Personalized copywriting generation method and device, computer equipment and storage medium

Abstract

The invention belongs to the field of artificial intelligence and finance, and relates to a personalized copywriting generation method, which comprises the following steps: classifying all customers according to historical customer feature data to obtain different customer groups, obtaining copywriting keyword combinations of the customer groups according to historical copywriting of each customer group, and generating personalized copywriting keywords of the customer groups according to the copywriting keyword combinations; and constructing a customer group portrait according to the customer group characteristics and the customer group document keyword combination, matching a target customer group portrait according to the customer characteristics of the target customer, and generating a product recommendation scheme according to the customer characteristics and the target customer group portrait through a product recommendation model. And generating a personalized copywriting according to the customer features, the customer group portrait features and the product recommendation scheme through the copywriting generation large model. The invention further provides a personalized copywriting generation device, computer equipment and a storage medium. In addition, the invention also relates to a block chain technology, and historical customer feature data can be stored in a block chain. According to the invention, the client can be accurately positioned, and the generated copywriting can be accurately put.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence and financial technology, and in particular to a method, device, computer equipment and storage medium for generating personalized copy. Background Art

[0002] With the booming pet economy, the pet insurance market, as an important part of it, has shown a rapid growth trend in recent years. However, the competition in this market is also increasingly fierce. The traditional marketing methods that rely on widespread advertising, the use of fixed vocabulary and manual keyword screening are unable to cope with the increasingly diversified needs of pet owners and the highly segmented market. Traditional marketing methods have obvious limitations in accurately positioning target customer groups, achieving personalized recommendations and improving user experience, making it difficult to effectively stimulate market potential and meet the growing market demand.

[0003] Currently, online marketing in the pet insurance industry mainly relies on pre-set word libraries and manually intervened keyword selection strategies. This approach is not only inefficient, but also lacks real-time data analysis and user behavior insight capabilities, resulting in marketing content that is often lacking in pertinence and innovation, making it difficult to accurately reach potential customers' interests. In addition, with the rapid development of natural language processing (NLP) and artificial intelligence (AI) technologies, the market demand for automated and intelligent marketing solutions is becoming more urgent. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a method, device, computer equipment and storage medium for generating personalized copy, so as to solve the technical problems in the related art that the copy generation efficiency is low and the target customer group cannot be accurately located to achieve personalized recommendation.

[0005] In order to solve the above technical problems, the present application embodiment provides a method for generating personalized copy, which adopts the following technical solution:

[0006] Acquire historical customer characteristic data, and classify all customers according to the historical customer characteristic data to obtain different customer groups;

[0007] Obtaining historical copywriting of each customer group, extracting semantic features from the historical copywriting, and obtaining copywriting keyword combinations corresponding to the customer group;

[0008] Extracting customer group characteristics of the customer group, and constructing a customer group portrait based on the customer group characteristics and the copywriting keywords;

[0009] Obtain customer characteristics of target customers, and match target customer group portraits according to the customer characteristics;

[0010] According to the customer characteristics and the target customer group portrait, obtaining a product recommendation plan for the target customer through a product recommendation model;

[0011] The customer characteristics, the target customer group portrait and the product recommendation plan are input into a pre-trained copy generation model to generate personalized copy for the target customer.

[0012] In order to solve the above technical problems, the embodiment of the present application further provides a device for generating personalized copy, which adopts the following technical solution:

[0013] A classification module is used to obtain historical customer feature data, classify all customers according to the historical customer feature data, and obtain different customer groups;

[0014] An extraction module, used to obtain historical copywriting of each customer group, extract semantic features from the historical copywriting, and obtain copywriting keyword combinations corresponding to the customer group;

[0015] A construction module, used to extract customer group characteristics of the customer group, and construct a customer group portrait based on the customer group characteristics and the copywriting keyword combination;

[0016] A matching module is used to obtain customer characteristics of target customers and match target customer group portraits according to the customer characteristics;

[0017] A recommendation module, used to obtain a product recommendation plan for the target customer through a product recommendation model according to the customer characteristics and the target customer group portrait;

[0018] The copywriting generation module is used to input the customer characteristics, the target customer group portrait and the product recommendation plan into the pre-trained copywriting generation model to generate personalized copywriting for the target customer.

[0019] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0020] The computer device includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the method for generating personalized copy as described above when executing the computer-readable instructions.

[0021] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0022] The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the method for generating personalized copy as described above are implemented.

[0023] Compared with the prior art, this application has the following beneficial effects:

[0024] The present application provides a method for generating personalized copywriting, which classifies all customers according to historical customer characteristic data to obtain different customer groups, and obtains the copywriting keyword combination of the corresponding customer group according to the historical copywriting of each customer group, constructs a customer group portrait according to the customer group characteristics and the customer group copywriting keyword combination, matches the target customer group portrait according to the customer characteristics of the target customers, and can achieve accurate positioning of the target customers, achieve accurate delivery of generated copywriting, improve marketing effects, and reduce costs; through a product recommendation model, according to customer characteristics and target customer group portraits, a personalized product recommendation plan can be formed to provide data support for generating targeted copywriting content; by inputting customer characteristics, customer group portrait characteristics and product recommendation plans into a pre-trained copywriting generation model, a personalized copywriting that is highly matched with the target customers and attractive is generated, thereby achieving accurate push and efficient dissemination of marketing information, improving marketing effects and customer satisfaction, which is of great significance for increasing pet insurance sales, enhancing brand competitiveness and expanding market share. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] 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.

[0026] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0027] Figure 2 is a flow chart of an embodiment of a method for generating personalized copy according to the present application;

[0028] Figure 3 yes Figure 2 A flowchart of a specific implementation of step S201;

[0029] Figure 4 is a structural schematic diagram of an embodiment of a device for generating personalized text according to the present application;

[0030] Figure 5 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0031] 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.

[0032] 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.

[0033] 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.

[0034] 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.

[0035] 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.

[0036] 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.

[0037] 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 .

[0038] It should be noted that the method for generating personalized text provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the device for generating personalized text is generally arranged in the server / terminal device.

[0039] It should be understood that Figure 1 The 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 the implementation requirements.

[0040] Continue to refer Figure 2 , shows a flow chart of an embodiment of a method for generating personalized copy according to the present application, comprising the following steps:

[0041] Step S201, obtaining historical customer characteristic data, and classifying all customers according to the historical customer characteristic data to obtain different customer groups.

[0042] Obtain customer data of all historical customers visiting the pet insurance platform, including multi-dimensional data such as customer age, information retention, use of pet insurance services, customer claims record, pet raising experience, customer professional background, customer spending power, preference for pet insurance products, and purchasing patterns.

[0043] Data cleaning technology is used to pre-process the acquired historical customer feature data, remove missing values, outliers and other noise data, and obtain pre-processed historical customer feature data. Through feature engineering methods, key customer features that have a significant impact on customer group segmentation are extracted from the pre-processed historical customer feature data, such as customer age, information retention time, frequency of purchasing pet insurance, etc. Based on the extracted key customer features, a clustering algorithm is used to divide the historical customer feature data into different customer groups, each of which has similar characteristics and purchase tendencies.

[0044] In this embodiment, the method for generating personalized copy is executed on the electronic device (eg Figure 1 The server / terminal device shown in the figure can obtain historical customer feature 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.

[0045] It should be emphasized that in order to further ensure the privacy and security of historical customer characteristic data, the above historical customer characteristic data can also be stored in a blockchain node.

[0046] 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.

[0047] In some optional implementations of this embodiment, see Figure 3 As shown, the step of classifying all customers according to historical customer feature data to obtain different customer groups includes:

[0048] Step S301, standardizing the historical customer feature data to obtain normalized customer feature data, and extracting customer key features from the normalized customer feature data;

[0049] Step S302, determining K initial cluster centers according to the key features of the customers through a K-means clustering algorithm, where K is a positive integer;

[0050] Step S303, traverse each element in the customer key feature, calculate the distance between the element and each initial cluster center, and determine the group corresponding to the element;

[0051] Step S304, the cluster center of each group, obtaining the actual cluster center of each group;

[0052] Step S305, repeatedly calculating the distance between each element in the customer key feature and each actual cluster center, and continuously updating the cluster center of each group, until the grouping no longer changes, and different customer groups are obtained.

[0053] The standardization process includes normalizing the numerical features so that the features are distributed between 0 and 1. In this embodiment, clustering is performed based on the extracted key features of the customers.

[0054] In step S303, each element represents a key feature of a customer. Step S305 indicates that steps S303 to S304 are repeatedly executed until the grouping does not change and each grouping is a cluster. In this embodiment, the distance between each element and the cluster center can be calculated using the Euclidean distance algorithm.

[0055] In this embodiment, after the customers are divided into different customer groups, personalized copywriting can be generated according to the customer group to which the customer belongs. For example, if the customer belongs to group A (such as young people, no information retained, and has never purchased insurance), an entry-level copywriting will be pushed; if the customer belongs to group B (such as middle-aged people, recently retained information, and occasionally purchases insurance), a cost-effective copywriting will be pushed.

[0056] By dividing customer groups into different groups, we can accurately locate the groups, provide personalized services for different customer groups, and improve marketing accuracy and conversion rates.

[0057] In some optional implementations, the step of classifying all customers according to historical customer feature data includes:

[0058] Step A, standardizing the historical customer feature data to obtain normalized customer feature data, and extracting customer key features from the normalized customer feature data;

[0059] Step B: Input the key features of the customer into the decision tree rules, and use a feature selection algorithm based on information gain or Gini index to select the feature that contributes most to customer classification as the splitting node of the decision tree;

[0060] Step C: divide the customer key features into different subsets according to the selected split features, and recursively repeat step B on each subset until the preset stopping conditions are met, such as the minimum number of samples of leaf nodes, the maximum depth of the tree, etc.;

[0061] Step D: At the leaf node of the decision tree, according to the distribution of key features of the customers falling into the node, determine the customer group represented by the node and label it accordingly.

[0062] In this embodiment, after obtaining different customer groups, different customer groups and their corresponding feature combination rules are obtained by traversing all leaf nodes of the decision tree to form a customer group knowledge base; when new customer data arrives, according to its feature vector, starting from the root node, traverse down along the branch path of the decision tree until a certain leaf node is reached, and then it can be determined which customer group the customer belongs to. Through customer feedback and business data analysis, the decision tree rules are continuously optimized, and the customer group knowledge base is regularly updated to adapt to market changes and the dynamic evolution of customer needs.

[0063] Step S202, obtaining historical texts of each customer group, extracting semantic features from the historical texts, and obtaining text keyword combinations of the corresponding customer group.

[0064] Historical copy refers to the historical marketing copy data of each customer group. The historical marketing copy data of each customer group is obtained from the pet insurance platform through database query, API interface call or crawler technology, and semantic features are extracted from the historical marketing copy data. Based on the extracted semantic features, the copy keyword preferences of different customer groups are determined, and personalized copy keyword combinations are automatically generated.

[0065] In some optional implementations of this embodiment, the step of extracting semantic features from historical texts to obtain text keyword combinations for corresponding customer groups includes:

[0066] Input the historical copy into the semantic feature extraction model, which includes a vector embedding layer, an attention mechanism layer, and a keyword output layer;

[0067] Preprocess the historical documents through the vector embedding layer and convert the preprocessed historical documents into text vectors;

[0068] The attention mechanism layer is used to perform semantic understanding and feature extraction on the text vector to obtain the semantic features of the copy.

[0069] The keyword output layer is used to predict and judge the semantic features of the copy and generate copy keyword combinations corresponding to the customer groups.

[0070] In this embodiment, a trained semantic feature extraction model is used to extract semantic features, and the historical copy is preprocessed through the vector embedding layer, including word segmentation, part-of-speech tagging, etc., and then the preprocessed historical copy is converted into a text vector that the model can understand. Through the attention mechanism layer, a multi-head self-attention mechanism is used to perform semantic understanding and feature extraction on the text vector, capture the deep semantic information contained in the text, and obtain the semantic features of the copy. Finally, through the keyword output layer, the copy keyword preferences of different customer groups are predicted according to the semantic features of the copy, and the prediction probability values ​​are sorted from high to low, and a preset number of keywords ranked in front are selected as the copy keyword combination of the customer group.

[0071] By deeply analyzing historical copywriting and generating customer group keyword combinations, we can more accurately grasp customer needs, provide content that is more in line with customer interests and needs, and improve customer experience and satisfaction. At the same time, by generating customer group copywriting keywords, we can accurately locate customer groups, allocate resources more reasonably, and improve marketing efficiency and effectiveness.

[0072] Step S203, extracting customer group characteristics of the customer group, and constructing a customer group portrait based on the customer group characteristics and the copywriting keyword combination.

[0073] In this embodiment, customer group characteristics of the customer group are extracted, including customer age information, information retention, and pet insurance usage, and the customer group characteristics are associated with customer group copywriting keyword combinations to construct a customer group portrait.

[0074] Specifically, based on the customer's age information, the age segmentation method is used to divide the customer into different age groups to obtain the customer's age group attributes; the customer's information retention status is obtained, and the customer is judged as a new customer, old customer or lost customer by the length of information retention time, and the customer's information retention attribute is determined; the customer's pet insurance usage status is obtained, and based on the frequency and time of use, it is judged whether the customer is a high-frequency user, a low-frequency user or has never used pet insurance, and the customer's pet insurance usage attribute is determined; the customer's age group attribute, information retention attribute and pet insurance usage attribute are used as the three dimensions of customer portraits to construct the basic framework of customer portraits; in In the basic framework of customer portrait, according to the customer's age group attributes and the copy keyword combination, the corresponding pet feeding knowledge, common pet diseases and other information are associated to enrich the information in the age group dimension; in the basic framework of customer portrait, according to the customer's information retention and the copy keyword combination, the customer's contact information, pet information, historical communication records, etc. are associated to enrich the information in the information retention dimension; in the basic framework of customer portrait, according to the customer's pet insurance usage and the copy keyword combination, the pet insurance products purchased by the customer, the claims record, service evaluation and other information are associated to enrich the information in the pet insurance usage dimension to obtain the final customer group portrait.

[0075] Among them, the knowledge graph embedding technology is used to vectorize the structured information in the customer group portrait, and obtain a low-dimensional semantic representation of the customer group portrait, which is used as the input of the model.

[0076] Step S204, obtaining customer characteristics of target customers, and matching target customer group portraits according to the customer characteristics.

[0077] In this embodiment, the target customer is the customer who is currently consulting on the pet insurance platform, and the customer characteristics of the customer are obtained, including the customer's age, information retention, and use of pet insurance.

[0078] Through the feature matching algorithm, the customer features are matched with each customer group portrait, and the customer group portrait with the highest matching degree is used as the standard customer group portrait. Among them, the feature matching algorithm includes Jaccard similarity coefficient, cosine similarity, Euclidean distance, etc.

[0079] Step S205, obtaining product recommendation solutions for target customers through a product recommendation model based on customer characteristics and target customer group portraits.

[0080] In this embodiment, based on customer characteristics and target customer group portraits, the target customer's potential needs and purchasing tendencies are predicted through a product recommendation model to obtain suitable recommended products, wherein the recommended products are pet insurance products suitable for the target customers. For example, if the target customer has high purchasing power and risk awareness, a suitable high-security pet insurance product is recommended; if the target customer has a high frequency of claims, suitable additional insurance products are recommended, such as pet medical insurance, surgical insurance, etc. According to the target customer's potential needs and purchasing tendencies, the most suitable pet insurance product combination is matched to form a personalized product recommendation plan.

[0081] In some optional implementations of this embodiment, the step of obtaining a product recommendation solution for a target customer through a product recommendation model according to customer characteristics and target customer group portraits includes:

[0082] Input customer features and target customer group portraits into the product recommendation model, where the product recommendation model includes a feature embedding layer, an attention multi-layer perception layer, a matrix decomposition layer, and a prediction layer;

[0083] The customer features and target customer group portraits are embedded through the feature embedding layer to obtain the customer feature embedding vector;

[0084] Perform deep feature extraction on the customer feature embedding vector through the attention multi-layer perception layer to obtain multi-layer perception customer features;

[0085] The customer feature embedding vector is extracted through the matrix decomposition layer to obtain hidden customer features;

[0086] The multi-layer perceived customer features and hidden customer features are input into the prediction layer for prediction to obtain product recommendation solutions.

[0087] Among them, the feature embedding layer is used to convert customer features and target customer group portraits into vectors to obtain customer feature embedding vectors; the attention multi-layer perception layer combines the attention mechanism and the multi-layer perceptron (MLP) to enhance the ability to capture key information in the customer feature embedding vector; the matrix decomposition layer can extract implicit information and potential structures in the customer feature embedding vector to alleviate the sparsity problem of customer feature data. The prediction layer can use a fully connected layer to concatenate the learned multi-layer perception customer features and hidden customer features to obtain their joint representation, which is input into the fully connected layer, and different weights are assigned to the features of different parts of the joint representation, and then the prediction results, i.e., the prediction scores, are output, and product recommendation solutions are obtained based on the prediction results.

[0088] In some optional implementations, the association rule mining algorithm Apriori can be used to obtain product recommendation solutions suitable for target customers based on customer characteristics and target customer group portraits.

[0089] By combining the customer characteristics of target customers and the target customer group portrait to predict product recommendation plans, we can further explore the potential connection between customers and products, generate product recommendation plans that meet the needs of target customers based on their actual conditions, make the recommendation results closer to the hidden needs of customers, improve the accuracy of recommendations, and thereby improve conversion rates and customer satisfaction, and promote business growth.

[0090] Step S206, input customer characteristics, customer group portrait characteristics and product recommendation solutions into the pre-trained copy generation model to generate personalized copy for target customers.

[0091] In this embodiment, the pre-trained copy generation model includes a text position embedding layer, a multi-layer decoder and a copy output layer, wherein the decoder performs autoregressive generation on personalized copy, and each step predicts the next word based on the previously generated personalized copy and the output of the encoder.

[0092] Specifically, customer characteristics, customer group portrait characteristics and product recommendation plans are input into the pre-trained copy generation model; the customer characteristics, target customer group portrait and product recommendation plans are respectively embedded in text and position through the text position embedding layer to obtain the input vector; the input vector is input into the multi-layer decoder, calculated through the multi-head attention mechanism and the cross-attention mechanism to obtain the predicted text sequence; the predicted text sequence is decoded through the copy output layer to obtain personalized copy for the target customers.

[0093] Each encoder includes a masked multi-head self-attention layer, an encoder-decoder attention layer, and a feedforward network layer. The masked multi-head self-attention layer performs a multi-head attention mechanism calculation on the input vector to obtain the masked attention feature; the encoder-decoder attention layer performs a cross-attention mechanism calculation on the masked attention feature to obtain the global attention feature; the global attention feature is input into the feedforward network layer for copywriting generation to obtain the predicted text sequence.

[0094] The prediction output layer includes a fully connected layer and a softmax layer. Since the last decoder outputs a predicted text sequence, the output predicted text sequence is mapped into a longer sequence through the fully connected layer. The text at each position represents the score of the corresponding text. The score is converted into probabilities through the softmax layer, and the preset number of words with the highest probability are output to obtain the probability distribution of each predicted text.

[0095] Next, a sequence decoding technology such as Beam Search is used to search and score multiple possible predicted texts, and the text with the highest score is selected as the personalized text for the final output.

[0096] In this embodiment, the customer group portrait is updated based on the customer characteristics of the target customers and the corresponding personalized copy.

[0097] By adopting a large copy generation model to automatically generate personalized copy for target customers, we can achieve accurate analysis of customer characteristics and intelligent generation of personalized copy, thereby improving the efficiency and accuracy of copy generation and enhancing customer satisfaction and loyalty.

[0098] This application classifies all customers according to historical customer characteristic data to obtain different customer groups, and obtains the corresponding customer group copy keyword combination based on the historical copy of each customer group, builds a customer group portrait according to the customer group characteristics and the customer group copy keyword combination, matches the target customer group portrait according to the customer characteristics of the target customers, and can achieve accurate positioning of the target customers, achieve accurate delivery of generated copy, improve marketing effects, and reduce costs; through the product recommendation model, according to customer characteristics and target customer group portraits, it can form personalized product recommendation plans, provide data support for generating targeted copy content; by inputting customer characteristics, customer group portrait characteristics and product recommendation plans into the pre-trained copy generation model, it can generate personalized copy that is highly matched with the target customers and attractive, thereby achieving accurate push and efficient dissemination of marketing information, and improving marketing effects and customer satisfaction.

[0099] In some optional implementations, after the step of generating personalized copy for target customers, the following steps are further included:

[0100] Obtain real-time interaction data for each customer in the customer base, and continuously adjust and optimize the copy generation strategy of the copy generation model based on the real-time interaction data.

[0101] Among them, real-time interactive data includes customer feedback data and customer behavior data. Customer feedback data includes feedback information such as satisfaction or dissatisfaction with the generated copy content. Customer behavior data includes browsing time, number of clicks, purchasing behavior, etc.

[0102] After generating personalized copy, we obtain the target customer's real-time feedback and behavior data, dynamically adjust the copy generation model based on the real-time feedback and behavior data, and then adjust the copy generation strategy to obtain keywords, tone and content that are more in line with the target customer's preferences. For example, if the target customer responds positively to the current personalized copy, the copy generation model will increase the weight of the strategy and generate more copy in a similar style; if the target customer responds negatively, the strategy weight will be reduced and other copy generation strategies will be tried; through continuous trials and optimization, the copy generation model will gradually find the copy generation strategy that can maximize the customer's positive feedback and improve the accuracy and conversion rate of marketing.

[0103] Apply the optimized copy generation strategy to new customers, automatically match the optimal strategy according to customer characteristics, generate personalized copy, and attract customers to purchase pet insurance; continue to collect customer feedback and behavioral data, and regularly retrain and optimize the copy generation model to adapt to changes in customer needs and maintain the effectiveness and accuracy of copy generation.

[0104] In this embodiment, the step of continuously adjusting and optimizing the copywriting generation strategy of the copywriting generation model based on real-time interactive data includes:

[0105] Extract customer feedback data and customer behavior data from real-time interactive data, perform semantic analysis on customer feedback data, extract keywords and sentiment tendencies, and obtain customer preference characteristics;

[0106] Use customer characteristic data, customer preference characteristics and customer behavior data to adjust and optimize the copy generation model.

[0107] Among them, emotional tendencies include positive feedback, negative feedback, etc. The sentiment analysis of customer feedback data is performed through the commonly used semantic analysis algorithm to obtain customer preference characteristics; through the copy generation big model, new personalized copy is generated according to customer feature data, customer preference characteristics and customer behavior data, and the similarity between the new personalized copy and the previous personalized copy is calculated. The similarity is used as the evaluation result of the model to adjust the model parameters of the copy generation big model.

[0108] By continuously adjusting and optimizing the copywriting generation model, we can improve the pertinence and attractiveness of the copywriting, continuously improve marketing effectiveness and conversion rate, and thus attract more potential customers.

[0109] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0110] The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning. A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed 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 may include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium may be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0111] 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.

[0112] Further references Figure 4 , as a response to the above Figure 2 The present application provides an embodiment of a device for generating personalized text, 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.

[0113] like Figure 4 As shown, the personalized copy generation device 400 described in this embodiment includes: a classification module 401, an extraction module 402, a construction module 403, a matching module 404, a recommendation module 405 and a copy generation module 406. Among them:

[0114] The classification module 401 is used to obtain historical customer feature data, and classify all customers according to the historical customer feature data to obtain different customer groups;

[0115] The extraction module 402 is used to obtain the historical copywriting of each customer group, extract semantic features from the historical copywriting, and obtain the copywriting keyword combination corresponding to the customer group;

[0116] The construction module 403 is used to extract the customer group characteristics of the customer group, and construct a customer group portrait according to the customer group characteristics and the copywriting keyword combination;

[0117] The matching module 404 is used to obtain customer characteristics of target customers and match target customer group portraits according to the customer characteristics;

[0118] The recommendation module 405 is used to obtain a product recommendation plan for the target customer through a product recommendation model according to the customer characteristics and the target customer group portrait;

[0119] The copywriting generation module 406 is used to input the customer characteristics, the target customer group portrait and the product recommendation plan into the pre-trained copywriting generation model to generate personalized copywriting for the target customer.

[0120] It should be emphasized that in order to further ensure the privacy and security of historical customer characteristic data, the above historical customer characteristic data can also be stored in a blockchain node.

[0121] Based on the above-mentioned personalized copy generation device 400, by dividing the historical customer feature data into different customer groups and obtaining the corresponding customer group copy keyword combination according to the historical copy of each customer group, a customer group portrait is constructed according to the customer group characteristics and the customer group copy keyword combination, and the target customer group portrait is matched according to the customer characteristics of the target customer, so as to achieve accurate positioning of the target customer, achieve accurate delivery of generated copy, improve marketing effect and reduce costs; through the product recommendation model, according to customer characteristics and target customer group portrait, a personalized product recommendation plan can be formed to provide data support for generating targeted copy content; by inputting customer characteristics, customer group portrait characteristics and product recommendation plans into the pre-trained copy generation model, a personalized copy that is highly matched with the target customer and attractive is generated, thereby achieving accurate push and efficient dissemination of marketing information, and improving marketing effect and customer satisfaction.

[0122] In some optional implementations of this embodiment, the classification module 401 is further used to:

[0123] Standardizing the historical customer feature data to obtain normalized customer feature data, and extracting customer key features from the normalized customer feature data;

[0124] Determine K initial cluster centers according to the customer key features by using a K-means clustering algorithm, where K is a positive integer;

[0125] Traversing each element in the customer key feature, calculating the distance between the element and each of the initial cluster centers, and determining the group corresponding to the element;

[0126] Updating the cluster center of each of the groups to obtain the actual cluster center of each of the groups;

[0127] The distance between each element in the customer key feature and each actual cluster center is repeatedly calculated, and the cluster center of each group is continuously updated until the grouping no longer changes, thereby obtaining different customer groups.

[0128] By dividing customer groups into different groups, we can accurately locate the groups, provide personalized services for different customer groups, and improve marketing accuracy and conversion rates.

[0129] In some optional implementations of this embodiment, the extraction module 402 includes:

[0130] An input submodule, used for inputting the historical copy into a semantic feature extraction model, wherein the semantic feature extraction model includes a vector embedding layer, an attention mechanism layer, and a keyword output layer;

[0131] An embedding submodule, used for preprocessing the historical copy by the vector embedding layer, and converting the preprocessed historical copy into a text vector;

[0132] The semantic feature submodule is used to perform semantic understanding and feature extraction on the text vector through an attention mechanism layer to obtain the semantic features of the text;

[0133] The prediction submodule is used to predict and judge the semantic features of the copy through the keyword output layer to generate a copy keyword combination corresponding to the customer group.

[0134] By deeply analyzing historical copywriting and generating customer group keyword combinations, we can more accurately grasp customer needs, provide content that is more in line with customer interests and needs, and improve customer experience and satisfaction. At the same time, by generating customer group copywriting keywords, we can accurately locate customer groups, allocate resources more reasonably, and improve marketing efficiency and effectiveness.

[0135] In some optional implementations, the recommendation module 405 is further configured to:

[0136] Inputting the customer features and the target customer group portrait into the product recommendation model, wherein the product recommendation model includes a feature embedding layer, an attention multi-layer perception layer, a matrix decomposition layer, and a prediction layer;

[0137] Embedding the customer features and the target customer group portrait through the feature embedding layer to obtain a customer feature embedding vector;

[0138] Performing deep feature extraction on the customer feature embedding vector through the attention multi-layer perception layer to obtain multi-layer perception customer features;

[0139] Extracting hidden features from the customer feature embedding vector through the matrix decomposition layer to obtain hidden customer features;

[0140] The multi-layer perceived customer features and the hidden customer features are input into the prediction layer for prediction to obtain a product recommendation solution.

[0141] By combining the customer characteristics of target customers and the target customer group portrait to predict product recommendation plans, we can further explore the potential connection between customers and products, and generate product recommendation plans that meet the needs of target customers based on their actual conditions, so that the recommendation results are closer to the hidden needs of customers and the accuracy of recommendations is improved.

[0142] In some optional implementations, the copywriting generation module 406 is further used to:

[0143] Inputting the customer characteristics, the target customer group portrait and the product recommendation scheme into a pre-trained copywriting generation model, wherein the copywriting generation model includes a text position embedding layer, a multi-layer decoder and a copywriting output layer;

[0144] Through the text position embedding layer, text embedding and position embedding are performed on the customer features, the target customer group portrait, and the product recommendation scheme, respectively, to obtain an input vector;

[0145] Input the input vector into the multi-layer decoder, calculate through the multi-head attention mechanism and the cross attention mechanism, and obtain the predicted text sequence;

[0146] The predicted text sequence is decoded through the copy output layer to obtain personalized copy for the target customer.

[0147] By adopting a large copy generation model to automatically generate personalized copy for target customers, we can achieve accurate analysis of customer characteristics and intelligent generation of personalized copy, thereby improving the efficiency and accuracy of copy generation.

[0148] In some optional implementations, the personalized copy generation device 400 further includes an optimization module, which is used to:

[0149] Real-time interaction data of each customer in the customer group is obtained, and the copy generation strategy of the copy generation model is continuously adjusted and optimized based on the real-time interaction data.

[0150] By continuously collecting customer feedback and behavioral data, we regularly retrain and optimize the copy generation model to adapt to changes in customer needs and maintain the effectiveness and accuracy of copy generation.

[0151] In this embodiment, the optimization module includes:

[0152] A feature extraction submodule is used to extract customer feedback data and customer behavior data from the real-time interactive data, and perform semantic analysis on the customer feedback data to extract keywords and sentiment tendencies to obtain customer preference features;

[0153] The adjustment and optimization submodule is used to adjust and optimize the copywriting generation model based on the customer feature data, the customer preference features and the customer behavior data.

[0154] By continuously adjusting and optimizing the copywriting generation model, we can improve the pertinence and attractiveness of the copywriting, continuously improve marketing effectiveness and conversion rate, and thus attract more potential customers.

[0155] To solve the above technical problems, the present application also provides a computer device. Figure 5 , Figure 5 This is a basic structural block diagram of the computer device in this embodiment.

[0156] The computer device 5 includes a memory 51, a processor 52, and a network interface 53 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 5 with a memory 51, a processor 52, and a network interface 53, 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 (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0157] The computer device may be a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with the user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0158] The memory 51 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 51 can be an internal storage unit of the computer device 5, such as a hard disk or memory of the computer device 5. In other embodiments, the memory 51 can also be an external storage device of the computer device 5, such as a plug-in hard disk equipped on the computer device 5, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 51 can also include both the internal storage unit of the computer device 5 and its external storage device. In this embodiment, the memory 51 is generally used to store the operating system and various application software installed on the computer device 5, such as computer-readable instructions of the method for generating personalized copywriting, etc. In addition, the memory 51 can also be used to temporarily store various types of data that have been output or are to be output.

[0159] The processor 52 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 52 is generally used to control the overall operation of the computer device 5. In this embodiment, the processor 52 is used to run the computer-readable instructions stored in the memory 51 or process data, such as running the computer-readable instructions of the method for generating the personalized copy.

[0160] The network interface 53 may include a wireless network interface or a wired network interface. The network interface 53 is generally used to establish a communication connection between the computer device 5 and other electronic devices.

[0161] By dividing historical customer characteristic data into different customer groups, and obtaining the corresponding customer group copy keyword combination based on the historical copy of each customer group, building a customer group portrait based on customer group characteristics and customer group copy keyword combination, and matching the target customer group portrait according to the customer characteristics of the target customers, it is possible to accurately locate the target customers, achieve accurate delivery of generated copy, improve marketing effects, and reduce costs; through the product recommendation model, according to customer characteristics and target customer group portraits, it is possible to form personalized product recommendation plans, provide data support for generating targeted copy content; by inputting customer characteristics, customer group portrait characteristics and product recommendation plans into the pre-trained copy generation model, personalized copy that is highly matched with the target customers and attractive is generated, thereby achieving accurate push and efficient dissemination of marketing information, and improving marketing effects and customer satisfaction.

[0162] The present application also provides another embodiment, namely, providing a computer-readable storage medium, 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 method for generating personalized copy as described above.

[0163] By dividing historical customer characteristic data into different customer groups, and obtaining the corresponding customer group copy keyword combination based on the historical copy of each customer group, building a customer group portrait based on customer group characteristics and customer group copy keyword combination, and matching the target customer group portrait according to the customer characteristics of the target customers, it is possible to accurately locate the target customers, achieve accurate delivery of generated copy, improve marketing effects, and reduce costs; through the product recommendation model, according to customer characteristics and target customer group portraits, it is possible to form personalized product recommendation plans, provide data support for generating targeted copy content; by inputting customer characteristics, customer group portrait characteristics and product recommendation plans into the pre-trained copy generation model, personalized copy that is highly matched with the target customers and attractive is generated, thereby achieving accurate push and efficient dissemination of marketing information, and improving marketing effects and customer satisfaction.

[0164] 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, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0165] Obviously, the embodiments described above are only some embodiments of the present application, rather than all 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 implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for generating personalized copy, characterized in that: The steps include: Acquire historical customer characteristic data, and classify all customers according to the historical customer characteristic data to obtain different customer groups; Obtaining historical copywriting of each customer group, extracting semantic features from the historical copywriting, and obtaining copywriting keyword combinations corresponding to the customer group; Extracting customer group characteristics of the customer group, and constructing a customer group portrait based on the customer group characteristics and the copywriting keyword combination; Obtain customer characteristics of target customers, and match target customer group portraits according to the customer characteristics; According to the customer characteristics and the target customer group portrait, obtaining a product recommendation plan for the target customer through a product recommendation model; The customer characteristics, the target customer group portrait and the product recommendation plan are input into a pre-trained copy generation model to generate personalized copy for the target customer.

2. The method for generating personalized copy according to claim 1, characterized in that: The step of classifying all customers according to the historical customer feature data to obtain different customer groups includes: Standardizing the historical customer feature data to obtain normalized customer feature data, and extracting customer key features from the normalized customer feature data; Determine K initial cluster centers according to the customer key features by using a K-means clustering algorithm, where K is a positive integer; Traversing each element in the customer key feature, calculating the distance between the element and each of the initial cluster centers, and determining the group corresponding to the element; Updating the cluster center of each of the groups to obtain the actual cluster center of each of the groups; The distance between each element in the customer key feature and each actual cluster center is repeatedly calculated, and the cluster center of each group is continuously updated until the grouping no longer changes, thereby obtaining different customer groups.

3. The method for generating personalized copy according to claim 1, characterized in that: The step of extracting semantic features from the historical copywriting to obtain the copywriting keyword combination corresponding to the customer group includes: Inputting the historical copy into a semantic feature extraction model, wherein the semantic feature extraction model includes a vector embedding layer, an attention mechanism layer, and a keyword output layer; Preprocessing the historical documents through the vector embedding layer, and converting the preprocessed historical documents into text vectors; The text vector is semantically understood and features extracted through an attention mechanism layer to obtain semantic features of the text; The keyword output layer predicts and judges the semantic features of the text, and generates a text keyword combination corresponding to the customer group.

4. The method for generating personalized copy according to claim 3, characterized in that: The step of obtaining a product recommendation solution for the target customer through a product recommendation model according to the customer characteristics and the target customer group portrait comprises: Inputting the customer features and the target customer group portrait into the product recommendation model, wherein the product recommendation model includes a feature embedding layer, an attention multi-layer perception layer, a matrix decomposition layer, and a prediction layer; Embedding the customer features and the target customer group portrait through the feature embedding layer to obtain a customer feature embedding vector; Performing deep feature extraction on the customer feature embedding vector through the attention multi-layer perception layer to obtain multi-layer perception customer features; Extracting hidden features from the customer feature embedding vector through the matrix decomposition layer to obtain hidden customer features; The multi-layer perceived customer features and the hidden customer features are input into the prediction layer for prediction to obtain a product recommendation solution.

5. The method for generating personalized copy according to claim 1, characterized in that: The step of inputting the customer characteristics, the target customer group portrait and the product recommendation scheme into the pre-trained copywriting generation model to generate personalized copywriting for the target customer includes: Inputting the customer characteristics, the target customer group portrait and the product recommendation scheme into a pre-trained copywriting generation model, wherein the copywriting generation model includes a text position embedding layer, a multi-layer decoder and a copywriting output layer; Through the text position embedding layer, text embedding and position embedding are performed on the customer features, the target customer group portrait, and the product recommendation scheme, respectively, to obtain an input vector; Input the input vector into the multi-layer decoder, calculate through the multi-head attention mechanism and the cross attention mechanism, and obtain the predicted text sequence; The predicted text sequence is decoded through the copy output layer to obtain personalized copy for the target customer.

6. The method for generating personalized copy according to claim 1, characterized in that: After the step of generating personalized copy for the target customer, the method further includes: Real-time interaction data of each customer in the customer group is obtained, and the copy generation strategy of the copy generation model is continuously adjusted and optimized based on the real-time interaction data.

7. The method for generating personalized copy according to claim 6, characterized in that: The step of continuously adjusting and optimizing the copywriting generation strategy of the copywriting generation model based on the real-time interactive data includes: Extracting customer feedback data and customer behavior data from the real-time interactive data, and performing semantic analysis on the customer feedback data to extract keywords and sentiment tendencies to obtain customer preference features; The customer feature data, the customer preference features and the customer behavior data are used to adjust and optimize the copywriting generation model.

8. A device for generating personalized copy, characterized in that: include: A classification module is used to obtain historical customer feature data, classify all customers according to the historical customer feature data, and obtain different customer groups; An extraction module, used to obtain historical copywriting of each customer group, extract semantic features from the historical copywriting, and obtain copywriting keyword combinations corresponding to the customer group; A construction module, used to extract customer group characteristics of the customer group, and construct a customer group portrait based on the customer group characteristics and the copywriting keyword combination; A matching module is used to obtain customer characteristics of target customers and match target customer group portraits according to the customer characteristics; A recommendation module, used to obtain a product recommendation plan for the target customer through a product recommendation model according to the customer characteristics and the target customer group portrait; The copywriting generation module is used to input the customer characteristics, the target customer group portrait and the product recommendation plan into the pre-trained copywriting generation model to generate personalized copywriting for the target customer.

9. A computer device, characterized in that: The invention comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the method for generating personalized copy as claimed in 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 method for generating personalized copywriting according to any one of claims 1 to 7 are implemented.

Images