An intelligent customer relationship management system based on multi-modal data fusion

Abstract

The application relates to the technical field of customer management, in particular to an intelligent customer relationship management system based on multi-modal data fusion, which comprises a multi-modal data acquisition module, a state vector generation module, a customer relationship network topology construction module and a self-adaptive relationship repair strategy generation module; the multi-modal data acquisition module extracts a semantic contradiction index of a text work order of a voice dialogue, a service request frequency change rate and a historical performance quality score from a customer interaction log; the state vector generation module generates a customer relationship state vector based on a health degree score; the customer relationship network topology construction module constructs a causal correlation edge of a customer-service defect-relationship attenuation based on the customer relationship state vector, outputs a customer relationship network topology structure with an interpretable path, and forms a relationship maintenance knowledge graph; and the self-adaptive relationship repair strategy generation module generates a self-adaptive relationship repair strategy. Through identification of a high-weight relationship attenuation path, the application can locate key nodes and paths of customer relationship degradation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of customer management, and in particular to an intelligent customer relationship management system based on multi-modal data fusion. BACKGROUND

[0002] With the rapid development of information technology and the continuous change of customer needs, enterprises are facing unprecedented challenges in improving customer satisfaction and loyalty. Traditional customer relationship management (CRM) methods mainly rely on single-channel customer interaction data, usually based on historical transaction records or customer feedback for analysis, but this method has the following shortcomings:

[0003] Only relying on a single interaction data source (such as telephone, email, online chat, etc.), ignoring the integration and deep mining of cross-channel, multi-modal data, this single data source method cannot fully understand the customer's needs, behavior and emotional fluctuations, and it is also difficult to accurately assess the overall customer relationship.

[0004] Existing customer relationship management systems mostly rely on customer satisfaction surveys, historical purchase records, and other surface data to assess the health of customer relationships, lacking effective early warning mechanisms. When customer relationships go wrong, it often takes a long time to accumulate before being discovered, causing the enterprise to miss the best opportunity to take timely remedial measures. SUMMARY

[0005] The present application provides an intelligent customer relationship management system based on multi-modal data fusion.

[0006] An intelligent customer relationship management system based on multi-modal data fusion, comprising:

[0007] A multi-modal data acquisition module: extracting the semantic contradiction index of the text order of the voice dialogue, the service request frequency change rate, and the historical performance quality score from the customer interaction log, wherein:

[0008] Semantic contradiction index extraction: double-channel semantic encoding of voice dialogue text and corresponding order text, respectively extracting sentiment polarity and responsibility attribution labels, comparing sentiment polarity and responsibility attribution labels to determine if it is a contradictory event, and marking it as a contradictory event when the result is a contradictory event. According to the frequency of contradictory events, a semantic contradiction index is generated;

[0009] Service request frequency change rate extraction: according to the preset period (7 days / 30 days), the total amount of customer cross-channel service requests is calculated, and the time series difference algorithm is used to calculate the service request frequency change rate of the adjacent period request amount ;

[0010] Historical performance quality score extraction: extract performance indicators from the service order library, and generate a comprehensive performance quality score by weighted summation ;

[0011] State vector generation module: based on the semantic contradiction index, service request frequency change rate and historical performance quality score, generate a customer relationship state vector based on health score ;

[0012] Customer relationship network topology construction module: based on the customer relationship state vector , create a triple structure of customer node, service resource node and relationship degradation incentive node, based on the triple structure, construct the causal association edge of "customer-service defect-relationship attenuation", output the customer relationship network topology structure with interpretable path, and form the relationship maintenance knowledge graph;

[0013] Adaptive relationship repair strategy generation module: according to the relationship maintenance knowledge graph, the following is executed:

[0014] Identify high-weight relationship attenuation path;

[0015] Match the countermeasures in the preset strategy library.

[0016] Optionally, in the semantic contradiction index extraction, the natural language processing technology is used to perform double-channel semantic encoding on the voice dialogue text and the corresponding order text, and the sentiment polarity and responsibility attribution label are extracted respectively, the sentiment polarity vector includes positive sentiment and negative sentiment, and the responsibility attribution label is used to identify the designated responsible party in the order, including the customer, the customer service or the third party;

[0017] For each service event, directly compare the sentiment polarity in the voice text with the responsibility attribution label in the order text, when the sentiment polarity is positive sentiment (such as customer service voice promise solution) and the order responsibility attribution is customer or third party, mark it as a contradiction event, according to the occurrence frequency of the contradiction event, generate a semantic contradiction index, which is used to measure the potential relationship contradiction existing in the customer service process, and then affect the customer relationship health assessment.

[0018] Optionally, the service request frequency change rate is calculated as:

[0019] , wherein is the service request frequency change rate, representing the relative change percentage of the service request volume in the adjacent two periods, is the total service request volume in the current period , that is, the total number of customer cross-channel service requests in the time window, is the total service request volume in the previous period The total number of service requests in the inner time window, i.e., the total number of cross-channel service requests of the customer in the previous time window.

[0020] Optionally, the set of fulfillment indexes includes a problem resolution rate and a customer satisfaction score, and the comprehensive score of the fulfillment quality is represented as:

[0021] , wherein:

[0022] is the comprehensive score of the fulfillment quality, representing a comprehensive quality score of customer service fulfillment, is the response time score, is the problem resolution rate score, is the customer satisfaction score, is a weight coefficient of each index.

[0023] Optionally, the state vector generation module specifically includes:

[0024] a semantic contradiction index fusion unit: based on the semantic contradiction index, quantifying the inconsistency of emotional conflict and responsibility attribution in customer interaction, obtaining a sentiment health score of the customer relationship The higher the semantic contradiction index, the lower the sentiment health score of the customer.

[0025] a service request frequency change rate fusion unit: according to the value of the service request frequency change rate , assessing the volatility of customer service demand, the greater the change rate , the greater the fluctuation of customer demand, which may be a manifestation of the instability of the customer relationship, after normalizing the service request frequency change rate to the interval [0, 1], obtaining a customer demand fluctuation score .

[0026] a fulfillment quality score fusion unit: based on the historical fulfillment quality score , combining the past service fulfillment of the customer to assess the fulfillment health of the customer relationship, the fulfillment quality score directly reflects the customer's satisfaction with the service and the effectiveness of problem solving, the fulfillment health score directly uses the historical fulfillment quality score after standardization , the standardized fulfillment health score is: , wherein, and are the minimum and maximum values of the historical fulfillment score, respectively;

[0027] the final customer relationship state vector , wherein each dimension represents the health status of the customer relationship.

[0028] Optionally, the creation of the customer node includes based on the customer relationship state vector. The extracted health scores are used to create customer nodes, including the following information:

[0029] Basic customer identification;

[0030] Customer's emotional health score ;

[0031] Customer demand volatility score ;

[0032] Customer performance health score ;

[0033] The creation of the service resource node includes creating service resource nodes based on various service resources in the service system, including customer service personnel, service platforms, and problem solutions. The service resource node represents the interaction between the customer and the service resource, including the service history of the service resource.

[0034] The creation of the relationship degradation trigger nodes involves analyzing customer historical interaction data to create relationship degradation trigger nodes and identify potential factors that lead to the decline of customer relationships, including shirking responsibility (corresponding to emotional health), service defects (corresponding to the rate of change in service request frequency), and performance issues (corresponding to performance quality).

[0035] Triple structure creation: Based on the creation of customer nodes, service resource nodes, and relationship degradation cause nodes, a triple structure is constructed to represent the relationship between customers, service resources, and service defects. The triple form is: [customer node, service resource node, relationship degradation cause node].

[0036] Optionally, the causal association edges are based on the customer state vector. Construct a causal link that forms a customer-service-relationship decay path, where each causal link also includes an edge weight. edge weight Used to quantify the severity of customer relationship degradation and the impact of different degradation factors on customer relationships;

[0037] The customer nodes, service resource nodes, relationship degradation cause nodes, and causal connection edges form a complete relationship maintenance knowledge graph, outputting a customer relationship network topology with interpretable paths.

[0038] Optionally, the edge weights are calculated as follows:

[0039] ,in, This is the service request frequency variation rate. The higher the rate, the more unstable the customer's demand, which in turn leads to the deterioration of customer relationships. The absolute value is taken. Customer emotional health score, the lower the emotional health, the stronger the customer's dissatisfaction, leading to the customer relationship more easily degenerate, in order to reflect the severity of degradation, using Is the historical performance quality score, the lower the past service performance quality, the worse the customer experience, leading to customer churn, using Historical performance quality score, Is the weight coefficient, respectively control service request frequency change rate, emotional health score and performance quality score on the influence of edge weight.

[0040] Optionally, the adaptive relationship repair strategy generation module specifically comprises:

[0041] Identify high weight relationship attenuation path: in the relationship maintenance knowledge graph, the high weight relationship attenuation path represents the path with high customer relationship degradation probability, which is composed of multiple causal association edges, using Dijkstra algorithm to start from the customer node, traverse all causal relationship paths (i.e. the causal path of customer-service defect-relationship attenuation) in the graph, calculate the total weight of each path, the total weight The weighted sum of the edge weight on the path is: Wherein, Is the weight of the Edge in the path, The number of edges in the path, select the path with total weight greater than the weight threshold as the high weight relationship attenuation path.

[0042] Match the countermeasures in the preset strategy library: according to the identified high weight relationship attenuation path, the countermeasures in the preset strategy library will be used to repair the customer relationship, and the countermeasures include the pre-defined strategies to cope with the customer relationship recession.

[0043] Optionally, the preset strategy library specifically comprises:

[0044] Automatic trigger compensation agreement: if there is a problem of responsibility dodging on the path, automatically trigger the compensation measures (such as gift points, coupons, etc.);

[0045] Assign a dedicated customer manager: if there is a performance problem on the path, and there is a long-term negative interaction history, assign a dedicated customer manager to the customer to ensure more accurate and personalized communication with the customer;

[0046] Prioritize service problems: if there is a service defect problem on the path, immediately prioritize its service request to repair the problem.

[0047] The beneficial effects of the present application are:

[0048] ​​The application, through multi-modal data fusion, combines voice dialogue text, service request frequency change rate and performance quality score, establishes a customer relationship health degree dynamic evaluation system, through comprehensive analysis of these data, potential factors of customer relationship degradation can be accurately identified, through calculation of semantic contradiction index, conflict events of emotion and responsibility attribution are found and marked, which provides key clues and early warning information for subsequent relationship repair, the technology makes customer relationship management more intelligent, can reflect micro changes of customer relationship in real time, and identifies key factors affecting customer satisfaction in time.

[0049] The application constructs a relationship maintenance knowledge graph, through identification of high weight relationship attenuation path, key nodes and paths of customer relationship degradation can be located, through dynamic calculation of node and edge weight in causal link, actual influence of different relationship degradation inducements on customer relationship can be reflected, through in-depth analysis of these causal association paths, the system can accurately match corresponding countermeasures, such as automatically triggering service compensation agreement or assigning exclusive customer manager, so as to efficiently formulate relationship maintenance strategy and preferentially implement the most urgent repair measures, greatly improving the pertinence and response speed of customer relationship repair. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only a part of the application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0051] Fig. 1 The figure is a management system function module schematic diagram of the embodiment of the application.

[0052] Fig. 2 The figure is a state vector generation module schematic diagram of the embodiment of the application. DETAILED DESCRIPTION

[0053] The application will be described in detail below in combination with the drawings and specific embodiments. It should be noted that in order to make the embodiments more detailed, the following embodiments are the best, preferred embodiments, and other alternative ways can also be used by those skilled in the art to implement; and the drawings are only used to more specifically describe the embodiments, and are not intended to specifically limit the application.

[0054] It is noted that the recitations "one embodiment," "an embodiment,” “some embodiments,” “one example,” “an example,” “some examples,” and the like mean an embodiment that can include a particular feature, structure, or characteristic but every embodiment can not necessarily include the particular feature, structure, or characteristic. Moreover, the appearances of such phrases in various places in the specification are not necessarily intended to be referred to the same particular feature, structure, or characteristic but are intended to cover each individual implementation.

[0055] Generally, the terminology can be understood at least in part from usage in context. For example, the term “one or more” as used herein, depending at least in part upon context, can be used to describe any feature, structure, or characteristic in a singular sense or can be used to describe combinations of features, structures or characteristics in a plural sense. Similarly, terms, such as “based on,” can be understood as not necessarily delimiting a set of factors, but instead can be understood as allowing for existence of other factors not expressly described, at least in some contexts.

[0056] As shown in Figs. 1-2 A multi-modal data fusion based intelligent customer relationship management system, comprising:

[0057] A multi-modal data acquisition module: extracting a semantic contradiction index of a text work order of a voice dialogue, a service request frequency change rate, and a historical performance quality score from a customer interaction log, wherein:

[0058] Semantic contradiction index extraction: performing double-channel semantic encoding on the voice dialogue text and the corresponding work order text, respectively extracting sentiment polarity and responsibility attribution labels, determining whether it is a contradictory event by comparing the sentiment polarity and the responsibility attribution labels, marking it as a contradictory event when the determination result is a contradictory event, and generating a semantic contradiction index according to the contradictory event frequency;

[0059] Service request frequency change rate extraction: counting the total amount of cross-channel service requests of a customer according to a preset period (7 days / 30 days), and calculating the service request frequency change rate of the request amount of adjacent periods using a time series difference algorithm , attributing the abnormal fluctuation (absolute value of change rate ≥ 50%) to the channel source, and marking the main fluctuation channel;

[0060] Historical performance quality score extraction: extracting a performance index set from a service work order library and generating a performance quality comprehensive score by weighted summation ;

[0061] A state vector generation module: generating a customer relationship state vector based on the health score based on the semantic contradiction index, the service request frequency change rate, and the historical performance quality score ;

[0062] Customer relationship network topology construction module: based on customer relationship state vector , create a triple structure of customer node, service resource node, and relationship degradation incentive node, based on the triple structure, construct the causal association edge of "customer-service defect-relationship attenuation", output the customer relationship network topology structure with interpretable path, and form the relationship maintenance knowledge graph;

[0063] Adaptive relationship repair strategy generation module: according to the relationship maintenance knowledge graph, the following is executed:

[0064] Identify high-weight relationship attenuation paths (such as response delay of a certain customer service representative leading to a significant increase in relationship entropy of a specific customer group);

[0065] Match countermeasures in the preset strategy library (such as automatically triggering service compensation agreement or assigning a dedicated customer manager).

[0066] In the semantic contradiction index extraction, based on natural language processing technology, the voice dialogue text and the corresponding work order text are double-channel semantic encoded, and the sentiment polarity and responsibility attribution label are extracted respectively. The sentiment polarity vector includes positive sentiment and negative sentiment, and the responsibility attribution label is used to identify the designated responsible party in the work order, including the customer, the customer service or the third party;

[0067] For each service event, directly compare the sentiment polarity in the voice text with the responsibility attribution label in the work order text. When the sentiment polarity is positive (such as customer voice commitment to solution) and the work order responsibility attribution is customer or third party, it is marked as a contradiction event. According to the occurrence frequency of the contradiction event, the semantic contradiction index is generated, which is used to measure the potential relationship contradiction in the customer service process and thus affect the customer relationship health assessment.

[0068] Natural language processing technology includes sentiment analysis model (BERT) to extract sentiment polarity from voice dialogue text. The sentiment polarity vector represents the direction and intensity of emotion in the text, which is positive (such as solution commitment) or negative (such as service evasion): , wherein, is the sentiment polarity vector, which represents the sentiment polarity of the voice dialogue, is the voice dialogue text, Sentiment represents the calculation process of the sentiment analysis model;

[0069] The purpose of the sentiment analysis model is to extract the sentiment polarity and intensity of emotion in the voice dialogue text. The sentiment analysis model processes the text input and outputs the sentiment polarity vector. The calculation process is as follows:

[0070] Text preprocessing:

[0071] The voice dialogue text is preprocessed, including:

[0072] Word segmentation, stop word removal, stemming, or word form restoration; for example, for the text "Customer service has promised to solve the problem", after processing, we get: {customer service, promise, solve, problem}.

[0073] Sentiment Classification: The text is classified into sentiments using a pre-trained sentiment analysis model. The sentiment analysis model maps the input text to sentiment polarity and outputs a sentiment score or sentiment category.

[0074] The sentiment analysis model Sentiment(·) outputs a sentiment polarity score s and the corresponding label:

[0075] If s > 0, then the label is "positive";

[0076] If s≤0, then the label is "negative".

[0077] The responsibility attribution label is extracted from the work order text to identify the responsible party (such as customer service, a third party, or the customer) in the service incident. The responsibility attribution label is a discrete value.

[0078] ,in, This is a responsibility attribution label, indicating the attribution of responsibility in the work order text. It's a work order text, Responsibility This describes the calculation process of the responsibility attribution extraction model. The purpose of the responsibility attribution extraction model is to extract the responsible party label (such as customer service, third party, or customer) from the work order text. It relies on named entity recognition technology, and the calculation process is as follows:

[0079] Text preprocessing: Similarly, the work order text is first preprocessed, following steps similar to those in sentiment analysis. For example, the text "Third-party supplier failed to deliver on time" is processed to obtain: {Third-party, Supplier, Failed to deliver}

[0080] Named Entity Recognition: Using a named entity recognition model, the relevant responsible entities in the text are identified. In this example, the named entity recognition model identifies "third-party supplier" as a responsible party.

[0081] Responsibility Attribution Classification: Classifying the extracted entities to determine whether they are "customer service", "third party" or "customer" is a simple classification task that can be based on keyword matching.

[0082] The semantic contradiction index is normalized based on the frequency of contradictory events to obtain an index value that reflects the degree of contradiction in the relationship.

[0083] Service request frequency change rate The calculation is as follows:

[0084] wherein, is the service request frequency change rate, representing the relative change percentage of the service request volume in the adjacent two periods, is the total number of service requests in the current period , i.e., the total number of cross-channel service requests of customers in the time window (for example: the total number of requests in the past 7 days or 30 days), is the total number of service requests in the previous period , i.e., the total number of cross-channel service requests of customers in the previous time window.

[0085] The performance indicator set includes the problem resolution rate, the customer satisfaction score, and the performance quality comprehensive score, which is represented as:

[0086] wherein:

[0087] is the performance quality comprehensive score, representing the comprehensive quality score of customer service performance, for evaluating the overall service quality obtained by the customer in the service process, is the response time score, measuring the response speed of the customer service after the customer request, the faster the response speed, the higher the score, the service with shorter response time can better meet the customer's expectations, and the score is also higher, is the problem resolution rate score, measuring whether the customer's problem is effectively solved. The higher the problem resolution rate, the better the service quality, and the higher the score, is the customer satisfaction score, measuring the overall satisfaction of the customer to the service, the feedback of the customer (such as satisfaction survey, score, etc.) determines this score, is the weight coefficient of each indicator, for adjusting the relative importance of different indicators in the comprehensive score, the sum of the weight coefficients is 1, according to the business requirements, different indicators can be given different weights, in the present application, can take the values 0.4, 0.3, 0.3.

[0088] The state vector generation module specifically includes:

[0089] The semantic contradiction index fusion unit: based on the semantic contradiction index, quantifying the inconsistency of emotional conflict and responsibility attribution in customer interaction, obtaining the emotional health score of customer relationship The higher the semantic contradiction index, the lower the emotional health score of the customer, and the emotional health score can be calculated by the following formula:

[0090] wherein, contradiction is the occurrence frequency of the contradiction event, total is the total number of all customer interaction events, and the emotional health score The closer to 1 indicates that the emotional relationship is healthier, and the closer to 0 indicates that the emotional relationship is less stable.

[0091] The service request frequency change rate fusion unit: according to the numerical value of the service request frequency change rate , assess the volatility of customer service demand, the change rate The greater the change rate, the greater the fluctuation of customer demand, which may be a manifestation of the instability of customer relationship. After normalizing the service request frequency change rate to the interval [0, 1], the customer demand fluctuation score is obtained, and the specific calculation method is: , wherein is a preset maximum change rate threshold, which can be set to 50%, indicating extreme service request fluctuation.

[0092] The performance quality score fusion unit: based on the historical performance quality score , combined with the past service performance of the customer, to assess the performance health of the customer relationship. The performance quality score directly reflects the customer's satisfaction with the service and the effectiveness of problem solving, and the performance health score directly uses the standardized historical performance quality score , and the standardized performance health score is: , wherein and are the minimum and maximum values of the historical performance score, respectively;

[0093] The final customer relationship state vector , wherein each dimension represents the health status of the customer relationship.

[0094] The creation of the customer node includes the creation of the customer node based on the health scores extracted from the customer relationship state vector , including the following information:

[0095] The basic identity of the customer;

[0096] The emotional health score of the customer ;

[0097] The demand fluctuation score of the customer ;

[0098] The performance health score of the customer ;

[0099] These scores reflect the health status of each dimension (emotion, demand, performance) of the customer relationship. Each customer node will have information on these dimensions for subsequent relationship analysis;

[0100] The creation of service resource nodes includes creating service resource nodes based on various types of service resources in the service system, including customer service personnel, service platforms, and problem-solving solutions. Service resource nodes represent the interaction between customers and service resources, including the service history of service resources.

[0101] The creation of relationship degradation trigger nodes includes creating relationship degradation trigger nodes by analyzing customer historical interaction data, identifying potential factors that lead to customer relationship decline, including responsibility avoidance (corresponding to emotional health), service defects (corresponding to service request frequency change rate), and compliance problems (corresponding to compliance quality). These trigger nodes are directly related to the decline of customer relationships.

[0102] Triplet structure creation: Based on the creation of customer nodes, service resource nodes, and relationship degradation trigger nodes, a triplet structure is constructed to represent the relationship between customers and service resources, service defects, and relationship degradation trigger nodes. The triplet form is: [customer node, service resource node, relationship degradation trigger node], for example:

[0103] (Customera, Customer Service Personnelb, Response Delay), which indicates that the service relationship of Customera is affected by the response delay of Customer Service Personnelb.

[0104] The creation of relationship degradation trigger nodes is as follows:

[0105] 1. Collect historical interaction data:

[0106] From the customer's historical interaction data, the following key data points are extracted:

[0107] Customer service request records: including customer complaints, consultations, and problem-solving records.

[0108] 2. Identify potential relationship degradation triggers:

[0109] Based on the collected historical interaction data, identify potential relationship degradation trigger nodes:

[0110] a. Service defects: By analyzing customer feedback records (complaints, refund requests, service interruptions, etc.), identify frequent service defect types, such as:

[0111] Function failure: system crash, service interruption.

[0112] Operation error: customer service personnel or system error leading to service failure.

[0113] b. Compliance problems: Extract the time difference between customer requests and responses to identify long-unanswered customer service requests, which may be an important trigger for customer dissatisfaction. Based on emotional analysis of chat records or voice conversations, identify inappropriate communication problems.

[0114] c. Responsibility Shifting: There is no clear responsibility in communication.

[0115] 3. Create Relationship Deterioration Cause Nodes

[0116] Convert the above potential factors into relationship deterioration cause nodes, and create each cause node based on the following information:

[0117] Node Name: Identify the specific content of the cause (service defect, response delay, poor communication).

[0118] Node Attributes: Specific attributes of the node, such as defect type, delay time, sentiment analysis results.

[0119] 4. Identify the relationship between cause nodes: After creating cause nodes, we need to consider the association between these nodes and how they affect customer relationships, consider the following types of associations:

[0120] Causal Relationship: Some causes directly lead to the deterioration of customer relationships.

[0121] Joint Influence: Multiple causes may act together, increasing the risk of customer relationship deterioration.

[0122] Temporal Relationship: The order of occurrence of some causes may affect the degree of customer relationship deterioration, for example, long service response delay may be the direct cause of customer complaints.

[0123] 5. Finally, through the above steps, relationship deterioration cause nodes can be generated, which can form part of the knowledge graph nodes, and when building the relationship network topology, connect with other customer nodes, service resource nodes through edges, forming a complete relationship maintenance graph.

[0124] Causal Association Edge Based on Customer State Vector Construction, forming a causal link between customer-service-relationship decay, the health status of customer relationship will affect the relationship strength between customers and service defects, causal association edges also include edge weights , edge weights used to quantify the severity of customer relationship deterioration and the degree of influence of different deterioration factors on customer relationship;

[0125] Customer nodes, service resource nodes, relationship deterioration cause nodes, and causal association edges form a complete relationship maintenance knowledge graph, output customer relationship network topology with explainable paths, paths can help identify key factors that lead to customer relationship deterioration and provide guidance for subsequent relationship repair. For example, the path may show that the relationship deterioration of customer A is due to the chain reaction of response delay, service defect, etc.

[0126] The triplet structure (customer node, service resource node, relationship degradation cause node) defines the basic relationship between customers and service resources and service defects. The causal relationship edge, based on these triplets, further represents the causal chain of customer relationship decline. For example, the triplet represents the service defect caused by the response delay between customer A and customer service representative B, while the causal relationship edge represents the specific impact and intensity of this service defect on the decline of customer relationship. Therefore, the triplet structure is the basis of the causal relationship edge, which quantifies the impact of service defects on customer relationship.

[0127] The edge weights are calculated as follows:

[0128] ,in, This is the service request frequency variation rate. The higher the rate, the more unstable the customer's demand, which in turn leads to the deterioration of customer relationships. The absolute value is taken. , Customer emotional health score: the lower the emotional health score, the stronger the customer's dissatisfaction, leading to a greater likelihood of customer relationship degradation. To reflect the severity of degradation, a [missing information - likely a specific parameter or measure] is used. , This is a historical performance quality score; the lower the score, the worse the past service performance, resulting in a poorer customer experience and customer churn. Historical performance rating These are weighting coefficients that control the impact of service request frequency change rate, emotional health score, and performance quality score on edge weights, and can be adjusted according to actual business conditions.

[0129] The adaptive relationship repair strategy generation module specifically includes:

[0130] Identifying High-Weight Relationship Decay Paths: In a relationship maintenance knowledge graph, high-weight relationship decay paths represent paths with a high probability of customer relationship degradation. These paths consist of multiple causal edges. Using Dijkstra's algorithm, starting from the customer node, all causal relationship paths in the graph (i.e., the causal path of customer-service defect-relationship decay) are traversed, and the total weight of each path is calculated. The weighted sum of the weights of each edge on the path: ,in, It is the first in the path The weight of the edge, It is the number of edges in the path. Paths with a total weight greater than the weight threshold are selected as high-weight relationship decay paths. The weight threshold is based on the percentile of historical data. Paths with a total weight above the 90th percentile are selected as high-weight relationship decay paths.

[0131] Match the countermeasures in the preset policy library: according to the identified high-weight relationship attenuation path, the countermeasures in the preset policy library are adopted to repair the customer relationship, and the countermeasures include the pre-defined strategies for coping with the customer relationship recession.

[0132] The preset policy library specifically includes:

[0133] Automatic triggering of compensation agreement: if there is a problem of responsibility shirking on the path, an automatic compensation measure (such as giving points, coupons, etc.) is triggered;

[0134] Assigning a dedicated customer manager: if there is a performance problem on the path and there is a long-term negative interaction history, a dedicated customer manager is assigned to the customer to ensure more accurate and personalized communication with the customer;

[0135] Prioritize service problems: if there is a service defect problem on the path, immediately prioritize its service request to fix the problem.

[0136] Total weight It refers to the sum of all edge weights of a complete path starting from the customer node, passing through one or more relationship degradation inducement nodes and service resource nodes, and the total weight of the path reflects the comprehensive influence of all factors on the customer relationship degradation.

[0137] The relationship degradation inducement node is a key node in the path, representing a potential factor that leads to customer relationship degradation, and each node is connected to one or more causal association edges, indicating that a certain factor has affected the health of the customer relationship. The identification method of these nodes includes:

[0138] Identify nodes from causal links: when building a relationship maintenance knowledge graph, nodes correspond to specific events or service problems.

[0139] Determine the influence of nodes according to the causal links of the path: by analyzing the path from the customer node to the relationship attenuation node, determine which nodes have a greater impact on the customer relationship, for example, if a certain node (service defect node) in the path has a strong causal relationship with the subsequent node (relationship attenuation node), then this node can be considered as the main relationship degradation inducement node.

[0140] In summary, it is first based on the total weight to find a high-weight relationship attenuation path, and then based on the high-weight relationship attenuation path to find the edge that leads to the high-weight relationship attenuation path, and based on the edge weight of this edge to find the corresponding customer relationship state vector one of them, and then based on the corresponding customer relationship state vector to find the relationship degradation inducement node, so as to match the strategy of the preset policy library.

[0141] The present application encompasses any alternatives, modifications, equivalent methods and solutions made to the essence and scope of the present application. In order to make the public have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be fully understood without the description of these details to those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, procedures, elements and circuits, etc. are not described in detail.

[0142] The above is only the preferred embodiment of the present application, and it should be pointed out that for ordinary skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can also be made, which should be considered as the protection scope of the present application.

Claims

1. An intelligent customer relationship management system based on multi-modal data fusion, characterized in that, Comprise: Multi-modal data acquisition module: extract the semantic contradiction index of the text ticket of the voice dialogue from the customer interaction log, the service request frequency change rate and the historical performance quality score, wherein: Semantic contradiction index extraction: double-channel semantic encoding is performed on the voice dialogue text and the corresponding ticket text, and the sentiment polarity and responsibility attribution label are extracted respectively. By comparing the sentiment polarity and responsibility attribution label, it is judged whether it is a contradictory event. When the judgment result is a contradictory event, it is marked as a contradictory event. The semantic contradiction index is generated according to the contradictory event frequency; Service request frequency change rate extraction: according to a preset period, the total amount of customer cross-channel service requests is counted, and a time series difference algorithm is used to calculate the service request frequency change rate of the request amount of adjacent periods ; Historical performance quality score extraction: extract performance indicators set from service work order library, weighted sum to generate performance quality comprehensive score ; The state vector generation module generates a customer relationship state vector based on the health degree score based on the semantic contradiction index, service request frequency change rate, and historical performance quality score ; A customer relationship network topology construction module: based on the customer relationship state vector , create a triple structure of customer nodes, service resource nodes, and relationship degradation incentive nodes, based on the triple structure, construct the causal association edge of "customer-service defect-relationship attenuation", output the customer relationship network topology structure with an interpretable path, and form a relationship maintenance knowledge graph; Adaptive relationship repair strategy generation module: according to the relationship maintenance knowledge graph, the following is performed: Identify high-weight relationship attenuation path; Match the countermeasures in the preset strategy library; In the semantic contradiction index extraction, the voice dialogue text and the corresponding ticket text are double-channel semantic encoded based on natural language processing technology, and the sentiment polarity and responsibility attribution label are extracted respectively. The sentiment polarity vector includes positive sentiment and negative sentiment. The responsibility attribution label is used to identify the designated responsible party in the ticket, including the customer, the customer service or the third party; For each service event, directly compare the sentiment polarity in the voice text with the responsibility attribution label in the ticket text. When the sentiment polarity is positive sentiment and the ticket responsibility attribution is customer or third party, mark it as a contradictory event. According to the contradictory event frequency, the semantic contradiction index is generated for statistical generation, which is used to measure the potential relationship contradiction in the customer service process; The adaptive relationship repair strategy generation module specifically includes: Identify high-weight relationship attenuation path: the high-weight relationship attenuation path represents the path with high customer relationship degradation probability, which is composed of multiple causal correlation edges. Dijkstra algorithm is used to traverse all causal relationship paths from the customer node in the graph, calculate the total weight of each path, and the total weight is the weighted sum of the weights of the edges on the path: wherein, is the weight of the th edge in the path, is the number of edges in the path, and the path with a total weight greater than the weight threshold is selected as the high-weight relationship attenuation path; Match the countermeasures in the preset strategy library: according to the identified high-weight relationship attenuation path, the countermeasures in the preset strategy library are used to repair the customer relationship. The countermeasures include predefined strategies to deal with customer relationship decline.

2. The intelligent customer relationship management system based on multi-modal data fusion according to claim 1, characterized in that, The service request frequency change rate The calculation is: wherein, is the service request frequency change rate, representing the relative change percentage of the service request volume in the adjacent two periods, is the total service request volume in the current period , i.e., the total number of cross-channel service requests of customers in the time window, is the total service request volume in the previous period , i.e., the total number of cross-channel service requests of customers in the previous time window.

3. The intelligent customer relationship management system based on multi-modal data fusion as claimed in claim 1, wherein, The performance index set includes problem solving rate and customer satisfaction score. The performance quality comprehensive score is represented as: wherein: is a comprehensive score of the performance quality, indicating the comprehensive quality score of the performance of the customer service, is a response time length score, is a problem resolution rate score, is a customer satisfaction score, is a weight coefficient of each index.

4. The intelligent customer relationship management system based on multi-modal data fusion according to claim 1, characterized in that, The state vector generation module specifically includes: a semantic contradiction index fusion unit: based on the semantic contradiction index, quantifying the inconsistency of emotional conflict and responsibility attribution in customer interaction, obtaining a score of emotional health degree of customer relationship , the higher the semantic contradiction index, the lower the score of emotional health degree of customer The service request frequency change rate fusion unit: according to the numerical value of the service request frequency change rate , assess the volatility of customer service demand, the change rate The greater, the greater the fluctuation of customer demand, after normalizing the service request frequency change rate to the interval [0, 1], the customer demand fluctuation degree score ; The performance quality score fusion unit: based on historical performance quality scores , combined with the past service performance of the customer to evaluate the performance health of the customer relationship, the performance quality score Directly reflects the customer's satisfaction with the service and the effectiveness of problem solving, the performance health score Directly use the historical performance quality score after standardization , the standardized performance health score is: , wherein And The minimum and maximum values of the historical performance score, respectively Final customer relationship state vector where each dimension represents a health state of the customer relationship.

5. The intelligent customer relationship management system based on multi-modal data fusion as claimed in claim 1, wherein, The creation of the customer node includes creating a customer node based on the health score extracted from the customer relationship state vector, including the following information: Basic identity of the customer; Client emotional well-being score ; Customer demand volatility score ; Customer compliance health score ; The creation of the service resource node includes creating a service resource node according to various service resources in the service system, including customer service personnel, service platform and problem solving solution. The service resource node represents the interaction between the customer and the service resource, including the service history of the service resource; The creation of the relationship degradation incentive node includes creating a relationship degradation incentive node by analyzing the customer historical interaction data, identifying the potential factors leading to customer relationship decline, including responsibility avoidance, service defects and performance problems; Triplet structure creation: based on the creation of the customer node, the service resource node and the relationship degradation incentive node, a triplet structure is constructed to represent the relationship between the customer and the service resource, the service defect, and the triplet form is: [customer node, service resource node, relationship degradation incentive node].

6. The intelligent customer relationship management system based on multi-modal data fusion as claimed in claim 5, wherein, The causal association edge is according to the customer state vector constructing, forming a causal chain of customer-service-relationship attenuation, the causal association edge further comprises an edge weight The edge weight is used to quantify the severity of customer relationship degradation and the influence degree of different degradation factors on customer relationship; The customer node, service resource node, relationship degradation incentive node and causal association edge form a complete relationship maintenance knowledge graph, and output a customer relationship network topology structure with an interpretable path.

7. The intelligent customer relationship management system based on multi-modal data fusion as claimed in claim 6, wherein, The edge weight calculation is: wherein, is the service request frequency change rate, the larger the more unstable the customer's demand, is the customer emotional health score, the lower the emotional health score, the stronger the customer's dissatisfaction, leading to the customer relationship more easily degrading is the historical performance quality score, the lower the past service performance quality, the worse the customer experience, leading to the customer churn is the weight coefficient, respectively controlling the influence of the service request frequency change rate, the emotional health score and the performance quality score on the edge weight.

8. The intelligent customer relationship management system based on multi-modal data fusion as claimed in claim 1, wherein, The preset strategy library specifically includes: Automatic trigger compensation agreement: if there is a problem of responsibility shirking on the path, automatically trigger compensation measures; Assign a dedicated customer manager: if there is a problem of performance on the path, and there is a long-term negative interaction history, assign a dedicated customer manager to the customer; Prioritize service problems: if there is a service defect problem on the path, immediately prioritize its service request and fix the problem.

Images