Product recommendation content generation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a product recommendation content generation method and device, electronic equipment and a storage medium, and belongs to the technical field of text processing. The method comprises the following steps: acquiring sample identity information, insurance product information, browsing record information and purchase record information; training a preset product demand prediction model based on the sample identity information, the browsing record information, the purchase record information and the insurance product information to obtain a target product demand prediction model; obtaining a target identity portrait of the target object, and performing demand prediction on the target object through the target product demand prediction model and the target identity portrait to obtain a target recommended product; sorting the target recommended products according to the target identity portraits to obtain a product recommendation sequence; and according to the target identity portrait and the product recommendation sequence, performing recommendation content generation on the target object to obtain target recommendation content. According to the embodiment of the invention, the product recommendation content better meeting user requirements can be generated.

Description

Technical Field

[0001] The present application relates to the field of text processing technology, and in particular to a method and device for generating product recommendation content, an electronic device, and a storage medium. Background Art

[0002] Product recommendation content is used to recommend or introduce products. For example, in the insurance field, product recommendation content helps users understand the characteristics and terms of insurance products and clarify the coverage of insurance. At present, a common technology for generating product recommendation content is to use a large language model. However, since the large language model mainly relies on public knowledge on the Internet, it does not have a deep understanding of insurance products and it is difficult to generate recommendation text that meets user needs, resulting in a low success rate of recommendation. Therefore, how to generate product recommendation content that better meets user needs has become an urgent problem to be solved. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose a method and device for generating product recommendation content, an electronic device and a storage medium, aiming to generate product recommendation content that better meets user needs.

[0004] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a method for generating product recommendation content, the method comprising:

[0005] Obtaining sample identity information of a sample object, obtaining insurance product information of an insurance product, and obtaining browsing record information and purchase record information of the sample object for the insurance product;

[0006] Training a preset product demand prediction model based on the sample identity information, the browsing record information, the purchase record information, and the insurance product information to obtain a target product demand prediction model;

[0007] Obtaining a target identity portrait of a target object, performing demand forecasting on the target object through the target product demand forecasting model and the target identity portrait, and obtaining a target recommended product;

[0008] Sort the target recommended products according to the target identity portrait to obtain a product recommendation sequence;

[0009] Recommended content is generated for the target object according to the target identity portrait and the product recommendation sequence to obtain target recommended content.

[0010] In some embodiments, the training of a preset product demand prediction model based on the sample identity information, the browsing record information, the purchase record information, and the insurance product information to obtain a target product demand prediction model includes:

[0011] Constructing a user portrait based on the sample identity information, the browsing record information, and the purchase record information to obtain user portrait information;

[0012] Perform preference clustering based on the sample identity information, the browsing record information, and the purchase record information to obtain user preference information;

[0013] The preset product demand forecasting model is trained based on the insurance product information, the user portrait information and the user preference information to obtain the target product demand forecasting model.

[0014] In some embodiments, the training of the preset product demand forecasting model based on the insurance product information, the user portrait information, and the user preference information to obtain the target product demand forecasting model includes:

[0015] Merging data based on the user portrait information and the user preference information to obtain portrait preference merged data;

[0016] Based on the portrait preference merged data, the data set is divided to obtain portrait preference data and insurance product labels;

[0017] Perform product forecasting on the preset product demand forecasting model based on the portrait preference data and the insurance product information to obtain a forecast product label;

[0018] The preset product demand forecasting model is optimized in parameters based on the insurance product label and the predicted product label to obtain the target product demand forecasting model.

[0019] In some embodiments, the step of optimizing the parameters of the preset product demand forecasting model based on the insurance product label and the forecast product label to obtain the target product demand forecasting model includes:

[0020] Calculate the cross entropy loss value according to the predicted product label and the insurance product label to obtain cross entropy loss value data;

[0021] The parameters of the preset product demand forecasting model are optimized according to the cross entropy loss value data to obtain the target product demand forecasting model.

[0022] In some embodiments, the preset product demand forecasting model includes an embedding layer, a portrait preference encoding layer, a feature extraction layer, and a decoding layer. The product forecasting of the preset product demand forecasting model based on the portrait preference data and the insurance product information to obtain a predicted product label includes:

[0023] Performing text embedding on the portrait preference data through the embedding layer to obtain a portrait preference embedding;

[0024] Performing text embedding on the insurance product information through the embedding layer to obtain insurance product embedding;

[0025] Text encoding is performed on the image preference embedding through the image preference encoding layer to obtain an image preference encoding;

[0026] Extract product features from the insurance product embedding through the feature extraction layer to obtain insurance product features;

[0027] The insurance product features and the portrait preference codes are text-decoded through the decoding layer to obtain the predicted product label.

[0028] In some embodiments, performing preference clustering based on the sample identity information, the browsing record information, and the purchase record information to obtain user preference information includes:

[0029] Clustering browsing records according to the sample identity information and the browsing record information to obtain browsing clustering areas;

[0030] Clustering the purchase records according to the sample identity information and the purchase record information to obtain a purchase clustering area;

[0031] Performing association relationship clustering according to the sample identity information, the purchase record information, and the browsing record information to obtain an association relationship clustering area;

[0032] The user preference information is obtained by merging information based on the browsing cluster area, the purchasing cluster area and the association relationship cluster area.

[0033] In some embodiments, the step of sorting the target recommended products according to the target identity portrait to obtain a product recommendation sequence includes:

[0034] Extracting features of the target identity portrait and the target recommended products to obtain user-product aggregate features;

[0035] Performing product recommendation evaluation on the user product aggregation features to obtain product recommendation evaluation data; wherein the product recommendation evaluation data is used to evaluate the rationality of the recommendation of the target recommended product;

[0036] A recommendation sequence is generated for the target recommended product according to the product recommendation evaluation data to obtain the product recommendation sequence.

[0037] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application proposes a device for generating product recommendation content, the device comprising:

[0038] An information acquisition module is used to acquire sample identity information of a sample object, acquire insurance product information of an insurance product, and acquire browsing record information and purchase record information of the sample object for the insurance product;

[0039] A model training module, used to train a preset product demand prediction model based on the sample identity information, the browsing record information, the purchase record information and the insurance product information to obtain a target product demand prediction model;

[0040] A demand prediction module is used to obtain a target identity portrait of a target object, perform demand prediction on the target object through the target product demand prediction model and the target identity portrait, and obtain a target recommended product;

[0041] A product sorting module, used to sort the target recommended products according to the target identity portrait to obtain a product recommendation sequence;

[0042] The content generation module is used to generate recommended content for the target object according to the target identity portrait and the product recommendation sequence to obtain target recommended content.

[0043] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0044] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0045] The present application proposes a method and device for generating product recommendation content, an electronic device and a storage medium, which obtain sample identity information of a sample object, obtain insurance product information of an insurance product, and obtain browsing record information and purchase record information of the sample object on the insurance product, and then train a preset product demand prediction model based on the sample identity information, browsing record information, purchase record information and insurance product information to obtain a target product demand prediction model that can accurately predict a target product demand prediction model that meets user needs from a variety of insurance product information; then sort the target recommended products according to the target identity portrait to obtain a product recommendation sequence, thereby further predicting the insurance product most needed by the user; further, generate recommendation content for the target object according to the target identity portrait and the product recommendation sequence to obtain the target recommendation content, thereby generating product recommendation content that better meets user needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flowchart of a method for generating product recommendation content provided by an embodiment of the present application;

[0047] Figure 2 yes Figure 1 Flow chart of step S102 in FIG.

[0048] Figure 3 yes Figure 2 Flow chart of step S202 in FIG.

[0049] Figure 4 yes Figure 2 Flow chart of step S203 in FIG.

[0050] Figure 5 yes Figure 4 Flow chart of step S403 in FIG.

[0051] Figure 6 yes Figure 4 Flow chart of step S404 in FIG.

[0052] Figure 7 yes Figure 1 Flow chart of step S104 in FIG.

[0053] Figure 8 It is a structural diagram of a device for generating product recommendation content provided in an embodiment of the present application;

[0054] Fig. 9 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0056] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0058] First, some nouns involved in this application are analyzed:

[0059] Ranking Neural Network (RankNet): is a neural network model used to learn ranking tasks. It is mainly used in information retrieval, recommendation systems and other scenarios where items need to be ranked. Ranking neural network is an application branch of artificial intelligence. It learns the relative ranking between items by simulating the human judgment process. RankNet is mainly trained through paired input samples. Each pair of samples will be evaluated by the model and generate a relative importance score. Then, the model is adjusted through the optimization algorithm to accurately reflect the ranking relationship between samples.

[0060] Large-scale Natural Language Model (LNM): A deep learning model trained with a large corpus, designed to understand and generate human language. Large-scale natural language models process and understand texts ranging from simple sentences to complex texts by simulating the human language understanding and generation process. Large-scale natural language models are widely used in technical fields such as machine translation, text summarization, sentiment analysis, question-answering systems, and dialogue systems. In addition, these models also show excellent performance in tasks such as information extraction, knowledge graph construction, and language generation.

[0061] User Profiling: User profiling is a user model constructed by extracting features from user behavior data, personal attributes, social network and other information using data analysis technology. User profiling helps the system better understand user characteristics by analyzing and summarizing information on user interests, preferences, consumption habits, lifestyles, etc., so as to provide more personalized services or product recommendations. In terms of technical implementation, user profiling involves machine learning, data mining, statistical analysis and other methods to process and analyze large amounts of user data.

[0062] Preference Information: It is data that represents user preferences and interests by analyzing and aggregating user choices, behaviors, and feedback. It is an important component for understanding and predicting user behavior. Preference information is constructed by collecting data such as user evaluations of products, services, or content, purchase history, and browsing behavior, aiming to reveal user consumption tendencies and interests. Preference information involves methods such as data mining, machine learning, and statistical analysis to identify and quantify user preferences. This information is widely used in fields such as recommendation systems, advertising targeting, market analysis, and user experience design, helping companies or systems provide more personalized and accurate services.

[0063] Product recommendation content is used to recommend or introduce products. For example, in the insurance field, product recommendation content helps users understand the characteristics and terms of insurance products and clarify the coverage of insurance. At present, a common technology for generating product recommendation content is to use a large language model. However, since the large language model mainly relies on public knowledge on the Internet, it does not have a deep understanding of insurance products and it is difficult to generate recommendation text that meets user needs, resulting in a low success rate of recommendation. Therefore, how to generate product recommendation content that better meets user needs has become an urgent problem to be solved.

[0064] Based on this, the embodiments of the present application provide a method and device for generating product recommendation content, an electronic device and a storage medium, aiming to generate product recommendation content that better meets user needs.

[0065] A method and device for generating product recommendation content, an electronic device, and a storage medium provided in an embodiment of the present application are specifically described through the following embodiments. First, a method for generating product recommendation content in an embodiment of the present application is described.

[0066] 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.

[0067] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0068] The product recommendation content generation method provided in the embodiment of the present application relates to the field of text processing technology. The product recommendation content generation method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the product recommendation content generation method, etc., but is not limited to the above forms.

[0069] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0070] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0071] Figure 1 is an optional flowchart of the method for generating product recommendation content provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S105.

[0072] Step S101, obtaining sample identity information of a sample object, obtaining insurance product information of an insurance product, and obtaining browsing record information and purchase record information of the sample object for the insurance product;

[0073] Step S102, training a preset product demand prediction model based on sample identity information, browsing record information, purchase record information and insurance product information to obtain a target product demand prediction model;

[0074] Step S103, obtaining a target identity portrait of the target object, performing demand forecasting on the target object through the target product demand forecasting model and the target identity portrait, and obtaining a target recommended product;

[0075] Step S104, sorting the target recommended products according to the target identity portrait to obtain a product recommendation sequence;

[0076] Step S105, generating recommended content for the target object according to the target identity portrait and the product recommendation sequence to obtain target recommended content.

[0077] Steps S101 to S105 shown in the embodiment of the present application are performed by obtaining sample identity information of the sample object, obtaining insurance product information of the insurance product, and obtaining browsing record information and purchase record information of the sample object on the insurance product, and then training the preset product demand prediction model based on the sample identity information, browsing record information, purchase record information and insurance product information to obtain a target product demand prediction model that can accurately predict the target product demand that meets the user's needs from a variety of insurance product information; then, the target recommended products are sorted according to the target identity portrait to obtain a product recommendation sequence, thereby further predicting the insurance product most needed by the user; further, recommendation content is generated for the target object according to the target identity portrait and the product recommendation sequence to obtain the target recommendation content, thereby generating product recommendation content that better meets the user's needs.

[0078] In step S101 of some embodiments, the sample object refers to a current user or potential user of an insurance product, and the sample identity information is a feature of the sample object, such as age, gender, occupation, or place of residence. Insurance product information is all detailed data of an insurance product, including but not limited to insurance type, coverage, insurance terms, price, and compensation conditions. Browsing record information is a data record left by a sample object when browsing a website or application, including but not limited to the time, frequency, and duration of viewing various insurance products. Purchase record information is a specific record of a sample object's purchase of insurance products in the past, including but not limited to the type of product purchased, purchase date, and payment method.

[0079] See also Figure 2 In some embodiments, step S102 may include but is not limited to steps S201 to S203:

[0080] Step S201, constructing a profile based on the sample identity information, browsing record information and purchase record information to obtain user profile information;

[0081] Step S202, performing preference clustering based on sample identity information, browsing record information and purchase record information to obtain user preference information;

[0082] Step S203, training a preset product demand prediction model based on insurance product information, user portrait information and user preference information to obtain a target product demand prediction model.

[0083] In steps S201 to S203 shown in the embodiment of the present application, a portrait is constructed based on the sample identity information, browsing history information and purchase history information to obtain user portrait information, and then preference clustering is performed based on the sample identity information, browsing history information and purchase history information to obtain user preference information. Finally, a preset product demand prediction model is trained on the insurance product information, user portrait information and user preference information to obtain a target product demand prediction model that can accurately predict the target product demand that meets the user's needs from a variety of insurance product information.

[0084] In step S201 of some embodiments, a portrait is constructed based on the sample identity information, browsing record information, and purchase record information. The portrait construction is first classified based on the sample identity information, browsing record information, and purchase record information. For example, the sample user information is a male aged 40, and the browsing record information and purchase record information include browsing record information or purchase record information of children's insurance products, and the sample object is determined to be a parent. For example, the sample user information is a male aged 40, and the browsing record information and purchase record information include browsing record information or purchase record information of elderly insurance products, and the sample object is determined to be a child. For example, the sample user information is a male aged 30, and the browsing record information and purchase record information include browsing record information or purchase record information of auto insurance products, and the sample object is determined to be a car owner. It should be noted that the identity of the sample object can include multiple, for example, the identity of the sample object is both a car owner and a parent. Then, the purchasing power information of the sample object is determined based on the occupation information, browsing record information, and purchase record information in the sample identity information. Then, the health awareness information of the sample object is determined based on the age, browsing record information, and purchase record information of the sample user information. Then, based on the age, browsing history and purchase history of the sample user, the asset management intention information of the sample object is determined. Based on the identity information, purchasing power information, health awareness information and asset management intention information of the sample object as the portrait of the sample object, the user portrait information is obtained.

[0085] See also Figure 3 In some embodiments, step S202 may include but is not limited to steps S301 to S304:

[0086] Step S301, clustering browsing records according to sample identity information and browsing record information to obtain browsing clustering areas;

[0087] Step S302, clustering the purchase records according to the sample identity information and the purchase record information to obtain a purchase clustering area;

[0088] Step S303, clustering the association relationships according to the sample identity information, purchase record information and browsing record information to obtain an association relationship clustering area;

[0089] Step S304: Merge information based on the browsing clustering area, the purchasing clustering area, and the association relationship clustering area to obtain user preference information.

[0090] In steps S301 to S304 shown in the embodiment of the present application, browsing records are clustered according to sample identity information and browsing record information to obtain browsing clustering areas, and then purchase records are clustered according to sample identity information and purchase record information to obtain purchase clustering areas. Then, association clustering is performed according to sample identity information, purchase record information and browsing record information to obtain association clustering areas. Finally, information is merged based on browsing clustering areas, purchase clustering areas and association clustering areas to obtain user preference information, thereby achieving accurate analysis of user browsing preferences, thereby obtaining user personalized preference information, and subsequently generating more accurate recommended content based on user preferences.

[0091] In step S301 of some embodiments, browsing record clustering is performed based on sample identity information and browsing record information. In one embodiment, KNN clustering is performed based on sample identity information and browsing record information to obtain multiple clustering areas, and clustering areas are screened according to a preset browsing screening threshold to obtain browsing clustering areas. For example: the sample identity information is divided based on age groups to obtain user group information of multiple different age groups, and then KNN clustering is performed on the browsing record information based on the user group information. For example, users aged between 30 and 40 prefer to browse products classified as million-level medical products in terms of medical insurance products. For example, users in the south prefer to browse claims solutions classified as one million-level in terms of auto insurance claims. For example, users whose occupation is a train driver prefer to browse claims solutions classified as five million-level in terms of auto insurance claims.

[0092] In step S302 of some embodiments, purchase record clustering is performed based on the sample identity information and the browsing record information. In one embodiment, KNN clustering is performed based on the sample identity information and the purchase record information to obtain multiple clustering areas, and the clustering areas are screened according to a preset purchase screening threshold to obtain the purchase clustering areas. For example, the sample identity information is divided based on age groups to obtain user group information of multiple different age groups, and then KNN clustering is performed on the browsing record information based on the user group information. For example, users aged between 30 and 40 prefer to purchase products in the category of million-level medical insurance products.

[0093] In step S303 of some embodiments, first, association mining is performed based on the purchase record information and the browsing record information to obtain the purchase-browse association relationship. For example, after browsing a million-level auto insurance claim insurance product, the user tends to purchase a three-million-level auto insurance claim product. For example, after browsing a travel insurance product, the user tends to purchase a medical insurance product. Then, KNN clustering is performed based on the sample identity information and the purchase-browse association relationship. For example, a female user, after browsing a travel insurance product, tends to purchase a medical insurance product. For example, a male user, after browsing a million-level auto insurance claim insurance product, tends to purchase a three-million-level auto insurance claim product.

[0094] In step S304 of some embodiments, information is merged based on browsing cluster areas, purchasing cluster areas and association cluster areas to obtain user preference information. The information is merged in a key-value pair manner. For example, the key is users between the ages of 30 and 40, and the value is users who prefer to browse products in the category of million-level medical care, users who prefer to purchase products in the category of million-level medical care, and users who tend to purchase three million-level auto insurance claims products after browsing million-level auto insurance claims products.

[0095] See also Figure 4 In some embodiments, step S203 may include but is not limited to steps S401 to S404:

[0096] Step S401, merging data based on user portrait information and user preference information to obtain portrait preference merged data;

[0097] Step S402, dividing the data set based on the portrait preference merge data to obtain the portrait preference data and insurance product labels;

[0098] Step S403, performing product forecasting on a preset product demand forecasting model based on the portrait preference data and the insurance product information to obtain a forecasted product label;

[0099] Step S404, optimizing the parameters of the preset product demand forecasting model based on the insurance product label and the forecast product label to obtain the target product demand forecasting model.

[0100] In steps S401 to S404 shown in the embodiment of the present application, data is merged based on user portrait information and user preference information to obtain portrait preference merged data, and then data sets are divided based on the portrait preference merged data to obtain portrait preference data and insurance product labels. Then, product prediction is performed on a preset product demand prediction model based on the portrait preference data and insurance product information to obtain predicted product labels. Finally, parameters of the preset product demand prediction model are optimized based on the insurance product labels and predicted product labels to obtain a target product demand prediction model that can accurately predict the target product demand that meets user needs from a variety of insurance product information.

[0101] In step S401 of some embodiments, information is merged based on user portrait information and user preference information to obtain portrait preference merged data, and the information is merged in a manner of associated information merging. For example, users aged between 30 and 40 prefer to browse products classified as million-level medical products, prefer to purchase products classified as million-level medical products, and like to purchase three million-level auto insurance claims products after browsing million-level auto insurance claims products. Users aged between 30 and 40 include first user portrait information and second user portrait information. The first portrait information is (identity: parent, child, purchasing power information is: annual purchasing power in the first tier, health awareness information is: lacking, asset management intention is: moderate), and the second portrait information is (identity: individual, child, purchasing power information is: annual purchasing power in the second tier, health awareness information is: sufficient, asset management intention is: sufficient), and the insurance products purchased by users aged between 30 and 40 include: the first product, the second product, and the third product.

[0102] In step S402 of some embodiments, data set division is performed based on the portrait preference merged data, insurance product information in the portrait preference merged data is extracted, insurance product labels are obtained, and information other than insurance product information in the portrait preference merged data is used as portrait preference data.

[0103] In one embodiment, the portrait preference merged data is that users aged between 30 and 40 prefer to browse products classified as million-level medical products, prefer to buy products classified as million-level medical products, and like to buy three million-level auto insurance claims products after browsing million-level auto insurance claims products, and users aged between 30 and 40 include first user portrait information and second user portrait information. The first portrait information is (identity: parent, child, purchasing power information is: annual purchasing power in the first tier, health awareness information is: lack, asset management intention is: moderate), the second portrait information is (identity: individual, child, purchasing power information is: annual purchasing power in the second tier, health awareness information is: sufficient, asset management intention is: sufficient), and the insurance products purchased by users aged between 30 and 40 include: the first product, the second product, and the third product.

[0104] After the data set is divided, the portrait preference data is the portrait preference merged data for users aged between 30 and 40, who prefer to browse products classified as million-level medical products, prefer to buy products classified as million-level medical products, and like to browse million-level auto insurance claims products. After that, they will tend to buy three million-level auto insurance claims products, and users aged between 30 and 40 include the first user portrait information and the second user portrait information. The first portrait information is (identity: parents, children, purchasing power information is: annual purchasing power is in the first tier, health awareness information is: lacking, asset management intention is: moderate), and the second portrait information is (identity: individual, child, purchasing power information is: annual purchasing power is in the second tier, health awareness information is: sufficient, asset management intention is: sufficient).

[0105] Insurance products are labeled as first product, second product, and third product.

[0106] See also Figure 5 In some embodiments, the preset product demand prediction model includes an embedding layer, a portrait preference encoding layer, a feature extraction layer, and a decoding layer. Step S403 includes but is not limited to steps S501 to S505:

[0107] Step S501, text embedding the portrait preference data through an embedding layer to obtain a portrait preference embedding;

[0108] Step S502, text embedding the insurance product information through the embedding layer to obtain insurance product embedding;

[0109] Step S503, text encoding the portrait preference embedding through the portrait preference encoding layer to obtain the portrait preference code;

[0110] Step S504, extracting product features from the insurance product embedding through the feature extraction layer to obtain insurance product features;

[0111] Step S505, text decoding is performed on the insurance product features and profile preference encoding through the decoding layer to obtain a predicted product label.

[0112] In steps S501 to S505 shown in the embodiment of the present application, the portrait preference data is text-embedded through the embedding layer to obtain the portrait preference embedding, and the insurance product information is text-embedded through the embedding layer to obtain the insurance product embedding, and then the portrait preference embedding is text-encoded through the portrait preference encoding layer to obtain the portrait preference encoding, and at the same time, the product features of the insurance product embedding are extracted through the feature extraction layer to obtain the insurance product features, and finally, the insurance product features and the portrait preference encoding are text-decoded through the decoding layer to obtain the predicted product label, thereby achieving accurate prediction of user needs and providing a data basis for the subsequent generation of product recommendation content that better meets user needs.

[0113] In step S501 of some embodiments, the embedding layer is used to convert the text into a more dense and low-dimensional real number vector, and the portrait preference embedding is obtained by inputting the portrait preference data into the embedding layer for text embedding. In one embodiment, the portrait preference data is embedded in the text using the embedding layer in the Transformer deep learning network to obtain the portrait preference embedding.

[0114] In step S502 of some embodiments, the insurance product embedding is obtained by inputting the insurance product information into the embedding layer. In one embodiment, the insurance product information is text-embedded using the embedding layer in the Transformer deep learning network to obtain the insurance product embedding.

[0115] In step S503 of some embodiments, the portrait preference encoding layer is a network structure for converting portrait preference embeddings into text encodings. In one embodiment, the portrait preference encoding layer is composed of a position encoding layer and an encoder layer in a Transformer deep learning network. The portrait preference code is first input into the position encoding layer to obtain a portrait preference position code. The portrait preference position code and the portrait preference embedding are then weightedly summed to obtain a portrait preference weighted code. The portrait preference weighted code is then input into the encoder layer to obtain a portrait preference code.

[0116] In step S504 of some embodiments, the feature extraction layer is a network structure for extracting features from the insurance product embedding. In one embodiment, the feature extraction layer is composed of a sparse autoencoder. It should be noted that since the insurance product embedding includes a large amount of complex and redundant information, and the sparse autoencoder can activate only a few neurons related to key features by forcing the hidden layer to have low activity, it can effectively extract the most representative and distinguishing features from the insurance product embedding, while ignoring irrelevant or repeated information, so as to accurately extract representative feature information from the insurance product embedding and obtain insurance product features.

[0117] It should be noted that during the training process of the preset product demand forecasting model, the insurance product information needs to be input for training to obtain the target product demand forecasting model. During the application of the target product demand forecasting model, the feature extraction layer does not need to be input again, and will directly output the output of the last iteration during the training process of the preset product demand forecasting model, thereby improving the prediction speed of the target product demand forecasting model.

[0118] In step S505 of some embodiments, the decoding layer is a network structure that performs text decoding on the insurance product features and portrait preference encoding, converting the vector representation into a text representation. In one embodiment, the decoding layer is composed of a decoder in a Transformer, and the insurance product features and portrait preference features are input into the decoder in the Transformer to obtain a predicted product label.

[0119] See also Figure 6 In some embodiments, step S404 includes but is not limited to steps S601 to S602:

[0120] Step S601, calculating the cross entropy loss value according to the predicted product label and the insurance product label to obtain cross entropy loss value data;

[0121] Step S602, optimizing the parameters of the preset product demand forecasting model according to the cross entropy loss value data to obtain the target product demand forecasting model.

[0122] In steps S601 to S602 shown in the embodiment of the present application, cross entropy loss values ​​are calculated based on the predicted product labels and insurance product labels to obtain cross entropy loss value data, and then parameters of the preset product demand prediction model are optimized based on the cross entropy loss value data to obtain a target product demand prediction model that can accurately predict the target product demand that meets the user's needs from a variety of insurance product information.

[0123] In step S601 of some embodiments, the cross entropy loss value is a loss function used in deep learning to measure the difference between the model's prediction and the true label. The cross entropy loss value is calculated based on the predicted product label and the insurance product label to obtain the cross entropy loss value data.

[0124] In step S602 of some embodiments, parameter optimization is to perform gradient descent optimization on the network parameters in the preset product demand forecasting model so that the predicted result of the preset product demand forecasting model is close to the insurance product label, thereby improving the prediction ability of the preset product demand forecasting model and obtaining the target product demand forecasting model.

[0125] In step S103 of some embodiments, the target object is a potential product purchase object, and a portrait is constructed based on the target identity information of the target object to obtain a target identity portrait. It should be noted that the principle of constructing a portrait based on the target identity information of the target object to obtain a target identity portrait is consistent with the principle of constructing a portrait based on a sample object, which will not be repeated here. The target identity portrait is input into the target product demand prediction model to realize demand prediction for the target object and obtain a target recommended product.

[0126] It should be noted that, in one embodiment, the target browsing history and target purchase history of the target object can also be obtained, and preference clustering can be performed to obtain target preference information. The target identity portrait and target preference information can then be input into the target product demand forecasting model to forecast the demand for the target object and obtain target recommended products.

[0127] See also Figure 7 In some embodiments, step S104 may include but is not limited to steps S701 to S703:

[0128] Step S701, extracting features of the target identity portrait and the target recommended products to obtain user-product aggregate features;

[0129] Step S702, performing product recommendation evaluation on the user product aggregation features to obtain product recommendation evaluation data; wherein the product recommendation evaluation data is used to evaluate the rationality of the recommendation of the target recommended product;

[0130] Step S703: Generate a recommendation sequence for the target recommended product according to the product recommendation evaluation data to obtain a product recommendation sequence.

[0131] Steps S701 to S703 shown in the embodiment of the present application extract features of the target identity portrait and the target recommended products to obtain user-product aggregation features, then perform product recommendation evaluation on the user-product aggregation features to obtain product recommendation evaluation data, and finally generate a recommendation sequence for the target recommended products based on the product recommendation evaluation data to obtain a product recommendation sequence, thereby further sorting the target recommended products based on the target recommended products to obtain product recommendations that are more in line with the target object, thereby providing a more accurate data basis for the follow-up and generating product recommendation content that better meets user needs.

[0132] In step S701 of some embodiments, feature extraction is to extract key information from the target identity portrait and the target recommended products. In one embodiment, feature extraction is composed of an input layer and a hidden layer in a ranking neural network (RankNet). The target identity portrait and the target recommended products are input into the input layer, and the output of the input layer is input into the hidden layer to obtain user-product aggregation features.

[0133] In step S702 of some embodiments, product recommendation evaluation is used to output evaluation data of each product for target recommended products based on user product aggregation features. In one embodiment, product recommendation evaluation is performed through a paired output layer in a sorting neural network to obtain product recommendation evaluation data.

[0134] In step S703 of some embodiments, the target recommended products are sorted according to the product recommendation evaluation data to obtain a product recommendation sequence. For example, the target recommended products include a first product, a second product, and a third product, and the product recommendation evaluation data are (first product: 1, second product: 0), (first product: 1, third product: 0), and (second product: 0, third product: 1). The product recommendation evaluation data is sorted to obtain a product recommendation sequence of the first product, the third product, and the second product.

[0135] In step S105 of some embodiments, the target identity portrait and product recommendation sequence are input into a pre-trained large-scale natural language model (LNM), and the large-scale natural language model generates recommendation content to obtain target recommendation content.

[0136] See also Figure 8 The embodiment of the present application also provides a product recommendation content generation device, which can implement the above-mentioned product recommendation content generation method, and the device includes:

[0137] The information acquisition module 801 is used to acquire the sample identity information of the sample object, acquire the insurance product information of the insurance product, and acquire the browsing record information and purchase record information of the insurance product by the sample object;

[0138] Model training module 802, used to train a preset product demand prediction model based on sample identity information, browsing record information, purchase record information and insurance product information to obtain a target product demand prediction model;

[0139] Demand prediction module 803, used to obtain a target identity portrait of a target object, perform demand prediction on the target object through a target product demand prediction model and the target identity portrait, and obtain a target recommended product;

[0140] The product sorting module 804 is used to sort the target recommended products according to the target identity portrait to obtain a product recommendation sequence;

[0141] The content generation module 805 is used to generate recommended content for the target object according to the target identity portrait and the product recommendation sequence to obtain target recommended content.

[0142] The specific implementation of the product recommendation content generating device is substantially the same as the specific implementation of the product recommendation content generating method described above, and will not be described in detail herein.

[0143] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above product recommendation content generation method when executing the computer program. The electronic device can be any smart terminal including a tablet computer, a car computer, etc.

[0144] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0145] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0146] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the product recommendation content generation method of the embodiment of this application;

[0147] Input / output interface 903, used to implement information input and output;

[0148] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0149] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0150] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0151] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned product recommendation content generation method is implemented.

[0152] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0153] The product recommendation content generation method, product recommendation content generation device, electronic device and storage medium provided in the embodiments of the present application obtain sample identity information of a sample object, obtain insurance product information of an insurance product, and obtain browsing record information and purchase record information of the sample object on the insurance product, and then train a preset product demand prediction model based on the sample identity information, browsing record information, purchase record information and insurance product information to obtain a target product demand prediction model that can accurately predict a target product demand prediction model that meets user needs from a variety of insurance product information; then sort the target recommended products according to the target identity portrait to obtain a product recommendation sequence, thereby further predicting the insurance product most needed by the user; further, generate recommendation content for the target object according to the target identity portrait and the product recommendation sequence to obtain the target recommendation content, thereby generating product recommendation content that better meets user needs.

[0154] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0155] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0156] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0157] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0158] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0159] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0160] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0161] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0162] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0163] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0164] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for generating product recommendation content, characterized in that: The method comprises: Obtaining sample identity information of a sample object, obtaining insurance product information of an insurance product, and obtaining browsing record information and purchase record information of the sample object for the insurance product; Training a preset product demand prediction model based on the sample identity information, the browsing record information, the purchase record information, and the insurance product information to obtain a target product demand prediction model; Obtaining a target identity portrait of a target object, performing demand forecasting on the target object through the target product demand forecasting model and the target identity portrait, and obtaining a target recommended product; Sort the target recommended products according to the target identity portrait to obtain a product recommendation sequence; Recommended content is generated for the target object according to the target identity portrait and the product recommendation sequence to obtain target recommended content.

2. The method according to claim 1, characterized in that The training of a preset product demand prediction model based on the sample identity information, the browsing record information, the purchase record information and the insurance product information to obtain a target product demand prediction model includes: Constructing a user portrait based on the sample identity information, the browsing record information, and the purchase record information to obtain user portrait information; Perform preference clustering based on the sample identity information, the browsing record information, and the purchase record information to obtain user preference information; The preset product demand forecasting model is trained based on the insurance product information, the user portrait information and the user preference information to obtain the target product demand forecasting model.

3. The method according to claim 2, characterized in that The step of training the preset product demand forecasting model based on the insurance product information, the user portrait information, and the user preference information to obtain the target product demand forecasting model includes: Merging data based on the user portrait information and the user preference information to obtain portrait preference merged data; Based on the portrait preference merged data, the data set is divided to obtain portrait preference data and insurance product labels; Perform product forecasting on the preset product demand forecasting model based on the portrait preference data and the insurance product information to obtain a forecast product label; The preset product demand forecasting model is optimized in parameters based on the insurance product label and the predicted product label to obtain the target product demand forecasting model.

4. The method according to claim 3, characterized in that The step of optimizing the parameters of the preset product demand forecasting model based on the insurance product label and the forecast product label to obtain the target product demand forecasting model includes: Calculate the cross entropy loss value according to the predicted product label and the insurance product label to obtain cross entropy loss value data; The parameters of the preset product demand forecasting model are optimized according to the cross entropy loss value data to obtain the target product demand forecasting model.

5. The method according to claim 3, characterized in that: The preset product demand prediction model includes an embedding layer, a portrait preference encoding layer, a feature extraction layer, and a decoding layer. The preset product demand prediction model is used to perform product prediction based on the portrait preference data and the insurance product information to obtain a predicted product label, including: Performing text embedding on the portrait preference data through the embedding layer to obtain a portrait preference embedding; Performing text embedding on the insurance product information through the embedding layer to obtain insurance product embedding; Text encoding is performed on the image preference embedding through the image preference encoding layer to obtain an image preference encoding; Extract product features from the insurance product embedding through the feature extraction layer to obtain insurance product features; The insurance product features and the portrait preference codes are text-decoded through the decoding layer to obtain the predicted product label.

6. The method according to claim 2, characterized in that The performing preference clustering based on the sample identity information, the browsing record information and the purchase record information to obtain user preference information includes: Clustering browsing records according to the sample identity information and the browsing record information to obtain browsing clustering areas; Clustering the purchase records according to the sample identity information and the purchase record information to obtain a purchase clustering area; Performing association relationship clustering according to the sample identity information, the purchase record information, and the browsing record information to obtain an association relationship clustering area; The user preference information is obtained by merging information based on the browsing cluster area, the purchasing cluster area and the association relationship cluster area.

7. The method according to claim 1, characterized in that The step of sorting the target recommended products according to the target identity portrait to obtain a product recommendation sequence includes: Extracting features of the target identity portrait and the target recommended products to obtain user-product aggregate features; Performing product recommendation evaluation on the user product aggregation features to obtain product recommendation evaluation data; wherein the product recommendation evaluation data is used to evaluate the rationality of the recommendation of the target recommended product; A recommendation sequence is generated for the target recommended product according to the product recommendation evaluation data to obtain the product recommendation sequence.

8. A device for generating product recommendation content, characterized in that: The device comprises: An information acquisition module is used to acquire sample identity information of a sample object, acquire insurance product information of an insurance product, and acquire browsing record information and purchase record information of the sample object for the insurance product; A model training module, used to train a preset product demand prediction model based on the sample identity information, the browsing record information, the purchase record information and the insurance product information to obtain a target product demand prediction model; A demand prediction module is used to obtain a target identity portrait of a target object, perform demand prediction on the target object through the target product demand prediction model and the target identity portrait, and obtain a target recommended product; A product sorting module, used to sort the target recommended products according to the target identity portrait to obtain a product recommendation sequence; The content generation module is used to generate recommended content for the target object according to the target identity portrait and the product recommendation sequence to obtain target recommended content.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the product recommendation content generating method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for generating product recommendation content according to any one of claims 1 to 7 is implemented.

Images