Content recommendation method and device, equipment and medium

Abstract

The invention belongs to the field of artificial intelligence and the field of financial science and technology, and relates to a content recommendation method, which comprises the following steps: when an interaction operation of a target customer is detected, obtaining target interaction data of the target customer; performing feature extraction on the target interaction data by adopting a deep learning model to obtain a target feature vector of the target customer; adopting a reinforcement learning model to perform strategy generation processing on the target feature vector to obtain a target service strategy of the target customer; obtaining a historical customer portrait of a historical customer, and obtaining a target customer portrait of the target customer based on the target feature vector and the historical customer portrait; determining to-be-recommended content of the target customer from historical content recommendation information corresponding to the historical customer portrait; and adopting the target service strategy to recommend the to-be-recommended content to the target customer. The invention further provides a device, equipment and a medium. In addition, the invention also relates to a block chain technology, and the target interaction data can be stored in a block chain. The accuracy of content recommendation can be improved.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence technology and financial technology, and in particular to a content recommendation method, device, equipment and medium. Background Art

[0002] With the increasingly fierce market competition, especially at a time when the traffic dividend is gradually fading, the insurance industry is facing unprecedented challenges. In order to seize the market opportunity, companies have turned to refined operation strategies, striving to accurately capture and meet the diverse needs of consumers. In the insurance service scenario, the importance of customer relationship management systems has become increasingly prominent, among which the application of natural language processing technology provides strong support for understanding the content of user communication.

[0003] However, existing customer relationship management systems based on natural language processing technology still have obvious deficiencies in semantic understanding. Most of them can only perform shallow semantic analysis, and it is difficult to deeply explore the key demands of customers, resulting in a significant reduction in the accuracy of content recommendations. In addition, these systems lack flexibility and pertinence in the formulation of service strategies, and are unable to provide customized services based on the different characteristics of customers, thus affecting the service success rate and customer satisfaction. Especially in the field of auto insurance services, customer demands and emotions have a decisive influence on service results. Unfortunately, current systems often cannot accurately capture and respond to these nuances, resulting in inaccurate content recommendations and unsatisfactory service results.

[0004] In summary, the main problems with existing technologies include: insufficient semantic understanding of user communication content, lack of customization and flexibility in service strategies, and low accuracy of content recommendations. Summary of the invention

[0005] The purpose of the embodiments of the present application is to propose a content recommendation method, device, equipment and medium to solve the existing problems of insufficient semantic understanding depth of user communication content, lack of customization and flexibility of service strategies, and low accuracy of content recommendation.

[0006] In the first aspect, a content recommendation method is provided, which adopts the following technical solution:

[0007] When an interactive operation of a target customer is detected, the target interaction data of the target customer is obtained; a preset deep learning model is used to perform feature extraction on the target interaction data to obtain a target feature vector of the target customer; a preset reinforcement learning model is used to perform strategy generation processing on the target feature vector to obtain a target service strategy for the target customer; historical customer portraits of historical customers are obtained, and a target customer portrait of the target customer is obtained based on the target feature vector and the historical customer portraits; the target content to be recommended for the target customer is determined from the historical content recommendation information corresponding to the historical customer portraits; and the target service strategy is used to recommend the content to be recommended to the target customer.

[0008] In a second aspect, a content recommendation device is provided, which adopts the following technical solution:

[0009] A detection module, used for acquiring target interaction data of a target customer when an interaction operation of a target customer is detected;

[0010] An extraction module is used to extract features from target interaction data using a preset deep learning model to obtain a target feature vector of a target customer;

[0011] A generation module is used to use a preset reinforcement learning model to perform strategy generation processing on the target feature vector to obtain a target service strategy for the target customer;

[0012] An acquisition module is used to acquire historical customer portraits of historical customers, and obtain a target customer portrait of a target customer based on the target feature vector and the historical customer portraits;

[0013] A determination module, used to determine the content to be recommended for the target customer from the historical content recommendation information corresponding to the historical customer profile;

[0014] The recommendation module is used to adopt a target service strategy to recommend content to target customers.

[0015] In a third aspect, a computer device is provided, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the above-mentioned content recommendation method are implemented.

[0016] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned content recommendation method.

[0017] In the solution implemented by the above-mentioned content recommendation method, device, equipment and medium, the deep learning model is used to deeply mine the needs and emotional characteristics in the target customer interaction data, which significantly improves the accuracy of semantic understanding and enables customer service personnel to accurately grasp the core demands of customers. With the help of the reinforcement learning model, service strategies can be dynamically adjusted and optimized to achieve customization and flexibility of service strategies, which is especially suitable for complex and changeable customer needs in auto insurance services. Combining historical customer portraits with target feature vectors, the generated target customer portraits further refine customer segmentation to ensure the accuracy and personalization of content recommendations. This not only enhances the customer service staff's speech guidance and customer Q&A capabilities, but also promotes customized content recommendations, greatly improving the conversion success rate of customer service staff's online communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0020] Figure 2 It is a flowchart of a content recommendation method provided by this application;

[0021] Figure 3 It is a structural diagram of a content recommendation device provided by the present application;

[0022] Figure 4 It is a structural schematic diagram of a computer device provided by this application. DETAILED DESCRIPTION

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0024] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0025] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0026] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0027] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0028] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to a laptop computer 1011, a tablet computer 1012 or a mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0029] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .

[0030] It should be noted that the content recommendation method provided in the embodiment of the present application is generally executed by a server, and accordingly, the content recommendation device is generally arranged in the server.

[0031] It should be understood that Figure 1The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0032] Continue to refer Figure 2 , shows a flow chart of an embodiment of a content recommendation method according to the present application. The content recommendation method comprises the following steps:

[0033] Step S201, when an interactive operation of a target customer is detected, target interaction data of the target customer is obtained.

[0034] In this embodiment, the content recommendation method is executed on the electronic device (eg Figure 1 The server shown in the figure) can obtain the target interaction data through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0035] Among them, target customers refer to users who interact with the insurance service system and show potential service needs or purchase intentions. These users are connected to the system through specific channels (such as web pages, apps, customer service hotlines, etc.), and their interactive behaviors and needs are the basis for the system to provide services. For example, in the auto insurance service system, a user who is consulting about auto insurance purchases or claims is a target customer. The data and behavioral characteristics of target customers are the main basis for the system to analyze and recommend.

[0036] Interactive operations refer to various behaviors performed by target customers in the insurance service system, including but not limited to clicking, inputting, selecting, sliding, etc. These operations reflect customers' interests and needs for system functions and services, and are an important way for the system to obtain customers' intentions and preferences. For example, when a customer clicks the "Learn more about auto insurance" button in the system and enters personal information to query a quote, these are all interactive operations. By monitoring and analyzing these operations, the system can understand customers' needs and expectations.

[0037] Among them, target interaction data refers to the relevant data captured and recorded by the system when the target customer performs an interactive operation. This data includes but is not limited to the customer's input content, click records, browsing history, etc. Target interaction data is the basis for the system to understand customer intentions, perform semantic analysis, and formulate strategies. For example, "I want to buy cost-effective car insurance" entered by a customer in the system is a target interaction data.

[0038] Step S202: Use a preset deep learning model to extract features from the target interaction data to obtain a target feature vector of the target customer.

[0039] Among them, the deep learning model is a machine learning algorithm based on a neural network structure that can automatically learn and extract features from a large amount of data. In the insurance service system, the deep learning model is used to extract features from the target interaction data to obtain the customer's potential needs and preferences. For example, the system can use the deep learning model to perform semantic analysis on the customer's text input and extract key features such as the intention to purchase auto insurance and price sensitivity.

[0040] The target feature vector is a vector representation obtained by the deep learning model after extracting features from the target interaction data. This vector contains the key needs and preference information of the customer, and is the basis for subsequent strategy generation and target customer profile construction. For example, a feature vector indicating that a customer is sensitive to car insurance prices and prefers high-security products is the target feature vector.

[0041] Step S203, using a preset reinforcement learning model to perform strategy generation processing on the target feature vector to obtain a target service strategy for the target customer.

[0042] Among them, the reinforcement learning model is a machine learning algorithm that learns the best strategy through continuous trial and error and optimization. In the insurance service system, the reinforcement learning model is used to generate appropriate words and service strategies based on the target feature vector. In the model training and iteration stage, based on the historical interaction data of multiple historical customers, the model simulates different service scenarios and words, observes customer feedback and satisfaction, and continuously optimizes the service strategy. For example, for price-sensitive customers, the reinforcement learning model may generate words and recommendations that emphasize the cost-effectiveness advantage.

[0043] Among them, strategy generation processing refers to the process of using reinforcement learning models to analyze and process target feature vectors to generate the optimal service strategy for target customers. These strategies include recommended content (for example, specific products), service methods, communication skills, etc., aiming to improve service success rate and customer satisfaction. For example, based on the customer's feature vector, the system may generate a set of service strategies that recommend specific auto insurance packages and provide exclusive customer service support.

[0044] Among them, the target service strategy refers to the personalized service plan formulated for the target customers, including recommended products, service methods, communication skills, etc. This strategy aims to improve service quality and customer satisfaction by accurately matching customer needs and expectations. For example, for a young customer who pays attention to service and claim settlement speed, the system may generate a set of target service strategies that include fast claim settlement service and exclusive customer service support.

[0045] Step S204, obtaining the historical customer portrait of the historical customer, and obtaining the target customer portrait of the target customer based on the target feature vector and the historical customer portrait.

[0046] Among them, historical customers refer to customers who have interacted with or purchased from the insurance service system in the past. The data and behavioral characteristics of these customers are important references for the system to analyze and recommend. For example, customers who have purchased auto insurance in the past year are historical customers. By analyzing the data of historical customers, the system can understand the customer's purchasing habits, preferences and other information, providing a basis for formulating service strategies for new customers.

[0047] Among them, the historical customer portrait refers to the user portrait obtained by comprehensively analyzing the characteristics, behaviors, preferences and other information of historical customers. This portrait reflects the overall characteristics and demand trends of historical customers and is an important basis for building new customer portraits and making content recommendations. For example, by analyzing historical customer data, the system can build a historical customer portrait of "middle-aged male, preferring high-security auto insurance."

[0048] The target customer portrait refers to the user portrait obtained after comprehensive analysis of the target customer based on the target feature vector and historical customer portrait. It is the key basis for the system to make content recommendations and formulate service strategies. For example, by combining the customer's feature vector and historical customer portrait, the system can construct a target customer portrait of "young women, focusing on service and cost-effectiveness".

[0049] Step S205, determining the content to be recommended for the target customer from the historical content recommendation information corresponding to the historical customer portrait.

[0050] Among them, historical content recommendation information refers to the content information recommended by historical customers corresponding to the inward historical customer profiles in the historical time, including product recommendations, service suggestions, etc. This information reflects the needs and preferences of historical customers and provides a reference for content recommendations for new customers.

[0051] Among them, the content to be recommended refers to the products or services that the system has selected for the target customers based on the target customer profile and historical content recommendation information. These contents are designed to meet the personalized needs of the target customers and improve service quality and customer satisfaction.

[0052] Step S206: adopting the target service strategy to recommend the content to be recommended to the target customer.

[0053] In one embodiment, based on the generated target service strategy, the selected content to be recommended can be recommended to the target customer through multiple channels such as intelligent customer service, email push, and SMS notification. For example, for a young customer who values ​​service experience and after-sales support, the system may use friendly communication techniques to recommend a car insurance product that includes 24-hour online customer service and fast claims service. By accurately matching customer needs and service strategies, the system can significantly improve service satisfaction and conversion rate.

[0054] The embodiments of the present application can deeply mine the needs and emotional characteristics in the target customer interaction data through a deep learning model, significantly improving the accuracy of semantic understanding, so that customer service personnel can accurately grasp the core demands of customers. With the help of a reinforcement learning model, it is possible to dynamically adjust and optimize service strategies to achieve customization and flexibility of service strategies, which is particularly suitable for complex and changeable customer needs in auto insurance services. Combining historical customer portraits with target feature vectors, the generated target customer portraits further refine customer segmentation to ensure the accuracy and personalization of content recommendations. This not only enhances the customer service staff's ability to provide guidance and answer customer questions, but also promotes customized content recommendations, greatly improving the conversion success rate of customer service staff's online communication.

[0055] In some optional implementations of this embodiment, step 202, using a preset deep learning model to extract features from the target interaction data to obtain a target feature vector of the target customer, specifically includes the following steps:

[0056] Preprocess the target interaction data to obtain preprocessed text data; use the sentiment analysis algorithm in the preset deep learning model to determine the sentiment tendency in the text data and obtain the emotional state information of the target customers; use the topic modeling algorithm in the deep learning model to cluster the text data and obtain the topic information that the target customers are concerned about; use the keyword extraction algorithm in the deep learning model to extract keywords from the text data and obtain the preference information of the target customers; construct the target feature vector of the target customers based on the emotional state information, topic information and preference information.

[0057] Among them, preprocessing includes text cleaning, word segmentation, removal of stop words, etc., to preprocess the target interaction data for subsequent analysis and processing.

[0058] Among them, text data refers to the carrier of target customer interaction information after preprocessing, which is presented in plain text form. It comes directly from the customer's communication content, such as chat records, consulting questions, evaluation feedback, etc. Text data reflects the customer's true intention and emotional state, and is an important basis for constructing the target customer feature vector. For example, "I am not satisfied with the auto insurance claim process" is a typical text data, which contains the customer's dissatisfaction with the auto insurance service.

[0059] Among them, the sentiment analysis algorithm is a component in the deep learning model used to judge the emotional tendency of text data. It analyzes the words, phrases and sentence structures in the text to identify positive, negative or neutral emotional colors, thereby obtaining the emotional state information of the target customers.

[0060] Among them, emotional state information is the output result obtained after the sentiment analysis algorithm processes the text data. It describes the emotional state of the target customer at a specific moment of interaction, such as happiness, sadness, anger, satisfaction or dissatisfaction. Emotional state information is one of the important dimensions for constructing the target customer feature vector, which reflects the customer's psychological needs and emotional reactions.

[0061] Among them, the topic modeling algorithm is a component in the deep learning model used to perform topic clustering on text data. It automatically discovers potential topics in a text collection by analyzing features such as word co-occurrence relationships and semantic similarity in the text, and gives the probability distribution of words under each topic. The topic modeling algorithm can reveal the hot topics and points of interest that customers are concerned about, and provide information on the topic dimension for constructing the target customer feature vector. For example, in the field of auto insurance services, the topic modeling algorithm may find that customers generally pay attention to topics such as "claims process", "premium price" and "value-added services".

[0062] Among them, topic clustering is one of the core functions of the topic modeling algorithm, which refers to the process of merging text data with similar content or semantics into the same topic. Through topic clustering, the system can automatically identify and divide the subject areas that customers are concerned about, providing a basis for subsequent content recommendations and service strategy formulation.

[0063] Among them, topic information is one of the output results obtained after the topic modeling algorithm processes the text data. It describes the subject areas that the target customers are concerned about and their degree of importance. Topic information is one of the important dimensions for constructing the target customer feature vector, which reflects the customer's focus of interest and hot spots.

[0064] Among them, preference information is the target customer's preference and tendency information for specific products or services extracted from text data through keyword extraction algorithms. It reflects the customer's personal preferences and priorities in purchasing decisions and is one of the key dimensions for constructing the target customer feature vector.

[0065] In one example, the method of this embodiment is applied to the customer relationship management system of the insurance industry, especially the online service platform for products such as auto insurance and life insurance. The target interaction data mainly comes from channels such as customer online consultation, evaluation feedback, and social media interaction. The system first preprocesses these data, including steps such as removing irrelevant characters, word segmentation, and removing stop words, to obtain preprocessed text data. For example, the customer input "I am not satisfied with the speed of auto insurance claims" is processed into "I, auto insurance, claims, speed, dissatisfied", which provides a basis for subsequent analysis. The sentiment analysis algorithm in the preset deep learning model is used to judge the sentiment tendency of the preprocessed text data. The algorithm analyzes the vocabulary, phrases, and sentence structures in the text to identify positive, negative, or neutral emotional colors and obtain the emotional state information of the target customer. For example, the above text data is judged to be negative, indicating that the customer is dissatisfied with the speed of auto insurance claims. The topic modeling algorithm in the deep learning model is used to perform topic clustering on the text data. The algorithm automatically discovers potential topics in the text collection by analyzing features such as word co-occurrence and semantic similarity in the text, and gives the probability distribution of words under each topic. Through topic clustering, the system can identify hot topics and points of interest that customers are concerned about. For example, in the field of auto insurance services, the system may find that customers generally pay attention to topics such as "claims process", "premium price" and "value-added services". The keyword extraction algorithm in the deep learning model is used to extract keywords that can reflect customer preferences and needs from the text data. These keywords are an important basis for constructing the target customer feature vector. For example, keywords such as "auto insurance" and "claims speed" are extracted from the above text data, indicating that customers have a high concern about the claims speed of auto insurance products. Finally, the target feature vector of the target customer can be constructed based on the emotional state information, topic information and preference information. The feature vector contains multiple dimensions, each of which corresponds to a characteristic attribute of the customer. For example, emotional state information can be represented by three values: positive, negative or neutral; topic information can be represented by a vector containing multiple topic labels; preference information can be represented by a vector containing keywords and their weights. By combining these feature attributes, a complete target feature vector is formed.

[0066] The embodiment of the present application can perform multi-dimensional analysis of the target interaction data through a deep learning model, thereby accurately capturing customer emotions, topics of concern, and preferences. The preprocessing step ensures data quality and lays a solid foundation for subsequent analysis. The sentiment analysis algorithm can keenly capture the emotional color in the text and provide a powerful tool for understanding the emotional state of customers. The topic modeling algorithm effectively reveals the hot topics that customers are concerned about through cluster analysis, which helps companies grasp market dynamics. The keyword extraction algorithm deeply explores the customer's preference information and provides a key basis for personalized services. The target feature vector constructed based on the above information can fully reflect the customer's real needs and interests.

[0067] In some optional implementations of this embodiment, in step S203, before the target service strategy of the target customer is obtained by using a preset reinforcement learning model to perform strategy generation processing on the target feature vector, the following steps are specifically included:

[0068] Obtain historical interaction data of multiple historical customers; use a deep learning model to extract features from the historical interaction data to obtain historical feature vectors of multiple historical customers; based on the historical feature vectors, preset interaction scripts, and preset service results, train and iterate the preset initial reinforcement learning model to obtain a preset reinforcement learning model.

[0069] Among them, multiple historical customers refer to a collection of customer entities that have interacted with the insurance company in the past period of time. The historical data of these customers is used to train and optimize subsequent service models. For example, in the system of a car insurance company, it may contain the data of all customers who have purchased car insurance and had claims consultation in the past year, which is used to analyze the needs and preferences of customers in the claims process.

[0070] Among them, historical interaction data refers to various information records generated by multiple historical customers in the process of interacting with insurance companies, including but not limited to text chats, voice calls, email exchanges, etc. It is used to reflect the customer's historical communication content and behavioral characteristics. For example, the vehicle information, accident description, and previous communication records provided by customers when consulting auto insurance claims are all part of historical interaction data.

[0071] The historical feature vector is a numerical vector that can characterize the customer's historical behavior patterns and preferences, obtained by extracting features from the deep learning model of historical interaction data. For example, by extracting and encoding information such as keywords, emotional tendencies, and topic categories in the customer's historical communication content, a historical feature vector containing multiple dimensions can be obtained to describe the customer's comprehensive characteristics.

[0072] Interactive speech refers to the language expression and skills used by insurance companies when interacting with customers, including but not limited to opening remarks, questioning methods, answer templates, etc. In the training of reinforcement learning models, interactive speech is used as part of the input to simulate real customer service scenarios and help the model learn how to choose the most appropriate communication method based on different customer characteristics. For example, for emotionally agitated customers, the system may use soothing language and more detailed explanations to respond to their questions.

[0073] Among them, service results refer to the effects or results achieved by insurance companies after providing services to customers, including but not limited to customer satisfaction, claims processing efficiency, complaint rate and other indicators. These results are an important basis for measuring service quality and effectiveness, and are also the key basis for reinforcement learning models to optimize and iterate strategies. For example, in a car insurance claims service, if the customer is satisfied with the processing results and the claims process is efficient and smooth, then the result of this service can be marked as positive and used for subsequent model training and optimization.

[0074] The initial reinforcement learning model refers to a machine learning model used to generate service strategies after no training or only preliminary training. The model receives customer feature vectors and interactive dialogues as input, and continuously optimizes strategy selection according to the preset reward function to achieve the best service effect.

[0075] Among them, service strategy refers to a service solution or action plan generated based on the reinforcement learning model and targeting the specific needs and behavioral characteristics of target customers. It aims to accurately capture and meet the diverse needs of customers and improve service quality and customer satisfaction. For example, when it is detected that the target customer shows anxiety and dissatisfaction when consulting about auto insurance claims, a service strategy containing detailed explanations, fast claims channels, and additional care measures can be generated to ease customer emotions and improve service effectiveness.

[0076] In one example, first, the historical interaction data of multiple historical customers can be obtained from the customer relationship management system of an insurance company. These data include but are not limited to customer consultation records, claims applications, complaint feedback, etc., covering various behaviors and needs of customers in the insurance service process. Next, a deep learning model (such as a convolutional neural network CNN or a long short-term memory network LSTM) can be used to extract features from the historical interaction data. Through the learning of the model, the customer's interaction content is converted into a high-dimensional feature vector, which can accurately reflect the customer's historical behavior patterns, preferences and needs. Next, based on the extracted historical feature vectors, the preset interaction words and the preset service results, the preset initial reinforcement learning model is trained and iterated. During the training process, the model will select different interaction words according to the customer's feature vector and predict possible service results. By comparing with the actual service results, the model continuously adjusts the strategy selection to maximize the preset reward function (such as customer satisfaction, claims processing efficiency, etc.). Finally, the trained reinforcement learning model can generate personalized service strategies based on the customer's real-time feature vector. For example, when the model detects that a customer is anxious and dissatisfied when consulting about a car insurance claim, it generates a service strategy that includes detailed explanations, a fast claims channel, and additional care measures to ease the customer's emotions and improve service outcomes.

[0077] The embodiment of the present application can extract features from historical interaction data through a deep learning model, accurately capture the potential needs and preferences of multiple historical customers, and generate highly representative historical feature vectors. These vectors are used as inputs for reinforcement learning model training, effectively improving the model's ability to recognize customer behavior patterns. At the same time, combined with the preset interactive words and service results, the initial reinforcement learning model is fully trained and iterated, so that it can generate personalized service strategies based on the real-time characteristics of customers. This not only significantly improves the accuracy of service recommendations, but also enhances the customization and flexibility of service strategies.

[0078] In some optional implementations of this embodiment, step S204, obtaining a historical customer portrait of a historical customer, specifically includes the following steps:

[0079] A preset clustering algorithm is used to cluster the historical feature vectors to obtain multiple cluster centers. For each cluster center, the historical customers corresponding to the customer feature vectors around each cluster center are classified into the same category, and the customer category corresponding to each cluster center is determined to obtain multiple customer categories. For each customer category, based on the historical feature vectors, the common features of the historical customers in each customer category are counted. Based on the common features, the preset customer portrait generation rules are used to generate the customer portrait corresponding to each customer category to obtain the historical customer portrait of the historical customers.

[0080] Among them, the clustering algorithm is an unsupervised learning method used to divide the data set into multiple subsets or clusters, so that the data points in the same cluster are more similar to each other, while the data points in different clusters are less similar. In this embodiment, the clustering algorithm automatically discovers potential customer grouping patterns from the historical feature vector set without pre-defining category labels. The algorithm classifies customers with similar characteristics into one category by performing steps such as distance calculation and similarity evaluation on the historical feature vectors.

[0081] The cluster center refers to a point or vector representing the center position of each cluster or subset in the clustering algorithm. In this embodiment, the cluster center is obtained after the clustering algorithm divides the historical feature vectors, which characterizes the core features or average values ​​of each customer category. The cluster center is obtained by calculating the mean or weighted mean of all historical feature vectors in the cluster, and is used to represent the typical features of the customer category.

[0082] Among them, customer categories refer to different groups into which customers with similar characteristics or behaviors are divided according to the clustering results of historical feature vectors. In this embodiment, customer categories are obtained by dividing historical feature vectors through clustering algorithms. Customers in each category have high similarity in characteristics, while customers in different categories have great differences.

[0083] The common features refer to the information shared by all historical customer feature vectors in the same customer category, which can represent the overall features of customers in this category. In this embodiment, the common features are obtained by counting the feature vectors of historical customers in each customer category, which characterizes the typical behavior patterns or demand preferences of customers in this category.

[0084] Among them, customer portrait generation rules refer to the principles or methods for converting the characteristic information of customer groups into specific and describable customer portraits based on historical feature vectors and common characteristics.

[0085] In one example, a preset clustering algorithm (such as K-means algorithm) is used to cluster historical feature vectors. During the clustering process, the algorithm divides the feature vectors into multiple clusters according to their similarity, and the center of each cluster is the cluster center. By continuously adjusting the position of the cluster center, the sum of the distances from the feature vectors in each cluster to the cluster center is minimized, thereby obtaining the optimal clustering result. For each cluster center, the historical customers corresponding to the customer feature vectors around it can be classified into the same category, and the customer category corresponding to each cluster center is determined. In this way, multiple customer categories with different characteristics are obtained, and the customers in each category have high similarity in characteristics. For each customer category, based on the historical feature vectors, the common features of the historical customers in the category are counted. These common features include but are not limited to age, gender, occupation, purchase preferences, etc., which can represent the overall characteristics and behavior patterns of customers in this category. Finally, based on the common features, the preset customer portrait generation rules are used to generate specific customer portraits for each customer category. These portraits include descriptions of the customer's basic information, consumption habits, interests and hobbies, etc., which help companies to understand the characteristics of customer groups more intuitively.

[0086] The embodiment of the present application can efficiently group historical feature vectors through a clustering algorithm, accurately locate multiple cluster centers, and then scientifically divide customer categories. The customer feature vectors around each cluster center are reasonably classified to ensure the accuracy and representativeness of the customer categories. On this basis, for each customer category, the common features are deeply mined and counted, and these features accurately reflect the unique needs and preferences of customers in each category. Through the preset customer portrait generation rules, the common features are successfully converted into intuitive and specific customer portraits.

[0087] In some optional implementations of this embodiment, step S204, obtaining a target customer profile of the target customer based on the target feature vector and the historical customer profile, specifically includes the following steps:

[0088] Obtain the cluster center vector of each cluster center; calculate the similarity between the target feature vector and the cluster center vector to obtain a similarity value; determine the target cluster center of the cluster center vector corresponding to the highest similarity value from multiple cluster centers; determine the customer category corresponding to the target cluster center as the customer category of the target customer; obtain the target historical customer portrait corresponding to the customer category to which the target customer belongs from the historical customer portraits, and use the target historical customer portrait as the target customer portrait of the target customer.

[0089] The cluster center vector refers to the feature vector corresponding to each cluster center when the clustering algorithm is used to group the historical feature vectors. This vector can be obtained by calculating the mean of all historical feature vectors in the cluster or using other statistical methods to characterize the typical characteristics of the cluster center.

[0090] The target cluster center refers to the cluster center with the highest similarity value after calculating the similarity between the target feature vector of the target customer and all cluster center vectors.

[0091] The target historical customer portrait refers to the customer portrait corresponding to the category to which the target customer belongs, obtained from the historical customer portrait. This portrait integrates the common characteristics of all historical customers in the category, such as age, gender, purchase preferences, etc., and is used to characterize the potential characteristics and needs of the target customer.

[0092] In one example, first, a preset clustering algorithm is used to cluster the historical feature vectors to obtain multiple cluster centers. Each cluster center represents a typical group in the historical customer data and has distinct characteristics. Subsequently, the mean or median of all historical feature vectors in each cluster center is calculated to obtain the cluster center vector of each cluster center. When the target customer's interactive operation is detected, the target interactive data is obtained, and the feature extraction is performed through the deep learning model to obtain the target feature vector. Next, the similarity between the target feature vector and all cluster center vectors is calculated. The similarity calculation can use measurement methods such as cosine similarity, Euclidean distance, and Manhattan distance. In this embodiment, cosine similarity can be used as a metric because it can reflect the degree of similarity between two vectors in direction and is suitable for comparison between feature vectors. By comparing the similarity values ​​of the target feature vector and each cluster center vector, the cluster center vector corresponding to the highest similarity value is determined, which is the target cluster center. The cluster center represents the customer category closest to the target customer. According to the target cluster center, the corresponding customer category is determined to be the customer category of the target customer. Then, the target historical customer portrait corresponding to the category is obtained from the historical customer portrait. The target historical customer portrait combines the common characteristics of all historical customers in the category, such as age, gender, consumption habits, etc., and can reflect the potential needs and preferences of the target customers. Finally, the target historical customer portrait is used as the target customer portrait of the target customer.

[0093] The embodiment of the present application can accurately obtain the cluster center vector of each cluster center as a typical feature representation of the customer category. When calculating the similarity between the target feature vector and the cluster center vector, the degree of association between the target customer and each customer category can be efficiently and accurately reflected. By comparing the similarity values, the cluster center with the highest similarity is determined as the target cluster center, so as to accurately classify the target customer into the corresponding customer category. This classification process not only improves the accuracy of customer classification, but also effectively avoids the subjectivity and uncertainty in traditional classification methods. Furthermore, the target historical customer portrait of the category to which the target customer belongs is obtained from the historical customer portrait as the target customer portrait of the target customer. This step makes full use of the common characteristics of historical customers and provides a comprehensive and in-depth feature description for the target customer.

[0094] In some optional implementations of this embodiment, step S205, determining the content to be recommended for the target customer from the historical content recommendation information corresponding to the historical customer profile, specifically includes the following steps:

[0095] Obtain content recommendation information corresponding to historical customer portraits within a historical time period; based on the content recommendation information, determine target content recommendation information corresponding to the target historical customer portrait; based on the target content recommendation information, obtain the content to be recommended for the target customer.

[0096] Among them, content recommendation information refers to the specific content information recommended to historical customers corresponding to historical customer profiles within a historical time period.

[0097] The target content recommendation information refers to specific content information recommended to historical customers corresponding to the target historical customer profile within the historical time period. This information is included in the content recommendation information.

[0098] The content to be recommended refers to a specific set of content that is selected based on the target content recommendation information and is ultimately prepared to be recommended to the target customers. These contents are optimized and screened by algorithms to best meet the interests and needs of the target customers.

[0099] In one example, content recommendation information of historical customer profiles in a historical time period (such as the past year) can be extracted from the database. This information includes but is not limited to news information, movie recommendations, shopping discounts, insurance type recommendations, etc. Based on the content recommendation information, the target content recommendation information corresponding to the target historical customer profile (i.e., the historical customer profile similar to or of the same category as the target customer) is determined. Finally, based on the target content recommendation information, the specific content set that is ultimately prepared to be recommended to the target customer, i.e., the content to be recommended, is screened and optimized.

[0100] The embodiment of the present application can ensure the richness and timeliness of data by comprehensively collecting and analyzing content recommendation information within the historical time period. Then, the specific content needs of the target historical customer portrait are accurately locked, effectively improving the personalization and accuracy of content recommendations. Finally, based on the target content recommendation information, highly matched content to be recommended is intelligently generated and presented to the target customer, which not only significantly enhances the user experience satisfaction, but also promotes the efficiency and conversion rate of content distribution.

[0101] In some optional implementations of this embodiment, after the target service strategy is adopted to recommend the content to be recommended to the target customer in step S206, the following steps may be specifically included:

[0102] Obtain feedback data from target customers regarding the content to be recommended; based on the feedback data, perform parameter tuning on the deep learning model and the reinforcement learning model to obtain an optimized deep learning model and an optimized reinforcement learning model.

[0103] Among them, feedback data refers to the direct or indirect response information collected from the target customer about the recommended content after the recommended content is recommended to the target customer. This data usually comes from the customer's clicks, browsing time, purchase behavior, evaluation or satisfaction survey on the recommended content. Feedback data comes from the behavior and attitude of the target customer, characterizing the customer's acceptance and preference for the recommended content. It is used to evaluate the effectiveness of the recommendation system and serves as an important basis for subsequent model optimization. For example, if the target customer shows a high click-through rate and positive evaluation of a certain recommended content, these feedback data will be used for parameter tuning to improve the accuracy of future recommendations.

[0104] In one example, the system monitors the target customer's interactive operations such as clicking and browsing on the user interface, and then obtains target interactive data such as click frequency and dwell time from these interactive operations. Then, the pre-trained deep learning model is used to extract features from the target interactive data to obtain the target feature vector of the target customer. This vector contains key information such as customer interests and preferences. Then, the preset reinforcement learning model is used to perform strategy generation processing based on the target feature vector to generate a target service strategy for the target customer. This strategy aims to maximize customer satisfaction with the recommended content. At the same time, the system obtains the historical customer portrait of the historical customer, combines the target feature vector, and generates the target customer portrait of the target customer through methods such as similarity calculation. Based on the target customer portrait, the content to be recommended that matches the target customer portrait is screened out from the historical content recommendation information. Then, according to the target service strategy, the screened content to be recommended is recommended to the target customer. The system records the target customer's feedback data on these contents, such as click-through rate, conversion rate, etc. Finally, based on these feedback data, the deep learning model and reinforcement learning model are optimized using optimization methods such as back propagation algorithm to obtain more accurate recommendation effects.

[0105] The embodiment of the present application can ensure the directness and relevance of the data source by collecting instant feedback data from target customers on the content to be recommended, providing a solid foundation for model tuning. Subsequently, these feedback data are used to implement refined parameter adjustments for the deep learning model and the reinforcement learning model. The deep learning model can thereby understand user preferences more deeply, while the reinforcement learning model, under the guidance of user feedback, continuously optimizes the decision-making strategy and realizes dynamic adaptation of the recommendation strategy, so that subsequent recommended content is more in line with the actual needs of users, effectively improving user experience and satisfaction.

[0106] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned target interaction data, historical content recommendation information and historical interaction data, the above-mentioned target interaction data, historical content recommendation information and historical interaction data can also be stored in a node of a blockchain.

[0107] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0108] The embodiments of the present application can build and optimize related models and networks based on artificial intelligence technology, such as deep learning models, reinforcement learning models, etc. Among them, artificial intelligence (AI) models are the crystallization of theory and practice that simulate the decision-making process of human intelligence through algorithms and data analysis to solve complex problems, predict future trends, or realize automation tasks. These models use a large amount of historical data and real-time information, and are trained and optimized through a specific algorithm framework to achieve efficient, accurate and reliable performance.

[0109] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0110] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0111] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a content recommendation device, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0112] like Figure 3 As shown, the content recommendation device 400 of this embodiment includes: a detection module 401, an extraction module 402, a generation module 403, an acquisition module 404, a determination module 405 and a recommendation module 406. Among them:

[0113] The detection module 401 is used to obtain target interaction data of the target customer when an interaction operation of the target customer is detected;

[0114] An extraction module 402 is used to extract features from the target interaction data using a preset deep learning model to obtain a target feature vector of the target customer;

[0115] A generation module 403 is used to use a preset reinforcement learning model to perform strategy generation processing on the target feature vector to obtain a target service strategy for the target customer;

[0116] An acquisition module 404 is used to acquire a historical customer portrait of a historical customer, and obtain a target customer portrait of a target customer based on the target feature vector and the historical customer portrait;

[0117] A determination module 405 is used to determine the content to be recommended for the target customer from the historical content recommendation information corresponding to the historical customer portrait;

[0118] The recommendation module 406 is used to recommend the content to be recommended to the target customer by adopting the target service strategy.

[0119] The embodiments of the present application can deeply mine the needs and emotional characteristics in the target customer interaction data through a deep learning model, significantly improving the accuracy of semantic understanding, so that customer service personnel can accurately grasp the core demands of customers. With the help of a reinforcement learning model, it is possible to dynamically adjust and optimize service strategies to achieve customization and flexibility of service strategies, which is particularly suitable for complex and changeable customer needs in auto insurance services. Combining historical customer portraits with target feature vectors, the generated target customer portraits further refine customer segmentation to ensure the accuracy and personalization of content recommendations. This not only enhances the customer service staff's ability to provide guidance and answer customer questions, but also promotes customized content recommendations, greatly improving the conversion success rate of customer service staff's online communication.

[0120] In one embodiment, the extraction module 402 includes:

[0121] A preprocessing submodule is used to preprocess the target interaction data to obtain preprocessed text data;

[0122] The judgment submodule is used to use the sentiment analysis algorithm in the preset deep learning model to judge the sentiment tendency in the text data and obtain the emotional state information of the target customer;

[0123] The topic clustering submodule is used to use the topic modeling algorithm in the deep learning model to perform topic clustering on text data and obtain topic information that target customers are concerned about;

[0124] The extraction submodule is used to extract keywords from text data using the keyword extraction algorithm in the deep learning model to obtain the preference information of target customers;

[0125] The construction submodule is used to construct the target feature vector of the target customer based on the emotional state information, topic information and preference information.

[0126] The embodiment of the present application can perform multi-dimensional analysis of the target interaction data through a deep learning model, thereby accurately capturing customer emotions, topics of concern, and preferences. The preprocessing step ensures data quality and lays a solid foundation for subsequent analysis. The sentiment analysis algorithm can keenly capture the emotional color in the text and provide a powerful tool for understanding the emotional state of customers. The topic modeling algorithm effectively reveals the hot topics that customers are concerned about through cluster analysis, which helps companies grasp market dynamics. The keyword extraction algorithm deeply explores the customer's preference information and provides a key basis for personalized services. The target feature vector constructed based on the above information can fully reflect the customer's real needs and interests.

[0127] In one embodiment, the acquisition module 404 includes:

[0128] The clustering submodule is used to cluster the historical feature vectors using a preset clustering algorithm to obtain multiple cluster centers;

[0129] The first determination submodule is used to classify the historical customers corresponding to the customer feature vectors around each cluster center into the same category for each cluster center, determine the customer category corresponding to each cluster center, and obtain multiple customer categories;

[0130] A statistical submodule is used to count the common features of historical customers in each customer category based on the historical feature vectors;

[0131] The generation submodule is used to generate a customer portrait corresponding to each customer category according to common characteristics and adopt preset customer portrait generation rules to obtain historical customer portraits of historical customers.

[0132] The embodiment of the present application can efficiently group historical feature vectors through a clustering algorithm, accurately locate multiple cluster centers, and then scientifically divide customer categories. The customer feature vectors around each cluster center are reasonably classified to ensure the accuracy and representativeness of the customer categories. On this basis, for each customer category, the common features are deeply mined and counted, and these features accurately reflect the unique needs and preferences of customers in each category. Through the preset customer portrait generation rules, the common features are successfully converted into intuitive and specific customer portraits.

[0133] In one embodiment, the acquisition module 404 includes:

[0134] The first acquisition submodule is used to obtain the cluster center vector of each cluster center;

[0135] A calculation submodule is used to calculate the similarity between the target feature vector and the cluster center vector to obtain a similarity value;

[0136] A second determination submodule is used to determine a target cluster center of a cluster center vector corresponding to a cluster center vector having the highest similarity value from among multiple cluster centers;

[0137] A third determination submodule is used to determine the customer category corresponding to the target cluster center as the customer category of the target customer;

[0138] The second acquisition submodule is used to obtain a target historical customer portrait corresponding to the customer category to which the target customer belongs from the historical customer portraits, and use the target historical customer portrait as the target customer portrait of the target customer.

[0139] The embodiment of the present application can accurately obtain the cluster center vector of each cluster center as a typical feature representation of the customer category. When calculating the similarity between the target feature vector and the cluster center vector, the degree of association between the target customer and each customer category can be efficiently and accurately reflected. By comparing the similarity values, the cluster center with the highest similarity is determined as the target cluster center, so as to accurately classify the target customer into the corresponding customer category. This classification process not only improves the accuracy of customer classification, but also effectively avoids the subjectivity and uncertainty in traditional classification methods. Furthermore, the target historical customer portrait of the category to which the target customer belongs is obtained from the historical customer portrait as the target customer portrait of the target customer. This step makes full use of the common characteristics of historical customers and provides a comprehensive and in-depth feature description for the target customer.

[0140] In one embodiment, the determination module 405 includes:

[0141] The third acquisition submodule is used to obtain content recommendation information corresponding to historical customer portraits in historical time periods;

[0142] A fourth determination submodule is used for a submodule, which is used to determine target content recommendation information corresponding to a target historical customer profile based on the content recommendation information;

[0143] The submodule is used to obtain the content to be recommended for the target customer based on the target content recommendation information.

[0144] The embodiment of the present application can ensure the richness and timeliness of data by comprehensively collecting and analyzing content recommendation information within the historical time period. Then, the specific content needs of the target historical customer portrait are accurately locked, effectively improving the personalization and accuracy of content recommendations. Finally, based on the target content recommendation information, highly matched content to be recommended is intelligently generated and presented to the target customer, which not only significantly enhances the user experience satisfaction, but also promotes the efficiency and conversion rate of content distribution.

[0145] In one embodiment, the content recommendation device 400 further includes:

[0146] A first data acquisition module, used to acquire historical interaction data of multiple historical customers;

[0147] A feature extraction module is used to extract features from historical interaction data using a deep learning model to obtain historical feature vectors of multiple historical customers;

[0148] The training module is used to train and iterate the preset initial reinforcement learning model based on historical feature vectors, preset interactive scripts and preset service results to obtain the preset reinforcement learning model.

[0149] The embodiment of the present application can extract features from historical interaction data through a deep learning model, accurately capture the potential needs and preferences of multiple historical customers, and generate highly representative historical feature vectors. These vectors are used as inputs for reinforcement learning model training, effectively improving the model's ability to recognize customer behavior patterns. At the same time, combined with the preset interactive words and service results, the initial reinforcement learning model is fully trained and iterated, so that it can generate personalized service strategies based on the real-time characteristics of customers. This not only significantly improves the accuracy of service recommendations, but also enhances the customization and flexibility of service strategies.

[0150] In one embodiment, the content recommendation device 400 further includes:

[0151] The second data acquisition module is used to acquire the target customer's feedback data on the content to be recommended;

[0152] The tuning module is used to tune the parameters of the deep learning model and the reinforcement learning model based on the feedback data to obtain the optimized deep learning model and the optimized reinforcement learning model.

[0153] The embodiment of the present application can ensure the directness and relevance of the data source by collecting instant feedback data from target customers on the content to be recommended, providing a solid foundation for model tuning. Subsequently, these feedback data are used to implement refined parameter adjustments for the deep learning model and the reinforcement learning model. The deep learning model can thereby understand user preferences more deeply, while the reinforcement learning model, under the guidance of user feedback, continuously optimizes the decision-making strategy and realizes dynamic adaptation of the recommendation strategy, so that subsequent recommended content is more in line with the actual needs of users, effectively improving user experience and satisfaction.

[0154] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0155] The computer device 6 includes a memory 61, a processor 62, and a network interface 63 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 6 having a memory 61, a processor 62, and a network interface 63, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASIC), programmable gate arrays (FPGA), digital signal processors (DSP), embedded devices, etc.

[0156] Computer devices can be computing devices such as desktop computers, notebooks, PDAs, and cloud servers. Computer devices can interact with users through keyboards, mice, remote controls, touch pads, or voice control devices.

[0157] The memory 61 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 61 can be an internal storage unit of the computer device 6, such as a hard disk or memory of the computer device 6. In other embodiments, the memory 61 can also be an external storage device of the computer device 6, such as a plug-in hard disk equipped on the computer device 6, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 61 can also include both the internal storage unit of the computer device 6 and its external storage device. In this embodiment, the memory 61 is generally used to store the operating system and various application software installed on the computer device 6, such as computer-readable instructions of the content recommendation method, etc. In addition, the memory 61 can also be used to temporarily store various types of data that have been output or are to be output.

[0158] In some embodiments, the processor 62 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 62 is generally used to control the overall operation of the computer device 6. In this embodiment, the processor 62 is used to run computer-readable instructions stored in the memory 61 or process data, such as computer-readable instructions for running the content recommendation method.

[0159] The network interface 63 may include a wireless network interface or a wired network interface, and the network interface 63 is generally used to establish a communication connection between the computer device 6 and other electronic devices.

[0160] The embodiments of the present application can deeply mine the needs and emotional characteristics in the target customer interaction data through a deep learning model, significantly improving the accuracy of semantic understanding, so that customer service personnel can accurately grasp the core demands of customers. With the help of a reinforcement learning model, it is possible to dynamically adjust and optimize service strategies to achieve customization and flexibility of service strategies, which is particularly suitable for complex and changeable customer needs in auto insurance services. Combining historical customer portraits with target feature vectors, the generated target customer portraits further refine customer segmentation to ensure the accuracy and personalization of content recommendations. This not only enhances the customer service staff's ability to provide guidance and answer customer questions, but also promotes customized content recommendations, greatly improving the conversion success rate of customer service staff's online communication.

[0161] The present application also provides another implementation, namely, providing a computer-readable storage medium, which stores computer-readable instructions. The computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the content recommendation method as described above.

[0162] The embodiments of the present application can deeply mine the needs and emotional characteristics in the target customer interaction data through a deep learning model, significantly improving the accuracy of semantic understanding, so that customer service personnel can accurately grasp the core demands of customers. With the help of a reinforcement learning model, it is possible to dynamically adjust and optimize service strategies to achieve customization and flexibility of service strategies, which is particularly suitable for complex and changeable customer needs in auto insurance services. Combining historical customer portraits with target feature vectors, the generated target customer portraits further refine customer segmentation to ensure the accuracy and personalization of content recommendations. This not only enhances the customer service staff's ability to provide guidance and answer customer questions, but also promotes customized content recommendations, greatly improving the conversion success rate of customer service staff's online communication.

[0163] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods of each embodiment of the present application.

[0164] Obviously, the embodiments described above are only a part of the embodiments of the present application, rather than all the embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific embodiments, or to replace some of the technical features therein with equivalents. All equivalent structures made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, are similarly within the scope of patent protection of this application. The non-company software tools or components that appear in the embodiments of this application are merely examples and do not represent actual use.

Claims

1. A content recommendation method, characterized in that: The steps include: When an interactive operation of a target customer is detected, acquiring target interaction data of the target customer; Using a preset deep learning model, feature extraction is performed on the target interaction data to obtain a target feature vector of the target customer; Using a preset reinforcement learning model, the target feature vector is processed by strategy generation to obtain a target service strategy for the target customer; Acquire a historical customer portrait of a historical customer, and obtain a target customer portrait of the target customer based on the target feature vector and the historical customer portrait; Determining the content to be recommended for the target customer from the historical content recommendation information corresponding to the historical customer portrait; The target service strategy is adopted to recommend the content to be recommended to the target customer.

2. The method according to claim 1, characterized in that The step of using a preset deep learning model to extract features from the target interaction data to obtain a target feature vector of the target customer specifically includes: Preprocessing the target interaction data to obtain preprocessed text data; Using a sentiment analysis algorithm in a preset deep learning model to determine the sentiment tendency in the text data, and obtain the emotional state information of the target customer; Using the topic modeling algorithm in the deep learning model to perform topic clustering on the text data, and obtain topic information that the target customers are concerned about; Using the keyword extraction algorithm in the deep learning model to extract keywords from the text data, and obtain the preference information of the target customers; A target feature vector of the target customer is constructed according to the emotional state information, the topic information and the preference information.

3. The method according to claim 1, characterized in that Before the step of using a preset reinforcement learning model to perform strategy generation processing on the target feature vector to obtain the target service strategy for the target customer, the step further includes: Obtain historical interaction data of multiple historical customers; Using the deep learning model, extracting features from the historical interaction data to obtain historical feature vectors of the multiple historical customers; Based on the historical feature vectors, preset interactive words and preset service results, the preset initial reinforcement learning model is trained and iterated to obtain the preset reinforcement learning model.

4. The method according to claim 3, characterized in that The step of obtaining the historical customer portrait of the historical customer specifically includes: Using a preset clustering algorithm to cluster the historical feature vectors to obtain multiple cluster centers; For each cluster center, classify the historical customers corresponding to the customer feature vectors around each cluster center into the same category, determine the customer category corresponding to each cluster center, and obtain multiple customer categories; For each customer category, based on the historical feature vector, counting the common features of historical customers in each customer category; According to the common characteristics, a preset customer portrait generation rule is adopted to generate a customer portrait corresponding to each of the customer categories, thereby obtaining a historical customer portrait of the historical customer.

5. The method according to claim 4, characterized in that The step of obtaining a target customer portrait of the target customer based on the target feature vector and the historical customer portrait specifically includes: Obtaining the cluster center vector of each cluster center; Calculating the similarity between the target feature vector and the cluster center vector to obtain a similarity value; Determine the target cluster center of the cluster center vector corresponding to the highest similarity value from multiple cluster centers; Determining the customer category corresponding to the target cluster center as the customer category of the target customer; A target historical customer portrait corresponding to the customer category to which the target customer belongs is obtained from the historical customer portraits, and the target historical customer portrait is used as the target customer portrait of the target customer.

6. The method according to claim 5, characterized in that The step of determining the content to be recommended for the target customer from the historical content recommendation information corresponding to the historical customer portrait specifically includes: Obtaining content recommendation information corresponding to the historical customer profile within a historical time period; Based on the content recommendation information, determining target content recommendation information corresponding to the target historical customer profile; Based on the target content recommendation information, the content to be recommended for the target customer is obtained.

7. The method according to claim 1, characterized in that After the step of adopting the target service strategy to recommend the content to be recommended to the target customer, the method further includes: Acquiring feedback data from the target customer regarding the content to be recommended; Based on the feedback data, parameters of the deep learning model and the reinforcement learning model are tuned to obtain an optimized deep learning model and an optimized reinforcement learning model.

8. A content recommendation device, characterized in that: include: A detection module, configured to obtain target interaction data of a target customer when an interaction operation of the target customer is detected; An extraction module, used to use a preset deep learning model to perform feature extraction on the target interaction data to obtain a target feature vector of the target customer; A generation module, used to use a preset reinforcement learning model to perform strategy generation processing on the target feature vector to obtain a target service strategy for the target customer; An acquisition module, used to acquire a historical customer portrait of a historical customer, and obtain a target customer portrait of the target customer based on the target feature vector and the historical customer portrait; A determination module, used to determine the content to be recommended for the target customer from the historical content recommendation information corresponding to the historical customer portrait; The recommendation module is used to adopt the target service strategy to recommend the content to be recommended to the target customer.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the content recommendation method according to any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the content recommendation method according to any one of claims 1 to 7 are implemented.

Images