Vehicle insurance claim risk identification method and device, computer device and storage medium

Abstract

The present application relates to the field of speech analysis, and particularly relates to a vehicle insurance claim risk identification method and device, computer equipment and storage medium. The method comprises: obtaining vehicle insurance report voice data; converting the voice data into text data through speech recognition technology; obtaining scene text data corresponding to the text data from a vehicle insurance scene library; performing voice slicing processing on the report voice data according to the scene text data to obtain answer voice data; performing emotion recognition on the answer voice data to obtain emotion information of the reporter; inputting the scene text data and the emotion information into a risk identification model to obtain a risk identification result. The present application identifies the risk of the emotion information of the reporter in different scenes, so that the risk identification not only considers the emotion information of the reporter, but also considers the scene information of the emotion information, which can improve the accuracy of the risk identification result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of speech analysis, and in particular to a vehicle insurance claim risk identification method and device, a computer device and a storage medium. BACKGROUND

[0002] With the rapid development of China's economy, the number of vehicles on the road is increasing, and the vehicle insurance industry has also emerged. Generally, in the existing vehicle insurance claim process, the insured person initiates the vehicle insurance claim process to the insurance party, and then the insurance party confirms the loss assessment of the claim data submitted by the insured person.

[0003] In the process of submitting claim data, there may be some fraudulent behavior. For example, when the reporter has drunk driving or changed driving, it is difficult to know the true accident situation of the reporter through the claim data submitted by the insured person, and the insurance party needs to arrange staff to go to the scene to investigate in order to know the reporter's drunk driving or changed driving. Therefore, a large amount of manpower and time is needed to identify the fraud risk in the claim process. Therefore, the existing vehicle insurance claim method has poor recognition effect on fraud risk. SUMMARY

[0004] Therefore, it is necessary to provide a vehicle insurance claim risk identification method, device, computer device and storage medium to solve the problem of poor recognition effect of the existing vehicle insurance claim method on fraud risk.

[0005] A vehicle insurance claim risk identification method comprises:

[0006] Obtain vehicle insurance reporting voice data; the vehicle insurance reporting voice data comprises reporter voice data and agent voice data;

[0007] Perform voice conversion processing on the agent voice data through voice recognition technology to obtain agent text data;

[0008] Obtain scene text data corresponding to the agent text data from a vehicle insurance scene library;

[0009] Perform voice slicing processing on the reporter voice data according to the scene text data to obtain answer sentence voice data corresponding to the scene text data;

[0010] Perform emotion recognition on the answer sentence voice data to obtain emotion information of the reporter;

[0011] Input the scene text data and the emotion information into a risk identification model for risk identification to obtain a risk identification result corresponding to the vehicle insurance reporting voice data.

[0012] A vehicle insurance claim risk identification device comprises:

[0013] The vehicle insurance claim voice data module is configured to obtain vehicle insurance claim voice data; the vehicle insurance claim voice data includes claimant voice data and agent voice data;

[0014] The agent text data module is configured to perform voice conversion processing on the agent voice data by using a voice recognition technology to obtain agent text data.

[0015] The scene text data module is configured to obtain scene text data corresponding to the agent text data from a vehicle insurance scene library.

[0016] The answer voice data module is configured to perform voice slicing processing on the claimant voice data according to the scene text data to obtain answer voice data corresponding to the scene text data.

[0017] The emotion information module is configured to perform emotion recognition on the answer voice data to obtain emotion information of the claimant.

[0018] The risk identification result module is configured to input the scene text data and the emotion information into a risk identification model to perform risk identification and obtain a risk identification result corresponding to the vehicle insurance claim voice data.

[0019] A computer device includes a memory, a processor, and computer readable instructions stored in the memory and executable on the processor, and the processor executes the computer readable instructions to implement the vehicle insurance claim risk identification method.

[0020] One or more readable storage media storing computer readable instructions, and the computer readable instructions are executed by one or more processors to cause the one or more processors to perform the vehicle insurance claim risk identification method.

[0021] The aforementioned method, device, computer equipment, and storage medium for identifying risks in auto insurance claims involve acquiring auto insurance claim voice data. This voice data includes the claimant's voice data and the agent's voice data. The agent's voice data is processed using speech recognition technology to obtain agent text data. Scene text data corresponding to the agent's text data is retrieved from an auto insurance scene database. The claimant's voice data is then processed into speech segments based on the scene text data to obtain corresponding response voice data. Emotional recognition is performed on the response voice data to obtain the claimant's emotional information. The scene text data and the emotional information are input into a risk identification model for risk identification, resulting in a risk identification result corresponding to the auto insurance claim voice data. This invention determines scene text data from agent voice data, then determines the claimant's response voice data corresponding to the scene text data, and performs emotional analysis based on this response voice data to obtain the claimant's emotional information. The emotional information determined based on the response voice data can more accurately express the claimant's emotions, improving the accuracy of emotional information. Furthermore, risk identification is performed based on emotional information and contextual text data. Risk identification is carried out based on the emotional information of the person reporting the crime in different scenarios. This means that risk identification not only takes into account the emotional information of the person reporting the crime, but also the contextual information that generates that emotional information, which can improve the accuracy of risk identification results. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of an application environment for the vehicle insurance claims risk identification method according to an embodiment of the present invention;

[0024] Figure 2 This is a flowchart illustrating a method for identifying vehicle insurance claim risks in one embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the structure of a vehicle insurance claims risk identification device according to an embodiment of the present invention;

[0026] Figure 4 This is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0027] With reference to the drawings and the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by those skilled in the art without creative efforts belong to the scope of the present application.

[0028] The vehicle insurance claim risk identification method provided in the embodiments can be applied in the application environment as shown in Figure 1 , wherein the client and the server communicate. The client includes but is not limited to various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers.

[0029] In an embodiment, as shown in Figure 2 , a vehicle insurance claim risk identification method is provided. The method is applied to the server in Figure 1 for example, and includes the following steps.

[0030] S10, obtaining vehicle insurance report voice data; the vehicle insurance report voice data includes reporter voice data and agent voice data.

[0031] Understandably, the vehicle insurance report voice data refers to the voice data generated between the reporter and the agent in the vehicle insurance claim application process. The reporter voice data refers to the voice data generated by the reporter in the vehicle insurance claim application process, including the voice of the reporter describing the accident needing vehicle insurance claim and the voice of the reporter answering the questions of the agent. The agent voice data refers to the voice data actually generated by the agent in the process of handling the vehicle insurance claim of the reporter, including the voice of the agent asking questions to the reporter and the voice of the agent answering the questions of the reporter.

[0032] S20, converting the agent voice data by voice recognition technology to obtain agent text data.

[0033] Understandably, the voice recognition technology refers to a technology of converting voice data into text data. In this embodiment, the voice recognition technology can be Automatic Speech Recognition (automatic speech recognition technology). The voice conversion processing refers to converting the agent voice data into text data by voice recognition technology to obtain the content of the agent text data. That is, the agent text data is the converted data of the agent voice data.

[0034] S30, obtaining scene text data corresponding to the agent text data from a vehicle insurance scene library.

[0035] Understandably, the car insurance scenario library includes a plurality of scenario sentence data, which is a sentence text pre-set by a car insurance field expert for different dialogue scenarios in car insurance claim. Different scenarios correspond to different scenario sentence data. For example, the scenario sentence data can be "What is your license plate number?", "Are you the driver yourself?", and "What did you hit?" and the like. The agent text data includes a plurality of agent sentence data. For example, the agent sentence data can be "Please report the license plate number", "Are you the driver?" and the like. The scenario text data refers to the text data in the car insurance scenario library corresponding to the agent text data. The scenario text data includes a plurality of scenario sentence data corresponding to the agent text data. Specifically, the similarity between all agent sentence data in the agent text data and all scenario sentence data in the car insurance scenario library is calculated to obtain a similarity calculation result. Further, according to the similarity calculation result, the scenario sentence data corresponding to the agent sentence data in the agent text data is extracted one by one to obtain the scenario text data. For example, if the agent text data S includes the agent sentence data A "Please report the license plate number", the agent sentence data B "Are you the driver?", and the agent sentence data C "Are you sure your license plate number is XXX?", according to the similarity calculation result, the scenario sentence data a corresponding to the agent sentence data A is "What is your license plate number?", the scenario sentence data b corresponding to the agent sentence data B is "Are you the driver yourself?", and there is no scenario sentence data corresponding to the agent sentence data C; then the scenario text data M corresponding to the agent text data S is the text data including the scenario sentence data a and the scenario sentence data b.

[0036] S40, performing voice slicing processing on the claimant voice data according to the scenario text data to obtain answer voice data corresponding to the scenario text data.

[0037] Understandably, voice slicing processing refers to the process of performing voice slicing on the claimant voice data according to the scenario text data to extract the answer voice data. Specifically, the agent timestamp of the agent sentence corresponding to the scenario sentence in the scenario text data is obtained from the agent text data. Further, according to the agent timestamp, the claimant voice data is subjected to voice slicing processing to obtain the answer voice data corresponding to the scenario text data. The answer voice data refers to the voice data when the claimant answers the agent question.

[0038] S50, performing emotion recognition on the answer voice data to obtain the emotion information of the claimant.

[0039] Understandably, the answer sentence voice data refers to the voice data of the reporter corresponding to the scene text data. The emotion recognition refers to a process of analyzing the context semantics and tone in the answer sentence voice data to obtain the emotion information of the reporter. The emotion information refers to information of the relevant emotions of the reporter when answering the questions of the seat.

[0040] S60, inputting the scene text data and the emotion information into a risk identification model to perform risk identification, to obtain a risk identification result corresponding to the vehicle insurance reporting voice data.

[0041] Understandably, the risk identification model refers to a neural network model that has been trained and completed, which is used to perform risk identification and evaluation on the vehicle insurance claim application of the reporter according to the scene text data and the emotion information, to obtain a risk identification result corresponding to the vehicle insurance reporting voice data. The risk identification result is used to indicate whether the vehicle insurance claim application of the reporter has fraudulent behavior and risk level information, etc. The fraudulent behavior includes the behavior of the reporter existing drunk driving or changing driving and denying.

[0042] In steps S10-S60, the vehicle insurance reporting voice data is obtained; the vehicle insurance reporting voice data includes reporter voice data and seat voice data; the seat voice data is processed by voice recognition technology to obtain seat text data; the scene text data corresponding to the seat text data is obtained from a vehicle insurance scene library; the reporter voice data is processed by voice slicing according to the scene text data to obtain answer sentence voice data corresponding to the scene text data; the emotion information of the reporter is obtained by performing emotion recognition on the answer sentence voice data; the scene text data and the emotion information are input into a risk identification model to perform risk identification, to obtain a risk identification result corresponding to the vehicle insurance reporting voice data. In this embodiment, the scene text data is determined by the seat voice data, and then the answer sentence voice data of the reporter corresponding to the scene text data is determined, and emotion analysis is performed based on the answer sentence voice data to obtain the emotion information of the reporter. The emotion information determined based on the answer sentence voice data can more accurately express the emotions of the reporter, and improve the accuracy of the emotion information. Then, the risk identification is performed based on the emotion information and the scene text data, and the risk identification is performed for the emotion information of the reporter in different scenes, so that the risk identification not only considers the emotion information of the reporter, but also considers the scene information of the emotion information, which can improve the accuracy of the risk identification result.

[0043] Optionally, in step S30, i.e., obtaining the scene text data corresponding to the seat text data from the vehicle insurance scene library, includes:

[0044] S301, performing word vector conversion processing on the seat text data to obtain seat text vectors;

[0045] S302, calculate the similarity between the agent text vector and the scene text vector of the vehicle insurance scenario library, to obtain a text similarity value;

[0046] S303, according to the text similarity value, determine the scene text data corresponding to the agent text data.

[0047] Understandably, the word vector conversion process refers to the process of converting the agent text data from text data to vector data to obtain the agent text vector. The agent text vector is the vector expression of the agent text data. The scene text vector is the vector of all scene sentence data in the vehicle insurance scenario library. The text similarity value refers to the similarity between the agent text vector and the scene text vector of the vehicle insurance scenario library. That is, the text similarity value refers to the similarity between the agent text data and all scene sentence data in the vehicle insurance scenario library. The text similarity value includes several sentence similarity values. Among them, the sentence similarity value refers to the similarity between the agent sentence data in the agent text data and the scene sentence data in the vehicle insurance scenario library.

[0048] In steps S301-S303, according to the agent text data, the scene text data corresponding to the vehicle insurance claim voice data can be quickly determined.

[0049] Optionally, in step S301, that is, the word vector conversion process of the agent text data is performed to obtain the agent text vector, comprising:

[0050] S3011, calculate the word vector of each agent word in the agent text data by a word vector algorithm to obtain several agent word vectors;

[0051] S3012, according to the agent word vector, weight average the agent word contained in each agent sentence data in the agent text data to obtain several agent sentence vectors;

[0052] S3013, generate the agent text vector according to the several agent sentence vectors.

[0053] It can be understood that the word vector algorithm is used to convert each agent word in the agent text data into a vector, which can be an algorithm of word2vec. Among them, word2vec includes two algorithms, namely skip-gram and CBOW (Continuous Bag-of-Word Model). Among them, skip-gram is to predict the words around the center word through the center word, and CBOW is to predict the center word through the surrounding words. The agent word is a word in the agent text data. The agent word vector refers to the vector expression of the agent sentence data in the agent text data. Among them, the agent sentence data includes a plurality of agent words. The word weight of each agent word in the agent sentence data is calculated through the word vector algorithm, and the weighted average of all agent words contained in the agent sentence data is obtained according to the word weight of the agent word and the agent word vector of the agent word. The agent sentence vector includes a plurality of agent sentence vectors.

[0054] In steps S3011-S3013, the vector of each agent word in the agent text data is calculated through the word vector algorithm, so as to obtain the vector of the agent text data, which can improve the accuracy of vector conversion.

[0055] Optionally, the text similarity value includes a plurality of sentence similarity values.

[0056] In step S302, that is, the similarity between the agent text vector and the scene text vector of the car insurance scene library is calculated to obtain a text similarity value, including:

[0057] S3021, obtaining any agent sentence vector in the agent text vector;

[0058] S3022, calculating the sentence similarity between the agent sentence vector and each scene sentence vector in the scene text vector to obtain a plurality of sentence similarity values.

[0059] It can be understood that the sentence similarity value refers to the similarity between the agent sentence data in the agent text data and the scene sentence data in the car insurance scene library. That is, the sentence similarity value refers to the similarity between the agent sentence vector and the scene sentence vector. Among them, the scene sentence vector is the vector expression of the scene sentence data in the car insurance scene.

[0060] In steps S3021 and S3022, the sentence similarity between any agent sentence vector in the agent text vector and each scene sentence vector in the scene text vector is calculated, which considers the sentence similarity value between all agent sentence vectors and all scene sentence vectors, so that the scene text data determined according to the sentence similarity value is more accurate.

[0061] Optionally, the scene text data comprises a plurality of scene sentences; the agent text data comprises a plurality of agent sentences; and the answer voice data comprises a plurality of suspect answer sentences.

[0062] In step S40, the voice slicing processing of the suspect voice data according to the scene text data is performed to obtain answer voice data corresponding to the scene text data, which comprises:

[0063] S401, obtaining an agent timestamp of an agent sentence corresponding to the scene sentence from the agent text data;

[0064] S402, performing voice slicing processing on the suspect voice data according to the agent timestamp to obtain a suspect answer sentence voice corresponding to the scene sentence.

[0065] Understandably, the agent timestamp refers to the timestamp of the agent sentence in the agent text data. According to the agent timestamp, the suspect timestamp corresponding to the agent timestamp in the suspect voice data is determined. The suspect timestamp is the timestamp of the suspect speaking recorded in the suspect voice data. Here, the suspect timestamp can be the timestamps of the two closest sentence voices of the suspect generated after the agent timestamp in the suspect voice data corresponding to the agent timestamp. For example, when the agent sentence corresponding to the scene sentence is "Hello, what can I do for you", the agent timestamp corresponding to the agent sentence is "1.24:3.41"; the two closest sentence voices of the suspect generated after the agent timestamp in the suspect voice data are "Hello, my car was hit" and "Hello, my car was hit", and the timestamps of "Hello, my car was hit" and "Hello, my car was hit" are "3.67:5.55", then "3.67:5.55" is the suspect timestamp corresponding to the agent timestamp. The voice slicing processing of the suspect voice data refers to performing voice slicing processing on the suspect voice data according to the determined suspect timestamp corresponding to the agent timestamp to obtain a plurality of voice slicing data, and extracting the voice slicing data corresponding to the suspect timestamp as the suspect answer sentence voice to obtain the suspect answer sentence voice corresponding to the scene sentence.

[0066] In this embodiment, the suspect answer sentence voice corresponding to the scene sentence is obtained from the suspect voice data, so that the obtained suspect answer sentence voice has the scene, and the accuracy of emotion recognition can be improved.

[0067] Optionally, in step S50, the emotion recognition of the answer voice data is performed to obtain the suspect's emotional information, which comprises:

[0068] S501, performing voice conversion processing on the answer voice data by the voice recognition technology to obtain answer text data;

[0069] S502, input the answer sentence text data, the answer sentence voice data and the scene text data into a multi-modal emotion recognition model;

[0070] S503, perform multi-modal emotion analysis processing on the answer sentence text data, the answer sentence voice data and the scene text data through the multi-modal emotion recognition model to obtain emotion features of the reporter;

[0071] S504, according to the emotion features, identify and classify the emotion type of the reporter to obtain the emotion information.

[0072] Understandably, the multi-modal emotion recognition model is used for multi-modal emotion analysis on the reporter's emotion according to the answer sentence text data, the answer sentence voice data and the scene text data. The multi-modal emotion recognition model can be a model based on TensorFusion Network (TensorFusion Network). The emotion features refer to the emotion information of the reporter contained in the answer sentence voice data. The emotion features include different emotion information of the reporter in different dialogue scenes. The emotion information includes different emotion types of the reporter in different dialogue scenes. The emotion types include sadness, fear, happiness, disgust, surprise, anger, hesitation, neutrality and tension, etc.

[0073] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution. The execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0074] In an embodiment, a vehicle insurance claim risk identification device is provided, which corresponds to the vehicle insurance claim risk identification method in the above embodiment. As shown in the figure, the vehicle insurance claim risk identification device includes a vehicle insurance report voice data module 10, a seat text data module 20, a scene text data module 30, an answer sentence voice data module 40, an emotion information module 50 and a risk identification result module 60. The function modules are described in detail as follows: Figure 3

[0075] The vehicle insurance report voice data module 10 is used for obtaining vehicle insurance report voice data; the vehicle insurance report voice data includes reporter voice data and seat voice data;

[0076] The seat text data module 20 is used for performing voice conversion processing on the seat voice data through voice recognition technology to obtain seat text data;

[0077] The scene text data module 30 is used for obtaining scene text data corresponding to the seat text data from a vehicle insurance scene library;

[0078] ​The answer sentence voice data module 40 is configured to perform voice slicing processing on the reporter voice data according to the scene text data, so as to obtain answer sentence voice data corresponding to the scene text data.

[0079] The emotion information module 50 is configured to perform emotion recognition on the answer sentence voice data, so as to obtain emotion information of the reporter.

[0080] The risk identification result module 60 is configured to input the scene text data and the emotion information into a risk identification model to perform risk identification, so as to obtain a risk identification result corresponding to the vehicle insurance reporting voice data.

[0081] The scene text data module 30 comprises:

[0082] The agent text vector unit is configured to perform word vector conversion processing on the agent text data, so as to obtain an agent text vector.

[0083] The text similarity value unit is configured to calculate a similarity between the agent text vector and a scene text vector of the vehicle insurance scene library, so as to obtain a text similarity value.

[0084] The scene text data unit is configured to determine scene text data corresponding to the agent text data according to the text similarity value.

[0085] Optionally, the agent text vector unit comprises:

[0086] The agent word vector unit is configured to calculate a word vector of each agent word in the agent text data by using a word vector algorithm, so as to obtain a plurality of agent word vectors.

[0087] The agent sentence vector unit is configured to perform weighted average on agent words contained in each agent sentence data in the agent text data according to the agent word vectors, so as to obtain a plurality of agent sentence vectors.

[0088] The agent text vector generation unit is configured to generate the agent text vector according to the plurality of agent sentence vectors.

[0089] Optionally, the text similarity value comprises a plurality of sentence similarity values.

[0090] The text similarity value unit comprises:

[0091] The agent sentence vector unit is configured to obtain any agent sentence vector in the agent text vector.

[0092] The sentence similarity value unit is configured to calculate a sentence similarity between the agent sentence vector and each scene sentence vector in the scene text vector, so as to obtain a plurality of sentence similarity values.

[0093] Optionally, the scene text data comprises a plurality of scene sentences; the agent text data comprises a plurality of agent sentences; and the answer sentence voice data comprises a plurality of complainant answer sentence voices.

[0094] The answer sentence voice data module 40 comprises:

[0095] An agent timestamp unit, configured to acquire an agent timestamp of an agent sentence corresponding to the scene sentence from the agent text data;

[0096] A complainant answer sentence voice unit, configured to perform voice slicing processing on the complainant voice data according to the agent timestamp, to obtain a complainant answer sentence voice corresponding to the scene sentence.

[0097] Optionally, the emotion information module 50 comprises:

[0098] An answer sentence text data unit, configured to perform voice conversion processing on the answer sentence voice data by using the voice recognition technology, to obtain answer sentence text data;

[0099] An input data unit, configured to input the answer sentence text data, the answer sentence voice data and the scene text data into a multi-modal emotion recognition model;

[0100] An emotion feature unit, configured to perform multi-modal emotion analysis processing on the answer sentence text data, the answer sentence voice data and the scene text data by using the multi-modal emotion recognition model, to obtain emotion features of the complainant;

[0101] An emotion information unit, configured to identify and classify an emotion type of the complainant according to the emotion features, to obtain the emotion information.

[0102] The specific limitations of the vehicle insurance claim risk identification device can refer to the limitations of the vehicle insurance claim risk identification method described above, which will not be repeated here. Each module in the vehicle insurance claim risk identification device described above can be realized by software, hardware and their combinations in whole or in part. The above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so that the processor calls and executes the operations corresponding to each module.

[0103] In one embodiment, a computer device is provided, which can be a server, and its internal structure diagram can be as shown in Figure 4As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium, an internal memory. The readable storage medium stores an operating system, computer readable instructions and a database. The internal memory provides an environment for the operation of the operating system and computer readable instructions in the readable storage medium. The database of the computer device is used to store the data involved in the vehicle insurance claim risk identification method. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer readable instructions are executed by the processor to implement a vehicle insurance claim risk identification method. The readable storage medium provided by the embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

[0104] In one embodiment, a computer device is provided, comprising a memory, a processor, and computer readable instructions stored on the memory and executable on the processor, the processor executing the computer readable instructions to implement the following steps:

[0105] Obtaining vehicle insurance claim voice data; the vehicle insurance claim voice data includes claimant voice data and agent voice data;

[0106] Converting the agent voice data into text data through voice recognition technology;

[0107] Obtaining scene text data corresponding to the agent text data from a vehicle insurance scene library;

[0108] Processing the claimant voice data through voice slicing according to the scene text data, to obtain answer voice data corresponding to the scene text data;

[0109] Identifying the emotion of the claimant from the answer voice data, to obtain emotion information of the claimant;

[0110] Inputting the scene text data and the emotion information into a risk identification model to identify the risk, to obtain a risk identification result corresponding to the vehicle insurance claim voice data.

[0111] In one embodiment, one or more computer readable storage media having computer readable instructions stored thereon are provided. The readable storage medium provided by the embodiment includes a non-volatile readable storage medium and a volatile readable storage medium. The computer readable instructions are stored on the readable storage medium, and the computer readable instructions are executed by one or more processors to implement the following steps:

[0112] Obtaining vehicle insurance claim voice data; the vehicle insurance claim voice data includes claimant voice data and agent voice data;

[0113] The voice data of the agent is processed by voice recognition technology to obtain agent text data;

[0114] The scene text data corresponding to the agent text data is obtained from a vehicle insurance scene library;

[0115] The voice data of the claimant is processed by voice slicing according to the scene text data to obtain answer sentence voice data corresponding to the scene text data;

[0116] The emotion information of the claimant is obtained by performing emotion recognition on the answer sentence voice data;

[0117] The scene text data and the emotion information are input into a risk identification model to perform risk identification, and a risk identification result corresponding to the vehicle insurance voice data is obtained.

[0118] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by computer readable instructions instructing related hardware, and the computer readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium, and when the computer readable instructions are executed, the processes of the above-mentioned embodiments can be included. Wherein, any reference to memory, storage, database or other medium used in each embodiment provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0119] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0120] The above-described embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for identifying risks in vehicle insurance claims, characterized by, The method comprises the following steps: acquiring vehicle insurance claim voice data; the vehicle insurance claim voice data comprises claimant voice data and agent voice data; performing voice conversion processing on the agent voice data by using a voice recognition technology to obtain agent text data; obtaining scene text data corresponding to the agent text data from a vehicle insurance scene library; the vehicle insurance scene library comprises scene sentence data, wherein the scene sentence data is a sentence text pre-set by a vehicle insurance field expert for different dialogue scenes in vehicle insurance claim settlement; the scene text data comprises scene sentence data corresponding to the agent text data; performing voice slicing processing on the claimant voice data according to the scene text data to obtain answer sentence voice data corresponding to the scene text data; performing emotion recognition on the answer sentence voice data to obtain emotion information of the claimant; inputting the scene text data and the emotion information into a risk identification model to perform risk identification and obtain a risk identification result corresponding to the vehicle insurance claim voice data.

2. The motor insurance claim risk identification method of claim 1, wherein, The step of obtaining scene text data corresponding to the agent text data from the vehicle insurance scene library comprises the following steps: performing word vector conversion processing on the agent text data to obtain agent text vectors; calculating the similarity between the agent text vectors and scene text vectors of the vehicle insurance scene library to obtain a text similarity value; determining scene text data corresponding to the agent text data according to the text similarity value.

3. The motor insurance claim risk identification method of claim 2, wherein, The step of performing word vector conversion processing on the agent text data to obtain agent text vectors comprises the following steps: calculating the word vectors of each agent word in the agent text data by using a word vector algorithm to obtain a plurality of agent word vectors; performing weighted average on the agent words contained in each agent sentence data in the agent text data according to the agent word vectors to obtain a plurality of agent sentence vectors; generating the agent text vectors according to the plurality of agent sentence vectors.

4. The motor insurance claim risk identification method of claim 3, wherein, The text similarity value comprises a plurality of sentence similarity values. The step of calculating the similarity between the agent text vectors and the scene text vectors of the vehicle insurance scene library to obtain a text similarity value comprises the following steps: obtaining any agent sentence vector in the agent text vectors; calculating the sentence similarity between the agent sentence vector and each scene sentence vector in the scene text vectors to obtain a plurality of sentence similarity values.

5. The method of claim 1, wherein, The scene text data comprises a plurality of scene sentences; the agent text data comprises a plurality of agent sentences; and the answer sentence voice data comprises a plurality of claimant answer sentence voices. The step of performing voice slicing processing on the claimant voice data according to the scene text data to obtain answer sentence voice data corresponding to the scene text data comprises the following steps: obtaining the agent time stamp of the agent sentence corresponding to the scene sentence from the agent text data; performing voice slicing processing on the claimant voice data according to the agent time stamp to obtain the claimant answer sentence voice corresponding to the scene sentence.

6. The method of claim 1, wherein, The step of performing emotion recognition on the answer sentence voice data to obtain emotion information of the claimant comprises the following steps: The answer sentence voice data is subjected to voice conversion processing through the voice recognition technology, and answer sentence text data is obtained. The answer sentence text data, the answer sentence voice data and the scene text data are input into a multi-modal emotion recognition model. The multi-modal emotion recognition model is used to analyze and process the answer sentence text data, the answer sentence voice data and the scene text data, and the emotional characteristics of the reporter are obtained. According to the emotional characteristics, the emotional type of the reporter is identified and classified, and the emotional information is obtained.

7. An automobile insurance claim risk identification device characterized by comprising: It comprises: A vehicle insurance reporting voice data module is configured to obtain vehicle insurance reporting voice data, wherein the vehicle insurance reporting voice data comprises reporter voice data and agent voice data. An agent text data module is configured to convert the agent voice data into agent text data through voice recognition technology. A scene text data module is configured to obtain scene text data corresponding to the agent text data from a vehicle insurance scene library, wherein the vehicle insurance scene library comprises scene sentence data, the scene sentence data is a sentence text pre-set by a vehicle insurance field expert for different dialogue scenes in vehicle insurance claims, and the scene text data comprises scene sentence data corresponding to the agent text data. An answer sentence voice data module is configured to perform voice slicing processing on the reporter voice data according to the scene text data, and obtain answer sentence voice data corresponding to the scene text data. An emotional information module is configured to identify the emotional information of the reporter by performing emotion recognition on the answer sentence voice data. A risk identification result module is configured to input the scene text data and the emotional information into a risk identification model to identify the risk, and obtain a risk identification result corresponding to the vehicle insurance reporting voice data.

8. The vehicle insurance claim risk identification device of claim 7, wherein, The scene text data module comprises: An agent text vector unit is configured to convert the agent text data into an agent text vector. A text similarity value unit is configured to calculate the similarity between the agent text vector and a scene text vector of the vehicle insurance scene library, and obtain a text similarity value. A scene text data unit is configured to determine the scene text data corresponding to the agent text data according to the text similarity value.

9. A computer device comprising a memory, a processor, and computer readable instructions stored in the memory and executable on the processor, wherein, The processor executes the computer readable instructions to implement the vehicle insurance claims risk identification method of any one of claims 1 to 6.

10. One or more readable storage media having stored thereon computer- readable instructions, characterized in that, The computer readable instructions are executed by one or more processors to cause the one or more processors to perform the vehicle insurance claims risk identification method of any one of claims 1 to 6.

Images