Customer service semantic understanding model integration risk identification method, system, equipment and medium

Abstract

The invention relates to the technical field of intelligent customer service, in particular to a customer service semantic understanding model integration risk identification method, system and device and a medium. The customer service semantic understanding model integration risk identification method comprises the following steps: identifying a risk index; describing a trapezoidal fuzzy number; matter element conversion and classical domain and node domain determination; calculating an index weight; calculating the close degree; determining a risk level; performing comprehensive risk assessment; according to the method, by introducing the trapezoidal fuzzy number, the uncertainty and fuzziness of the risk index can be described more accurately. A matter-element extension model is introduced to define a classical domain and a node domain, and the closeness is calculated, so that quantitative evaluation of the risk level of the risk index is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent customer service technology, and more particularly to a customer service semantic understanding model integrated risk identification method, system, device, and medium. Background Art

[0002] With the rapid development of artificial intelligence and big data technologies, customer service systems are gradually transforming towards intelligence and automation. In this transformation process, semantic understanding models, as key technical components, play a core role in understanding user intentions and providing accurate services. However, it is not easy to integrate advanced semantic understanding models into the original customer service system. This integration process not only involves complex technical interface docking and data migration, but also may bring a series of potential risks, such as technical compatibility risks, data integrity risks, performance stability risks, and business integrity risks.

[0003] A common solution is to adopt a rule-based risk assessment method. This method checks each link in the integration process one by one by setting a series of predefined rules and conditions to identify possible risks. Although this method is relatively simple and easy to implement, the accuracy and comprehensiveness of its assessment results may be limited by the experience and knowledge level of the rule setter.

[0004] Another solution is to use existing risk assessment models or frameworks and make adaptive modifications. For example, risk assessment models in other industries or fields can be used for reference and appropriately adjusted and optimized according to the characteristics and needs of the customer service system. This approach can make full use of existing resources and reduce duplication of work, but it may require additional efforts to ensure that the modified model matches the actual situation of the customer service system.

[0005] In addition, some approximate solutions may involve using statistical methods or data mining techniques to analyze historical data to identify potential risks in the integration process. These methods can provide objective and quantitative evaluation results based on a large amount of data, but may be affected by factors such as data quality, sample size, and statistical method selection.

[0006] Existing technologies have obvious deficiencies in risk assessment of the integration of electric customer service semantic understanding models with original systems. Specifically, existing technologies are often based on traditional risk assessment methods, such as qualitative analysis or simple quantitative assessment, which have limitations in identifying and analyzing integration risks. Due to the lack of specialized design and in-depth analysis for the integration of customer service semantic understanding models, existing technologies may not be able to comprehensively and accurately assess various risks in the integration process, such as technical compatibility risks, data integrity risks, performance stability risks, etc.

[0007] The existence of these shortcomings may lead to problems such as inaccurate risk identification, incomplete risk assessment, and untimely risk response during the integration process, which in turn affects the stability and service quality of the customer service system. Summary of the invention

[0008] The purpose of this invention is to provide an innovative electric customer service semantic understanding model integrated quantitative assessment model. The model can comprehensively consider multiple risk factors, use mathematical and statistical methods to quantitatively assess integrated risks, and thus provide more accurate and comprehensive risk assessment results. The specific technical solution is as follows:

[0009] Through the present invention, decision makers can formulate risk response strategies more scientifically, reduce risks in the integration process, and ensure the stability and service quality of the power customer service system.

[0010] In the first aspect, a customer service semantic understanding model integrated risk identification method includes the following process:

[0011] S100: Identify risk indicators; specifically, clarify all risk factors that may be involved in the integration process of the customer service semantic understanding model and form a risk indicator set;

[0012] S200: Trapezoidal fuzzy number description; specifically includes:

[0013] For each risk indicator, trapezoidal fuzzy numbers are used to describe its possible value range and distribution;

[0014] Assumptions is a trapezoidal fuzzy number, which is defined over the real number field X, then it can be expressed as an ordered four-tuple Where a≤b≤c≤d, a and d represent the lower and upper bounds of the trapezoidal fuzzy number, respectively, and b and c represent the starting and ending points of the membership peak;

[0015] Trapezoidal fuzzy number The membership function It can be defined as:

[0016]

[0017] The membership function is 0 when x=a and x=d, and 1 when x=b and x=c; in the interval a≤x≤b, the membership increases linearly from 0 to 1; in the interval c≤x≤d, the membership decreases linearly from 1 to 0;

[0018] S300: Matter-element transformation and determination of classical domain and section domain; specifically including:

[0019] The risk index represented by trapezoidal fuzzy numbers is transformed into matter-element in matter-element extenics, namely, risk matter-element.

[0020] For each risk indicator, a thing element is established, where: thing (N): risk indicator itself; characteristic (C): the scope of influence of risk indicator, which can be described by the parameters a, b, c, d of trapezoidal fuzzy number; value (V): the specific value of risk indicator, i.e. trapezoidal fuzzy number itself; classical domain and section domain are used to define evaluation criteria and evaluation level; for risk assessment, the classical domain can be defined as the acceptable range of risk, and the section domain can be defined as the unacceptable range of risk;

[0021] S400: indicator weight calculation;

[0022] S500: Calculate the closeness; specifically including:

[0023] According to the principle of matter-element extenics, the closeness of each risk indicator to different risk levels is calculated;

[0024] The calculation of the closeness involves comparing the risk element with each risk level in the classical domain and the segment domain; for each risk level j, the risk element is calculated. The degree of closeness to this level K j ; calculated by the following formula:

[0025]

[0026] in, is the membership degree of the risk element at xi, is the jth risk level in x i The degree of membership of the place;

[0027] S600: Determine the risk level; specifically, including: based on the closeness criterion, determine the risk level to which each risk indicator belongs; if the closeness of a risk indicator is the highest at a certain level, determine that the indicator is at that risk level;

[0028] S700: Comprehensive risk assessment; specifically including: calculating the overall risk level using the weighted average method based on the weight of each risk indicator and the risk level to which it belongs.

[0029] Secondly, the customer service semantic understanding model integrates the risk identification system, including:

[0030] The risk indicator identification module is used to identify all risk factors that may be involved in the integration process of the customer service semantic understanding model and form a risk indicator set;

[0031] Trapezoidal fuzzy number description module, used to describe the possible value range and distribution of each risk indicator using trapezoidal fuzzy numbers;

[0032] The matter-element conversion and classical domain and section domain determination unit is used to convert the risk index represented by trapezoidal fuzzy numbers into matter-element in matter-element extenics, namely, risk matter-element;

[0033] An indicator weight calculation unit, used to calculate the indicator weight;

[0034] The closeness calculation unit is used to calculate the closeness of each risk indicator relative to different risk levels based on the principle of matter-element extenics;

[0035] The risk level unit is used to determine the risk level to which each risk indicator belongs based on the closeness criterion; if the closeness of a risk indicator is the highest at a certain level, the indicator is determined to be at that risk level;

[0036] The comprehensive risk assessment unit is used to calculate the overall risk level using the weighted average method based on the weight of each risk indicator and the risk level to which it belongs.

[0037] In a third aspect, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned customer service semantic understanding model integrated risk identification method when executing the computer program.

[0038] In a fourth aspect, a non-transitory computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the method for integrating risk identification with a customer service semantic understanding model is implemented.

[0039] The advantages of the present invention are as follows:

[0040] By introducing trapezoidal fuzzy numbers, the present invention can more accurately describe the uncertainty and ambiguity of risk indicators. Compared with traditional deterministic evaluation methods, trapezoidal fuzzy numbers provide richer information, including the possible value range of risk indicators, membership functions, etc., thereby improving the accuracy of risk assessment.

[0041] By introducing the matter-element extenics model to define the classical domain and the node domain, and calculating the closeness, the quantitative evaluation of the risk level of the risk indicator is realized. This method not only considers the actual value of the risk indicator, but also considers its closeness to different risk levels, which further improves the accuracy of risk assessment.

[0042] In the embodiment, a weight calculation method based on the C-OWA operator is adopted, which combines expert experience and quantitative analysis to make the weight of risk indicators more reasonable and scientific. This helps decision makers to more accurately grasp the importance of each risk indicator and thus formulate more effective risk management strategies.

[0043] Comprehensive risk assessment and response strategy formulation: Through weighted average or other appropriate aggregation methods, the present invention calculates the overall risk level and formulates targeted risk response strategies based on the risk assessment results. This helps decision makers grasp the risk situation from a global perspective, take effective measures to reduce risks in a timely manner, and enhance the practicality of risk assessment.

[0044] The present invention provides a model optimization and iteration method based on actual project progress and risk assessment results. This means that as the project progresses and risks change, the risk assessment model can be continuously adjusted and improved to meet new risk assessment needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flow chart of a customer service semantic understanding model integrated risk identification method of the present invention. DETAILED DESCRIPTION

[0046] Example 1

[0047] A customer service semantic understanding model integrated risk identification method includes the following process:

[0048] S100: Identify risk indicators; specifically include:

[0049] Identify all possible risk factors involved in the integration of the customer service semantic understanding model and form a set of risk indicators; the factors include: indicator 1 - technical compatibility, indicator 2 - data integrity, indicator 3 - performance stability;

[0050] S200: Trapezoidal fuzzy number description; specifically includes:

[0051] For each risk indicator, trapezoidal fuzzy numbers are used to describe its possible value range and distribution;

[0052] Assumptions is a trapezoidal fuzzy number, which is defined over the real number field X, then it can be expressed as an ordered four-tuple Where a≤b≤c≤d, a and d represent the lower and upper bounds of the trapezoidal fuzzy number, respectively, and b and c represent the starting and ending points of the membership peak;

[0053] Trapezoidal fuzzy number The membership function It can be defined as:

[0054]

[0055] The membership function is 0 when x=a and x=d, and 1 when x=b and x=c; in the interval a≤x≤b, the membership increases linearly from 0 to 1; in the interval c≤x≤d, the membership decreases linearly from 1 to 0;

[0056] For example, for technical compatibility, experts evaluated the possible range of values ​​as [70, 80, 90, 100] (a = 70, b = 80, c = 90, d = 100), indicating that the technical compatibility is at least 70% guaranteed, but most likely 90% and up to 100%;

[0057] S300: Matter-element transformation and determination of classical domain and section domain; specifically including:

[0058] The risk index represented by trapezoidal fuzzy numbers is transformed into matter-element in matter-element extenics, namely, risk matter-element.

[0059] For each risk indicator, a thing element is established, where: thing (N): risk indicator itself; characteristic (C): the scope of influence of risk indicator, which can be described by the parameters a, b, c, d of trapezoidal fuzzy number; value (V): the specific value of risk indicator, i.e. trapezoidal fuzzy number itself; classical domain and section domain are used to define evaluation criteria and evaluation level; for risk assessment, the classical domain can be defined as the acceptable range of risk, and the section domain can be defined as the unacceptable range of risk;

[0060] For example, the matter element of technology compatibility can be expressed as [technology compatibility, [70, 100]]; then, define the classical domain, such as the value range of technology compatibility in the normal state is [85, 95]; at the same time, determine the node domain, such as the possible larger range of technology compatibility values ​​is [70, 100];

[0061] S400: Indicator weight calculation; the indicator weight calculation in S400 adopts a weight calculation method based on a C-OWA operator to calculate the weight of each risk indicator; the specific process includes:

[0062] S410: Expert scoring; specifically includes:

[0063] S411: Invite 5 experts with relevant field knowledge and experience to participate in the scoring; these experts should have a deep understanding of the risk assessment of the integration of the power customer service semantic understanding model;

[0064] S412: Each expert uses the fuzzy language scoring method to score the importance of each risk factor; a nine-level language assessment scale is used, and the scoring range is as follows: S0 corresponds to: very low AL, S1 corresponds to: very low VL, S2 corresponds to: low L, S3 corresponds to: medium-low ML, S4 corresponds to: medium M, S5 corresponds to: medium-high MH, S6 corresponds to: high H, S7 corresponds to: very high VH, S8 corresponds to: extremely high AH; forming a complete assessment rating system;

[0065] S420: Data sorting and sorting; specifically including:

[0066] S421: Raw data collection: Specifically, based on the experts’ scores, collect and organize the raw data of each risk factor. Each risk factor will have k score values, corresponding to the evaluation results of k experts;

[0067] S422: Data sorting; specifically, for each risk factor, re-sort its k score values ​​in descending order; this step is intended to ensure that subsequent calculations can be performed based on the sorted data to reflect the consistency of expert opinions;

[0068] S423: Numbering to form a vector; specifically: after sorting, number each score value starting from 0 to form a vector; for risk factor R1, its sorted score value vector may be expressed as [V1, V2, …, V k ], where V i represents the score of the i-th expert, and V1≥V2≥…≥V k ;

[0069] S430: weight calculation based on C-OWA operator; specifically including:

[0070] S431: C-OWA operator definition; specifically:

[0071] The C-OWA operator is an ordered average weighted operator based on the number of combinations, which allows different weights to be assigned according to the ranking position of the data;

[0072] D-is defined as a function C:In→I, where I represents a set of interval numbers and n is the number of dimensions; this function maps n interval numbers to a single interval number; specifically, for an input set of interval numbers X = ([a1, b1], [a2, b2], …, [a n ,b n ]) and the corresponding weight vector w=(w1,w2,…,w n ), where w i ≥0 and The C-OWA operator is used to calculate the final weight of each risk factor;

[0073] S432: weight calculation; according to the definition of the C-OWA operator, the C-OWA operator is applied to the score value vector of each risk factor to calculate the final weight of the risk factor; the specific calculation process involves weighted summation of the ranked score values, where the weight is determined according to the number of combinations and the ranking position; the specific process is as follows:

[0074] First, sort the input interval numbers according to the lower or upper bound of the interval number; for simplicity, we assume that the lower bound is used for non-decreasing sorting; for each sorted interval number, calculate the midpoint;

[0075] Midpoint m i The calculation formula is:

[0076]

[0077] Among them, x i and i are the lower and upper bounds of the ith interval respectively;

[0078] Then, using the weight w i The sorted midpoints are weighted averaged, and the final output is the sum of these weighted midpoints;

[0079] Mathematical expression of C-OWA operator:

[0080]

[0081] Among them, x i and i is the lower and upper bounds of the sorted ith interval number, w i is the corresponding weight;

[0082] S433: Normalization: In order to ensure that the sum of the weights of all risk factors is 1, the calculated weights are normalized; this step is necessary because it ensures the rationality and comparability of the weight values;

[0083] S500: Calculate the closeness; specifically including:

[0084] According to the principle of matter-element extenics, the closeness of each risk indicator to different risk levels is calculated;

[0085] The calculation of the closeness involves comparing the risk element with each risk level in the classical domain and the segment domain; for each risk level j, the risk element is calculated. The degree of closeness to this level K j ; calculated by the following formula:

[0086]

[0087] in, is the membership degree of the risk element at xi, is the jth risk level in x i For example, the degree of closeness of technical compatibility (indicator 1) to the “low risk” level is 0.7 (assuming it is calculated by the formula);

[0088] S600: Determine the risk level; specifically including: based on the proximity criterion, determine the risk level to which each risk indicator belongs; if the proximity of a risk indicator is the highest at a certain level, then determine that the indicator is at that risk level; for example, the proximity of technical compatibility (indicator 1) is the highest at the "low risk" level, so it is determined to be "low risk".

[0089] S700: Comprehensive risk assessment; specifically including: calculating the overall risk level using the weighted average method based on the weight of each risk indicator and the risk level to which it belongs.

[0090] For example, if technical compatibility (indicator 1) is "low risk", data integrity (indicator 2) is "medium risk", and performance stability (indicator 3) is "low risk", and the weights are 0.3, 0.4, and 0.3 respectively, then the overall risk level is 0.3 "low risk" + 0.4 "medium risk" + 0.3* "low risk", and the specific value is obtained through quantitative evaluation.

[0091] Through the above steps, the present invention successfully solves the problem of determining the weights of risk assessment indicators when multiple experts make group decisions by using the weight calculation method based on the C-OWA operator. Compared with the traditional hierarchical analysis method, this method is not only more practical and operational, but also can more accurately reflect the consistency of expert opinions, thereby improving the accuracy and reliability of risk assessment.

[0092] Next, the present invention will further use the risk assessment model based on trapezoidal fuzzy numbers and matter-element extenics, combined with the weight values ​​calculated above, to quantitatively assess the risks integrated into the electric customer service semantic understanding model. This step will provide decision makers with more comprehensive and in-depth risk assessment results based on the fuzziness and uncertainty of risk factors and the extended analysis principle of matter-element extenics.

[0093] Table 1 Nine-level fuzzy language value scale

[0094]

[0095] For example, for technical compatibility, the scores of the five experts are [8, 7, 9, 8, 8], and the calculated weight is 0.3 (assuming it is calculated by the C-OWA operator).

[0096] Result analysis and response strategies:

[0097] Conduct detailed analysis of risk assessment results to identify potential high-risk areas and key factors.

[0098] For example, data integrity (Indicator 2) was found to be a major risk factor that requires significant attention.

[0099] Develop targeted risk response strategies based on risk assessment results.

[0100] For example, in response to the risk of data integrity (Indicator 2), strategies such as strengthening data backup and optimizing data verification are formulated.

[0101] Based on actual project progress and risk assessment results, feedback is continuously collected to optimize model parameters and structure.

[0102] For example, according to the newly collected data, the value range or weight distribution of the trapezoidal fuzzy number is adjusted to improve the accuracy and reliability of risk assessment.

[0103] Example 2

[0104] The difference between this embodiment and embodiment 1 is that the specific implementation of S400 is different, and the rest is the same.

[0105] The specific process of S400 is as follows:

[0106] If there are n indicators and m experts, the specific process of calculating the weights is as follows:

[0107] Arrange each expert's rating in descending order;

[0108] Calculate the number of combinations

[0109] Generate weight vector

[0110] For each indicator, calculate its absolute weight w j (:,j)=a.*x(:,j);

[0111] Calculating relative weights

[0112] Through the above steps, the present invention uses trapezoidal fuzzy numbers and matter-element extenics to construct a comprehensive risk assessment model, which not only effectively handles the uncertainty and ambiguity in the integration process of the power customer service semantic understanding model, but also provides a scientific risk management framework and tools for decision makers.

Claims

1. A customer service semantic understanding model integrated risk identification method, characterized in that: The process includes the following: S100: Identify risk indicators; specifically, clarify all risk factors that may be involved in the integration process of the customer service semantic understanding model and form a risk indicator set; S200: Trapezoidal fuzzy number description; specifically includes: For each risk indicator, trapezoidal fuzzy numbers are used to describe its possible value range and distribution; Assumptions is a trapezoidal fuzzy number, which is defined over the real number field X, then it can be expressed as an ordered four-tuple Where a≤b≤c≤d, a and d represent the lower and upper bounds of the trapezoidal fuzzy number, respectively, and b and c represent the starting and ending points of the membership peak; Trapezoidal fuzzy number The membership function It can be defined as: The membership function is 0 when x=a and x=d, and 1 when x=b and x=c; in the interval a≤x≤b, the membership increases linearly from 0 to 1; in the interval c≤x≤d, the membership decreases linearly from 1 to 0; S300: Matter-element transformation and determination of classical domain and section domain; specifically including: The risk index represented by trapezoidal fuzzy numbers is transformed into matter-element in matter-element extenics, namely, risk matter-element. For each risk indicator, a thing element is established, where: thing (N): risk indicator itself; characteristic (C): the scope of influence of risk indicator, which can be described by the parameters a, b, c, d of trapezoidal fuzzy number; value (V): the specific value of risk indicator, i.e. trapezoidal fuzzy number itself; classical domain and section domain are used to define evaluation criteria and evaluation level; for risk assessment, the classical domain can be defined as the acceptable range of risk, and the section domain can be defined as the unacceptable range of risk; S400: indicator weight calculation; S500: Calculate the closeness; specifically including: According to the principle of matter-element extenics, the closeness of each risk indicator to different risk levels is calculated; The calculation of the closeness involves comparing the risk element with each risk level in the classical domain and the segment domain; for each risk level j, the risk element is calculated. The degree of closeness to this level K j ; calculated by the following formula: in, is the membership degree of the risk element at xi, is the jth risk level in x i The degree of membership of the place; S600: Determine the risk level; specifically, including: based on the closeness criterion, determine the risk level to which each risk indicator belongs; if the closeness of a risk indicator is the highest at a certain level, determine that the indicator is at that risk level; S700: Comprehensive risk assessment; specifically including: calculating the overall risk level using the weighted average method based on the weight of each risk indicator and the risk level to which it belongs.

2. According to claim 1, a customer service semantic understanding model integrated risk identification method is characterized in that: The indicator weight calculation in S400 adopts a weight calculation method based on a C-OWA operator to calculate the weight of each risk indicator; if there are n indicators and m experts, the specific process of calculating the weight is as follows: Arrange each expert's rating in descending order; Calculate the number of combinations Generate weight vector For each indicator, calculate its absolute weight w j (∶,j)=a.*x(∶,j); Calculating relative weights 3. According to claim 2, a customer service semantic understanding model integrated risk identification method is characterized in that: In the S400, five experts are invited to evaluate the importance of risk indicators and score them using nine-level fuzzy language to form a set of evaluation indicators.

4. According to claim 1, a customer service semantic understanding model integrated risk identification method is characterized in that: The indicator weight calculation in S400 uses a weight calculation method based on a C-OWA operator to calculate the weight of each risk indicator; The specific process includes: S410: Expert scoring; specifically includes: S411: Invite k experts with relevant field knowledge and experience to participate in the scoring; S412: Each expert uses the fuzzy language scoring method to score the importance of each risk factor; S420: Data sorting and sorting; specifically including: S421: Raw data collection; specifically: according to the expert scores, collect and organize the raw data of each risk factor; each risk factor will have k score values, corresponding to the evaluation results of k experts; S422: data sorting; specifically: for each risk factor, re-sort its k score values ​​in descending order; S423: Numbering to form a vector; specifically: after sorting, number each score value starting from 0 to form a vector; for risk factor R1, its sorted score value vector may be expressed as [V1, V2, …, V k ], where V i represents the score of the i-th expert, and V1≥V2≥…≥V k ; S430: weight calculation based on C-OWA operator; specifically including: S431: C-OWA operator definition; specifically: The C-OWA operator is an ordered average weighted operator based on the number of combinations, which allows different weights to be assigned according to the ranking position of the data; D-is defined as a function C:In→I, where I represents a set of interval numbers and n is the number of dimensions; this function maps n interval numbers to a single interval number; specifically, for an input set of interval numbers X = ([a1, b1], [a2, b2], …, [a n ,b n ]) and the corresponding weight vector w=(w1,w2,…,w n ), where w i ≥0 and The C-OWA operator is used to calculate the final weight of each risk factor; S432: weight calculation: according to the definition of the C-OWA operator, the C-OWA operator is applied to the score value vector of each risk factor to calculate the final weight of the risk factor; S433: Normalization processing: In order to ensure that the sum of the weights of all risk factors is 1, the calculated weights are normalized.

5. According to claim 4, a customer service semantic understanding model integrated risk identification method is characterized in that: In S432, the specific calculation process involves weighted summing of the ranked score values, where the weight is determined according to the number of combinations and the ranked positions; the specific process is as follows: First, sort the input interval numbers according to the lower or upper bound of the interval number; for simplicity, we assume that the lower bound is used for non-decreasing sorting; for each sorted interval number, calculate the midpoint; Midpoint m i The calculation formula is: Among them, x i and i are the lower and upper bounds of the ith interval respectively; Then, using the weight w i The sorted midpoints are weighted averaged, and the final output is the sum of these weighted midpoints; Mathematical expression of C-OWA operator: E-where x i and i is the lower and upper bounds of the sorted ith interval number, w i is the corresponding weight.

6. A customer service semantic understanding model integrated risk identification method according to any one of claims 1 to 5, characterized in that: The factors in S100 include: technical compatibility (index 1), data integrity (index 2), and performance stability (index 3).

7. A customer service semantic understanding model integrated risk identification method according to claim 3 or 4, characterized in that: The nine-level language assessment scale is used, and the scoring range is as follows: S0 Very Low (AL) S1 is very low (VL) S2 low (L) S3 Medium Low (ML) S4 Medium (M) S5 medium-high (MH) S6 High (H) S7 Very High (VH) S8 Very High (AH) Form a complete evaluation level system.

8. A customer service semantic understanding model integrated risk identification system that implements the risk identification method described in claim 1, characterized in that: include: The risk indicator identification module is used to identify all risk factors that may be involved in the integration process of the customer service semantic understanding model and form a risk indicator set; Trapezoidal fuzzy number description module, used to describe the possible value range and distribution of each risk indicator using trapezoidal fuzzy numbers; The matter-element conversion and classical domain and section domain determination unit is used to convert the risk index represented by trapezoidal fuzzy numbers into matter-element in matter-element extenics, namely, risk matter-element; An indicator weight calculation unit, used to calculate the indicator weight; The closeness calculation unit is used to calculate the closeness of each risk indicator relative to different risk levels based on the principle of matter-element extenics; The risk level unit is used to determine the risk level to which each risk indicator belongs based on the proximity criterion; If the closeness of a risk indicator is the highest at a certain level, then the indicator is judged to be at that risk level; The comprehensive risk assessment unit is used to calculate the overall risk level using the weighted average method based on the weight of each risk indicator and the risk level to which it belongs.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, a customer service semantic understanding model integrated risk identification method as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, a customer service semantic understanding model integrated risk identification method is implemented as described in any one of claims 1 to 7.

Images