Claim settlement auditing collaborative agent system based on large model thinking chain

Through the claims review collaborative intelligent system based on the big model thinking chain, the existing claims process is inefficient and high labor costs are solved, efficient and accurate claims review is achieved, and customer satisfaction and system adaptability are improved.

CN119006173APending Publication Date: 2024-11-22上海栈略数据技术有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410959038.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The existing claims process is inefficient and labor costs, which leads to inconvenient customer service and prone to human errors.

Method used

Adopt a claim review collaborative agent system based on the big model thinking chain, including database, model library, rule library, processing unit, monitoring unit, optimization unit and configuration unit. By establishing a thinking chain, claims information processing and optimization are automatically processed and analyzed. Claims data to improve audit efficiency and accuracy.

Benefits of technology

It significantly improves the efficiency and accuracy of claims review, reduces the workload and error rate of manual review, reduces customer waiting time, improves customer satisfaction, and has strong adaptability and scalability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006173A_ABST
    Figure CN119006173A_ABST
Patent Text Reader

Abstract

The invention relates to the field of intelligent auditing systems, in particular to a claim settlement auditing collaborative agent system based on a large model thinking chain, which comprises a database, a model library and a rule library, and is characterized in that the rule library comprises a processing unit used for identifying the type of claim settlement information, determining a processing mode of the claim settlement information according to the type of the claim settlement information, and sending the processing mode to the database; calculating an accuracy degree evaluation value in a corresponding processing mode; the optimization unit is connected with the processing unit, and determines an optimization mode of the claim settlement auditing process according to a difference value between the accuracy degree evaluation value and a preset accuracy degree evaluation value; and the configuration unit is connected with the processing unit, and determines a configuration mode for reconfiguring the claim settlement information in a corresponding processing mode according to the complexity of the language reasoning process of the claim settlement auditing process. According to the invention, the problems of improving the claim settlement efficiency, reducing the labor cost and improving the customer satisfaction in the prior art are solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent audit systems, and in particular to a claims audit collaborative intelligent agent system based on a large model thinking chain. Background Art

[0002] The big model thinking chain is a way of thinking based on large-scale pre-trained language models. It applies natural language processing and machine learning techniques to natural language understanding and generation, and can be used in multiple fields such as text classification, sentiment analysis, machine translation, question-answering systems, etc. In the big model thinking chain, a pre-trained language model is used to extract the semantic information of the input text and map it to a low-dimensional vector space for further processing. This low-dimensional vector representation can be used as input to subsequent machine learning models to complete different tasks, such as classification, regression, sequence generation, etc. The big model thinking chain can achieve good performance without requiring too much training data by using large-scale pre-trained language models.

[0003] With the rapid development and popularization of the insurance industry, claims have become an important part of customer service. The traditional claims process usually involves a lot of manual operations and complex document applications, which is time-consuming and prone to human errors, causing inconvenience to customers.

[0004] The claims review collaborative intelligent agent system based on the big model thinking chain can greatly improve claims efficiency, reduce costs, and improve customer satisfaction.

[0005] Chinese Patent Publication No.: CN115049512A discloses an intelligent claims settlement system, including: an input module, a rule configuration module, an audit module and a settlement module, wherein the input module is used to collect and input case information, the rule configuration module is used to configure claims rules according to case information, the audit module is used to review case information and claims rules, and the settlement module is used to calculate claims results according to case information and claims rules. The present invention can improve claims efficiency and cost control and risk control capabilities, and improve customer claims experience and satisfaction through the full process of information input, configuration, audit and claims procedures.

[0006] It can be seen that the existing technology has problems of low claims settlement efficiency and high labor costs. Summary of the invention

[0007] To this end, the present invention provides a claims review collaborative intelligent system based on a large model thinking chain to overcome the problems in the prior art of improving claims efficiency, reducing labor costs and improving customer satisfaction.

[0008] To achieve the above objectives, the present invention provides a claims review intelligent agent system based on a large model thinking chain, comprising:

[0009] Databases to store, manage and retrieve certain data related to the claims review process;

[0010] A model library, which is connected to the database and is used to establish a thinking chain of the claim review process;

[0011] A rule base, which is connected to the database and the model base respectively, is used to store and manage the rules related to the claim review process, and the rule base includes:

[0012] A processing unit, used to identify the claim information, so as to determine the processing of the claim information according to the type of the claim information;

[0013] A monitoring unit, used to monitor the claim review process to determine the accuracy of the thought chain for the claim information review process;

[0014] an optimization unit connected to the processing unit, for determining the optimization of the claim review process by the thought chain according to the accuracy;

[0015] A configuration unit is connected to the processing unit and is used to determine whether to reconfigure the claim information according to the complexity of the language reasoning process of the claim review process.

[0016] Furthermore, the processing unit determines a method for processing the claim information based on the claim information type, including, under the condition that the claim information type exists in the database, directly executing a claim review process through the thinking chain to process the claim information based on the claim information type and the claim amount.

[0017] Furthermore, the processing unit determines a method for processing the claim information based on the claim information type, including determining an initial continuous prompt learning time of the thinking chain based on the information complexity of the claim information to process the claim information under the condition that the claim information type does not exist in the database.

[0018] Furthermore, the optimization unit determines the optimization method of the thought chain for the claim review process according to the difference between the accuracy evaluation value and the preset accuracy evaluation value, including:

[0019] Under the condition that the difference between the accuracy evaluation value and the preset accuracy evaluation value is less than the preset accuracy evaluation value relative difference, optimizing the claim review process in a first optimization manner;

[0020] Under the condition that the difference between the accuracy evaluation value and the preset accuracy evaluation value is greater than or equal to the relative difference of the preset accuracy evaluation value, the claim review process is optimized in a second optimization manner.

[0021] Furthermore, the accuracy evaluation value is calculated according to the following formula:

[0022]

[0023] Among them, R represents the accuracy evaluation value, W i represents the weight of the i-th claim item, A i Represents the accuracy of the i-th claim item.

[0024] Furthermore, the first optimization method includes adjusting the continuous prompt learning time in the second processing method; the second optimization method includes compressing or removing redundancy of the model language through information entropy.

[0025] Furthermore, the information entropy is calculated according to the following formula:

[0026] H = -∑P(x)×log2P(x)

[0027] Among them, H represents information entropy and P(x) represents the probability of information occurrence.

[0028] Furthermore, the configuration unit determines, based on the complexity of the language reasoning process of the claim review process, a configuration method for reconfiguring the claim information under the corresponding processing method, including: under the condition that the complexity of the language reasoning process is greater than a preset complexity, reconfiguring the claim information in a first configuration method; under the condition that the complexity of the language reasoning process is less than or equal to a preset complexity, reconfiguring the claim information in a second configuration method.

[0029] Furthermore, the first configuration method includes: removing repeated keywords in the claim information according to keyword density; the second configuration method includes: adding synonym sentences of the claim information keywords to train and learn the thinking chain.

[0030] Furthermore, the keyword density is calculated according to the following formula:

[0031]

[0032] Among them, Q represents the keyword density, n represents the frequency of keyword occurrence, and N represents the total number of keywords.

[0033] Compared with the prior art, the beneficial effect of the present invention lies in that, by setting up the database, various data related to the claims review process can be stored and managed, and the database can be updated in real time, and can provide new data for the model library for automatic learning and training, while also being able to avoid inaccurate claims review results due to data delays or expiration.

[0034] Furthermore, by setting up the model library to establish the thinking chain of the claims review process, setting up the model library, and training and optimizing the thinking chain of the claims review process by establishing specific rules, the efficiency and accuracy of the claims review process can be improved.

[0035] Furthermore, by setting up the processing unit, different processing methods can be selected for different types of claim information, which can save system internal resources and improve the efficiency of claim review. The formula for calculating the accuracy evaluation value combines the accuracy of different claim items in a weighted average manner and takes into account their weights, where the weights can be adjusted according to the importance of different claim items. Using this formula for calculation can more accurately reflect the accuracy of the system's review of the claim information.

[0036] Furthermore, by setting up the optimization unit and determining different optimization methods for the claims review processing according to the accuracy evaluation value, the efficiency of the claims review process can be improved. Selecting different optimization methods according to actual conditions can ensure that the optimization process is more stable and reliable.

[0037] Furthermore, by setting the configuration unit, the configuration method for reconfiguring the claim information is determined according to the complexity of the language reasoning process, which can avoid unnecessary reasoning processes in the language reasoning process, thereby improving the reasoning efficiency, and then improving the efficiency of claim review, while also saving system internal resources and making full use of system resources.

[0038] Furthermore, the present invention is based on a large model thinking chain, which can simulate the logical reasoning process of humans and quickly and accurately analyze the claims information. By automatically processing and analyzing a large amount of claims data, the collaborative intelligent system can significantly improve the efficiency of claims review, while reducing the workload and error rate of manual review. At the same time, the automated processing method can reduce customer waiting time and unnecessary troubles, and improve user satisfaction. The claims review collaborative intelligent system based on the large model thinking chain has strong adaptability and scalability. As the claims review requirements continue to change and upgrade, the system can be easily expanded and upgraded to adapt to new needs and challenges. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a structural diagram of a claims review collaborative intelligent agent system based on a large model thinking chain according to an embodiment of the present invention;

[0040] Figure 2 A flowchart of a processing unit according to an embodiment of the present invention;

[0041] Figure 3 This is a workflow diagram of the optimization unit described in the embodiment of the present invention;

[0042] Figure 4 This is a workflow diagram of the configuration unit described in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0045] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0046] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0047] Example 1

[0048] See also Figure 1-Figure 4 As shown,

[0049] The claims review collaborative intelligent agent system based on the big model thinking chain in the embodiment of the present invention includes:

[0050] Databases to store, manage and retrieve certain data related to the claims review process;

[0051] A model library, which is connected to the database and is used to establish a thinking chain of the claim review process;

[0052] A rule base, which is connected to the database and the model base respectively, is used to store and manage the rules related to the claim review process, and the rule base includes:

[0053] A processing unit, used to identify the claim information, and determine a processing method for the claim information according to the type of the claim information;

[0054] A monitoring unit, used for monitoring the claim review process, so as to determine an accuracy evaluation value R of the thought chain for reviewing the claim information under a corresponding processing mode;

[0055] an optimization unit connected to the processing unit, for determining an optimization method of the thought chain for the claim review process according to a difference ΔR between the accuracy evaluation value R and a preset accuracy evaluation value R0;

[0056] A configuration unit is connected to the processing unit and is used to determine a configuration method for reconfiguring the claim information under a corresponding processing method according to the complexity of the language reasoning process of the claim review process.

[0057] In the embodiment of the present invention, the plurality of data includes customer information, policy information, and claim records.

[0058] In the embodiment of the present invention, the database can be updated in real time, can provide new data for the model library for automatic learning and training, and can also avoid inaccurate claims review results due to data delay or expiration. The rule library trains and optimizes the thinking chain of the claims review process in the model library through specific rules, which can improve the efficiency and accuracy of the claims review process.

[0059] Specifically, the processing unit determines the processing method of the claim information according to the claim information type, including:

[0060] Under the condition that the claim information type exists in the database, processing the claim information in a first processing manner;

[0061] Under the condition that the claim information type does not exist in the database, the claim information is processed in a second processing manner.

[0062] Specifically, the first processing method includes directly executing the claim review process through the thought chain;

[0063] In the embodiment of the present invention, under the first processing mode, directly executing the claim review process through the thinking chain is to directly input the claim information and the claim amount into the thinking chain to perform the claim review.

[0064] Specifically, the second processing method determines the initial continuous prompt learning time T of the thought chain according to the information complexity S of the claim information, and compares the information complexity S with the preset information complexity S0:

[0065] If S<S0, the initial continuous prompt learning time length T is determined to be the first preset time length T1;

[0066] If S≥S0, determine the initial continuous prompt learning time length T to be the second preset time length T2;

[0067] In an embodiment of the present invention, technical personnel in this field can adjust the preset information complexity S0, the first preset time length T1, and the second preset time length T2 according to actual conditions, where the preset complexity value is 0.65, the first preset time length value is 10 minutes, and the second preset time length value is 15 minutes.

[0068] Specifically, the information complexity S is calculated according to the following formula:

[0069] S=Σ(D i ×B i )

[0070] Among them, S represents the complexity of information, D i Indicates the score of factors affecting the complexity of claim information, B i Represents the weight coefficient.

[0071] In an embodiment of the present invention, under the second processing method, the weight coefficient corresponding to the factor affecting the complexity of the claim information in the formula for calculating the complexity of the information is adjusted according to the importance of the factor. Assuming that the factors affecting the complexity of the claim information are information diversity, information relevance to the case, and information structuring, the corresponding weight coefficients are set to 0.15, 0.3, and 0.15, wherein the score of the information diversity is the ratio of the number of information types to the sum of the total number of each information type, the score of the information relevance to the case is the ratio of the amount of data associated with the claim case in the claim information to the amount of claim information data, the associated data is the same data, and the information structuring is the ratio of the amount of structured data to the total amount of data.

[0072] In an embodiment of the present invention, different processing methods are selected for different types of claim information. For the types of claim information existing in the database, the claim review process can be directly executed through the thinking chain, thereby improving the efficiency of claim review on the basis of saving internal system resources.

[0073] Specifically, the accuracy evaluation value R is calculated according to the following formula:

[0074]

[0075] Among them, R represents the accuracy evaluation value, W i represents the weight of the i-th claim item, Ai Represents the accuracy of the i-th claim item.

[0076] In an embodiment of the present invention, for example, in the process of medical claims, the claim items include medical expenses, loss of work wages, accompanying expenses, living allowances, transportation expenses, and accommodation expenses, and the corresponding weight coefficients are set to 0.5, 0.15, 0.15, 0.05, 0.05, and 0.05 according to the importance. For example, in the process of property claims, the claim items include property losses, facility damage, repair and reconstruction costs, and removal and disposal costs, and the corresponding weight coefficients are set to 0.4, 0.3, 0.15, and 0.1 according to the importance. The accuracy of each claim item is determined by comparing the data in the claim information with the external data. For example, the accuracy of medical expenses can be determined by comparing the data in the claim information with the data provided by the medical institution.

[0077] In an embodiment of the present invention, the formula for calculating the accuracy evaluation value R combines the accuracies of different claim items in a weighted average manner and takes into account their weights, wherein the weights can be adjusted according to the importance of different claim items. Using this formula for calculation can more accurately reflect the accuracy of the system's review of the claim information.

[0078] Specifically, the optimization unit determines the optimization method of the claim review process by the thought chain according to the difference ΔR between the accuracy evaluation value R and the preset accuracy evaluation value R0, including determining the optimization method according to the comparison result of the difference and the relative difference ΔR0 of the preset accuracy evaluation value;

[0079] If ΔR<ΔR0, the optimization unit optimizes the claim review process in a first optimization manner;

[0080] If ΔR≥ΔR0, the optimization unit optimizes the claim review process in a second optimization manner.

[0081] In the embodiment of the present invention, the preset accuracy evaluation value is 0.85, and the preset accuracy evaluation value relative difference is 0.15.

[0082] Specifically, the first optimization method includes: adjusting the coefficient Adjusting the initial continuous prompt learning time T in the second processing mode, wherein ΔR is the difference between the accuracy evaluation value R and the preset accuracy evaluation value R0, and the adjusted continuous prompt learning time is Tt=T×X;

[0083] The second optimization method includes: compressing or removing redundancy from the model language through information entropy.

[0084] In the embodiment of the present invention, different optimization methods are adopted for the claim review process according to different differences in the accuracy evaluation values, which can improve the efficiency of the claim review process. Different optimization methods are selected according to actual conditions to ensure that the optimization process is more stable and reliable. Those skilled in the art can adjust the preset accuracy evaluation value R0 and the preset accuracy evaluation value relative value ΔR0 according to actual conditions.

[0085] In the embodiment of the present invention, if the difference in the accuracy evaluation values ​​is small, it indicates that the learning and training time of the model is insufficient, and the initial continuous prompt learning time needs to be adjusted; if the difference in the accuracy evaluation values ​​is large, it indicates that there is some repeated information in the model language, and the model language needs to be refined.

[0086] Specifically, the information entropy is calculated according to the following formula, setting

[0087] H=-ΣP(x)×log2P(x)

[0088] Among them, H represents information entropy, and P(x) represents the probability of claim information appearing.

[0089] In the embodiment of the present invention, by calculating the information entropy of each information item in the model language, redundant information can be identified, thereby compressing or removing redundancy on the corresponding information, thereby achieving the purpose of refining the model language.

[0090] Specifically, the configuration unit determines, according to the complexity of the language reasoning process of the claim review process, to reconfigure the claim information under the corresponding processing method, including determining the configuration method according to the comparison result of the complexity and the preset complexity;

[0091] Under the condition that the complexity of the language reasoning process is greater than a preset complexity, reconfiguring the claim information in a first configuration manner;

[0092] Under the condition that the complexity of the language reasoning process is less than or equal to a preset complexity, the claim information is reconfigured in a second configuration manner.

[0093] In the embodiment of the present invention, the higher the complexity of the language reasoning process, the higher the complexity of the claim information, indicating that the claim information needs to be reconfigured. According to the different complexity of the language reasoning process, different configuration methods are selected for the information reconfiguration of the corresponding processing method, so that the claim information can be configured more accurately. Those skilled in the art can adjust the preset complexity according to actual conditions.

[0094] Specifically, the first configuration method includes refining the claim information and removing repeated keywords in the claim information according to keyword density;

[0095] The second configuration method includes: adding synonymous sentences of the claim information keywords to train and learn the thinking chain.

[0096] In the embodiment of the present invention, the first configuration method refines the claim information by calculating the keyword density. The higher the keyword density, the more repeated keywords appear, and the repeated keywords are removed. The second configuration method adds synonyms to the claim information, enriches the data for model training, and improves the accuracy of the claim review.

[0097] Specifically, the keyword density is calculated according to the following formula:

[0098]

[0099] Among them, Q represents the keyword density, n represents the frequency of keyword occurrence, and N represents the total number of keywords.

[0100] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A claim review collaborative intelligent agent system based on a large model thinking chain, characterized in that: include: Databases to store, manage and retrieve certain data related to the claims review process; A model library, which is connected to the database and is used to establish a thinking chain of the claim review process; A rule base, which is connected to the database and the model base respectively, is used to store and manage the rules related to the claim review process, and the rule base includes: A processing unit, used to identify the claim information, so as to determine the processing of the claim information according to the type of the claim information; A monitoring unit, used to monitor the claim review process to determine the accuracy of the thought chain for the claim information review process; an optimization unit connected to the processing unit, for determining the optimization of the claim review process by the thought chain according to the accuracy; A configuration unit is connected to the processing unit and is used to determine whether to reconfigure the claim information according to the complexity of the language reasoning process of the claim review process.

2. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 1 is characterized in that: The processing unit determines the processing method of the claim information according to the claim information type, including directly executing the claim review process through the thinking chain to process the claim information according to the claim information type and the claim amount, under the condition that the claim information type exists in the database.

3. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 2 is characterized in that: The processing unit determines the processing method of the claim information according to the claim information type, including determining the initial continuous prompt learning time of the thinking chain according to the information complexity of the claim information to process the claim information under the condition that the claim information type does not exist in the database.

4. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 1 is characterized in that: The optimization unit determines the optimization method of the claim review process by the thought chain according to the difference between the accuracy evaluation value and the preset accuracy evaluation value, including: Under the condition that the difference between the accuracy evaluation value and the preset accuracy evaluation value is less than the preset accuracy evaluation value relative difference, optimizing the claim review process in a first optimization manner; Under the condition that the difference between the accuracy evaluation value and the preset accuracy evaluation value is greater than or equal to the relative difference of the preset accuracy evaluation value, the claim review process is optimized in a second optimization manner.

5. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 1 is characterized in that: The accuracy evaluation value is calculated according to the following formula: Among them, R represents the accuracy evaluation value, W i represents the weight of the i-th claim item, A i Represents the accuracy of the i-th claim item.

6. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 5 is characterized in that: The first optimization method includes adjusting the continuous prompt learning time in the second processing method; the second optimization method includes compressing or removing redundancy of the model language through information entropy.

7. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 6 is characterized in that: The information entropy is calculated according to the following formula: H = -∑P(x)×log2P(x) Among them, H represents information entropy and P(x) represents the probability of information occurrence.

8. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 1 is characterized in that: The configuration unit determines, based on the complexity of the language reasoning process of the claim review process, a configuration method for reconfiguring the claim information under the corresponding processing method, including: under the condition that the complexity of the language reasoning process is greater than a preset complexity, reconfiguring the claim information in a first configuration method; under the condition that the complexity of the language reasoning process is less than or equal to a preset complexity, reconfiguring the claim information in a second configuration method.

9. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 8 is characterized in that: The first configuration method includes: removing repeated keywords in the claim information according to keyword density; the second configuration method includes: adding synonym sentences of the keywords in the claim information to train and learn the thinking chain.

10. The claim review collaborative intelligent agent system based on the big model thinking chain according to claim 9 is characterized in that: The keyword density is calculated according to the following formula: Among them, Q represents the keyword density, n represents the frequency of keyword occurrence, and N represents the total number of keywords.