrected text features are input into a pre-trained hybrid generation model combining a generative adversarial network and a variational autoencoder. The model consists of a generative adversarial network (GAN) and a variational autoencoder (VAE), wherein the generative adversarial network includes a generator. The data expansion submodule is used to expand the claim intention data of the input corrected insurance claim text features through a generator in the hybrid generation model. The generator will expand new claim intention data with similar intentions but different expressions based on the corrected insurance claim text features. Subsequently, the expanded claim intention data is reconstructed through the variational autoencoder, for example, the generated claim intention data is mapped back to the original feature space, and optimized and corrected by removing noise and redundancy, so as to obtain more accurate and stable expanded claim intention data.

[0122] The data expansion module of the embodiment of the present application, by using the generator in the hybrid generation model, generates new claim intention data according to the text features after error correction, and can expand more new claim intention data with similar intentions but different expressions on the basis of maintaining the original intention data, greatly increasing the size and diversity of the training data set, and effectively solving the problem of low accuracy of claim intention recognition caused by data scarcity in the insurance industry. By reconstructing the generated claim intention data using a variational autoencoder, it is possible to remove noise and redundant information, retain key features, and thus improve the quality of the expanded data.

[0123] In some optional implementations of this embodiment, the above-mentioned model training module may include a data partitioning submodule, a model training submodule and a model verification submodule. Among them:

[0124] A data partitioning submodule, used to divide the insurance claim intention dataset used for training into a training set and a validation set;

[0125] A model training submodule is used to obtain a preset neural network model, input the training set into the neural network model for training, and obtain an initial claim intention recognition model;

[0126] The model verification submodule is used to verify the initial claim intention recognition model based on the verification set, and adjust the parameters of the initial claim intention recognition model according to the verification result to obtain the trained claim intention recognition model.

[0127] Specifically, the data partitioning submodule is used to divide the insurance claim intention data set for training into a training set and a validation set, and the training set and the validation set both include a number of insurance claim intention data for training. The model training submodule is used to obtain a preset neural network model as an initial claim intention recognition model. The neural network model can be built based on a deep learning framework (such as TensorFlow, PyTorch, etc.) and has the ability to process text data and recognize claim intentions. The insurance claim intention data in the training set are input into the preset neural network model for training in sequence. During the training process, the neural network model will learn how to map the input insurance claim intention data to the corresponding claim intention label to obtain the initial claim intention recognition model. The model verification submodule is used to verify the initial claim intention recognition model based on the validation set. During the verification process, the initial claim intention recognition model will predict the data in the validation set and compare it with the actual label to obtain a verification result for evaluating the model performance. According to the verification results, the parameters of the initial claim intention recognition model are gradually adjusted, and the model is retrained. After multiple iterations of training and parameter adjustment, the trained claims intention recognition model is finally obtained.

[0128] In some optional implementations of this embodiment, the intention recognition module 402 may include a feature extraction submodule and a category determination submodule.

[0129] A feature extraction submodule is used to extract text features of the insurance claim text to be identified, so as to obtain text features to be identified;

[0130] A category determination submodule is used to input the text features to be identified into the trained claim intention recognition model, calculate the probability values ​​of the text features to be identified belonging to each preset claim intention category through the trained claim intention recognition model, and determine the claim intention category with the largest probability value as the claim intention category corresponding to the insurance claim text to be identified.

[0131] Specifically, the feature extraction submodule is used to perform text preprocessing on the insurance claim text to be identified, such as word segmentation, stop word removal, word embedding, word vector, etc., to convert the insurance claim text to be identified into text features in the form of vectors or matrices to obtain the text features to be identified. The category determination submodule is used to input the text features to be identified into the trained claim intention recognition model, and calculate the probability value of the text features to be identified belonging to each preset claim intention category through the trained claim intention recognition model. The probability value reflects the possibility of the text features belonging to each category. According to the calculated probability value, the claim intention category with the largest probability value is selected to determine the claim intention category corresponding to the insurance claim text to be identified.

[0132] In some optional implementations of this embodiment, the process determination module 403 may include a relationship acquisition submodule and a process matching submodule.

[0133] A relationship acquisition submodule is used to obtain the corresponding relationship between the preset claim intention category and the claim processing process;

[0134] The process matching submodule is used to determine the claim processing process that matches the claim intention category corresponding to the insurance claim text to be identified based on the corresponding relationship, and send the insurance claim text to be identified to the claim processing process for corresponding claim processing.

[0135] Specifically, the relationship acquisition submodule is used to obtain a preset correspondence table or rule base, which contains the correspondence between each claim intention category and the claim processing process. The process matching submodule is used to find the claim processing process that matches the claim intention category corresponding to the insurance claim text to be identified based on the correspondence between each claim intention category and the claim processing process, and send the insurance claim text to be identified to the matched claim processing process for corresponding claim processing.

[0136] The process determination module 403 of the embodiment of the present application can automatically match the corresponding claim processing process according to the claim intention category through the pre-set correspondence between the claim intention category and the claim processing process, without the need for human intervention. This not only improves the processing efficiency, but also reduces the possibility of human errors, thereby realizing the automation and intelligence of claim processing.

[0137] To solve the above technical problems, the present application also provides a computer device. Figure 5 , Figure 5 This is a basic structural block diagram of the computer device in this embodiment.

[0138] The computer device 6 includes a memory 61, a processor 62, and a network interface 63 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 6 with a memory 61, a processor 62, and a network interface 63, but it should be understood that it is not required to implement all the components shown, and more or fewer components can 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 calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0139] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0140] The memory 61 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, disk, optical disk, etc. In some embodiments, the memory 61 can be an internal storage unit of the computer device 6, such as a hard disk or memory of the computer device 6. In other embodiments, the memory 61 can also be an external storage device of the computer device 6, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 6. Of course, the memory 61 can also include both the internal storage unit of the computer device 6 and its external storage device. In this embodiment, the memory 61 is generally used to store the operating system and various application software installed on the computer device 6, such as computer-readable instructions of the claim intention recognition method, etc. In addition, the memory 61 can also be used to temporarily store various types of data that have been output or are to be output.

[0141] The processor 62 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip in some embodiments. The processor 62 is generally used to control the overall operation of the computer device 6. In this embodiment, the processor 62 is used to run the computer-readable instructions stored in the memory 61 or process data, such as computer-readable instructions for running the claim intention identification method.

[0142] The network interface 63 may include a wireless network interface or a wired network interface. The network interface 63 is generally used to establish a communication connection between the computer device 6 and other electronic devices.

[0143] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the claim intention identification method as described above.

[0144] The computer device, computer-readable storage medium and computer-readable instructions provided by the embodiment of the present application, through the processor executing the pointer network model based on the guidance of the confusion set to correct the insurance claim text data for training, can accurately identify and correct the erroneous text fragments in the claim text, thereby improving the accuracy and reliability of the data; by combining the hybrid generation model of the generative adversarial network and the variational autoencoder, the claim intention data is expanded according to the insurance claim text data after the error correction, the number and diversity of sample data are increased, which helps the model to learn broader and deeper features during the training process, thereby improving the generalization ability of the model; based on the high-quality data after error correction and the diversified data after expansion, the claim intention recognition model is trained to more accurately identify the intent category of the insurance claim text to be identified. By using the trained claim intention recognition model to identify the intent category of the insurance claim text to be identified, the corresponding claim processing process is automatically determined, which not only improves the efficiency of claim processing, but also reduces the need for manual intervention and reduces the risk of human error. Based on the present application scheme, the problem of low accuracy of claim intention recognition caused by data scarcity in the insurance industry is solved, the automation processing capability is improved, and the burden of manual review is reduced.

[0145] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0146] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to replace some of the technical features therein with equivalents. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

[0147] The non-Company software tools or components appearing in the embodiments of this application are merely examples and do not represent actual use.

Claims

1. A method for identifying claim intention, characterized in that: The steps include: Obtain the insurance claim text to be identified; Using the trained claim intention recognition model to perform intention recognition on the insurance claim text to be recognized, and obtaining a claim intention category corresponding to the insurance claim text to be recognized; Determining the claims processing flow corresponding to the insurance claims text to be identified according to the claims intention category; Among them, the claim intention recognition model is obtained by pre-correcting the insurance claim text data used for training using a pointer network model guided by a confusion set; using a hybrid generation model combining a generative adversarial network and a variational autoencoder to expand the claim intention data based on the corrected insurance claim text data; and training the model based on the corrected insurance claim text data and the expanded claim intention data.

2. The claim intention identification method according to claim 1, characterized in that: Before the step of obtaining the insurance claim text to be identified, the method further includes: Acquire insurance claim text data for training, wherein the insurance claim text data for training includes at least one claim record; Using a pre-built pointer network model based on confusion set guidance, error correction processing is performed on the insurance claim text data for training to obtain error-corrected insurance claim text data; Inputting the error-corrected insurance claim text data into a pre-trained hybrid generative model combining a generative adversarial network and a variational autoencoder, performing data expansion based on the error-corrected insurance claim text data through the hybrid generative model to obtain expanded claim intention data; Combining the corrected insurance claim text data with the expanded claim intention data to form an insurance claim intention data set for training; An initial claim intention recognition model is trained based on the insurance claim intention data set used for training to obtain a trained claim intention recognition model.

3. The claim intention identification method according to claim 2, characterized in that: The step of using a pre-built pointer network model based on confusion set guidance to perform error correction processing on the insurance claim text data for training to obtain the error-corrected insurance claim text data includes: The insurance claim text data for training is input into the pointer network model guided by the confusion set, and a pre-integrated confusion set is obtained through the pointer network model guided by the confusion set. The erroneous text fragments existing in the insurance claim text data for training are identified according to the confusion set, and the identified erroneous text fragments are corrected to obtain the corrected insurance claim text data.

4. The claim intention identification method according to claim 2, characterized in that: The step of inputting the error-corrected insurance claim text data into a pre-trained hybrid generative model combining a generative adversarial network and a variational autoencoder, and performing data expansion according to the error-corrected insurance claim text data through the hybrid generative model to obtain expanded claim intention data includes: Converting the corrected insurance claim text data into corrected text features, and inputting the corrected text features into the hybrid generative model, wherein the hybrid generative model is composed of a generative adversarial network and a variational autoencoder, and the generative adversarial network includes a generator; The generator is used to expand the claim intention data of the error-corrected insurance claim text data, and the expanded claim intention data is reconstructed by the variational autoencoder to obtain the expanded claim intention data.

5. The claim intention identification method according to claim 2, characterized in that: The step of training the initial claim intention recognition model based on the insurance claim intention data set for training to obtain the trained claim intention recognition model comprises: Dividing the insurance claim intention dataset for training into a training set and a validation set; Obtaining a preset neural network model, inputting the training set into the neural network model for training, and obtaining an initial claim intention recognition model; The initial claim intention recognition model is verified based on the verification set, and the parameters of the initial claim intention recognition model are adjusted according to the verification result to obtain the trained claim intention recognition model.

6. The claim intention identification method according to claim 1, characterized in that: The step of using the trained claim intention recognition model to perform intent recognition on the insurance claim text to be recognized to obtain the claim intention category corresponding to the insurance claim text to be recognized includes: Extracting text features of the insurance claim text to be identified to obtain text features to be identified; The text features to be identified are input into the trained claim intention recognition model, and the probability values ​​of the text features to be identified belonging to each preset claim intention category are calculated through the trained claim intention recognition model, and the claim intention category with the largest probability value is determined as the claim intention category corresponding to the insurance claim text to be identified.

7. The claim intention identification method according to claim 1, characterized in that: The step of determining the claim processing flow corresponding to the insurance claim text to be identified according to the claim intention category includes: Obtain the correspondence between the preset claim intention categories and the claim processing procedures; According to the corresponding relationship, a claim processing flow that matches the claim intention category corresponding to the insurance claim text to be identified is determined, and the insurance claim text to be identified is sent to the claim processing flow for corresponding claim processing.

8. A claim intention recognition device, characterized in that: The device comprises: A text acquisition module, used to acquire the insurance claim text to be identified; An intention recognition module, used to use the trained claim intention recognition model to perform intention recognition on the insurance claim text to be recognized, and obtain a claim intention category corresponding to the insurance claim text to be recognized; A process determination module, used to determine the claim processing process corresponding to the insurance claim text to be identified according to the claim intention category; Among them, the claim intention recognition model is obtained by pre-correcting the insurance claim text data used for training using a pointer network model guided by a confusion set; expanding the claim intention data based on the corrected insurance claim text data using a hybrid generation model combining a generative adversarial network and a variational autoencoder; and training the model based on the corrected insurance claim text data and the expanded claim intention data.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the claim intention identification method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the claim intention identification method according to any one of claims 1 to 7.

Images