An intelligent customer service system and method based on intent recognition

Abstract

The present invention relates to the field of customer service processing, and particularly to an intelligent customer service system and method based on intent recognition. The customer service system includes a response evaluation module for determining a response analysis strategy for a target to be responded to in response to a response evaluation condition; an execution analysis module for determining an execution analysis method based on a processing execution difference index and a response quality difference index to determine an execution evaluation coefficient, and determining a setting method of a response processing coefficient in response to a response setting condition; a reference analysis module for determining a setting method of a response processing coefficient for a secondary response target in response to a target reference condition; and a service processing module connected to the execution analysis module and the reference analysis module for periodically determining whether to schedule a target to be responded to based on the node processing coefficients of each target service node. The present invention improves the uniformity of the customer service processing task allocation result.

Description

Technical Field

[0001] The present invention relates to the field of customer service processing, and particularly to an intelligent customer service system and method based on intent recognition. Background Art

[0002] Determining the inquiry intent of a customer service object based on intent recognition can provide effective guidance for subsequent customer service and task allocation, thereby ensuring the overall efficiency of customer service processing. However, when faced with electronic devices or other types of technical devices, there are differences in the degree of cognition among different service objects. Existing intelligent customer service service solutions based on intent recognition often only make intent judgments based on the information provided by the service object to allocate processing tasks, without considering the differences in the actual processing efficiency caused by the different cognitive situations of the service objects regarding the devices being inquired about, resulting in uneven task allocation during the actual customer service process. Therefore, how to determine targeted analysis means based on the cognitive situation of the service object regarding the inquired device to ensure the accuracy of the estimated result of the processing difficulty of the service object is an urgent problem for those skilled in the art.

[0003] Chinese Patent Publication No. CN112784039A discloses a method, device, and storage medium for allocating online customer service. Specifically, it obtains the current user consultation question. Secondly, it inputs the user consultation question into a pre-trained intent prediction model to generate an intent classification label. Among them, the intent prediction model is trained based on the user's order information, operation information, and the intent classification label determined by the user's first user consultation question on the consultation page. Finally, according to the intent classification label, it determines the type of consultation entrance and allocates the corresponding online customer service to the user. The above solution has the following problems: It fails to consider the impact of the differences in the cognitive situations of different users regarding the consultation products on the processing process of the consultation tasks, resulting in relatively poor uniformity of the task allocation results. Summary of the Invention

[0004] To this end, the present invention provides an intelligent customer service system and method based on intent recognition to overcome the problem in the prior art that the impact of the differences in the cognitive situations of different users regarding the consultation products on the customer service processing difficulty is not considered, resulting in relatively poor uniformity of the task allocation results.

[0005] To achieve the above object, the present invention provides an intelligent customer service system based on intent recognition, including:

[0006] A response evaluation module, which is used to execute analysis or information reference analysis with the goal of determining the response analysis strategy corresponding to each target to be responded to in response to the response evaluation condition;

[0007] An execution analysis module, which is connected to the response evaluation module, is used to determine an execution analysis method based on a processing execution difference index and a response quality difference index to obtain an execution evaluation coefficient, and in response to a response setting condition, to determine the setting method of the response processing coefficient as determining the execution evaluation coefficient based on an information association index and an associated reference execution index, or determining the execution evaluation coefficient based on an information association index and an analysis interference index;

[0008] A reference analysis module, which is connected to the response evaluation module, is used to determine, in response to a target reference condition, the setting method of the response processing coefficient for a type-two response target as based on an information pointing parameter and a pointing execution parameter, or a target execution coefficient and an information pointing parameter to determine the response processing coefficient for the type-two response target;

[0009] A service processing module, which is respectively connected to the execution analysis module and the reference analysis module, is used to periodically determine whether to schedule a target to be responded based on the node processing coefficients of each target service node.

[0010] Further, if the response evaluation condition responded by the response evaluation module is that there is a type-one response target, it is determined that the execution analysis module performs target execution analysis on the type-one response target, and the target execution analysis includes:

[0011] The execution analysis module responds to the execution analysis condition and determines an execution analysis method based on the processing execution difference index and the response quality difference index of the type-one response target;

[0012] The execution analysis module responds to the execution evaluation completion condition and determines the setting method of the response processing coefficient for the type-one response target based on the execution evaluation coefficient;

[0013] The execution analysis condition is that the response evaluation module determines that the execution analysis module performs target execution analysis on any type-one response target, and the execution evaluation completion condition is that the execution evaluation coefficient of the type-one response target is determined;

[0014] The type-one response target is a target to be responded with a statement standard index greater than a preset statement standard index or a reference processing execution coefficient greater than a preset reference processing execution coefficient.

[0015] Further, if the analysis setting condition responded by the execution analysis module is that the execution stability coefficient is greater than a preset execution stability coefficient, it is determined to determine the execution evaluation coefficient of the type-one response target based on the information association index and the associated reference execution index;

[0016] The execution evaluation coefficient has a positive correlation with the information association index and the associated reference execution index respectively, and the execution stability coefficient is determined according to the processing execution difference index and the response quality difference index of the type-one response target.

[0017] Further, the analysis setting condition responded by the execution analysis module is that the execution stability coefficient is less than or equal to the preset execution stability coefficient, then it is determined to determine the execution evaluation coefficient of a type of response target based on the information association index and the analysis interference index, and it is determined whether to adjust the execution evaluation coefficient based on the relevant execution parameters;

[0018] The evaluation adjustment condition responded by the execution analysis module is that the relevant execution parameter is greater than the preset relevant execution parameter, then it is determined to increase and adjust the execution evaluation coefficient according to the relevant execution parameter.

[0019] Further, the response setting condition responded by the execution analysis module is that the execution evaluation coefficient is greater than the preset execution evaluation coefficient, then it is determined to determine the response processing coefficient of a type of response target based on the execution evaluation coefficient and the associated reference execution index;

[0020] The response setting condition responded by the execution analysis module is that the execution evaluation coefficient is less than or equal to the preset execution evaluation coefficient, then it is determined to determine the response processing coefficient of a type of response target based on the execution evaluation coefficient and the execution stability coefficient.

[0021] Further, the response evaluation condition responded by the response evaluation module is that there are type-two response targets, then it is determined that the reference analysis module performs information reference analysis on the type-two response targets. The information reference analysis includes:

[0022] The reference analysis module responds to the reference analysis condition and determines the setting method of the response processing coefficient of the type-two response target based on the information reference index and the information overlap index of the type-two response target;

[0023] The reference analysis condition is that the response evaluation module determines that the reference analysis module performs information reference analysis on any type-two response target;

[0024] The type-two response target is a to-be-responded target whose expression standard index is less than or equal to the preset expression standard index and the reference processing execution coefficient is less than or equal to the preset reference processing execution coefficient.

[0025] Further, the target reference condition responded by the reference analysis module is that the information reference index is greater than the preset information reference index or the information overlap index is greater than the preset information overlap index, then the response processing coefficient of the type-two response target is determined based on the information pointing parameter and the pointing execution parameter;

[0026] The target reference condition responded by the reference analysis module is that the information reference index is less than or equal to the preset information reference index and the information overlap index is less than or equal to the preset information overlap index, then the response processing coefficient of the type-two response target is determined based on the target execution coefficient and the information pointing parameter.

[0027] Further, the service analysis module periodically responds to the node processing difference coefficient of the target scheduling condition to determine whether to perform task scheduling for the target to be replied of the critical service node;

[0028] If the target scheduling condition to which the service analysis module responds is that the node processing difference coefficient is greater than the preset node processing difference coefficient, it is determined that task scheduling is performed for the scheduling reply target of the critical service node;

[0029] The node processing difference coefficient is determined based on the node processing coefficients of each target service node.

[0030] Further, the service analysis module responds to the scheduling reply coefficient of the scheduling setting condition to determine the scheduling reply target of each critical service node, and the scheduling reply coefficient is determined according to the reply processing coefficient and the analysis related index;

[0031] If the scheduling setting condition to which the service analysis module responds is that the scheduling reply coefficient of the target to be replied is greater than the preset scheduling reply coefficient, it is determined that the target to be replied is the scheduling reply target;

[0032] The service analysis module responds to the target scheduling condition and determines the transfer service node of each scheduling reply target based on the analysis related index, and the target scheduling condition is that the determination of the scheduling reply target of each critical service node is completed.

[0033] The present invention also provides a method applied to the intelligent customer service system based on intention recognition described above, including:

[0034] Determine the reply analysis strategy of the target to be replied as target execution analysis or information reference analysis for the target to be replied according to the expression standard index and the reference processing execution coefficient;

[0035] Under the target execution analysis, determine the execution analysis method of a type of reply target according to the processing execution difference index and the reply quality difference index, and the execution analysis method is to determine the execution evaluation coefficient of a type of reply target based on the information association index and the associated reference execution index, or, the information association index and the analysis interference index;

[0036] Under the condition that the execution evaluation is completed, determine the setting method of the reply processing coefficient of a type of reply target based on the execution evaluation coefficient as to determine the reply processing coefficient of a type of reply target based on the execution evaluation coefficient and the associated reference execution index, or, the execution evaluation coefficient and the execution stability coefficient;

[0037] Under the information reference analysis, determine the setting method of the reply processing coefficient of the second type of reply target according to the information reference index and the information overlap index as to determine the reply processing coefficient of the second type of reply target based on the information pointing parameter and the pointing execution parameter, or, the target execution coefficient and the information pointing parameter;

[0038] Periodically determine whether to perform task scheduling for the target to be replied of the target service node according to the node processing difference coefficient, determine the scheduling reply targets of each key service node based on the scheduling reply coefficient, and determine the transfer service nodes of each scheduling reply target based on the analysis correlation index.

[0039] Compared with the prior art, the beneficial effect of the present invention is that the technical solution of the present invention determines a targeted response analysis strategy according to the expression standard index of the target to be replied and the reference processing execution coefficient, so that the analysis process of the target to be replied takes into account the differences in the understanding of the problems faced by different targets to be replied, avoiding the influence on the response processing coefficient due to the understanding situation, and the present invention improves the effectiveness of the response processing coefficients of each target to be replied determined, and further improves the uniformity of the customer service processing task allocation result.

[0040] Further, in the present invention, the expression standard index and the reference processing execution coefficient are used to characterize the effectiveness of the problem description and the execution ability of the problem-solving measures of the target to be replied in the present and the past for the problems it faces, so as to determine the understanding degree of different targets to be replied for the questioned items, provide effective support for the subsequent determination of the targeted response analysis strategy, and further ensure the adaptation degree of the set response analysis strategy to the actual analysis process.

[0041] Further, in the present invention, for a type of response target, the execution stability coefficient is determined according to the processing execution difference index and the response quality difference index of the type of response target, and the execution analysis method is determined according to the execution stability coefficient. Based on the execution stability coefficient, the execution analysis method is determined, avoiding the differences in the processing records of the response targets with better understanding, and further improving the accuracy of the determination result of the response processing coefficient for a type of response target, and ensuring the uniformity of the customer service processing task allocation result.

[0042] Further, in the present invention, for the second type of response target, the setting method of the response processing coefficient of the second type of response target is determined according to the information reference index and the information overlap index, that is, the targeted setting method is determined based on the difficulty of inferring the root cause of the failure from the initial service information of the second type of response target, making the determination method of the response processing coefficient more in line with the actual situation of the second type of response target, and the present invention ensures the accuracy of the determined response processing coefficient.

[0043] Furthermore, in the present invention, it is determined whether to perform task scheduling for the to-be-answered targets of critical service nodes according to the node processing difference coefficient between target service nodes, and the scheduled response targets and their corresponding transfer service nodes are determined according to the response processing coefficient and the analysis correlation index, so as to ensure the uniformity of the processing intensity and the consumption time of each target service node. The present invention improves the overall efficiency in the process of processing the to-be-answered targets. Description of the Drawings

[0044] Figure 1 It is a module connection diagram of the intelligent customer service system based on intention recognition of the present invention;

[0045] Figure 2 It is a flowchart of the execution analysis module of the present invention for determining the execution analysis method in response to the analysis setting conditions;

[0046] Figure 3 It is a flowchart of the service analysis module of the present invention for determining whether to perform task scheduling for the to-be-answered targets of target service nodes in response to the target scheduling conditions;

[0047] Figure 4 It is a schematic diagram of the intelligent customer service method based on intention recognition of the present invention. Detailed Embodiments

[0048] In order to make the objectives and advantages of the present invention clearer and more understandable, the present invention will be 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.

[0049] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0050] It should be noted that in the description of the present invention, the terms indicating the direction or position relationship such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the direction or position relationship shown in the drawings. This is only 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, and therefore cannot be understood as a limitation of the present invention.

[0051] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and defined, the terms "installation", "connection", and "linkage" 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 an indirect connection through an intermediate medium, and it can be the communication inside 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 situations.

[0052] Please refer to Figures 1 to 3 as shown, the present invention provides an intelligent customer service system based on intention recognition, including:

[0053] A response evaluation module, which is used to respond to the response evaluation condition and perform analysis or information reference analysis with the goal of determining the response analysis strategy corresponding to each target to be responded.

[0054] An execution analysis module, which is connected to the response evaluation module and is used to determine the execution analysis method based on the processing execution difference index and the response quality difference index to obtain the execution evaluation coefficient, and respond to the response setting condition to determine the setting method of the response processing coefficient as determining the execution evaluation coefficient based on the information association index and the associated reference execution index, or determining the execution evaluation coefficient based on the information association index and the analysis interference index;

[0055] A reference analysis module, which is connected to the response evaluation module and is used to respond to the target reference condition to determine the setting method of the response processing coefficient of the second-class response target as based on the information pointing parameter and the pointing execution parameter, or the target execution coefficient and the information pointing parameter to determine the response processing coefficient of the second-class response target;

[0056] A service processing module, which is respectively connected to the execution analysis module and the reference analysis module and is used to periodically determine whether to schedule the target to be responded based on the node processing coefficient of each target service node.

[0057] Among them, the present invention is used to arrange and schedule customer service for a specific category of products. During the arrangement and scheduling process, the product category for which customer service is provided is recorded as the target service category. If a service request from any product purchaser is received, then the product purchaser is recorded as the target to be responded, and analysis is performed on the target to be responded to determine its response processing coefficient. The response analysis strategy used is determined according to the expression standard index of the target to be responded and the reference processing execution coefficient. There are also several target service nodes in the present invention to provide customer service for each target to be responded. The target service categories in the present invention include but are not limited to: smart phones, smart wearable devices, smart home systems, and laptop computers;

[0058] In the present invention, several service analysis records are applied. Any one of the service analysis records records at least one expression standard index, reference processing execution coefficient, execution stability coefficient, associated determination quantity, relevant execution parameter, information reference index, information overlap index, node processing difference coefficient, and scheduling response coefficient during the allocation process of the processing task for the target to be replied. And each service analysis record corresponds to a qualified mark, which records whether the uniformity of the customer service processing task allocation result meets the user's requirements. It can be understood that the user can determine whether the uniformity of the customer service processing task allocation result meets the requirements according to the self-set index. For example, the self-set index can be, but is not limited to, the service response difference parameter, and the service response difference parameter is the absolute value of the difference between the maximum value and the minimum value of the number of remaining targets to be replied of the target service node.

[0059] Specifically, the response evaluation condition for the response evaluation module is that there is a type of response target, then it is determined that the execution analysis module performs target execution analysis on a type of response target, and the target execution analysis includes:

[0060] The execution analysis module responds to the execution analysis condition and determines the execution analysis method based on the processing execution difference index and the response quality difference index of a type of response target;

[0061] The execution analysis module responds to the execution evaluation completion condition and determines the setting method of the response processing coefficient of a type of response target based on the execution evaluation coefficient;

[0062] The execution analysis condition is that the response evaluation module determines that the execution analysis module performs target execution analysis on any type of response target, and the execution evaluation completion condition is that the execution evaluation coefficient of a type of response target is determined;

[0063] The type of response target is a target to be replied with an expression standard index greater than the preset expression standard index or a reference processing execution coefficient greater than the preset reference processing execution coefficient.

[0064] Among them, historical service records are applied in the present invention, which contain records of each service process of each target service node, and each historical service record contains the initial service information corresponding to each service process and the finally determined problem location. Detect the expression standard index of the initial service information in each service process of the historical service record for the target to be answered. The historical service record is a set of several customer service processes for the target to be answered. The target service categories corresponding to each service process in the historical service record are the same. The text information used to describe the service requirements of the target to be answered provided for the first time after the target to be answered sends a service request in each service process is recorded as the initial service information. For a single initial service information, the expression standard index is the number of professional phrases of the target service category in the initial service information. The reference processing execution coefficient is the average value of the execution duration of the solution provided by the target service node by the target to be answered in each service process. For a single service process, the execution duration is the interval duration between the moment when the target service node provides the solution and the moment when the target to be answered ends the service process. How to determine the professional phrases of each target service category is content that can be easily understood by those skilled in the art. For example, if the target service category is a smart home system, its corresponding professional phrases include, but are not limited to: ZigBee protocol, gateway device, scene mode, automation rule, and device linkage;

[0065] If the expression standard index of the target to be answered is greater than the preset expression standard index or the reference processing execution coefficient is greater than the preset reference processing execution coefficient, it is determined that target execution analysis is to be performed on the target to be answered whose expression standard index is greater than the preset expression standard index or the reference processing execution coefficient is greater than the preset reference processing execution coefficient. The values of the preset expression standard index and the preset reference processing execution coefficient can be determined by the user according to the actual working scenario. For example, the user can set them according to the service analysis record. The higher the user's requirement for the uniformity of the customer service processing task allocation result, the smaller the value of the preset expression standard index and the smaller the value of the preset reference processing execution coefficient. Provide a method for obtaining the value of the preset expression standard index. The service analysis record for information reference analysis of the target to be answered is recorded as the evaluation reference record. The maximum value of the expression standard index in the evaluation reference record that meets the user's requirement for the uniformity of the customer service processing task allocation result is recorded as the preset expression standard index. Provide a method for obtaining the value of the preset reference processing execution coefficient. The maximum value of the reference processing execution coefficient in the evaluation reference record that meets the user's requirement for the uniformity of the customer service processing task allocation result is recorded as the preset reference processing execution coefficient.

[0066] Specifically, the analysis setting condition responded by the execution analysis module is that the execution stability coefficient is greater than the preset execution stability coefficient, and it is determined to evaluate the execution coefficient of a type of response target based on the information association index and the associated reference execution index;

[0067] The execution evaluation coefficient is positively correlated with the information association index and the associated reference execution index respectively, and the execution stability coefficient is determined according to the processing execution difference index and the response quality difference index of a type of response target.

[0068] Among them, for a single type of response target, the execution stability coefficient = 1 / (response quality difference index + processing execution difference index), and the response quality difference index , where n is the number of service processes in which this type of response target exists in the historical service record, yi is the expression standard index corresponding to the i-th service process of this type of response target in the historical service record, y0 is the average value of the expression standard indexes corresponding to each service process of this type of response target in the historical service record, and the processing execution difference index , where di is the execution duration corresponding to the i-th service process of this type of response target in the historical service record, and d0 is the average value of the execution durations corresponding to each service process of this type of response target in the historical service record;

[0069] For the value of the preset execution stability coefficient, the user can determine it according to the actual working scenario. For example, the user can set it according to the service analysis record. The higher the user's requirement for the uniformity of the customer service task allocation result, the larger the value of the preset execution stability coefficient. A method for obtaining the value of the preset execution stability coefficient is provided. The service analysis record for determining the execution evaluation coefficient of a type of response target based on the information association index and the associated reference execution index is recorded as the execution reference record, and the minimum value of the execution stability coefficient in the execution reference records that meet the user's requirement for the uniformity of the customer service task allocation result is recorded as the preset execution stability coefficient;

[0070] If the execution stability coefficient of a type of response target is greater than the preset execution stability coefficient, it indicates that the overall understanding of the target processing category by this type of response target is good and remains stable all the time, that is, it indicates that the understanding of this type of response target is relatively comprehensive. Therefore, it is necessary to evaluate the processing difficulty of the initial service information that needs to be processed by the type of response target currently, and determine the execution evaluation coefficient of the type of response target based on the information association index and the associated reference execution index. For a single type of response target, the execution evaluation coefficient = ln (information association index × associated reference execution index). The information association index is the number of associated problem locations corresponding to the professional phrases included in the initial service information of this type of response target. The associated reference execution index is the average value of the execution processing durations of the associated problem locations corresponding to the professional phrases included in the initial service information of this type of response target. For a single problem location of the target processing category, if the associated determination number of this problem location is greater than the preset associated determination number, then this problem location is recorded as the associated problem location of this type of response target. The associated determination number is the number of professional phrases jointly included in the initial service information involved in this problem location and the initial service information of this type of response target. If the fault factor finally determined in a service process in the historical service record is this problem location, it is determined that this problem location is involved in the initial service information corresponding to this service process. For a single associated problem location, the execution processing duration is the average value of the processing durations of each service process determined as this associated problem location after analysis in the historical service record. For a single service process, the processing duration is the interval duration between the moment when the service request of the corresponding target to be answered is received and the end moment of this service process. The problem location is the possible fault factor of the target processing category. How to determine the existing fault factor corresponding to the target processing category is easy for those skilled in the art to understand. For example, if the target processing category is a smart home system, the problem locations include but are not limited to: power adapter damage, power cord looseness or poor contact, internal power module failure, motherboard failure, memory failure, system file damage, network connection interruption, signal interference or attenuation, actuator failure, firmware version mismatch, and too high or too low temperature;

[0071] The value of the preset associated determination number can be determined by the user according to the actual working scenario. For example, the user can set it according to the service analysis record. The higher the user's requirement for the uniformity of the service analysis record of the customer service processing task allocation result, the larger the value of the preset associated determination number. A method for obtaining the value of the preset associated determination number is provided. The average value of the associated determination numbers of each associated problem location in the service analysis record that meets the user's requirement for the uniformity of the customer service processing task allocation result is recorded as the preset associated determination number.

[0072] Specifically, the analysis setting condition responded by the execution analysis module is that the execution stability coefficient is less than or equal to the preset execution stability coefficient, then it is determined to determine the execution evaluation coefficient of a type of response target based on the information association index and the analysis interference index, and it is determined whether to adjust the execution evaluation coefficient based on the relevant execution parameters;

[0073] The evaluation adjustment condition responded by the execution analysis module is that the relevant execution parameter is greater than the preset relevant execution parameter, then it is determined to increase and adjust the execution evaluation coefficient according to the relevant execution parameter.

[0074] Among them, if the execution stability coefficient of a type of response target is less than or equal to the preset execution stability coefficient, it indicates that although this type of response target has a good overall understanding of the target processing category, there is a one-sided lack of clarity in the understanding of this type of response target, which will cause interference to the processing process and information expression. Therefore, it is necessary to further analyze the initial service information for this type of response target to determine the understanding of this type of response target for the currently provided initial service information, to determine the execution evaluation coefficient of this type of response target, determine the execution evaluation coefficient of this type of response target based on the information association index and the analysis interference index, and determine whether it is necessary to adjust the execution evaluation coefficient according to the relevant execution parameters;

[0075] For a type of response target with a single execution stability coefficient less than or equal to the preset execution stability coefficient, the initially determined execution evaluation coefficient of this type of response target = ln (information association index × analysis interference index), the analysis interference index = the number of interference problem positions of this type of response target / the number of associated problem positions of this type of response target, the interference problem position is the associated problem position with the associated determination quantity greater than the preset interference index, the relevant execution parameter = the number of relevant execution phrase groups of this type of response target / the information association index of this type of response target. The professional phrase groups involved in the initial service information of the service process with the execution duration greater than the average value of the execution durations corresponding to each service process in the historical service record of this type of response target are recorded as relevant execution phrase groups. If the relevant execution parameter of this type of response target is greater than the preset relevant execution parameter, then increase and adjust the initially determined execution evaluation coefficient according to the relevant execution parameter. The increase value of the initially determined execution evaluation coefficient has a positive correlation with the relevant execution parameter, and determine the setting method of the response coefficient of this type of response target according to the execution evaluation coefficient after the increase adjustment. If the relevant execution parameter of this type of response target is less than or equal to the preset relevant execution parameter, then do not adjust the initially determined execution evaluation coefficient, that is, determine the setting method of the response coefficient of this type of response target according to the initially determined execution evaluation coefficient;

[0076] The values of the preset interference index and the preset related execution parameters can be determined by the user according to the actual working scenario. For example, the user can set them according to the service analysis record. The higher the user's requirement for the uniformity of the distribution result of the customer service processing tasks, the larger the value of the preset interference index and the smaller the value of the preset related execution parameters. A method for obtaining the value of the preset interference index is provided. The minimum value of the associated determination quantity of each interference problem location in the service analysis record that meets the user's requirement for the uniformity of the distribution result of the customer service processing tasks is recorded as the preset interference index. A method for obtaining the value of the preset related execution parameters is provided. The service analysis record in which the execution evaluation coefficient is increased and adjusted according to the related execution parameters is recorded as the adjustment reference record. The minimum value of the related execution parameters of a type of response target in the adjustment reference record that meets the user's requirement for the uniformity of the distribution result of the customer service processing tasks is recorded as the preset related execution parameters.

[0077] Specifically, when the response setting condition of the execution analysis module is that the execution evaluation coefficient is greater than the preset execution evaluation coefficient, it is determined that the response processing coefficient of a type of response target is determined based on the execution evaluation coefficient and the associated reference execution index;

[0078] When the response setting condition of the execution analysis module is that the execution evaluation coefficient is less than or equal to the preset execution evaluation coefficient, it is determined that the response processing coefficient of a type of response target is determined based on the execution evaluation coefficient and the execution stability coefficient.

[0079] Among them, for a single type of response target, if the execution evaluation coefficient of this type of response target is greater than the preset execution evaluation coefficient, the response processing coefficient of this type of response target is determined based on the execution evaluation coefficient and the associated reference execution index, and the response processing coefficient = ln (execution evaluation coefficient × associated reference execution index); if the execution evaluation coefficient of this type of response target is less than or equal to the preset execution evaluation coefficient, the response processing coefficient of this type of response target is determined based on the execution evaluation coefficient and the execution stability coefficient, and the response processing coefficient = ln (execution evaluation coefficient / execution stability coefficient).

[0080] Specifically, when the response evaluation condition of the response evaluation module is that there is a second type of response target, it is determined that the reference analysis module performs information reference analysis on the second type of response target. The information reference analysis includes:

[0081] The reference analysis module responds to the reference analysis condition and determines the setting method of the response processing coefficient of the second type of response target based on the information reference index and the information overlap index of the second type of response target;

[0082] The reference analysis condition is that the response evaluation module determines that the reference analysis module performs information reference analysis on any second type of response target;

[0083] The second - type response target is a target to be responded, whose expression standard index is less than or equal to the preset expression standard index and whose reference processing execution coefficient is less than or equal to the preset reference processing execution coefficient.

[0084] Among them, if there is a target to be responded whose expression standard index is less than or equal to the preset expression standard index and whose reference processing execution coefficient is less than or equal to the preset reference processing execution coefficient, then information reference analysis is carried out for the target to be responded whose expression standard index is less than or equal to the preset expression standard index and whose reference processing execution coefficient is less than or equal to the preset reference processing execution coefficient. For a single second - type response target, the information reference index is the number of analysis keywords included in the initial service information of this second - type response target. The information overlap index = the number of overlapping keywords of this second - type response target / the information reference index of this second - type response target. For a single analysis keyword, if the overlap index of this analysis keyword is greater than the preset overlap index, then it is determined that this analysis keyword is an overlapping keyword. The overlap index is the number of problem - location involvements simultaneously related to this analysis keyword and other analysis keywords included in the initial service information of this second - type response target. The value of the preset overlap index can be determined by the user according to the actual working scenario. For example, the user can set it according to the service analysis record. The higher the user's requirement for the uniformity of the customer service task assignment result, the larger the value of the preset overlap index. A method for obtaining the value of the preset overlap index is provided. The average value of the overlap indices of each overlapping keyword in the service analysis records that meet the user's requirement for the uniformity of the customer service task assignment result is recorded as the preset overlap index. The user can set the analysis keywords according to the actual working scenario. The analysis keyword is a phrase that can be used to judge the problem - location of the initial service information. How to set the analysis keywords according to the actual working scenario is easily understood by those skilled in the art. For example, if the target processing category is a smart home system, the analysis keywords include but are not limited to: the power indicator light is not on, unable to connect, weak Wi - Fi signal, slow response, unable to control;

[0085] The values of the preset information reference index and the preset information overlap index can be determined by the user according to the actual working scenario. For example, the user can set them according to the service analysis record. A method for obtaining the value of the preset information reference index is provided. The service analysis records for determining the response processing coefficient of the second - type response target based on the target execution coefficient and the information - pointing parameter are recorded as the setting reference records. The maximum value of the information reference index in the setting reference records that meet the user's requirement for the uniformity of the customer service task assignment result is recorded as the preset information reference index. A method for obtaining the value of the preset information overlap index is provided. The maximum value of the information overlap index in the setting reference records that meet the user's requirement for the uniformity of the customer service task assignment result is recorded as the preset information overlap index.

[0086] Specifically, when the target reference conditions responded by the reference analysis module are that the information reference index is greater than the preset information reference index or the information overlap index is greater than the preset information overlap index, the response processing coefficient of the secondary response target is determined based on the information pointing parameter and the pointing execution parameter;

[0087] When the target reference conditions responded by the reference analysis module are that the information reference index is less than or equal to the preset information reference index and the information overlap index is less than or equal to the preset information overlap index, the response processing coefficient of the secondary response target is determined based on the target execution coefficient and the information pointing parameter.

[0088] Among them, for a single secondary response target, if the information reference index of this secondary response target is greater than the preset information reference index or the information overlap index is greater than the preset information overlap index, it indicates that there is relatively rich content in the initial service information of this secondary response target for inferring the root cause of the failure corresponding to the secondary response target, or there is an overlap in the information content pointing to the root cause of the failure, and further indicates that it is relatively easy to determine the root cause of the failure of this secondary response target. At this time, the response processing coefficient is determined according to the information pointing parameter and the pointing execution parameter of this secondary response target, and the response processing coefficient = ln (information pointing parameter × pointing execution parameter). The information pointing parameter is the sum of the information reference index and the information overlap index of this secondary response target. The pointing execution parameter is the execution processing duration for locating the pointing problem of this secondary response target. For a single pointing problem location, the execution processing duration is the average value of the processing durations of each service process determined as this pointing problem location after analysis in the historical service record. The pointing problem location is the problem location involved in the overlapping keyword of this secondary response target. If any overlapping keyword exists in the initial service information of any service process in the historical service record, it is determined that the failure location corresponding to this service process is the pointing problem location;

[0089] For a single secondary response target, if the information reference index of this secondary response target is less than or equal to the preset information reference index and the information overlap index is less than or equal to the preset information overlap index, it indicates that the content included in this secondary response target has a poor directivity for the root cause of the failure. At this time, the response processing coefficient is determined according to the target execution coefficient and the information pointing parameter of this secondary response target, and the response processing coefficient = ln (target execution coefficient × information pointing parameter). The target execution coefficient is the sum of the expression standard index and the reference processing execution coefficient of this secondary response target.

[0090] Specifically, the service analysis module periodically responds to the node processing difference coefficient of the target scheduling condition to determine whether to perform task scheduling for the to-be-responded targets of the key service nodes;

[0091] If the target scheduling condition responded by the service analysis module is that the node processing difference coefficient is greater than the preset node processing difference coefficient, it is determined that task scheduling is performed for the scheduling response target of the key service node;

[0092] The node processing difference coefficient is determined based on the node processing coefficients of each target service node.

[0093] Among them, in the present invention, a cyclic response scheduling period is applied, and the duration of the response scheduling period can be determined by the user himself. The higher the user's requirement for the uniformity of the distribution result of the customer service processing task, the shorter the duration of the response scheduling period. A duration of the response scheduling period is provided. The response scheduling period is 15 minutes. At the end of each response scheduling period, it is determined whether to schedule the scheduling response target of the key service node according to the node processing difference coefficient. If the node processing difference coefficient is greater than the preset node processing difference coefficient, it is determined that task scheduling is performed for the scheduling response target of the key service node. If the node processing difference coefficient is less than or equal to the preset node processing difference coefficient, it is determined that task scheduling is not performed for the scheduling response target of the key service node;

[0094] The node processing difference coefficient , where m is the number of target service nodes, kt is the cumulative response processing coefficient of the t-th target service node, and k0 is the average value of the cumulative response processing coefficients of each target service node. For a single target service node, the cumulative response processing coefficient is the sum of the response processing coefficients of each target to be responded that the target service node needs to process at the current moment. The value of the preset node processing difference coefficient can be determined by the user according to the actual working scenario. For example, the user can set it according to the service analysis record. The higher the user's requirement for the uniformity of the distribution result of the customer service processing task, the smaller the value of the preset node processing difference coefficient. A method for obtaining the value of the preset node processing difference coefficient is provided. The service analysis record for task scheduling of the target to be responded for the target service node is recorded as the scheduling reference record, and the minimum value of the node processing difference coefficient in the scheduling reference record that meets the user's requirement for the uniformity of the distribution result of the customer service processing task is recorded as the preset node processing difference coefficient.

[0095] Specifically, the service analysis module responds to the scheduling response coefficient of the scheduling setting condition to determine the scheduling response target of each key service node, and the scheduling response coefficient is determined according to the response processing coefficient and the analysis related index;

[0096] If the scheduling response coefficient of the service analysis module for the scheduling setting condition is that there is a target to be responded whose scheduling response coefficient is greater than the preset scheduling response coefficient, it is determined that the target to be responded is the scheduling response target;

[0097] The service analysis module responds to the target scheduling condition and determines the transfer service nodes of each scheduling response target based on the analysis-related index. The target scheduling condition is that each key service node has completed the determination of the scheduling response target.

[0098] Among them, the key service node is the target service node whose cumulative response processing coefficient is greater than the equilibrium processing coefficient. The equilibrium processing coefficient is the average value of the cumulative response processing coefficients of each target service node. For a single key service node, the scheduling response coefficient of each pending response target of the key service node is determined according to the response processing coefficient and the analysis-related index. The scheduling setting condition for which the service analysis module responds is that there is a pending response target whose scheduling response coefficient is greater than the preset scheduling response coefficient, then it is determined that the pending response target with a scheduling response coefficient greater than the preset scheduling response coefficient is the scheduling response target. For a single pending response target, the scheduling response coefficient = the response processing coefficient of this pending response target / the analysis-related index of this pending response target. The analysis-related index is the number of overlapping professional phrases existing in the initial service information of this pending response target. For any professional phrase included in the initial service information of this pending response target, if the number of pending response targets in the initial service information of this key service node that include this professional phrase is greater than the preset overlapping number, then this professional phrase is recorded as an overlapping professional phrase. For a single scheduling response target, the transfer priority coefficient of each pending transfer service node is determined according to the analysis-related index. The pending transfer service node is the target service node whose cumulative response processing coefficient is less than or equal to the equilibrium processing coefficient. The transfer priority coefficient and the analysis-related index are in a positive correlation relationship. The pending transfer service node with the largest transfer priority coefficient is recorded as the transfer service node of this scheduling response target, that is, this scheduling response target is transferred to this transfer service node to complete the customer service;

[0099] The values of the preset overlapping number and the preset scheduling response coefficient can be determined by the user according to the actual working scenario. For example, the user can set according to the service analysis record. The higher the user's requirement for the uniformity of the customer service processing task allocation result, the larger the value of the preset overlapping number and the smaller the value of the preset scheduling response coefficient. A value of the preset overlapping number is provided, and the value of the preset overlapping number is 50% of the number of pending response targets in the target service node. A method for obtaining the value of the preset scheduling response coefficient is provided. The average value of the scheduling response coefficients of the scheduling response targets in the service analysis record that meets the user's requirement for the uniformity of the customer service processing task allocation result is recorded as the preset scheduling response coefficient.

[0100] Please refer to Figure 4 shown, which is a schematic diagram of the intelligent customer service method based on intent recognition of the present invention. The present invention also provides an intelligent customer service method based on intent recognition, including:

[0101] According to the expression standard index and the reference processing execution coefficient, the response analysis strategy for the target to be answered is to perform target execution analysis or information reference analysis on the target to be answered;

[0102] Under the target execution analysis, according to the processing execution difference index and the response quality difference index, determine the execution analysis method for a class of response targets. The execution analysis method is to determine the execution evaluation coefficient for a class of response targets based on the information association index and the associated reference execution index, or the information association index and the analysis interference index;

[0103] Under the condition that the execution evaluation is completed, the setting method of the response processing coefficient for a class of response targets based on the execution evaluation coefficient is to determine the response processing coefficient for a class of response targets based on the execution evaluation coefficient and the associated reference execution index, or the execution evaluation coefficient and the execution stability coefficient;

[0104] Under the information reference analysis, according to the information reference index and the information overlap index, determine the setting method of the response processing coefficient for the second class of response targets. The setting method is to determine the response processing coefficient for the second class of response targets based on the information pointing parameter and the pointing execution parameter, or the target execution coefficient and the information pointing parameter;

[0105] Periodically determine whether to perform task scheduling on the targets to be answered of the target service node according to the node processing difference coefficient, determine the scheduling response targets of each key service node based on the scheduling response coefficient, and determine the transfer service nodes of each scheduling response target based on the analysis correlation index.

[0106] So far, the technical solution of the present invention has been described in combination with the preferred embodiments shown in the 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 all fall within the protection scope of the present invention.

[0107] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, 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. An intelligent customer service system based on intent recognition, characterized in that, Including: A response evaluation module, which is used to respond to the response evaluation condition and determine the response analysis strategy corresponding to each target to be responded according to the expression standard index and the reference processing execution coefficient in the historical service record. The response analysis strategy is target execution analysis or information reference analysis; The expression standard index is the number of professional phrases of the target service category in the initial service information, and the reference processing execution coefficient is the average value of the execution duration of the solution provided by the target service node for the target to be responded during each service process. The expression standard index and the reference processing execution coefficient are used to characterize the effectiveness of the target to be responded in problem description and the execution ability for problem-solving measures in the current and historical service records; An execution analysis module, which is connected to the response evaluation module, and is used to determine the execution analysis method for a type of response target based on the processing execution difference index and the response quality difference index to obtain the execution evaluation coefficient, and respond to the response setting condition to determine the setting method of the response processing coefficient as determining the execution evaluation coefficient based on the information association index and the associated reference execution index, or determining the execution evaluation coefficient based on the information association index and the analysis interference index; For a single type-1 response target, the response quality difference index , where n is the number of service processes in which this type-1 response target exists in the historical service record, yi is the expression standard index corresponding to the i-th service process of this type-1 response target in the historical service record, y0 is the average value of the expression standard indices corresponding to each service process of this type-1 response target in the historical service record, and the processing execution difference index , where di is the execution duration corresponding to the i-th service process of this type-1 response target in the historical service record, and d0 is the average value of the execution durations corresponding to each service process of this type-1 response target in the historical service record; A reference analysis module, which is connected to the response evaluation module, and is used to respond to the target reference condition for a type of response target to determine the setting method of the response processing coefficient of the type of response target as based on the information pointing parameter and the pointing execution parameter, or the target execution coefficient and the information pointing parameter to determine the response processing coefficient of the type of response target; A service analysis module, which is respectively connected to the execution analysis module and the reference analysis module, and is used to periodically determine whether to schedule the target to be responded based on the node processing coefficient of each target service node.

2. The intelligent customer service system based on intent recognition according to claim 1, wherein The response evaluation condition responded by the response evaluation module is that there is a type of response target, then it is determined that the execution analysis module performs target execution analysis on the type of response target. The target execution analysis includes: The execution analysis module responds to the execution analysis condition and determines the execution analysis method based on the processing execution difference index and the response quality difference index of the type of response target; The execution analysis module responds to the execution evaluation completion condition and determines the setting method of the response processing coefficient of the type of response target based on the execution evaluation coefficient; The execution analysis condition is that the response evaluation module determines that the execution analysis module performs target execution analysis on any type of response target, and the execution evaluation completion condition is that the execution evaluation coefficient of the type of response target is determined; The type of response target is the target to be responded with an expression standard index greater than the preset expression standard index or a reference processing execution coefficient greater than the preset reference processing execution coefficient.

3. The intelligent customer service system based on intent recognition according to claim 2, wherein, The analysis setting condition responded by the execution analysis module is that the execution stability coefficient is greater than the preset execution stability coefficient, then it is determined to determine the execution evaluation coefficient of the type of response target based on the information association index and the associated reference execution index; The execution evaluation coefficients are positively correlated with the information association index and the associated reference execution index respectively. The execution stability coefficient is determined according to the processing execution difference index and the response quality difference index of a type of response target.

4. The intelligent customer service system based on intent recognition according to claim 3, wherein When the analysis setting condition responded by the execution analysis module is that the execution stability coefficient is less than or equal to the preset execution stability coefficient, it is determined to determine the execution evaluation coefficient of a type of response target based on the information association index and the analysis interference index, and determine whether to adjust the execution evaluation coefficient based on the relevant execution parameters. When the evaluation adjustment condition responded by the execution analysis module is that the relevant execution parameters are greater than the preset relevant execution parameters, it is determined to increase and adjust the execution evaluation coefficient according to the relevant execution parameters.

5. The intelligent customer service system based on intent recognition according to claim 4, wherein When the response setting condition responded by the execution analysis module is that the execution evaluation coefficient is greater than the preset execution evaluation coefficient, it is determined to determine the response processing coefficient of a type of response target based on the execution evaluation coefficient and the associated reference execution index. When the response setting condition responded by the execution analysis module is that the execution evaluation coefficient is less than or equal to the preset execution evaluation coefficient, it is determined to determine the response processing coefficient of a type of response target based on the execution evaluation coefficient and the execution stability coefficient.

6. The intelligent customer service system based on intent recognition according to claim 5, wherein When the response evaluation condition responded by the response evaluation module is that there is a type of response target, it is determined to perform information reference analysis on the type of response target by referring to the analysis module. The information reference analysis includes: The reference analysis module responds to the reference analysis condition and determines the setting method of the response processing coefficient of the type of response target based on the information reference index and the information overlap index of the type of response target. The reference analysis condition is that the response evaluation module determines that the reference analysis module performs information reference analysis on any type of response target. The type of response target is a to-be-responded target with a statement standard index less than or equal to the preset statement standard index and a reference processing execution coefficient less than or equal to the preset reference processing execution coefficient.

7. The intelligent customer service system based on intent recognition according to claim 6, characterized in that, When the target reference condition responded by the reference analysis module is that the information reference index is greater than the preset information reference index or the information overlap index is greater than the preset information overlap index, the response processing coefficient of the type of response target is determined based on the information pointing parameter and the pointing execution parameter. When the target reference condition responded by the reference analysis module is that the information reference index is less than or equal to the preset information reference index and the information overlap index is less than or equal to the preset information overlap index, the response processing coefficient of the type of response target is determined based on the target execution coefficient and the information pointing parameter.

8. The intelligent customer service system based on intent recognition according to claim 7, wherein The service analysis module periodically responds to the node processing difference coefficient of the target scheduling condition to determine whether to perform task scheduling on the to-be-responded target of the key service node. When the target scheduling condition responded by the service analysis module is that the node processing difference coefficient is greater than the preset node processing difference coefficient, it is determined to perform task scheduling on the scheduled response target of the key service node. The node processing difference coefficient is determined based on the node processing coefficients of each target service node.

9. The intelligent customer service system based on intent recognition according to claim 8, characterized in that, The service analysis module responds to the scheduled response coefficient of the scheduling setting condition to determine the scheduled response target of each key service node. The scheduled response coefficient is determined according to the response processing coefficient and the analysis correlation index. The scheduling setting condition responded by the service analysis module is that the scheduling response coefficient of the target to be answered is greater than the preset scheduling response coefficient, then it is determined that the target to be answered is a scheduling response target; The service analysis module responds to the target scheduling condition and determines the transfer service nodes of each scheduling response target based on the analysis of relevant indexes. The target scheduling condition is that each key service node has completed the determination of the scheduling response target.

10. An intelligent customer service method applied to the intelligent customer service system based on intention recognition according to any one of claims 1 to 9, characterized in that, Including: According to the expression standard index and the reference processing execution coefficient, determine the response analysis strategy for the target to be answered as target execution analysis or information reference analysis for the target to be answered; Under the target execution analysis, according to the processing execution difference index and the response quality difference index, determine the execution analysis method for a type of response target. The execution analysis method is to determine the execution evaluation coefficient for a type of response target based on the information association index and the associated reference execution index, or the information association index and the analysis interference index; Under the condition that the execution evaluation is completed, the setting method of the response processing coefficient for a type of response target based on the execution evaluation coefficient is to determine the response processing coefficient for a type of response target based on the execution evaluation coefficient and the associated reference execution index, or the execution evaluation coefficient and the execution stability coefficient; Under the information reference analysis, according to the information reference index and the information overlap index, determine the setting method of the response processing coefficient for the second type of response target as to determine the response processing coefficient for the second type of response target based on the information pointing parameter and the pointing execution parameter, or the target execution coefficient and the information pointing parameter; Periodically determine whether to perform task scheduling for the targets to be answered of the target service node according to the node processing difference coefficient, determine the scheduling response targets of each key service node based on the scheduling response coefficient, and determine the transfer service nodes of each scheduling response target based on the analysis of relevant indexes.

Images