Automatic auditing method, device and equipment for insurance complaint and medium

Abstract

The invention relates to an automatic auditing method and device for insurance complaints, equipment and a medium. According to the method, entity extraction is carried out on a complaint text, feature extraction is carried out on complaint voice, acoustic features are input into a preset risk analysis model to output a risk level, a corresponding audit decision rule is determined, feature extraction is carried out on the complaint text, text features and acoustic features are fused, and fused features are input into a preset prediction model. And generating an audit report. The method and the device can be applied to business systems such as financial insurance, through the target entity of the complaint text and the acoustic features of the complaint voice, the risk level is obtained according to the acoustic features, the audit decision rule is determined according to the risk level and the target entity, the text features are obtained according to the complaint text, and the target prediction result is obtained according to the text features and the acoustic features. And indicating a complaint processing mode according to the target prediction result. Therefore, the complaint information of the user is comprehensively analyzed to accurately obtain the complaint appeal of the user.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of artificial intelligence and insurance technology, and in particular to an automatic auditing method and device for insurance complaints, equipment and a medium. BACKGROUND

[0002] With the development of information technology, enterprises in the field of financial insurance business began to try to use an automated way to handle complaints when processing the complaint content of users. However, most of the complaint content contains rich information that can reflect the user's comprehensive expression. However, this information is often ignored in traditional automated processing, resulting in a large deviation in the processing of user complaints.

[0003] Currently, the most common way to handle complaints is to use uniform rules to handle user complaints, and to perform single text analysis on the complaint content of the user, which makes it difficult to accurately solve the user's appeal. For example, in the field of financial insurance business, when a user applies for car insurance claims, the user can only identify "pay" when the user uses an urgent tone to urge the payment, so that the complaint is classified as a normal complaint, resulting in an incorrect judgment of the user's payment progress.

[0004] Therefore, how to comprehensively analyze the complaint information of the user and accurately obtain the complaint appeal of the user has become a problem to be solved. SUMMARY

[0005] Therefore, the embodiments of the present application provide an automatic auditing method, device, equipment and medium for insurance complaints to solve the problem of how to comprehensively analyze the complaint information of the user and accurately obtain the complaint appeal of the user.

[0006] In a first aspect, the embodiments of the present application provide an automatic auditing method for insurance complaints, comprising: After receiving the complaint, the complaint voice and complaint text of the user initiating the complaint are obtained, the complaint text is subjected to entity extraction to obtain a target entity, and the complaint voice is subjected to feature extraction to obtain acoustic features; The acoustic features are input into a preset risk analysis model to output a risk level, and a corresponding auditing decision rule is determined according to the target entity and the risk level; The complaint text is subjected to feature extraction to obtain text features, the text features and the acoustic features are fused to obtain fused features, and the fused features are input into a preset prediction model to output a target prediction result; An auditing report is generated according to the auditing decision rule and the target prediction result, and the auditing report is used to indicate the processing mode of the complaint.

[0007] In a second aspect, an embodiment of the present application provides an automatic auditing device for insurance complaints, comprising: a feature extraction module configured to, after receiving a complaint, acquire complaint voice and complaint text of a user who initiates the complaint, perform entity extraction on the complaint text to obtain a target entity, and perform feature extraction on the complaint voice to obtain acoustic features; a rule determination module configured to input the acoustic features into a preset risk analysis model to output a risk level, and determine a corresponding auditing decision rule according to the target entity and the risk level; a target prediction module configured to perform feature extraction on the complaint text to obtain text features, fuse the text features with the acoustic features to obtain fused features, input the fused features into a preset prediction model, and output a target prediction result; an auditing report module configured to generate an auditing report according to the auditing decision rule and the target prediction result, wherein the auditing report is used to indicate a processing manner for the complaint.

[0008] In a third aspect, an embodiment of the present application provides a computer device, which comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, and the processor implements the automatic auditing method for insurance complaints when executing the computer program.

[0009] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the automatic auditing method for insurance complaints.

[0010] Compared with the prior art, the embodiment of the present application has the following beneficial effects: In the present application, after receiving a complaint, the complaint voice and complaint text of the user initiating the complaint are obtained, entity extraction is performed on the complaint text to obtain target entities, feature extraction is performed on the complaint voice to obtain acoustic features, the acoustic features are input into a preset risk analysis model to output a risk level, according to the target entities and the risk level, a corresponding audit decision rule is determined, feature extraction is performed on the complaint text to obtain text features, the text features and the acoustic features are fused to obtain fused features, the fused features are input into a preset prediction model to output a target prediction result, according to the audit decision rule and the target prediction result, an audit report is generated, and the audit report is used to indicate the processing mode of the complaint. The present application can be applied in financial insurance and other business systems, through the target entities of the complaint text, the acoustic features of the complaint voice, the risk level obtained according to the acoustic features, the audit decision rule determined according to the risk level and the target entities, the text features obtained according to the complaint text, the target prediction result obtained according to the text features and the acoustic features, and the processing mode of the complaint indicated according to the target prediction result. Thus, the complaint information of the user is comprehensively analyzed to accurately obtain the complaint demands of the user. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0012] Figure 1 is an application environment schematic diagram of an automatic audit method for insurance complaints provided by an embodiment of the present application; Figure 2 is a flow schematic diagram of an automatic audit method for insurance complaints provided by an embodiment of the present application; Figure 3 is a flow schematic diagram of an automatic audit method for insurance complaints provided by an embodiment of the present application; Figure 4 is a flow schematic diagram of an automatic audit method for insurance complaints provided by an embodiment of the present application; Figure 5 is a flow schematic diagram of an automatic audit method for insurance complaints provided by an embodiment of the present application; Figure 6 is a structure schematic diagram of an automatic audit device for insurance complaints provided by an embodiment of the present application; Figure 7 is a structure schematic diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0013] In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular architectures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods, devices, circuits, and

[0014] It will be understood that the terms "comprises" and / or "comprising," when used in this specification, include the presence of one or more features, integers, steps, operations, elements, and / or components but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0015] It will be understood that the term "and / or," when used in the specification and in the following claims, is intended to mean one or more of the associated listed items can be present, and, if not present, are not excluded.

[0016] As used in this specification and claims, the terms "if' and "when" can be interpreted to mean "upon determination" or "in response to a determination" or "upon detection" or "in response to a detection" depending on the context. Similarly, the phrase "if determined" or "if detected" can be interpreted to mean "upon determination" or "in response to a determination" or "upon detection" or "in response to a detection" depending on the context.

[0017] In addition, the terms "first," "second," "third," etc. as used in the description and the appended claims are used only to distinguish one element from another, and do not otherwise limit the elements or the claims.

[0018] Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" or "in a various embodiment" or "in some embodiments" in various places throughout this specification are not necessarily all referring to the same embodiment, unless otherwise specified. The terms "comprise," "comprising," "including," "containing," "have," "having," and "include" and their variations are meant to be open-ended terms that do not limit any of the claims to the components, compositions, or acts listed therein. Rather, these terms are meant to encompass one or more of the listed components, compositions, or acts, and any equivalents thereof.

[0019] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0020] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0021] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0022] To illustrate the technical solution of this application, specific embodiments are described below.

[0023] The first embodiment of this application provides an automated auditing method for insurance complaints, which can be applied to, for example... Figure 1 In this application environment, the client communicates with the server. The server, as the carrier device of the enterprise system, can connect to the client through the enterprise system network.

[0024] The aforementioned enterprise systems may include, but are not limited to, financial systems, such as insurance systems and auditing systems. The systems are developed using software development kits (SDKs). Specifically, application authentication modules, intelligent routing modules, and high availability management modules can be developed. These modules can interface with user management systems within the enterprise system to achieve corresponding management functions.

[0025] The client includes, but is not limited to, a palm computer, a desktop computer, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a cloud terminal device, a personal digital assistant (PDA), and the like computer device. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and basic cloud computing services such as big data and artificial intelligence platforms.

[0026] Referring to Figure 2 , FIG. 1 is a flowchart of an automatic auditing method for insurance complaints provided by Embodiment Two of the present application. The automatic auditing method for insurance complaints can be applied to a server in Figure 1 .

[0027] As shown in Figure 2 , the automatic auditing method for insurance complaints can include the following steps: Step S201, after receiving a complaint, obtaining a complaint voice and a complaint text of a user initiating the complaint, performing entity extraction on the complaint text to obtain a target entity, and performing feature extraction on the complaint voice to obtain acoustic features.

[0028] Among them, after receiving an insurance complaint, two types of key data generated by the complaint user need to be collected, namely the complaint voice and the complaint text. The complaint voice can be audio generated by the user during the telephone complaint process, and the complaint text can be the complaint content filled in by the user online, the text information of the customer service record, etc.

[0029] Among them, the entity refers to an object with specific meaning, and through entity extraction, these key information can be accurately positioned from the text to provide a basis for subsequent analysis, and the acoustic feature is a parameter reflecting the physical characteristics of the voice signal.

[0030] For example, in the financial insurance complaint business scenario, the target entity can include the name of the insurance company, the name of the insurance product, the name of the policyholder, the event occurrence time, the location, etc. The acoustic features can be pitch, duration, volume, speed, tone, etc. These features can reflect the emotional state, speaking style, and the like information of the complaint user.

[0031] Optionally, noise extraction is performed on the complaint text to determine unstructured noise; The unstructured noise is removed from the complaint text to obtain a de-noised complaint text; Perform entity extraction on the denoised complaint text to obtain target entities.

[0032] Among them, in the complaint text, unstructured noise refers to information that is irrelevant to the core complaint content and does not materially help understand the complaint demands. These information may include colloquial mood words, repetitive expressions, non-standard words, network popular language but irrelevant to the essence of the complaint, etc. After determining the unstructured noise, these noises can be removed from the original complaint text. It can be realized by programming to traverse each word or phrase in the text, and if it is marked as noise, it will be deleted from the text.

[0033] Optionally, input the complaint voice into a voice analysis model to obtain a speech rate feature; Input the complaint voice into an emotion prediction model to obtain an emotion feature; Fuse the speech rate feature and the emotion feature to obtain an acoustic feature.

[0035] Among them, in the complaint scene, speech rate is an important clue. Fast speech rate may indicate that the user is agitated, eager to express dissatisfaction with the problem, while slow speech rate may indicate that the user is thinking, calm or has expression barriers, etc. Insurance companies can preliminarily judge the user's state according to the speech rate feature and adjust the subsequent communication strategy.

[0036] The user's emotion feature is crucial for handling complaints. If it is detected that the user is in an angry or dissatisfied state, the customer service personnel need to communicate more carefully, adopt strategies to calm emotions, and prioritize the user's emotional demands before handling specific complaint problems. If the user's emotions are relatively calm, it may be more appropriate to directly enter the problem solving link.

[0037] The fused acoustic feature can provide more comprehensive reference for subsequent complaint handling. In predicting the risk level of complaints, determining the processing priority, and developing communication strategies, the acoustic feature can play an important role. For example, when the acoustic feature shows that the user's speech rate is fast and the emotion is angry, it means that the complaint problem may be more serious, and professional personnel need to be arranged as soon as possible to handle it and adopt a gentle and sincere communication approach.

[0038] Step S202, input the acoustic feature into a preset risk analysis model to output a risk level, and determine a corresponding audit decision rule according to the target entity and the risk level.

[0039] Among them, after inputting the acoustic feature into the preset risk analysis model, the model will analyze and judge the input acoustic feature according to the pattern it has learned, and finally output a corresponding risk level.

[0040] The risk level is generally classified, for example, into low, medium and high. Low risk may mean that the user is only calmly feeding back the problem and is less likely to cause serious follow-up problems, medium risk means that the problem has a certain severity, and high risk indicates that the user is very agitated and the problem may be complex and serious and needs to be handled with priority.

[0041] The audit decision rule is a series of processing strategies pre-established according to different combinations of target entities and risk levels. These rules are established by insurance companies based on business experience, compliance requirements and risk management needs.

[0042] For example, if the target entity is a common small financial insurance claim and the risk level is low, the corresponding audit decision rule can be to use a simple online audit process to quickly handle the complaint. If the target entity is a major insurance claim with a high amount, and the risk level is high, the audit decision rule may require the establishment of a special investigation team to conduct a comprehensive and in-depth investigation of the claim event, including verification of records, accident authenticity, etc., and the involvement of senior management in decision-making.

[0043] Optionally, according to the target entity, a rule clause corresponding to the target entity is determined in combination with a rule engine.

[0044] According to the risk level, the rule clause is filtered to obtain a filtered rule, and the filtered rule is determined as an audit decision rule.

[0045] The rule engine is a rule-based reasoning system that stores a large number of business rules and can quickly match and apply corresponding rules based on input data. In the insurance complaint audit scenario, the rule engine stores rule clauses related to various target entities, which are formulated according to insurance business laws and regulations, company policies and past handling experience. When the risk level is high, more stringent and comprehensive rule clauses may be selected.

[0046] For example, in the financial insurance business scenario, for high-risk complaints, key regulatory rules for insurance companies will be enabled, including detailed review of the authenticity and compliance of claim data, increased scrutiny of whether product clauses contain misleading statements, etc.

[0047] In step S203, text features are extracted from the complaint text to obtain text features, the text features are fused with the acoustic features to obtain fused features, the fused features are input into a pre-set prediction model, and a target prediction result is output.

[0048] The text features refer to attributes extracted from the complaint text that can reflect the content and semantic information of the text, and the frequency of each word in the complaint text is counted.

[0049] For example, in an insurance complaint text, keywords such as "claim delay" and "poor service attitude" appear multiple times, and these high-frequency words may reflect the key issues of the complaint.

[0050] Among them, according to the parts of speech of each word in the complaint text, such as nouns, verbs, adjectives, etc. For example, a large number of adjectives (such as "poor" and "serious") may indicate that the user's negative emotions are relatively strong. Through natural language processing technology, the semantic information of the complaint text is extracted. For example, use the word vector model to convert the words in the text into vector representation, so as to capture the semantic relationship between words. If the complaint text mentions "the insurance clause does not match the introduction during sales", semantic analysis can identify that this involves the issue of insurance sales compliance.

[0051] Among them, the text features and acoustic features alone can only reflect the complaint situation from one aspect, and the fusion of the two can comprehensively utilize the information of voice and text to more comprehensively and accurately describe the complaint event. For example, the text content may express specific problems, but the emotional information in the voice can further strengthen the judgment of the severity of the problem.

[0052] The text feature vector and the acoustic feature vector are directly connected together to form a longer feature vector. For example, the text feature vector has 100 dimensions, and the acoustic feature vector has 50 dimensions. After splicing, a 150-dimensional fusion feature vector is obtained.

[0053] According to the importance of different features, the corresponding weights are given, and then the weighted text features and acoustic features are combined. For example, if the acoustic features are considered to be more able to reflect the user's true attitude in the current scenario, a higher weight can be given to the acoustic features.

[0054] The pre-set prediction model is obtained by training a large amount of historical data, and common prediction models include neural network models (such as multilayer perceptron, long short-term memory network LSTM), random forest, etc. The target prediction result is the prediction information of the complaint event given by the model according to the input fusion features. In the insurance complaint scenario, the prediction result may include the time of complaint settlement, whether additional investigation is needed, whether it will trigger further complaints from customers, etc.

[0055] Step S204, generating an audit report according to the audit decision rule and the target prediction result, the audit report being used to indicate the processing mode of the complaint.

[0056] Among them, the content of the audit report can include the basic information of the complaint user, the summary of the complaint content, the risk assessment, the presentation of the prediction result, the basis of the audit decision, the processing suggestion, etc. The audit report provides clear processing guidance for relevant personnel of the insurance enterprise.

[0057] For example, in the field of financial insurance business, if the audit report suggests starting the investigation procedure immediately, the relevant department will carry out detailed verification of the insurance complaint event as required, and if it suggests friendly communication with the user and giving certain compensation, the customer service personnel will negotiate with the user according to the insurance service order.

[0058] In the present application, after receiving the complaint, the complaint voice and complaint text of the user initiating the complaint are obtained, the target entity is obtained by performing entity extraction on the complaint text, the acoustic feature is obtained by performing feature extraction on the complaint voice, the acoustic feature is input into a preset risk analysis model, and the risk level is output, the corresponding audit decision rule is determined according to the target entity and the risk level, the text feature is obtained by performing feature extraction on the complaint text, the text feature and the acoustic feature are fused to obtain the fusion feature, the fusion feature is input into a preset prediction model, and the target prediction result is output, the audit report is generated according to the audit decision rule and the target prediction result, and the audit report is used to indicate the processing mode of the complaint. The present application can be applied to financial insurance and other business systems, and through the target entity of the complaint text and the acoustic feature of the complaint voice, the risk level is obtained according to the acoustic feature, the audit decision rule is determined according to the risk level and the target entity, the text feature is obtained according to the complaint text, the target prediction result is obtained according to the text feature and the acoustic feature, and the processing mode of the complaint is indicated according to the target prediction result. Thus, the complaint information of the user is comprehensively analyzed to accurately obtain the complaint demands of the user.

[0059] Referring to Figure 3 FIG. 1 is a flowchart of an automatic audit method for an insurance complaint provided by an embodiment of the present application. As shown in FIG. 1, the method comprises the following steps: Figure 3 When the complaint voice and complaint text of the user initiating the complaint are obtained in step S201, the following steps can be included: Step S301: Obtain the complaint voice of the user initiating the complaint.

[0060] Step S302: Perform text conversion on the complaint voice to obtain the complaint text.

[0061] Among them, in the actual scene of insurance complaints, the user initiates complaints in various ways, and the way of obtaining complaint voice is also different. The complaint methods include telephone complaint, online voice message, on-site complaint recording, etc. Among them, the telephone complaint can refer to when the user calls the customer service hotline of the insurance company to complain, the system will automatically record the call process, so as to obtain the complaint voice. This method can capture the user's tone, emotion and other information in real time when complaining. Online voice message can refer to after the user records the voice and uploads it on the platform, the insurance company can obtain the complaint voice. On-site complaint recording can refer to in some cases, the user may go to the offline service network of the insurance company to complain, and the staff will use the recording equipment to record the complaint process to keep the complete complaint information.

[0062] Among them, the text conversion is usually realized by means of speech recognition technology. The speech recognition system will extract and analyze the features of the input voice signal, and then match with the pre-trained voice model, convert the sound information in the voice into the corresponding text sequence, and convert the voice into text, which can be more convenient for text analysis, information extraction and other operations.

[0063] For example, using natural language processing technology to extract keywords and analyze semantics of the complaint text, so as to quickly locate the core problem of the complaint.

[0064] The embodiments of the present application provide basic data and information support for subsequent risk analysis, decision making, etc. by acquiring and processing complaint information.

[0065] Referring to Figure 4 is a flowchart of an automatic audit method for insurance complaints provided by the fourth embodiment of the present application. As Figure 4 shown, the step S202 of inputting the acoustic features into the preset risk analysis model and outputting the risk level can include the following steps: Step S401, using a preset risk analysis model, evaluating the speech rate feature in the acoustic feature to obtain a speech rate evaluation result.

[0066] Step S402, using a preset risk analysis model, evaluating the emotion feature in the acoustic feature to obtain an emotion evaluation result.

[0067] Step S403, weighting and fusing the speech rate evaluation result and the emotion evaluation result to obtain a comprehensive evaluation result, and determining the risk level according to the comprehensive total price result.

[0068] The pre-set risk analysis model has a set of pre-set evaluation criteria for the speech speed feature. This criterion is summarized based on a large amount of historical complaint data and the corresponding processing results. The model compares the input speech speed feature with these criteria for comparative analysis. If the speech speed of the complaint voice is in the fast interval, the evaluation result can be "high risk tendency", because fast speech speed often implies that the user is agitated and anxious, and the complaint problem may be more serious; if it is in the normal speed interval, the evaluation result can be "medium risk tendency"; slow speed can represent "low risk tendency", but this may also vary depending on the specific circumstances, such as the user may be calmly stating a complex problem.

[0069] The risk analysis model evaluates the user's emotional state reflected by the emotional feature. The emotional feature is presented in the form of probability of different emotional categories, such as anger, dissatisfaction, anxiety, calmness, satisfaction, etc. The model judges the severity of emotion according to these probability distributions and pre-set rules. If the emotional feature shows that the user is in a state of high probability of anger, the risk analysis model will give an evaluation result of "high risk", because an angry user is often very dissatisfied with the complaint problem and will take further action (such as complaining to the regulatory department, exposing on social media, etc.).

[0070] The speech speed evaluation result and the emotional evaluation result are respectively given different weights, and the size of the weight depends on the importance of the two factors in the actual complaint processing. The emotional factor may have a greater impact on the risk, so the weight of the emotional evaluation result may be relatively high. The model will map the comprehensive evaluation result to different risk levels according to the numerical range of the comprehensive evaluation result.

[0071] The present application evaluates the speech speed and emotional features in the acoustic features respectively and weights the fusion, which can more comprehensively and accurately evaluate the risk level of the complaint, and helps the insurance enterprise to reasonably allocate resources, prioritize high-risk complaints, and improve the efficiency and quality of complaint processing.

[0072] Referring to Figure 5 is a flowchart of an automatic audit method for insurance complaints provided by an embodiment of the present application. As Figure 5 indicated, the pre-set prediction model includes a semantic analysis model and a violation prediction model; The step of inputting the fusion feature into the pre-set prediction model and outputting the target prediction result can include the following steps: Step S501, using the semantic analysis model, performing semantic analysis on the fusion feature to obtain a semantic analysis result.

[0073] Step S502, using the violation prediction model, predicting the violation rate in the semantic analysis result to obtain a target prediction result.

[0074] The fusion feature is a comprehensive feature obtained by integrating multiple related features, and contains data such as text content, voice features, and context information. These features can more comprehensively and accurately reflect the characteristics and state of the analyzed object after fusion processing. The violation prediction model is trained based on a machine learning or deep learning algorithm, which learns a large amount of historical data including semantic features and corresponding violation conditions, thereby establishing a mapping relationship between the violation rate and the semantic features. When the semantic analysis result is input, the model can predict the violation rate according to the learned pattern and rule.

[0075] In this embodiment, by sequentially inputting the fusion feature into the semantic analysis model and the violation prediction model, semantic analysis is performed first, and then the violation rate is predicted. The whole process can effectively extract valuable content from complex information and accurately assess potential violation risks, thereby providing strong support for decision-making.

[0076] The insurance complaint automatic auditing method corresponding to the above embodiment, Figure 6 The structure block diagram of the insurance complaint automatic auditing device provided by the sixth embodiment of the present application is shown, and the above-mentioned insurance complaint automatic auditing device can be applied to Figure 1 The server. For the sake of illustration, only the parts related to the present application are shown.

[0077] Referring to Figure 6 The insurance complaint automatic auditing device comprises: A feature extraction module 61 is configured to, after receiving a complaint, acquire complaint voice and complaint text of a user initiating the complaint, perform entity extraction on the complaint text to obtain a target entity, and perform feature extraction on the complaint voice to obtain acoustic features. A rule determination module 62 is configured to input the acoustic features into a preset risk analysis model to output a risk level, and determine a corresponding auditing decision rule according to the target entity and the risk level. A target prediction module 63 is configured to perform feature extraction on the complaint text to obtain text features, fuse the text features with the acoustic features to obtain fusion features, and input the fusion features into a preset prediction model to output a target prediction result. An auditing report module 64 is configured to generate an auditing report according to the auditing decision rule and the target prediction result, wherein the auditing report is used to indicate a processing mode of the complaint.

[0078] Optionally, the feature extraction module 61 comprises: A noise extraction unit is configured to perform noise extraction on the complaint text to determine unstructured noise. a noise removing unit configured to remove the unstructured noise from the complaint text to obtain a de-noised complaint text; an entity extraction unit configured to perform entity extraction on the de-noised complaint text to obtain a target entity.

[0079] Optionally, the feature extraction module 61 further comprises: a complaint voice obtaining unit configured to obtain a complaint voice of a user who initiates the complaint; a text conversion unit configured to perform text conversion on the complaint voice to obtain a complaint text.

[0080] Optionally, the feature extraction module 61 further comprises: a speech rate feature obtaining unit configured to input the complaint voice into a speech analysis model to obtain a speech rate feature; an emotion feature obtaining unit configured to input the complaint voice into an emotion prediction model to obtain an emotion feature; an acoustic feature obtaining unit configured to fuse the speech rate feature and the emotion feature to obtain an acoustic feature.

[0081] Optionally, the rule determination module 62 comprises: a speech rate evaluation unit configured to use a preset risk analysis model to perform speech rate evaluation on the speech rate feature in the acoustic feature to obtain a speech rate evaluation result; an emotion evaluation unit configured to use a preset risk analysis model to perform emotion evaluation on the emotion feature in the acoustic feature to obtain an emotion evaluation result; a comprehensive evaluation unit configured to perform weighted fusion on the speech rate evaluation result and the emotion evaluation result to obtain a comprehensive evaluation result, and determine a risk level according to the comprehensive evaluation result.

[0082] Optionally, the rule determination module 62 further comprises: a rule clause determination unit configured to determine a rule clause corresponding to the target entity according to the target entity in combination with a rule engine; a screening unit configured to screen the rule clause according to the risk level to obtain a screened rule, and determine the screened rule as an audit decision rule.

[0083] Optionally, the target prediction module 63 comprises: a semantic analysis unit configured to use the semantic analysis model to perform semantic analysis on the fused feature to obtain a semantic analysis result; a target prediction unit configured to use the violation prediction model to predict a violation rate in the semantic analysis result to obtain a target prediction result.

[0084] It should be noted that the information interaction, execution process and the like among the above modules, units and sub-units are based on the same concept as the method embodiments of the present application, and the specific functions and technical effects brought by the same can be referred to the method embodiments part, and will not be described here.

[0085] Figure 7 A structural schematic diagram of a computer device is provided for the seventh embodiment of the present application. As shown in the figure, Figure 7 the computer device of this embodiment includes at least one processor (only one is shown in the figure), Figure 7 memory and a computer program stored in the memory and executable on the at least one processor, and the processor implements the steps in the automatic audit method of insurance complaints or the automatic audit method of insurance complaints embodiment when executing the computer program.

[0086] The computer device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand, Figure 7 that the computer device is only an example and does not constitute a limitation on the computer device, and the computer device can include more or fewer components than those shown, or combine certain components, or different components, for example, it can also include a network interface, a display screen and an input device, etc.

[0087] The processor can be a CPU, and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), ready programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0088] The memory includes a readable storage medium, an internal memory, etc., wherein the internal memory can be a memory of the computer device, and the internal memory provides an environment for running of the operating system and the computer-readable instructions in the readable storage medium. The readable storage medium can be a hard disk of the computer device, and in other embodiments, can also be an external storage device of the computer device, for example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory can include both an internal storage unit of the computer device and an external storage device. The memory is used to store an operating system, an application program, a BootLoader, data, and other programs, such as program codes of computer programs, etc. The memory can also be used to temporarily store data that has been output or will be output.

[0089] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module 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. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of software functional unit. In addition, the specific names of each functional unit and module are only for easy distinction, and do not limit the protection scope of the present application. The specific working process of the units and modules in the above device can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here. If the integrated unit is realized in the form of software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, all or part of the processes in the above embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a computer readable storage medium. When the processor executes the computer program, the steps of the above method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable file or some intermediate form. The computer readable medium at least includes any entity or device that can carry computer program code, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, computer readable medium cannot be electrical carrier signal and telecommunication signal.

[0090] The above embodiment methods can also be completed by a computer program product, which can be run on a computer device to make the computer device execute the steps of the above method embodiments.

[0091] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0092] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0093] In the embodiments provided in the present application, it should be understood that the disclosed apparatus / computer device and method can be implemented in other ways. For example, the apparatus / computer device embodiments described above are merely schematic, for example, the division of the modules or units is only a logical function division, and there can be another division in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0094] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected to achieve the purpose of the embodiments according to actual needs.

[0095] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make 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 automated audit of insurance complaints, characterized by, The method comprises the following steps: After receiving a complaint, obtaining complaint voice and complaint text of a user who initiates the complaint, performing entity extraction on the complaint text to obtain a target entity, and performing feature extraction on the complaint voice to obtain acoustic features; inputting the acoustic features into a preset risk analysis model to output a risk level, determining a corresponding audit decision rule according to the target entity and the risk level; performing feature extraction on the complaint text to obtain text features, fusing the text features and the acoustic features to obtain fused features, inputting the fused features into a preset prediction model to output a target prediction result; generating an audit report according to the audit decision rule and the target prediction result, wherein the audit report is used to indicate a processing mode of the complaint.

2. The method of automated audit of insurance complaints as claimed in claim 1 wherein, The entity extraction on the complaint text to obtain a target entity comprises the following steps: performing noise extraction on the complaint text to determine unstructured noise; removing the unstructured noise from the complaint text to obtain a denoised complaint text; performing entity extraction on the denoised complaint text to obtain a target entity.

3. The method for automated audit of insurance complaints as claimed in claim 1 wherein, The obtaining of the complaint voice and the complaint text of the user who initiates the complaint comprises the following steps: obtaining complaint voice of a user who initiates the complaint; performing text conversion on the complaint voice to obtain complaint text.

4. The method for automated audit of insurance complaints as claimed in claim 1 wherein, The feature extraction on the complaint voice to obtain acoustic features comprises the following steps: inputting the complaint voice into a voice analysis model to obtain a speech rate feature; inputting the complaint voice into an emotion prediction model to obtain an emotion feature; fusing the speech rate feature and the emotion feature to obtain acoustic features.

5. The method for automated audit of insurance complaints as claimed in claim 1 wherein, The inputting of the acoustic features into a preset risk analysis model to output a risk level comprises the following steps: using a preset risk analysis model to perform speech rate evaluation on the speech rate feature in the acoustic features to obtain a speech rate evaluation result; using a preset risk analysis model to perform emotion evaluation on the emotion feature in the acoustic features to obtain an emotion evaluation result; performing weighted fusion on the speech rate evaluation result and the emotion evaluation result to obtain a comprehensive evaluation result, and determining a risk level according to the comprehensive evaluation result.

6. The method for automated audit of insurance complaints as claimed in claim 1 wherein, The determination of a corresponding audit decision rule according to the target entity and the risk level comprises the following steps: determining a rule clause corresponding to the target entity in combination with a rule engine according to the target entity; performing screening on the rule clause according to the risk level to obtain a screened rule, and determining the screened rule as an audit decision rule.

7. The method of automated audit of insurance complaints as claimed in any one of claims 1 to 6, wherein, The preset prediction model comprises a semantic analysis model and a violation prediction model. The inputting of the fused features into a preset prediction model to output a target prediction result comprises the following steps: using the semantic analysis model to perform semantic analysis on the fused features to obtain a semantic analysis result; using the violation prediction model to predict a violation rate in the semantic analysis result to obtain a target prediction result.

8. An automated auditing apparatus for insurance complaints, characterized by, The method comprises the following steps: a feature extraction module is configured to, after receiving a complaint, obtain complaint voice and complaint text of a user who initiates the complaint, perform entity extraction on the complaint text to obtain a target entity, and perform feature extraction on the complaint voice to obtain acoustic features; A rule determining module is configured to input the acoustic features into a preset risk analysis model, output a risk level, and determine a corresponding audit decision rule according to the target entity and the risk level; A target prediction module is configured to perform feature extraction on the complaint text to obtain text features, fuse the text features with the acoustic features to obtain fused features, input the fused features into a preset prediction model, and output a target prediction result. An audit report module is configured to generate an audit report according to the audit decision rule and the target prediction result, and the audit report is used to indicate a processing mode of the complaint.

9. A computer device, comprising: The computer device comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, and the processor implements the automatic audit method of the insurance complaint according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to implement the automatic audit method of the insurance complaint according to any one of claims 1 to 7.

Images