Insurance contract auditing method and device and electronic equipment

Abstract

The invention discloses an insurance contract auditing method and device and electronic equipment, and belongs to the technical field of artificial intelligence. The method comprises the following steps: inputting a part of an insurance contract to be audited into a first insurance contract auditing model and a second insurance contract auditing model to obtain a first auditing result output by the first insurance contract auditing model and a second auditing result output by the second insurance contract auditing model; determining the accuracy of the first auditing result and the second auditing result; and under the condition that the accuracy of the first auditing result is higher than that of the second auditing result, auditing the to-be-audited insurance contract through the first insurance contract auditing model. Through the method, the problem that the accuracy of auditing the insurance contract is reduced when the version of the insurance contract auditing model is replaced can be at least solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an insurance contract review method, device and electronic equipment. Background Art

[0002] Since there is a large demand for reviewing insurance contracts, the review efficiency is low if it is done manually. Therefore, an artificial intelligence model is used to review insurance contracts in relevant technologies. By using insurance contract sample data, the insurance contract review model is trained, and the trained insurance contract review model is used to review the insurance contracts, thereby improving the review efficiency of insurance contracts.

[0003] However, since the sample data of insurance contracts used for training is limited, in actual use, the insurance contracts input into the insurance contract review model for review may be completely different from the type of insurance contract sample data, and the errors in the reviewed insurance contracts may also be errors not included in the sample data of insurance contracts, which may lead to the low accuracy of the insurance contract review model in reviewing the insurance contracts. Therefore, when upgrading and replacing the insurance contract review model, the accuracy of the new version of the insurance contract review model may be lower than that of the old version of the insurance contract review model because it has not been evaluated in actual application. If the new version of the insurance contract review model is used directly, it may lead to the problem of reduced accuracy in reviewing the insurance contract. Summary of the invention

[0004] The embodiments of the present application provide a method, device and electronic device for reviewing an insurance contract, which can at least solve the problem of reduced accuracy in reviewing an insurance contract when a version of an insurance contract review model is replaced.

[0005] In order to solve the above technical problems, this application is implemented as follows: In a first aspect, an embodiment of the present application provides a method for reviewing an insurance contract, the method comprising: inputting part of the insurance contracts to be reviewed into a first insurance contract review model and a second insurance contract review model, obtaining a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model, wherein the first insurance contract review model and the second insurance contract review model are used to review the input insurance contract and obtain the review result of the insurance contract, and the second insurance contract review model is the previous version of the first insurance contract review model; determining the accuracy of the first review result and the second review result; and when the accuracy of the first review result is higher than the accuracy of the second review result, reviewing the insurance contract to be reviewed through the first insurance contract review model.

[0006] In a second aspect, an embodiment of the present application provides an insurance contract review device, the device comprising: an acquisition module, used to input part of the insurance contracts to be reviewed into a first insurance contract review model and a second insurance contract review model, to obtain a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model, wherein the first insurance contract review model and the second insurance contract review model are used to review the input insurance contract and obtain the review result of the insurance contract, and the second insurance contract review model is the previous version of the first insurance contract review model; a determination module, used to determine the accuracy of the first review result and the second review result; and a review module, used to review the insurance contract to be reviewed through the first insurance contract review model when the accuracy of the first review result is higher than the accuracy of the second review result.

[0007] In a third aspect, an embodiment of the present application provides an electronic device, which includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the method described in the first aspect are implemented.

[0008] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0009] In a fifth aspect, an embodiment of the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps of the method described in the first aspect.

[0010] In a sixth aspect, an embodiment of the present application provides a computer program product, which is stored in a storage medium and is executed by at least one processor to implement the steps of the method described in the first aspect.

[0011] The technical solution provided by this application may have the following beneficial effects: In an embodiment of the present application, part of the insurance contracts to be reviewed can be input into the first insurance contract review model and the second insurance contract review model to obtain a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model, wherein the first insurance contract review model and the second insurance contract review model are used to review the input insurance contracts and obtain the review results of the insurance contracts, and the second insurance contract review model is the previous version of the first insurance contract review model; determine the accuracy of the first review result and the second review result; and when the accuracy of the first review result is higher than the accuracy of the second review result, review the insurance contract to be reviewed through the first insurance contract review model. In this way, it can avoid the problem of reducing the accuracy of the review of the insurance contract by directly using the first insurance contract review model to replace the second insurance contract review to review the insurance contract to be reviewed.

[0012] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0014] Figure 1 A flowchart of an insurance contract review method provided by an exemplary embodiment of the present application is shown; Figure 2 A flow chart showing a model upgrade method for an insurance contract review model provided by an exemplary embodiment of the present application is shown; Figure 3 A schematic diagram of the structure of an insurance contract review device provided by an exemplary embodiment of the present application is shown; Figure 4 A structural block diagram of an electronic device provided by an exemplary embodiment of the present application is shown; Figure 5 A structural block diagram of another electronic device provided by an exemplary embodiment of the present application is shown. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0016] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0017] Text auditing is generally a check of the text, including checking the spelling correctness, grammatical correctness, compliance correctness and other contents in the text. The insurance contract audit model in the embodiment of the present application is for insurance contracts, especially the insurance clauses therein, to perform spelling, grammar and compliance checks. The characteristics of insurance contracts and insurance clauses are: first, the audit volume is very large, including the entire content of the insurance contract; second, it is highly professional and requires a very high audit accuracy rate; third, the model is updated frequently, and it is necessary to keep up with national policy and regulatory requirements, company product adjustments, etc., and timely and accurately iterate and optimize the online audit model and update it with stability and confidence.

[0018] In the relevant technology, it can be divided into two stages. First, in the training stage, in order to test the model effect, it is generally necessary to set the audit model indicators, such as accuracy, and construct a test data set to quickly and automatically verify the performance of the trained model on the test data set. During the model training iteration, the indicators are strived to be better and better; second, after the model is launched, the online accuracy indicators are monitored to ensure the online effect. Generally, the evaluation of the model training stage and the online evaluation are performed separately, lacking feedback and linkage between the two, and the online monitoring method is single, lacking comprehensiveness, pertinence and robustness.

[0019] The shortcomings of the relevant methods for evaluating insurance contract audit models are: first, there is a lack of effective linkage and feedback between the audit model training evaluation and the online evaluation, which reduces the online effect of the model; second, one solution for online evaluation of the audit model is to observe the comparison between the old and new strategies through a small traffic comparison experiment (ab experiment), which will result in a lack of comprehensiveness. Comprehensiveness means that the distribution of online input data is often related to the time period. Only observing small periods of time will result in incomplete online evaluation; third, another solution for online evaluation is to fully launch it in scenarios such as insurance that have high accuracy requirements. It requires long-term online evaluation to ensure online effects. This solution will invest a lot of evaluation resources and reduce the efficiency of model iteration, optimization and upgrade.

[0020] Based on the above problems, an embodiment of the present application provides an improved solution to solve at least one of the above problems.

[0021] The following is a detailed description of the insurance contract review scheme provided by the embodiment of the present application through specific embodiments and their application scenarios in conjunction with the accompanying drawings.

[0022] Figure 1 The flowchart of an insurance contract review method provided by an exemplary embodiment of the present application is shown. The method 100 can be executed by an electronic device. In other words, the method can be executed by software or hardware installed in the electronic device. Figure 1 As shown, the method mainly includes the following steps: S101: Input some insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model.

[0023] Among them, the first insurance contract review model and the second insurance contract review model are used to review the input insurance contract and obtain the review result of the insurance contract, and the second insurance contract review model is the previous version of the first insurance contract review model.

[0024] In an embodiment of the present application, part of the insurance contracts to be reviewed can be input into the first insurance contract review model and the second insurance contract review model to obtain a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model. Before the new insurance contract review model (i.e., the first insurance contract review model) is officially used, part of the insurance contracts to be reviewed can be input into the first insurance contract review model and the second insurance contract review model to obtain the corresponding output first review result and second review result, so as to facilitate the evaluation of the first insurance contract review model and determine whether to adopt the first insurance contract review model to review the insurance contracts to be reviewed.

[0025] In an optional implementation, before inputting the part of the insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain the first review result output by the first insurance contract review model and the second review result output by the second insurance contract review model, the method may further include the following steps: Step 1, iteratively training the first insurance contract review model to be trained using the first insurance contract sample data until a preset training completion condition is met, thereby obtaining the first insurance contract review model that has been trained, wherein the first insurance contract sample data includes the first insurance contract sample and first annotated data of the first insurance contract sample; In an embodiment of the present application, first insurance contract sample data required for model training can be collected and produced in a targeted manner according to demand conditions, online effects, etc., where the first insurance contract sample data includes a first insurance contract sample and first labeled data of the first insurance contract sample; the first insurance contract sample data is then used to iteratively train the first insurance contract review model to be trained until preset training completion conditions are met, thereby obtaining a trained first insurance contract review model. Step 2: inputting the second insurance contract sample in the second insurance contract sample data into the first insurance contract review model to obtain a review result of the second insurance contract sample; In an embodiment of the present application, after completing the training of the first insurance contract review model, the first insurance contract review model may be tested using second insurance contract sample data (these sample data may be referred to as sample data in a test set), wherein the second insurance contract sample data includes the second insurance contract sample and the second annotation data corresponding to the second insurance contract sample. In actual applications, the ratio of the first insurance contract sample data to the second insurance contract sample data may be 3:7 or 2:8, which is not specifically limited in the embodiment of the present application. The second insurance contract sample may be input into the first insurance contract review model to obtain the review result of the second insurance contract sample, and the test result of the first insurance contract review model may be determined based on the review result.

[0026] Step 3: Based on the difference between the audit result of the second insurance contract sample and the second labeled data, obtain the technical indicators of the first insurance contract audit model, wherein the second labeled data is the labeled data of the second insurance contract sample in the second insurance contract sample data, and the technical indicators are used to indicate the accuracy and recall rate of the first insurance contract audit model.

[0027] In an embodiment of the present application, the technical indicators of the first insurance contract audit model can be obtained based on the difference between the audit results of the second insurance contract sample and the second labeled data. In practical applications, the technical indicators can be the accuracy and recall of the first insurance contract audit model. The recall rate is usually used to evaluate the performance of the model, where the accuracy rate refers to how many items in the audit results retrieved by the first insurance contract audit model are correct, and the recall rate refers to how many of all correct items are retrieved. The recall rate and accuracy rate are usually used to evaluate the quality of the results. When the accuracy rate is the same, the higher the recall rate, the better the model performance. When the recall rate is the same, the higher the accuracy rate, the better the model performance.

[0028] Step 4: When the technical indicators of the first insurance contract review model are greater than or equal to the technical indicators of the second insurance contract review model and the technical indicators of the first insurance contract review model meet the expected requirements, obtain the first insurance contract review model.

[0029] Among them, the technical indicators of the second insurance contract review model can be based on the same test data as the first insurance contract review model (i.e., the second insurance contract sample data mentioned above) or different test data, and obtained in the same manner as the first insurance contract review model mentioned above, or can also be obtained when testing the second insurance contract review model before it goes online, which is not limited in the specific embodiments of the present application.

[0030] In an embodiment of the present application, the technical indicator of the first insurance contract review model may be recorded as Ratio_exp, and the technical indicator of the second insurance contract review model may be recorded as Ratio_base. Ratio_exp greater than or equal to Ratio_base may mean that the accuracy of the first insurance contract review model is greater than or equal to the accuracy of the second insurance contract review model and the recall rate of the first insurance contract review model is greater than or equal to the recall rate of the second insurance contract review model, that is, the technical indicator may include accuracy and / or recall rate. When the technical indicator of the first insurance contract review model is greater than or equal to the technical indicator of the second insurance contract review model and the technical indicator of the first insurance contract review model meets the expected requirements, it can be determined that the first insurance contract review model is superior to the second insurance contract review model in terms of technical indicators and the performance meets certain requirements.

[0031] In an optional implementation, the first audit result and the second audit result include: text errors and audit opinions corresponding to the text errors.

[0032] In an embodiment of the present application, the first audit result and the second audit result may include: text errors and audit opinions corresponding to the text errors. The text errors may be one or more of the three errors of spelling, grammar, and compliance. The audit opinions corresponding to the text errors may be correct spelling, correct grammar, correct regulations, or the type of error corresponding to the text errors. This embodiment of the present application does not make specific limitations.

[0033] In an optional implementation, the step of inputting part of the insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain the first review result output by the first insurance contract review model and the second review result output by the second insurance contract review model may include the following steps: Step 1: Input a first percentage of insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain a first initial review result output by the first insurance contract review model and a second initial review result output by the second insurance contract review model.

[0034] The first proportion may be a preset value, for example, 10%. That is, 10% of the insurance contracts to be reviewed may be input into the first insurance contract review model and the second insurance contract review model. For example, by controlling the access volume, 10% of the received insurance contract review requests may be transferred to the first insurance contract review model and the second insurance contract review model, and the insurance contracts requested for review by the 10% insurance contract review requests may be input into the first insurance contract review model and the second insurance contract review model.

[0035] In an embodiment of the present application, the insurance contracts of the first proportion can be first input into the first insurance contract review model and the second insurance contract review model to obtain the first initial review result output by the first insurance contract review model and the second initial review result output by the second insurance contract review model. In actual applications, if the first insurance contract review model detects 50 problems (which may include one or more of the three errors of spelling, grammar, and compliance) for the insurance contracts of the first proportion, of which 47 have problems with the text and give correct review opinions, 1 gives an incorrect opinion, and 2 have no problems in the text, the second insurance contract review model detects 51 problems, 44 give correct review opinions, 4 give incorrect opinions, and 3 have no problems in the text, it can be considered that the first insurance contract review model has a win in both accuracy and recall, and the accuracy of the first initial review result is higher than the accuracy of the second initial review result.

[0036] Step 2. When the accuracy of the first initial audit result is higher than the accuracy of the second initial audit result, the second percentage of insurance contracts in the insurance contracts to be audited are input into the first insurance contract audit model and the second insurance contract audit model to obtain the first audit result output by the first insurance contract audit model and the second audit result output by the second insurance contract audit model.

[0037] The first proportion and the second proportion are rational numbers greater than 0 and less than 1, and the first proportion is less than the second proportion. For example, the first proportion may be 10%, and the second proportion may be 50%.

[0038] In the embodiment of the present application, when the accuracy of the first initial audit result is higher than the accuracy of the second initial audit result, the second proportion of insurance contracts can be input into the first insurance contract audit model and the second insurance contract audit model to obtain the first audit result output by the first insurance contract audit model and the second audit result output by the second insurance contract audit model. This can not only save time, but also avoid wasting resources and improve efficiency.

[0039] S102: Determine the accuracy of the first audit result and the second audit result.

[0040] In the embodiment of the present application, the first audit result and the second audit result can be compared to obtain the accuracy of the first audit result and the second audit result. In practical applications, the accuracy can be expressed by accuracy or recall, or by comparing the first insurance contract audit model and the second insurance contract audit model for the same insurance contract to determine which one is more accurate. The embodiment of the present application does not specifically limit this.

[0041] In an optional implementation, after determining the accuracy of the first audit result and the second audit result, the method further includes: When the accuracy of the first audit result is lower than or equal to the accuracy of the second audit result, the first insurance contract audit model is retrained using third insurance contract sample data.

[0042] In an embodiment of the present application, when the accuracy of the first audit result is lower than or equal to the accuracy of the second audit result, it means that the accuracy of the first insurance contract audit model is not high, the audit effect is unsatisfactory, and it cannot be used in practice. The first insurance contract audit model can be retrained using the third insurance contract sample data. In practical applications, the third insurance contract sample data may include a third insurance contract sample and third annotation data corresponding to the third insurance contract sample. The third insurance contract sample data may be different from the first insurance contract sample data and is a reacquired training sample. By retraining the first insurance contract audit model with the third insurance contract sample data, a new first insurance contract audit model can be obtained, which is convenient for upgrading the insurance contract audit model. After retraining, the above steps S101 and S102 can be re-executed to compare the advantages and disadvantages of the retrained first insurance contract audit model and the second insurance contract audit model.

[0043] S103: When the accuracy of the first audit result is higher than the accuracy of the second audit result, the insurance contract to be audited is audited by using the first insurance contract audit model.

[0044] In the embodiment of the present application, the accuracy of the first audit result is higher than the accuracy of the second audit result, which can indicate that the first insurance contract audit model has a higher accuracy in auditing the insurance contract, and the first insurance contract audit model can be used to audit the insurance contract to be audited. This can avoid the problem of directly using the first insurance contract audit model to audit the insurance contract to be audited, resulting in a decrease in the accuracy of the audit of the insurance contract. The accuracy of the first audit result is higher than the accuracy of the second audit result, and the use of the first insurance contract audit model to audit the insurance contract to be audited can improve the accuracy of the audit of the insurance contract.

[0045] In an optional implementation, after the insurance contract to be reviewed is reviewed by using the first insurance contract review model, the method may further include the following steps: Step 1, obtaining the first business indicator corresponding to the review of the insurance contract to be reviewed through the first insurance contract review model within a preset time period, wherein the first business indicator includes at least one of the following: the drafting time of the insurance contract, the number of round trips of the insurance contract.

[0046] In an embodiment of the present application, at least one of the drafting time of the insurance contract and the number of round trips of the insurance contract corresponding to the review of the insurance contract to be reviewed by the first insurance contract review model can be obtained within a preset time, for example, 1 week, after the first insurance contract review model is applied. Among them, after the user drafts an insurance contract, the first insurance contract review model is used for review. If there is an error in the review result, it is returned to the user for modification. After the modification is completed, it is submitted to the first insurance contract review model for review until the first insurance contract review model is reviewed and approved, and then submitted to the next review link. In the embodiment of the present application, the drafting time of the insurance contract refers to the time from the user drafting a contract or clause to the time when the contract or clause is submitted to the next approval link after being reviewed and approved, and the number of round trips of the insurance contract refers to the number of times the user submits the same insurance contract to the first insurance contract review model for review, and the first insurance contract review model has errors in reviewing the insurance contract and returns the insurance contract to the user for modification.

[0047] In the embodiment of the present application, optionally, the preset duration can be calculated according to the following formula:

[0048] Among them, Week_num is the number of weeks to be monitored, and α is a hyperparameter that controls the proportion of the final number of weeks. , diff_ratio may be the ratio of the first audit result output by the first insurance contract audit model to the second audit result output by the second insurance contract audit model. In practical applications, other methods may be used to calculate the preset time length, which is not specifically limited in the embodiments of the present application.

[0049] Step 2: Obtain the business effect corresponding to the first insurance contract review model based on the first business indicator.

[0050] In the embodiment of the present application, the business effect corresponding to the first insurance contract review model can be obtained according to the first business indicator. For example, a corresponding threshold can be set and directly determined by comparison, or other methods can be used for determination, which is not specifically limited in the embodiment of the present application. Through the business effect corresponding to the first insurance contract review model, the first insurance contract review model can be more comprehensively evaluated, and can also be used as the focus of the next model upgrade or replacement.

[0051] In an optional implementation, obtaining the business effect corresponding to the first insurance contract review model according to the first business indicator may include the following steps: Step 1, based on historical data of insurance contracts reviewed by the second insurance contract review model, obtaining an average value of a second business indicator corresponding to the second insurance contract review model within the preset time period; In an embodiment of the present application, the average value of the second business indicator corresponding to the second insurance contract review model within a preset time period can be obtained based on the historical data of the insurance contract reviewed by the second insurance contract review model. For example, if the preset time period is 1 week, the average value of the second business indicator of the second insurance review model averaged to each week can be obtained based on the historical data of the insurance contract reviewed by the second insurance contract review model. For example, assuming that the total time for the second insurance contract review model to review the insurance contract is 5 weeks, during these 5 weeks, the value of the second business indicator corresponding to the second insurance contract review model is 60, then the average value of the second business indicator within 1 week is 12.

[0052] Step 2: Based on the difference between the first business indicator and the average value, obtain the business effect corresponding to the first insurance contract review model.

[0053] In an embodiment of the present application, the business effect corresponding to the first insurance contract review model can be obtained based on the difference between the first business indicator and the average value.

[0054] In actual applications, if the first business indicator is better than the average, it can be considered that the business effect corresponding to the first insurance contract review model is better than the second insurance contract review model. If the first business indicator is not better than the average, the reason can be further determined. For example, the number of round trips of the first business indicator is significantly higher than the average. It may be due to user modification problems, or it may be due to changes in certain requirements of the insurance contract. For example, some regulations have changed in a short period of time, and the user has not modified them. The audit of the insurance contract using the first insurance contract review model has not found it, but it has been found in the next audit link. It may be necessary to consider retraining the first insurance contract review model.

[0055] In an embodiment of the present application, part of the insurance contracts to be reviewed can be input into the first insurance contract review model and the second insurance contract review model to obtain a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model, wherein the first insurance contract review model and the second insurance contract review model are used to review the input insurance contracts and obtain the review results of the insurance contracts, and the second insurance contract review model is the previous version of the first insurance contract review model; determine the accuracy of the first review result and the second review result; and when the accuracy of the first review result is higher than the accuracy of the second review result, review the insurance contract to be reviewed through the first insurance contract review model. In this way, it can avoid the problem of reducing the accuracy of the review of the insurance contract by directly using the first insurance contract review model to replace the second insurance contract review to review the insurance contract to be reviewed.

[0056] Figure 2 A flow chart of a model upgrade method for an insurance contract review model provided by an exemplary embodiment of the present application is shown, and the method includes: S201: Collect training samples.

[0057] In this step, you can collect and produce training samples required for model training in a targeted manner based on demand conditions, online effects, etc.

[0058] S202: Training insurance contract review model.

[0059] In this step, the insurance contract review model is iteratively trained using the collected data to obtain a trained model.

[0060] S203: Testing set detection.

[0061] In this step, the offline annotated test set can be used to automatically detect and calculate the technical indicators of the newly trained model (the first insurance contract review model) on the test set. If the detection result of the technical indicator meets the conditions, S204 is performed; if the detection result does not meet the conditions, it returns to S201 and retrains. The technical indicator generally refers to the accuracy, recall rate and other technical indicators of the model. The data of the test set includes input and manually annotated output. The input data is passed through the latest model to obtain the model's predicted value. The difference between this predicted value and the manual standard calculation can be used to obtain the technical indicator.

[0062] S204: Online small flow data comparison experiment; In this step, part of the online traffic is passed through the new model and the old model (the second insurance contract review model) to obtain the output results of the new model and the output results of the old model, and then the two output results of each line of data are compared to determine whether the review accuracy of the new model is higher or the review accuracy of the old model is higher. If the review accuracy of the new model is higher than the review accuracy of the old model, proceed to S205; if the comparison experiment results do not meet the conditions, return to S201 and retrain the samples.

[0063] S205: Regular regression of all online data; In this step, the first insurance contract review model can be fully launched, that is, the new model is set on the monitoring platform to completely replace the old model, and the old model has no traffic. At this time, the new model is regressed on a weekly basis to obtain the business indicators corresponding to the new model within each week. The business indicators can be the drafting time of the insurance contract and the number of round trips of the insurance contract. Then, the business indicators are compared with the average value of the business indicators corresponding to the old model within a week. If the business indicators of the new model are better than the old model within each week, S206 can be entered. If the result does not meet the requirements, return to S201 and retrain the sample.

[0064] S206: Complete this model upgrade.

[0065] After the full online regression analysis has reached the specified time, it can be considered that the First Insurance Contract Review Model has the ability to outperform the previous generation model, and the maximum robustness test is carried out under the premise of minimizing resource consumption, which means that this evaluation is over.

[0066] Figure 3 The schematic diagram of the structure of the insurance contract review device provided by an exemplary embodiment of the present application is shown. The insurance contract review device can be implemented as follows: Figure 1 In the illustrated embodiment, all or part of the contents, the review device of the insurance contract includes: an acquisition module 301, a determination module 302 and a review module 303.

[0067] In an embodiment of the present application, an acquisition module 301 is used to input part of the insurance contracts to be reviewed into a first insurance contract review model and a second insurance contract review model to obtain a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model, wherein the first insurance contract review model and the second insurance contract review model are used to review the input insurance contracts and obtain the review results of the insurance contracts, and the second insurance contract review model is the previous version of the first insurance contract review model; a determination module 302 is used to determine the accuracy of the first review result and the second review result; and an audit module 303 is used to audit the insurance contract to be reviewed through the first insurance contract review model when the accuracy of the first review result is higher than the accuracy of the second review result.

[0068] In an optional implementation, when the acquisition module 301 is used to input part of the insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain the first review result output by the first insurance contract review model and the second review result output by the second insurance contract review model, it is specifically used to: Inputting a first percentage of insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain a first initial review result output by the first insurance contract review model and a second initial review result output by the second insurance contract review model; When the accuracy of the first initial audit result is higher than the accuracy of the second initial audit result, the insurance contracts of the second percentage in the insurance contracts to be audited are input into the first insurance contract audit model and the second insurance contract audit model to obtain the first audit result output by the first insurance contract audit model and the second audit result output by the second insurance contract audit model; The first proportion and the second proportion are rational numbers greater than 0 and less than 1, and the first proportion is smaller than the second proportion.

[0069] In an optional implementation, the device further includes: a training module, configured to iteratively train a first insurance contract review model to be trained using first insurance contract sample data until a preset training completion condition is met, thereby obtaining the first insurance contract review model that has been trained, wherein the first insurance contract sample data includes the first insurance contract sample and first annotated data of the first insurance contract sample; In an optional implementation, the device further includes: a testing module, configured to input the second insurance contract sample in the second insurance contract sample data into the first insurance contract review model to obtain a review result of the second insurance contract sample; Based on the difference between the audit result of the second insurance contract sample and the second annotated data, obtaining the technical indicators of the first insurance contract audit model, wherein the second annotated data is the annotated data of the second insurance contract sample in the second insurance contract sample data, and the technical indicators are used to indicate the accuracy and recall rate of the first insurance contract audit model; When the technical indicators of the first insurance contract review model are greater than or equal to the technical indicators of the second insurance contract review model and the technical indicators of the first insurance contract review model meet the expected requirements, the first insurance contract review model is obtained.

[0070] In an optional implementation, the test module is further used to: When the accuracy of the first audit result is lower than or equal to the accuracy of the second audit result, the first insurance contract audit model is retrained using third insurance contract sample data.

[0071] In an optional implementation, the first audit result and the second audit result include: text errors and audit opinions corresponding to the text errors.

[0072] In an optional implementation, the acquisition module 301 is further used to: Obtaining a first business indicator corresponding to the preset time for reviewing the insurance contract to be reviewed through the first insurance contract review model, wherein the first business indicator includes at least one of the following: the drafting time of the insurance contract, the number of round trips of the insurance contract; According to the first business indicator, the business effect corresponding to the first insurance contract review model is obtained.

[0073] In an optional implementation, when the acquisition module 301 is used to acquire the business effect corresponding to the first insurance contract review model according to the first business indicator, it is specifically used to: Based on historical data of insurance contracts reviewed by the second insurance contract review model, obtaining an average value of a second business indicator corresponding to the second insurance contract review model within the preset time period; Based on the difference between the first business indicator and the average value, the business effect corresponding to the first insurance contract review model is obtained.

[0074] The review device of the insurance contract in the embodiment of the present application can be an electronic device, or a component in the electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or other devices other than a terminal. Exemplarily, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a car-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc. It can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., and the embodiment of the present application does not make specific limitations.

[0075] The insurance contract review device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0076] The insurance contract review device provided in the embodiment of the present application can achieve Figure 1 To avoid repetition, the various processes implemented in the illustrated method embodiment will not be described again here.

[0077] Alternatively, if Figure 4 As shown, an embodiment of the present application also provides an electronic device 400, including a processor 401 and a memory 402, wherein the memory 402 stores programs or instructions that can be executed on the processor 401, and when the program or instructions are executed by the processor 401, the various steps of the above-mentioned insurance contract review method are implemented, and the same technical effect can be achieved. To avoid repetition, they will not be repeated here.

[0078] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0079] Figure 5FIG. 1 is a block diagram of another electronic device 500 according to an exemplary embodiment of the present application. The electronic device 500 may be implemented as a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart watch, a TV, etc. The electronic device 500 may also be referred to as a user device, a portable terminal, a laptop terminal, a desktop terminal, etc.

[0080] Typically, the electronic device 500 includes a processor 501 and a memory 502 .

[0081] The processor 501 may include one or more processing cores, such as a 4-core processor, a 10-core processor, etc. The processor 501 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 501 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 501 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 501 may also include an AI processor, which is used to process computing operations related to machine learning.

[0082] The memory 502 may include one or more computer-readable storage media, which may be non-transitory. The memory 502 may also include a high-speed random access memory and a non-volatile memory, such as one or more disk storage devices and flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 502 is used to store at least one instruction, which is used to be executed by the processor 501 to implement all or part of the steps in the insurance contract review method shown in the method embodiment of the present application.

[0083] In some embodiments, the electronic device 500 may further optionally include: a peripheral device interface 503 and at least one peripheral device. The processor 501, the memory 502 and the peripheral device interface 503 may be connected via a bus or a signal line. Each peripheral device may be connected to the peripheral device interface 503 via a bus, a signal line or a circuit board. Specifically, the peripheral device includes: at least one of a radio frequency circuit 504, a display screen 505, a camera assembly 506, an audio circuit 507 and a power supply 508.

[0084] In some embodiments, the electronic device 500 further includes one or more sensors 509 , including but not limited to: an acceleration sensor 510 , a gyroscope sensor 511 , a pressure sensor 512 , an optical sensor 513 , and a proximity sensor 514 .

[0085] Those skilled in the art will understand that Figure 5 The structure shown in the figure does not constitute a limitation on the electronic device 500, and may include more or less components than those shown in the figure, or combine some components, or adopt a different component arrangement.

[0086] An embodiment of the present application also provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the various processes of the above-mentioned insurance contract review method are implemented and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0087] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0088] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned insurance contract review method, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0089] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0090] An embodiment of the present application provides a computer program product, which is stored in a storage medium. The computer program product is executed by at least one processor to implement the various processes of the insurance contract review method as described above, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0091] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0092] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0093] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A method for reviewing an insurance contract, characterized in that: include: Inputting part of the insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model, obtaining a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model, wherein the first insurance contract review model and the second insurance contract review model are used to review the input insurance contracts and obtain the review results of the insurance contracts, and the second insurance contract review model is the previous version of the first insurance contract review model; Determining the accuracy of the first audit result and the second audit result; When the accuracy of the first audit result is higher than the accuracy of the second audit result, the insurance contract to be audited is audited using the first insurance contract audit model.

2. The method according to claim 1, characterized in that The step of inputting part of the insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model includes: Inputting a first percentage of insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain a first initial review result output by the first insurance contract review model and a second initial review result output by the second insurance contract review model; When the accuracy of the first initial audit result is higher than the accuracy of the second initial audit result, the insurance contracts of the second percentage in the insurance contracts to be audited are input into the first insurance contract audit model and the second insurance contract audit model to obtain the first audit result output by the first insurance contract audit model and the second audit result output by the second insurance contract audit model; The first proportion and the second proportion are rational numbers greater than 0 and less than 1, and the first proportion is smaller than the second proportion.

3. The method according to claim 1, characterized in that Before inputting the part of the insurance contracts to be reviewed into the first insurance contract review model and the second insurance contract review model to obtain the first review result output by the first insurance contract review model and the second review result output by the second insurance contract review model, the method further includes: Iteratively training the first insurance contract review model to be trained using the first insurance contract sample data until a preset training completion condition is met, thereby obtaining the first insurance contract review model that has been trained, wherein the first insurance contract sample data includes the first insurance contract sample and first annotated data of the first insurance contract sample; Inputting the second insurance contract sample in the second insurance contract sample data into the first insurance contract review model to obtain a review result of the second insurance contract sample; Based on the difference between the audit result of the second insurance contract sample and the second annotated data, obtaining the technical indicators of the first insurance contract audit model, wherein the second annotated data is the annotated data of the second insurance contract sample in the second insurance contract sample data, and the technical indicators are used to indicate the accuracy and recall rate of the first insurance contract audit model; When the technical indicators of the first insurance contract review model are greater than or equal to the technical indicators of the second insurance contract review model and the technical indicators of the first insurance contract review model meet the expected requirements, the first insurance contract review model is obtained.

4. The method according to claim 3, characterized in that After determining the accuracy of the first audit result and the second audit result, the method further includes: When the accuracy of the first audit result is lower than or equal to the accuracy of the second audit result, the first insurance contract audit model is retrained using third insurance contract sample data.

5. The method according to claim 1, characterized in that The first audit result and the second audit result include: text errors and audit opinions corresponding to the text errors.

6. The method according to any one of claims 1 to 5, characterized in that: After the insurance contract to be reviewed is reviewed by using the first insurance contract review model, the method further includes: Obtaining a first business indicator corresponding to the review of the insurance contract to be reviewed by the first insurance contract review model within a preset time period, wherein the first business indicator includes at least one of the following: the drafting time of the insurance contract, the number of round trips of the insurance contract; According to the first business indicator, the business effect corresponding to the first insurance contract review model is obtained.

7. The method according to claim 6, characterized in that The obtaining, according to the first business indicator, a business effect corresponding to the first insurance contract review model includes: Based on historical data of the insurance contracts reviewed by the second insurance contract review model, obtaining an average value of a second business indicator corresponding to the second insurance contract review model within the preset time period; Based on the difference between the first business indicator and the average value, the business effect corresponding to the first insurance contract review model is obtained.

8. An insurance contract review device, characterized in that: include: an acquisition module, used for inputting part of the insurance contracts to be reviewed into a first insurance contract review model and a second insurance contract review model, and obtaining a first review result output by the first insurance contract review model and a second review result output by the second insurance contract review model, wherein the first insurance contract review model and the second insurance contract review model are used to review the input insurance contracts and obtain the review results of the insurance contracts, and the second insurance contract review model is a previous version of the first insurance contract review model; A determination module, determining the accuracy of the first audit result and the second audit result; An audit module is used to audit the insurance contract to be audited through the first insurance contract audit model when the accuracy of the first audit result is higher than the accuracy of the second audit result.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the insurance contract review method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the steps of the insurance contract review method as described in any one of claims 1 to 7 are implemented.

Images