Customer service method, device, equipment, medium and program product

Abstract

The invention provides a customer service method which can be applied to the technical field of artificial intelligence. The customer service method comprises the following steps: acquiring behavior data of a customer in a preset time period; obtaining fused semantic features of the behavior data, extracting sequence features from the fused semantic features, and obtaining customer emotion features and customer demands within a preset time period by using a multilayer neural network based on the sequence features; incremental data appears in response to the behavior data, a knowledge graph is updated based on the behavior data and the incremental data, and the knowledge graph is pre-constructed based on the behavior data; obtaining a target product based on the customer emotional characteristics, the customer demands and the knowledge graph, and generating a communication strategy based on the customer emotional characteristics; and in response to the received semantic input of the customer, communicating with the customer according to the communication strategy, and recommending the target product in the communication process. The invention further provides a customer service device, equipment, a storage medium and a program product.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and more specifically to a customer service method, apparatus, device, medium, and program product. Background Technology

[0002] Driven by digitalization, banks are increasingly shifting their communication with customers to online platforms. Online communication, with its efficiency and convenience, has become a crucial channel for banks to provide daily services and recommend financial products to customers.

[0003] To improve online service quality, customer emotional characteristics are often analyzed during communication. Existing analysis methods mainly rely on simple keyword matching or classification models. While these can provide basic emotion recognition, their emotion monitoring is not accurate enough to provide targeted services, thus affecting service quality and user experience.

[0004] Furthermore, most existing technologies rely on static rules or simple labels to predict customer needs. While this rule-based approach is easy to implement, it lacks dynamic and personalized analysis capabilities. For example, a customer might express interest in high-yield investments in a conversation, but historical behavioral data might reveal a preference for stable-yield products. If existing recommendation systems rely solely on this initial expression to serve customers, they may fail to accurately capture the customer's potential needs, thus reducing conversion rates.

[0005] In summary, existing technologies have significant shortcomings in emotion recognition and demand prediction, which limit the improvement of the efficiency and quality of online platform communication with customers. Summary of the Invention

[0006] In view of the above problems, this application provides customer service methods, apparatus, devices, media and program products that improve the accuracy of emotion recognition and demand prediction recognition.

[0007] According to a first aspect of this application, a customer service method is provided, the method comprising: acquiring customer behavior data within a preset time period; acquiring fused semantic features of the behavior data, extracting sequence features from the fused semantic features, and obtaining customer sentiment features and customer needs within the preset time period based on the sequence features using a multi-layer neural network; updating a knowledge graph based on the behavior data and the incremental data in response to the occurrence of incremental data in the behavior data, the knowledge graph being pre-constructed based on the behavior data; obtaining a target product based on the customer sentiment features, the customer needs, and the knowledge graph, and generating a communication strategy based on the customer sentiment features; and communicating with the customer according to the communication strategy in response to receiving semantic input from the customer, and recommending the target product during the communication process.

[0008] According to an embodiment of this application, the behavioral data is multimodal data, and obtaining the fused semantic features of the behavioral data includes: identifying the multimodal data and obtaining semantic features corresponding to each modality of data, wherein the multimodal data includes at least two of text data, speech data, and image data; assigning corresponding weights to the semantic features corresponding to each modality of data and performing weighted feature fusion to obtain the fused semantic features.

[0009] According to an embodiment of this application, updating a knowledge graph based on the behavioral data and the incremental data includes: generating a target dataset based on the behavioral data and the incremental data, wherein the data in the target dataset is used to indicate entities and entity relationships, and the entities include the customer and the product; and constructing a knowledge graph based on the entities and entity relationships using the target dataset as sample data.

[0010] According to an embodiment of this application, obtaining a target product based on the customer's emotional characteristics, the customer's needs, and the knowledge graph includes: obtaining a target product type using reinforcement learning based on the customer's needs and the knowledge graph; obtaining the customer's product preferences according to a preset matching rule based on the customer's emotional characteristics; and obtaining the target product from the target product type based on the product preferences.

[0011] According to an embodiment of this application, after recommending the target product during the communication process, the method further includes: collecting feedback from the customer based on the target product to obtain feedback data; and dynamically adjusting the model parameters of the reinforcement learning based on the feedback data.

[0012] According to an embodiment of this application, after obtaining customer behavior data within a preset time period, the method further includes: encrypting and / or de-identifying the behavior data; and performing differential privacy processing on the behavior data based on a preset noise level.

[0013] According to an embodiment of this application, after obtaining the customer emotional characteristics within the preset time period, the method further includes: generating a warning message in response to the customer emotional characteristics continuously exceeding a warning value within the preset time period.

[0014] A second aspect of this application provides a customer service device, comprising: an acquisition module for acquiring customer behavior data within a preset time period; a fusion module for acquiring fused semantic features of the behavior data, extracting sequence features from the fused semantic features, and obtaining customer emotional features and customer needs within the preset time period based on the sequence features using a multi-layer neural network; an increment module for updating a knowledge graph based on the behavior data and the increment data in response to the occurrence of incremental data in the behavior data, wherein the knowledge graph is pre-constructed based on the behavior data; an output module for obtaining a target product based on the customer emotional features, the customer needs, and the knowledge graph, and generating a communication strategy based on the customer emotional features; and a response module for communicating with the customer according to the communication strategy in response to receiving semantic input from the customer, and recommending the target product during the communication process.

[0015] A third aspect of this application provides an electronic device comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0016] A fourth aspect of this application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0017] The fifth aspect of this application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method. Attached Figure Description

[0018] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0019] Figure 1 The illustrations depict application scenarios of customer service methods, apparatus, devices, media, and program products according to embodiments of this application.

[0020] Figure 2 A flowchart illustrating a customer service method according to an embodiment of this application is shown schematically.

[0021] Figure 3 A flowchart illustrating the fusion semantic features of acquiring behavioral data according to an embodiment of this application is shown.

[0022] Figure 4 A flowchart illustrating the updating of a knowledge graph based on behavioral data and incremental data according to an embodiment of this application is shown schematically.

[0023] Figure 5 A flowchart illustrating the process of obtaining the target product according to an embodiment of this application is shown schematically;

[0024] Figure 6 This illustration schematically shows a deep learning model architecture diagram for obtaining customer emotional characteristics and customer needs within a preset time period according to an embodiment of this application;

[0025] Figure 7 A schematic diagram illustrating the structure of a customer service device according to an embodiment of this application is shown; and

[0026] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing a customer service method according to an embodiment of this application. Detailed Implementation

[0027] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0028] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0029] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0030] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0031] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0032] In scenarios involving automated decision-making using personal information, the methods, devices, and systems provided in this application all offer users corresponding entry points for choosing to agree to or reject the automated decision-making results. If the user chooses to reject, the process proceeds to the expert decision-making stage. Here, "automated decision-making" refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests, or economic, health, and credit status through computer programs, and then making a decision. Here, "expert decision-making" refers to the activity of making decisions by personnel who specialize in a particular field, possess specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0033] In related technologies, when customers consult online, the analysis of customer emotional characteristics mainly relies on simple keyword matching or classification models. However, the accuracy of emotion monitoring is insufficient, impacting service quality and user experience. Furthermore, customer demand prediction is often based on static rules or simple tags, lacking dynamic and personalized analysis capabilities. This makes it difficult to accurately capture potential customer needs, reducing recommendation conversion rates and limiting the improvement of the efficiency and quality of communication between online platforms and customers.

[0034] Embodiments of this application provide a customer service method, apparatus, device, medium, and program product. The customer service method includes: acquiring customer behavior data within a preset time period; acquiring fused semantic features of the behavior data, extracting sequence features from the fused semantic features, and using a multi-layer neural network based on the sequence features to obtain customer emotional features and customer needs within the preset time period; updating a knowledge graph based on the behavior data and the incremental data in response to the occurrence of incremental data, wherein the knowledge graph is pre-constructed based on the behavior data; obtaining a target product based on the customer emotional features, customer needs, and the knowledge graph; and generating a communication strategy based on the customer emotional features; and communicating with the customer according to the communication strategy in response to receiving semantic input from the customer, and recommending the target product during the communication process.

[0035] According to embodiments of this application, by acquiring the fused semantic features of behavioral data, long-range dependencies and implicit emotions can be captured. Further extraction of sequence features enhances the emotional understanding of customers with fluctuating emotions. Finally, multi-layer neural network mapping is used to obtain customer emotional characteristics and demand categories, thereby achieving accurate sentiment analysis and demand prediction to provide more personalized and precise services. In addition, updating the knowledge graph based on behavioral data and incremental data ensures the timeliness and accuracy of customer information. Furthermore, based on customer emotional characteristics and customer needs, combined with the rich entities and relationships in the knowledge graph, the target product is accurately predicted and matched. That is, the prediction process not only considers the customer's current behavioral data but also historical behavioral data, which helps to improve the accuracy of recommendations, thereby increasing customer satisfaction and recommendation conversion rate.

[0036] Figure 1 The illustration shows an application scenario diagram of a customer service method according to an embodiment of this application.

[0037] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0038] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0039] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0040] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0041] It should be noted that the customer service method provided in this application embodiment can generally be executed by server 105. Correspondingly, the customer service device provided in this application embodiment can generally be located in server 105. The customer service method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the customer service device provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0042] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0043] The following will be based on Figure 1 The described scene, through Figures 2-6 A customer service method according to an embodiment of this application will be described in detail.

[0044] Figure 2 A flowchart illustrating a customer service method according to an embodiment of this application is shown schematically.

[0045] like Figure 2 As shown, the customer service method of this embodiment includes operations S210 to S250.

[0046] In operation S210, customer behavior data within a preset time period is obtained.

[0047] In embodiments of this application, customer consent or authorization can be obtained before acquiring customer behavior data. For example, a request to acquire behavior data can be sent to the customer before operation S210. Operation S210 is performed if the customer consents or authorizes the acquisition of behavior data.

[0048] According to embodiments of this application, customer behavior data, including chat logs between customer service representatives and customers, is captured from a bank's customer interaction platform. The behavior data is then cleaned to remove redundant text, special symbols, Hypertext Markup Language (HTML) tags, and other irrelevant content. The behavior data is also standardized in spelling and abbreviation. The cleaned behavior data is then labeled using manual methods or a large language model to add sentiment tags and financial demand tags, providing high-quality input for subsequent sentiment and demand analysis.

[0049] According to embodiments of this application, behavioral data is multimodal data, which includes at least two of the following: text data, voice data, and image data. By processing information from different data sources (text, voice, and images), it is possible to analyze customer needs and emotions from multiple dimensions, thereby improving the accuracy of the analysis results.

[0050] According to embodiments of this application, behavioral data is encrypted and / or de-identified. For example, fields in the behavioral data that have identifying characteristics (such as a customer's name, contact information, etc.) are encrypted and / or de-identified to remove individual characteristics.

[0051] According to embodiments of this application, behavioral data is subjected to differential privacy processing based on a preset noise level. For example, when a customer provides sensitive information such as income level, differential privacy technology is enabled to encrypt the customer's sensitive information through a noise addition mechanism, making the data unassociatable with a specific customer identity.

[0052] For example, during a conversation with customer service, a customer might disclose personal information (such as family member details and income level) and inquire about family financial planning services. Since this data involves sensitive information, it's crucial to ensure customer privacy is protected. This can be achieved through encryption, de-identification, and differential privacy processing methods to safeguard the security and privacy of personal information.

[0053] In operation S220, the fused semantic features of behavioral data are obtained, sequence features are extracted from the fused semantic features, and based on the sequence features, a multi-layer neural network is used to obtain customer sentiment features and customer needs within a preset time period.

[0054] According to embodiments of this application, fused semantic features of behavioral data are obtained. Taking text data as an example, the behavioral data is fed into a Bidirectional Encoder Representations from Transformers (BERT) model for context encoding. The BERT model can capture long-range dependencies and implicit sentiments in the context. Next, sequence features are extracted from the fused semantic features. The output features of the BERT model are passed to a Bidirectional Long Short-Term Memory (BiLSTM) network layer to capture the temporal nature of emotions. Especially for customers with significant emotional fluctuations, BiLSTM can further analyze the contextual information to enhance emotional understanding. Finally, a Multilayer Perceptron (MLP) maps the features output by BiLSTM to emotional categories to obtain customer emotional characteristics within a preset time period, and simultaneously obtain customer needs within the preset time period. For example, if a customer asks, "How much interest will I get on a 100,000 yuan deposit over three years?", the system will... Figure 6 As shown, chat logs are converted into word embedding vectors and fed into a BERT model for contextual encoding. The output features of the BERT model are then passed to a BiLSTM layer to enhance sentiment understanding. Finally, the MLP maps the features output by the BiLSTM to one of three categories: positive, neutral, or negative, and identifies the customer's need as "fixed deposit." By using the BERT model to contextually encode text data, capturing long-range dependencies and implicit sentiment, and utilizing BiLSTM to extract sequence features, the emotional understanding of customers experiencing mood fluctuations is enhanced. Ultimately, the MLP maps the customer's emotional characteristics and need category, thereby achieving accurate sentiment analysis and need prediction.

[0055] According to embodiments of this application, a warning message is generated in response to a customer's emotional characteristics continuously exceeding a warning value within a preset time period. Specifically, based on customer behavioral data, the trend of their emotional changes is analyzed to identify whether the customer experiences emotional fluctuations and whether there is a long-term accumulation of negative emotions such as anxiety and disappointment. When the emotional fluctuation exceeds the warning value (e.g., the emotional index is negative for three consecutive days), a warning message is generated, notifying customer service for intervention. By timely identifying customer emotional fluctuations, potential customer emotions can be effectively captured, providing an opportunity for customer service intervention, preventing further deterioration of customer emotions, and thus improving customer satisfaction and loyalty.

[0056] In operation S230, in response to the emergence of incremental data in the behavioral data, the knowledge graph is updated based on the behavioral data and the incremental data. The knowledge graph is pre-built based on the behavioral data.

[0057] According to embodiments of this application, when new behavioral data emerges, semantic analysis is first performed on the incremental data to extract key information. Then, the existing knowledge graph is dynamically updated based on this key information, adding new nodes or relationships, or updating the attributes of existing nodes. For example, if a customer mentions that they are planning to buy a property and inquires about mortgage interest rates, repayment methods, etc., after identifying the customer's "housing purchase needs," this information is recorded in the customer's dynamic knowledge graph. After the knowledge graph is updated, the customer node is associated with the housing purchase-related demand nodes. By dynamically updating the knowledge graph, customer needs can be updated in real time, ensuring that the knowledge graph can reflect the latest behavioral changes in real time and maintain a consistent service experience throughout multi-turn dialogues.

[0058] When operating S240, based on customer emotional characteristics, customer needs, and knowledge graphs, the target product is obtained, and a communication strategy is generated based on customer emotional characteristics.

[0059] According to embodiments of this application, a reinforcement learning model is used to predict target products based on customer emotional characteristics, customer needs, and knowledge graphs. For example, based on customer needs and the knowledge graph, "buying a house" and "mortgage" are identified as core issues of concern to customers. Further, by combining the hesitant emotional characteristics exhibited by customers, the model predicts the financial products that customers may need in the future, such as low-risk, stable mortgage products. Simultaneously, additional financial advice (such as post-purchase financial planning) is provided to meet the customer's housing needs while alleviating their concerns about financial risks.

[0060] According to embodiments of this application, communication strategies are generated based on customer emotional characteristics, aiming to adopt corresponding communication methods for customers in different emotional states in order to improve customer experience and communication effectiveness. For example, when a customer shows anxiety, the communication strategy should focus on reassurance, with a slower speaking speed, gentler language, avoidance of complex terminology, and more affirmation and encouragement to make the customer feel understood and supported; if the customer is in a positive mood, communication can be more direct and enthusiastic, and relevant products or services can be recommended in a timely manner to enhance customer satisfaction and loyalty.

[0061] The above recommendation process takes into account the customer's emotional state, specific needs, and the correlation between financial products in order to provide customers with personalized and accurate product recommendations.

[0062] When operating the S250, in response to receiving semantic input from the customer, it communicates with the customer according to the communication strategy and recommends target products during the communication process.

[0063] According to an embodiment of this application, in response to receiving semantic input from a customer, communication is conducted with the customer in accordance with the communication strategy generated in step S240, and the target product obtained in step S240 is recommended during the communication process.

[0064] According to embodiments of this application, after recommending a target product during the communication process, the method further includes: collecting customer feedback based on the target product to obtain feedback data; and dynamically adjusting the parameters of the reinforcement learning model based on the feedback data. For example, if a mortgage product is recommended to a customer, the customer's feedback might be clicking the product link, asking for more details, or explicitly rejecting the recommendation. The system uses this feedback data as input to the reinforcement learning model in step S240, adjusting the model parameters based on the customer's positive or negative feedback to optimize future recommendation strategies. Through dynamic adjustment, the recommendation system can continuously learn and adapt to customer preferences, improving the accuracy and success rate of recommendations.

[0065] Figure 3 A flowchart illustrating the fusion semantic features of acquiring behavioral data according to an embodiment of this application is shown.

[0066] like Figure 3 As shown, the fusion semantic features for acquiring behavioral data in this embodiment include operations S310 to S320.

[0067] In operation S310, multimodal data is identified and semantic features corresponding to each modality are obtained. The multimodal data includes at least two of the following: text data, speech data, and image data.

[0068] In operation S320, the semantic features corresponding to each modality data are assigned corresponding weights, and weighted feature fusion is performed to obtain fused semantic features.

[0069] According to embodiments of this application, the BERT model is used to identify text data to obtain semantic features corresponding to the text data; the automatic speech recognition technology is used to transcribe the speech data into text data, the BERT model is further used to identify the text data, and the audio sentiment analysis technology (e.g., Mel frequency cepstral coefficients) is used to extract emotional features from the speech data to obtain semantic features corresponding to the speech data; and the image recognition technology is used to perform content recognition on the image data to obtain visual semantic features corresponding to the image data.

[0070] According to the embodiments of this application, the importance weights of each modality data are determined based on the actual application scenario and requirements. The semantic features extracted from each modality are weighted and fused. An attention mechanism is used to dynamically adjust the weights of each modality so as to flexibly adjust according to different situations, and finally obtain the fused semantic features.

[0071] For example, a customer inquires with a bank about opening a pension account and uploads a screenshot of a policy document about pensions. Through alternating voice and text, the customer expresses concern about future pension returns and confusion regarding specific account operations. First, image recognition technology (such as convolutional neural networks) can analyze the uploaded screenshot of the pension policy document to identify relevant content. Based on the image analysis results, key information related to the pension account is extracted, and appropriate interpretations are provided according to the customer's needs. Second, automatic speech recognition technology converts the voice data into text data, and audio sentiment analysis technology identifies the emotions in the voice. The customer's tone is found to be slightly confused and anxious. Further processing of the text using a sentiment analysis model determines the customer's emotional state to be "negative." Finally, based on the actual application scenario and needs, appropriate weights are assigned to each modality of data, and these weighted features are fused to obtain fused semantic features. By processing multimodal data and performing weighted feature fusion, information from various modalities can be integrated. Considering the importance of different modalities in different application scenarios, this approach not only allows for a more comprehensive capture and understanding of customer behavior and needs but also a more accurate understanding of customer needs and emotions, thereby improving the accuracy of recommendations and predictions.

[0072] Figure 4 The flowchart illustrating the updating of a knowledge graph based on behavioral data and incremental data according to an embodiment of this application is shown in the illustration.

[0073] like Figure 4 As shown, updating the knowledge graph based on behavioral data and incremental data in this embodiment includes operations S410 to S420.

[0074] In operation S410, a target dataset is generated based on behavioral data and incremental data. The data in the target dataset is used to indicate entities and entity relationships. Entities include customers and products.

[0075] In operation S420, a knowledge graph is constructed based on entities and entity relationships, using the target dataset as sample data.

[0076] According to embodiments of this application, behavioral data can be historical chat logs, and incremental data can be recent search behavior and clicked product links, etc. Integrating this data generates a target dataset. Entities in the target dataset include basic information about customers and products, and entity relationships include interaction relationships between customers and products, such as customer purchase history, browsing behavior, and product interest. Each time a customer generates a new interaction, the knowledge graph is updated in real time. For example, when a customer mentions "buying a house," a relevant node for buying a house is automatically added to the knowledge graph and associated with the customer node. This target dataset can serve as the foundation for building a knowledge graph, enabling a more accurate understanding of customer needs and preferences, thereby providing more personalized services and product recommendations.

[0077] According to embodiments of this application, the data type format of a knowledge graph can be represented by triples. Taking customers and products as entities, and the interaction behavior between customers and products as entity relationships as an example, the triple can be represented as (customer, interaction behavior, product). For example, (customer A, interested, product B) can represent the entity relationship of "interested" between customer A and product B. Customer A can be the head entity, and product B can be the tail entity. Based on the entity relationship of "interested", an association edge can be constructed from the head entity customer A to the tail entity product B. The direction of this association edge is the direction from the head entity to the tail entity.

[0078] Knowledge graphs built from behavioral and incremental data can intuitively display the complex relationships between customers and products. Furthermore, the knowledge graphs can be updated as incremental data is updated, ensuring the timeliness and accuracy of information. This provides strong support for precision marketing, personalized recommendations, and customer relationship management.

[0079] Figure 5 A flowchart illustrating the process of obtaining the target product according to an embodiment of this application is shown.

[0080] like Figure 5 As shown, obtaining the target product in this embodiment includes operations S510 to S530.

[0081] When operating the S510, the target product type is obtained based on customer needs and knowledge graphs using reinforcement learning.

[0082] When operating the S520, customer product preferences are obtained based on customer emotional characteristics and according to preset matching rules.

[0083] In operating S520, the target product is obtained from the target product type based on product preferences.

[0084] According to embodiments of this application, a reinforcement learning framework is constructed based on a state space, action space, reward function, and algorithm. The state space includes basic customer information, behavioral data, emotional characteristics, and other data. The action space refers to all financial products or services that can be recommended, such as fixed deposits and wealth management products. The reward function is designed based on customer feedback; for example, if a customer clicks on a recommended financial product link, it is considered positive feedback; if the customer rejects the recommendation, it is considered negative feedback. The strategy is updated according to the reward value, enabling the model to gradually learn customer preferences and recommend the most relevant products. In this embodiment, a Deep Q Network (DQN) algorithm can be used for training. The model generates the optimal recommendation strategy by continuously optimizing the Q-value function.

[0085] According to embodiments of this application, the target product type is obtained using reinforcement learning based on customer needs and a knowledge graph. For example, when a customer communicates with customer service, mentioning that they are planning to buy a property and inquiring about mortgage interest rates, repayment methods, etc., the system first identifies "buying a house" and "mortgage" as the core issues of concern to the customer. This information is then recorded in the customer's dynamic knowledge graph. After the knowledge graph is updated, the customer's demand nodes related to buying a house (such as "mortgage") are associated with the customer nodes. Finally, combining the knowledge graph, a reinforcement learning model is used to predict the target product type the customer may need (such as mortgage products, wealth management products, etc.).

[0086] According to embodiments of this application, preset matching rules are established based on customer emotional characteristics and core product attributes. Specifically, customer emotional characteristics are first categorized, such as positive, neutral, and negative; then, key product attributes sensitive to emotions are extracted to form a tag library, such as high risk, medium risk, and low risk, ensuring they correspond to emotional types; finally, a mapping between customer emotional characteristics and products is established to obtain preset matching rules. Furthermore, based on customer emotional characteristics, customer product preferences are obtained according to the preset matching rules. For example, if a customer's emotional characteristic is identified as negative, then the customer's product preference is determined to be low risk.

[0087] According to embodiments of this application, a target product is obtained from the target product type by combining the target product type with product preferences. For example, a low-risk product is selected as the target product from the wealth management product type. By accurately locating the target product type through reinforcement learning, and combining customer emotional characteristics and preset rules to match product preferences, the most suitable specific product is selected from the target product type to achieve personalized recommendations, which helps improve customer satisfaction and recommendation efficiency.

[0088] Based on the above customer service methods, this application also provides a customer service device. The following will be combined with... Figure 6 The device is described in detail.

[0089] Figure 7 A schematic block diagram of a customer service device according to an embodiment of this application is shown.

[0090] like Figure 7 As shown, the customer service device 700 of this embodiment includes an acquisition module 710, a fusion module 720, an incremental module 730, an output module 740, and a response module 750.

[0091] The acquisition module 710 is used to acquire customer behavior data within a preset time period. In one embodiment, the acquisition module 710 can be used to perform the operation S210 described above, which will not be repeated here.

[0092] The fusion module 720 is used to acquire fused semantic features of behavioral data, extract sequence features from the fused semantic features, and use a multi-layer neural network based on the sequence features to obtain customer sentiment features and customer needs within a preset time period. In one embodiment, the fusion module 720 can be used to perform the operation S220 described above, which will not be repeated here.

[0093] The incremental module 730 is used to update the knowledge graph based on the behavioral data and the incremental data in response to the occurrence of incremental data in the behavioral data. The knowledge graph is pre-constructed based on the behavioral data. In one embodiment, the incremental module 730 can be used to perform the operation S230 described above, which will not be repeated here.

[0094] The output module 740 is used to obtain the target product based on customer emotional characteristics, customer needs, and a knowledge graph, and to generate a communication strategy based on customer emotional characteristics. In one embodiment, the output module 740 can be used to perform the operation S240 described above, which will not be repeated here.

[0095] The response module 750 is used to respond to received semantic input from the customer, communicate with the customer according to the communication strategy, and recommend target products during the communication process. In one embodiment, the response module 750 can be used to perform the operation S250 described above, which will not be repeated here.

[0096] According to an embodiment of this application, the customer service device 700 further includes a weighting module for identifying multimodal data and obtaining semantic features corresponding to each modality. The multimodal data includes at least two of the following: text data, voice data, and image data. The module assigns corresponding weights to the semantic features corresponding to each modality and performs weighted feature fusion to obtain fused semantic features. In one embodiment, the weighting module can be used to perform the operations S310 to S320 described above, which will not be repeated here.

[0097] According to an embodiment of this application, the customer service device 700 further includes a knowledge graph module, used to generate a target dataset based on behavioral data and incremental data. The data in the target dataset is used to indicate entities and entity relationships, and the entities include the customers and products. A knowledge graph is constructed based on the entities and entity relationships using the target dataset as sample data. In one embodiment, the knowledge graph module can be used to perform the operations S410 to S420 described above, which will not be repeated here.

[0098] According to an embodiment of this application, the customer service device 700 further includes a product module, used to obtain a target product type based on customer needs and a knowledge graph using reinforcement learning; obtain customer product preferences based on customer emotional characteristics according to a preset matching rule; and obtain a target product from the target product type based on the product preferences. In one embodiment, the product module can be used to perform the operations S510 to S530 described above, which will not be repeated here.

[0099] According to embodiments of this application, any multiple modules among the acquisition module 710, fusion module 720, incremental module 730, output module 740, and response module 750 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the acquisition module 710, fusion module 720, incremental module 730, output module 740, and response module 750 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the acquisition module 710, fusion module 720, incremental module 730, output module 740, and response module 750 may be implemented at least partially as a computer program module that can perform corresponding functions when the computer program module is run.

[0100] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing a customer service method according to an embodiment of this application.

[0101] like Figure 8As shown, an electronic device 1500 according to an embodiment of this application includes a processor 1501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1502 or a program loaded from a storage portion 1508 into a random access memory (RAM) 1503. The processor 1501 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1501 may also include onboard memory for caching purposes. The processor 1501 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0102] RAM 1503 stores various programs and data required for the operation of electronic device 1500. Processor 1501, ROM 1502, and RAM 1503 are interconnected via bus 1504. Processor 1501 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 1502 and / or RAM 1503. It should be noted that the programs may also be stored in one or more memories other than ROM 1502 and RAM 1503. Processor 1501 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0103] According to embodiments of this application, the electronic device 1500 may further include an input / output (I / O) interface 1505, which is also connected to a bus 1504. The electronic device 1500 may also include one or more of the following components connected to the input / output (I / O) interface 1505: an input section 1506 including a keyboard, mouse, etc.; an output section 1507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1508 including a hard disk, etc.; and a communication section 1509 including a network interface card such as a LAN card, modem, etc. The communication section 1509 performs communication processing via a network such as the Internet. A drive 1510 is also connected to the input / output (I / O) interface 1505 as needed. A removable medium 1511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 1510 as needed so that computer programs read from it can be installed into the storage section 1508 as needed.

[0104] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0105] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 1502 and / or RAM 1503 and / or one or more memories other than ROM 1502 and RAM 1503 described above.

[0106] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this application.

[0107] When the computer program is executed by the processor 1501, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0108] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 1509, and / or installed from the removable medium 1511. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0109] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 1509, and / or installed from the removable medium 1511. When the computer program is executed by the processor 1501, it performs the functions defined in the system of this application embodiment. According to the embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0110] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0111] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0112] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

[0113] The embodiments of this application have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of this application. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. The scope of this application is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this application, and all such substitutions and modifications should fall within the scope of this application.

Claims

1. A customer service method, characterized in that, The method includes: Obtain customer behavior data within a preset time period; The fused semantic features of the behavioral data are obtained, sequence features are extracted from the fused semantic features, and customer emotion features and customer needs within the preset time period are obtained based on the sequence features using a multi-layer neural network. In response to the emergence of incremental data in the behavioral data, the knowledge graph is updated based on the behavioral data and the incremental data, the knowledge graph being pre-built based on the behavioral data; Based on the customer emotional characteristics, the customer needs, and the knowledge graph, a target product is obtained, and a communication strategy is generated based on the customer emotional characteristics. In response to receiving semantic input from the customer, the system communicates with the customer according to the communication strategy and recommends the target product during the communication process.

2. The customer service method according to claim 1, characterized in that, The behavioral data is multimodal data, and the fused semantic features of the behavioral data are obtained, including: Identify the multimodal data and obtain the semantic features corresponding to each modality of data. The multimodal data includes at least two of the following: text data, speech data, and image data. The semantic features corresponding to each modality data are assigned corresponding weights, and weighted feature fusion is performed to obtain the fused semantic features.

3. The customer service method according to claim 1, characterized in that, Updating the knowledge graph based on the behavioral data and the incremental data includes: A target dataset is generated based on the behavioral data and the incremental data. The data in the target dataset is used to indicate entities and entity relationships. The entities include the customers and products. Using the target dataset as sample data, a knowledge graph is constructed based on the entities and their relationships.

4. The customer service method according to claim 1, characterized in that, Based on the customer sentiment characteristics, customer needs, and the knowledge graph, the target product is obtained, including: Based on the customer needs and the knowledge graph, the target product type is obtained using reinforcement learning. Based on the customer's emotional characteristics, the customer's product preferences are obtained according to a preset matching rule; Based on the product preference, the target product is obtained from the target product type.

5. The customer service method according to claim 4, characterized in that, After recommending the target product during the communication process, it also includes: Collect customer feedback based on the target product to obtain feedback data; The reinforcement learning model parameters are dynamically adjusted based on the feedback data.

6. The customer service method according to claim 1, characterized in that, After obtaining customer behavior data within a preset time period, the process also includes: The behavioral data is encrypted and / or de-identified. The behavioral data is subjected to differential privacy processing based on a preset noise level.

7. The customer service method according to claim 1, characterized in that, After obtaining the customer emotional characteristics within the preset time period, the method further includes: generating a warning message in response to the customer emotional characteristics continuously exceeding the warning value within the preset time period.

8. A customer service device, characterized in that, include: The acquisition module is used to acquire customer behavior data within a preset time period; The fusion module is used to obtain the fusion semantic features of the behavioral data, extract sequence features from the fusion semantic features, and obtain customer emotion features and customer needs within the preset time period based on the sequence features using a multi-layer neural network. An incremental module is used to update a knowledge graph based on the behavioral data and the incremental data in response to the occurrence of incremental data in the behavioral data. The knowledge graph is pre-built based on the behavioral data. The output module is used to obtain the target product based on the customer emotional characteristics, the customer needs and the knowledge graph, and to generate a communication strategy based on the customer emotional characteristics. The response module is used to respond to the semantic input received from the customer, communicate with the customer according to the communication strategy, and recommend the target product during the communication process.

9. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

Images