Object recommendation method and device, electronic equipment and readable storage medium

Abstract

The invention relates to the technical field of object recommendation, and provides an object recommendation method and device, electronic equipment and a readable storage medium. The method comprises the following steps: inputting each historical session vector corresponding to user data and an attribute vector corresponding to a to-be-recommended object into a trained object recommendation model for attention mechanism processing to obtain a weight value corresponding to each historical session vector; determining each weight session vector corresponding to the user data according to each session weight value corresponding to the to-be-recommended object and each historical session vector corresponding to the user data; performing alignment processing on the user data to obtain an aligned session vector corresponding to the user data; determining a recommendation score corresponding to the to-be-recommended object according to an attribute vector corresponding to the user data, each alignment session vector corresponding to the user data and an attribute vector corresponding to the to-be-recommended object; and determining target object data displayed to the user based on a preset recommendation score threshold and the recommendation score corresponding to the to-be-recommended object.

Description

Technical Field

[0001] The present disclosure relates to the technical field of object recommendation, and in particular to an object recommendation method, device, electronic device and readable storage medium. Background Art

[0002] With the development of Internet technology, user behavior data is becoming increasingly abundant. Accurately mining user interests from a large amount of data has become a research hotspot in the field of recommendation systems. Conventional recommendation methods are based on user behavior sequence models that use attention mechanisms to model user behavior. However, this method mainly focuses on single-type behaviors and fails to fully utilize the temporal information in user behavior sequences. When processing user interests, existing technologies treat user behaviors as isolated events and ignore the user's behavior patterns in specific sessions, which limits the model's ability to capture dynamic changes in user interests.

[0003] It can be seen that the existing technology has the problem of low accuracy of recommendations and poor understanding and capture of user interests because the types of behavioral data concerned are too single and cannot reflect the overall characteristics. Summary of the invention

[0004] In view of this, the embodiments of the present disclosure provide an object recommendation method, device, electronic device and readable storage medium to solve the problem in the prior art that the type of behavioral data concerned is too single and cannot reflect the overall characteristics, resulting in low recommendation accuracy and poor understanding and capture of user interests.

[0005] In a first aspect of an embodiment of the present disclosure, a method for object recommendation is provided, including: obtaining an attribute vector corresponding to user data, at least one historical session vector corresponding to the user data, and an attribute vector corresponding to an object to be recommended; inputting each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended into a trained object recommendation model for attention mechanism processing to obtain a weight value corresponding to each historical session vector; determining each weighted session vector corresponding to the user data according to each session weight value corresponding to the object to be recommended and each historical session vector corresponding to the user data; aligning each weighted session vector corresponding to the user data to obtain an aligned session vector corresponding to the user data; determining a recommendation score corresponding to the object to be recommended according to the attribute vector corresponding to the user data, each aligned session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended; determining target object data to be displayed to the user based on a preset recommendation score threshold and the recommendation score corresponding to the object to be recommended, and sending the target object data to a target terminal device for displaying the target object data in a current graphical user interface of the target terminal device.

[0006] According to a second aspect of an embodiment of the present disclosure, there is provided an object recommendation device, comprising: an acquisition module, configured to acquire an attribute vector corresponding to user data, at least one historical session vector corresponding to the user data, and an attribute vector corresponding to an object to be recommended; a first processing module, configured to input each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended into a trained object recommendation model for attention mechanism processing, so as to obtain a weight value corresponding to each historical session vector; a first determination module, configured to determine each weighted session vector corresponding to the user data according to each session weight value corresponding to the object to be recommended and each historical session vector corresponding to the user data; a second processing module, configured to align each weighted session vector corresponding to the user data, so as to obtain an aligned session vector corresponding to the user data; a second determination module, configured to determine a recommendation score corresponding to the object to be recommended according to the attribute vector corresponding to the user data, each aligned session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended; and a third processing module, configured to determine target object data to be displayed to the user based on a preset recommendation score threshold and the recommendation score corresponding to the object to be recommended, and to send the target object data to a target terminal device, so as to display the target object data in a current graphical user interface of the target terminal device.

[0007] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0008] According to a fourth aspect of an embodiment of the present disclosure, a readable storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0009] Compared with the prior art, the disclosed embodiments have the following beneficial effects: by acquiring the attribute vector corresponding to the user data, at least one historical session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended, the trained object recommendation model is used to perform attention mechanism processing on each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended, so as to obtain the weight value corresponding to each historical session vector, and then the weight value can be combined with each historical session vector corresponding to the user data to determine each weighted session vector corresponding to the user data, and the dimension alignment processing is performed on each weighted session vector corresponding to the user data to obtain the aligned session vector corresponding to the user data, and the attribute vector corresponding to the user data is calculated. The recommendation score corresponding to the object to be recommended is obtained, and the recommendation score corresponding to the object to be recommended is numerically compared with a preset recommendation score threshold. When the recommendation score corresponding to the object to be recommended is When it is greater than or equal to a preset recommendation score threshold, the object to be recommended is determined as the target object data to be displayed to the user, and the target object data can be sent to the target terminal device. The target object data is used to be displayed on the current graphical user interface of the target terminal device. By inputting the attribute vector corresponding to the user data, the historical session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended into the trained object recommendation model and processing them using the attention mechanism, the importance of each historical session to the current recommendation task can be dynamically evaluated, thereby improving the accuracy of the recommendation results. The application of the attention mechanism allows the object recommendation model to assign different weights when processing different users and historical sessions, so that it can adapt to different recommendation scenarios and user behavior patterns, thereby enhancing the flexibility of the object recommendation model. By aligning the weighted session vectors corresponding to the user data to obtain the aligned session vector, the subsequent calculation process can be simplified, the calculation efficiency and response speed can be improved, and the understanding of user interests can be enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1 is a schematic diagram of an application scenario of an embodiment of the present disclosure;

[0012] Figure 2 is a flowchart of an object recommendation method provided by an embodiment of the present disclosure;

[0013] Figure 3 is a structural schematic diagram of an object recommendation device provided by an embodiment of the present disclosure;

[0014] Figure 4 It is a structural schematic diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0015] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.

[0016] It should be noted that the user information (including but not limited to terminal device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0017] A method and device for recommending an object according to an embodiment of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0018] Figure 1 Schematic diagram of an application scenario of an embodiment of the present disclosure. The application scenario may include terminal devices 1, 2 and 3, a server 4 and a network 5.

[0019] Terminal devices 1, 2 and 3 can be hardware or software. When terminal devices 1, 2 and 3 are hardware, they can be various electronic devices with display screens and supporting communication with server 4, including but not limited to smart phones, tablet computers, laptop portable computers and desktop computers, etc.; when terminal devices 1, 2 and 3 are software, they can be installed in the above electronic devices. Terminal devices 1, 2 and 3 can be implemented as multiple software or software modules, or as a single software or software module, and the embodiments of the present disclosure are not limited to this. Furthermore, various applications can be installed on terminal devices 1, 2 and 3, such as data processing applications, instant messaging tools, social platform software, search applications, shopping applications, etc.

[0020] The server 4 may be a server that provides various services, for example, a background server that receives a request sent by a terminal device that establishes a communication connection with the server, and the background server may receive and analyze the request sent by the terminal device, and generate a processing result. The server 4 may be a single server, or a server cluster composed of several servers, or a cloud computing service center, which is not limited in the embodiments of the present disclosure.

[0021] It should be noted that the server 4 can be hardware or software. When the server 4 is hardware, it can be various electronic devices that provide various services for the terminal devices 1, 2, and 3. When the server 4 is software, it can be multiple software or software modules that provide various services for the terminal devices 1, 2, and 3, or it can be a single software or software module that provides various services for the terminal devices 1, 2, and 3, and the embodiments of the present disclosure are not limited to this.

[0022] The network 5 can be a wired network connected by coaxial cable, twisted pair and optical fiber, or it can be a wireless network that can interconnect various communication devices without wiring, such as Bluetooth, Near Field Communication (NFC), infrared, etc., which is not limited in the embodiments of the present disclosure.

[0023] The user can establish a communication connection with the server 4 via the network 5 through the terminal devices 1, 2 and 3 to receive or send information, etc. Specifically, the server 4 can obtain the attribute vector corresponding to the user data, at least one historical session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended through the terminal devices 1, 2 and 3, and perform attention mechanism processing on each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended through the trained object recommendation model to obtain the weight value corresponding to each historical session vector, and then determine each weighted session vector corresponding to the user data according to the weight value in combination with each historical session vector corresponding to the user data, perform dimension alignment processing on each weighted session vector corresponding to the user data, and obtain the aligned session vector corresponding to the user data, and obtain the recommendation score corresponding to the object to be recommended by calculating the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended, and perform numerical comparison between the recommendation score corresponding to the object to be recommended and the preset recommendation score threshold, and when the recommendation score corresponding to the object to be recommended is greater than or equal to the preset recommendation score threshold, the object to be recommended is determined as the target object data to be displayed to the user, and the target object data can be sent to the target terminal device, and the target object data is used to be displayed on the current graphical user interface of the target terminal device.

[0024] It should be noted that the specific types, quantities and combinations of the terminal devices 1, 2 and 3, the server 4 and the network 5 can be adjusted according to the actual needs of the application scenario, and the embodiments of the present disclosure are not limited to this.

[0025] Figure 2 It is a flowchart of an object recommendation method provided by an embodiment of the present disclosure. Figure 2 The object recommendation method can be Figure 1The server executes. Figure 2 As shown, the object recommendation method includes:

[0026] Step 201: Obtain an attribute vector corresponding to user data, at least one historical session vector corresponding to the user data, and an attribute vector corresponding to an object to be recommended.

[0027] Specifically, the data input by the user to the object recommendation model through the terminal device in the current user graphical interface can be extracted from data sources such as historical behavior records, personal information, and / or preferences authorized by the user, and then the input data obtained by the object recommendation model can be encoded into an attribute vector through deep learning, and at least one historical session vector can be constructed based on the user's historical interaction behavior, where the user's historical interaction behavior includes but is not limited to clicks, browsing, or purchases, etc., and the attributes of the object to be recommended can be vectorized to obtain the attribute vector corresponding to the object to be recommended. The object to be recommended can be a product or content, etc., which is not limited here, thereby enhancing the structured representation capability of the data, improving data processing efficiency, and improving the accuracy and personalization level of the object recommendation model.

[0028] Among them, user data can be user-related information, including but not limited to the user's age, gender, geographic location, historical purchase records, and / or browsing behavior, etc.; the attribute vector corresponding to the user data can be a numerical vector converted into the user data through feature extraction and vectorization, and the attribute vector corresponding to the user data can be used for efficient calculation and pattern recognition, etc.; the historical session vector corresponding to the user data can be a vector obtained by encoding the session information through the object recommendation model based on the user's historical interactive behavior, including but not limited to browsing, purchasing, and / or evaluation, etc., according to the time sequence or logical structure of the session, and the historical session vector corresponding to the user data can be used to characterize the user's long-term or short-term interests and behavior patterns.

[0029] The object to be processed may be an object to be recommended to the user, including but not limited to commodities on an e-commerce platform or video content on a streaming platform; the attribute vector corresponding to the object to be processed may be a vector obtained by converting the attributes of the object to be processed through a similar vectorization method, and the attributes of the object to be processed include but are not limited to the price, category, brand of the commodity, the type, label, and / or popularity of the content.

[0030] For example, in an e-commerce platform, when a user logs in and starts browsing products, the user's registration information, historical purchase records, recent browsing behavior and other data can be processed through vectorization methods to obtain the user's corresponding attribute vector and the historical session vector corresponding to the user data. The products on the platform can be generated into corresponding attribute vectors based on the corresponding price range, brand, and keywords in the product description.

[0031] Step 202: Input each historical conversation vector corresponding to the user data and the attribute vector corresponding to the object to be recommended into the trained object recommendation model for attention mechanism processing to obtain the weight value corresponding to each historical conversation vector.

[0032] Specifically, the historical conversation vectors corresponding to the user data and the attribute vectors corresponding to the object to be recommended can be input into the trained object recommendation model. The object recommendation model can process the input vectors through the built-in attention mechanism and calculate the weight value corresponding to each historical conversation vector, thereby enhancing the object recommendation model's attention to the user's historical behavior, improving the personalization of the recommendation results, and improving the accuracy of the recommendation results.

[0033] Among them, the object recommendation model can be a machine learning model. The object recommendation model can determine the user's interest in the recommended object by analyzing the historical behavior data corresponding to the user data and the attributes corresponding to the object to be recommended. The historical behavior data corresponding to the user data includes but is not limited to historical sessions, clicks, and / or purchase records, etc. The attributes corresponding to the object to be recommended can be category, price, and / or brand, etc., which are not limited here.

[0034] The weight value corresponding to the historical conversation vector may refer to the value calculated by the attention mechanism in the object recommendation model for the historical conversation vector. The weight value corresponding to the historical conversation vector may be used to represent the importance of the historical conversation vector for recommending the object to be recommended.

[0035] For example, in the task of recommending products to users on an e-commerce platform, the object recommendation model can obtain the user's historical shopping session data, such as browsed products or products added to the shopping cart, and convert the historical shopping session data into a historical session vector. The platform can extract the attribute vector of the product to be recommended, such as product category, price, and / or brand, and then input the historical session vector and the attribute vector of the product to be recommended into the trained object recommendation model. The object recommendation model can calculate the weight value corresponding to each historical session vector through the attention mechanism. If the user has frequently browsed and purchased electronic products recently, the historical session vector related to electronic products can be assigned a higher weight value.

[0036] Step 203 : determining each weighted session vector corresponding to the user data according to each session weight value corresponding to the to-be-recommended object and each historical session vector corresponding to the user data.

[0037] Specifically, the session weight value corresponding to the object to be recommended and the historical session vector corresponding to the user data can be weighted summed or weighted averaged to obtain the weighted session vectors corresponding to the user data, thereby enhancing the correlation between the user's historical session and the object to be recommended, improving the object recommendation model's ability to capture user interests, and improving the pertinence and effectiveness of the recommendation results.

[0038] The weighted session vector corresponding to the user data may be a vector obtained by weighted processing of each session weight value corresponding to the recommended object and each historical session vector corresponding to the user data. The weighted session vector corresponding to the user data may include information about the user's historical sessions and characteristics of the recommended object.

[0039] For example, in an e-commerce platform, the object recommendation model can calculate the weight value of each historical session based on the user's historical session data and the currently browsed product category, and then weight the user's historical session vector according to the weight value of each historical session to obtain a weighted session vector.

[0040] Step 204 , aligning the weighted session vectors corresponding to the user data to obtain an aligned session vector corresponding to the user data.

[0041] Specifically, the lengths or dimensions of each weighted session vector corresponding to the user data can be aligned through summation and pooling processing, and the elements in each weighted session vector corresponding to the user data can be summed and converted into aligned session vectors with the same length or dimension, thereby enhancing the representation ability of user data, improving the efficiency of the object recommendation model in utilizing the user's historical session information, and improving the robustness and generalization ability of the object recommendation model.

[0042] Among them, the aligned session vector corresponding to the user data can be a vector with the same length or dimension obtained by summing and pooling the weighted session vectors corresponding to the user data. The aligned session vector corresponding to the user data retains important information of the user's historical session, and through alignment processing, the session vectors of each user are kept consistent in dimension and length.

[0043] For example, on a social media platform, when users browse or interact with different content, such as posts, videos, or comments, the interaction behavior can be recorded and a corresponding historical session vector can be generated. Then, based on the user's current interests and needs, such as if they are looking for a specific type of content, the historical session vector can be weighted to obtain a weighted session vector. The weighted session vector can then be aligned to obtain an aligned session vector.

[0044] Step 205 , determining a recommendation score corresponding to the object to be recommended according to the attribute vector corresponding to the user data, each aligned session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended.

[0045] Specifically, the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended can be fused, including but not limited to feature interaction through vector concatenation, dot product, or deep learning models, and then the fused feature vector can be learned and calculated through a multi-layer perceptron to obtain a recommendation score corresponding to the object to be recommended. The recommendation score corresponding to the object to be recommended can be used to characterize the degree of match between the object to be recommended and the user data, thereby enhancing the personalization ability of the object recommendation model, improving the accuracy of the recommendation, and enhancing the user experience.

[0046] Among them, the recommendation score corresponding to the object to be recommended can be a quantitative indicator, and the recommendation score corresponding to the object to be recommended can be used to characterize the matching degree or correlation between the object to be recommended and the user data. The recommendation score corresponding to the object to be recommended can be obtained by feature fusion and calculation of the attribute vector corresponding to the user data, the various aligned session vectors corresponding to the user data, and the attribute vector of the object to be recommended. The recommendation score corresponding to the object to be recommended can also be used to sort or filter objects in the recommendation list.

[0047] For example, in a recommendation system of an e-commerce platform, user data may include the user's purchase history and browsing history, and the recommended objects may be products on the platform. The attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to each product may be fused and calculated to obtain the recommendation score of each product for the user.

[0048] Step 206, based on the preset recommendation score threshold and the recommendation score corresponding to the object to be recommended, determine the target object data to be displayed to the user, and send the target object data to the target terminal device for displaying the target object data on the current graphical user interface of the target terminal device.

[0049] Specifically, the recommendation score corresponding to the object to be recommended can be compared with the preset recommendation score threshold, and objects with recommendation scores greater than or equal to the preset recommendation score threshold can be screened out as target object data. The screened target object data can then be sent to the corresponding target terminal device through network transmission or other means, so that the target object data can be displayed on the current graphical user interface of the target terminal device, wherein the preset recommendation score threshold can be a pre-set score standard, which can be used to screen the objects to be recommended. The target terminal device can be a device currently being used by the user, including but not limited to a smartphone, tablet computer, or computer, etc. The current graphical user interface can be the interface currently being viewed by the user, which can include a variety of visual elements and operation options, thereby enhancing the screening ability of the object recommendation model and improving the pertinence and effectiveness of the recommendation results. By pre-setting the recommendation score threshold, objects that are more closely matched with the user's interests or needs can be automatically screened out.

[0050] Among them, the preset recommendation score threshold may be a score standard set based on historical data, business needs, or user feedback, and the preset recommendation score threshold may be used to screen objects to be recommended; the target object data may refer to the data of objects to be recommended whose recommendation scores are greater than or equal to the preset recommendation score threshold; the target terminal device may refer to the device currently being used by the user, and the target terminal device may be used to receive and display the target object data; the current graphical user interface may refer to the interface currently being viewed by the user, and the current graphical user interface may include a variety of visual elements and operation options, and the current graphical user interface may be used to interact with the user.

[0051] For example, in a recommendation system of an online video platform, a user can watch videos using a smartphone. The object recommendation model can filter out videos that match the user's interests as target object data based on a preset recommendation score threshold, and then send the target object data to the user's smartphone. After receiving the target object data, the smartphone can display the target object data on the current graphical user interface, such as a video playback page or a recommendation list page.

[0052] According to the technical solution provided by the embodiment of the present disclosure, an attribute vector corresponding to user data, at least one historical session vector corresponding to the user data, and an attribute vector corresponding to the object to be recommended are obtained, and through the trained object recommendation model, attention mechanism processing is performed on each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended to obtain a weight value corresponding to each historical session vector, and then, based on the weight value and each historical session vector corresponding to the user data, each weighted session vector corresponding to the user data is determined, and dimension alignment processing is performed on each weighted session vector corresponding to the user data to obtain an aligned session vector corresponding to the user data, and a recommendation score corresponding to the object to be recommended is obtained by calculating the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended, and the recommendation score corresponding to the object to be recommended is numerically compared with a preset recommendation score threshold, and when the recommendation score corresponding to the object to be recommended is greater than or equal to the value of the threshold, the recommendation score is calculated. When the preset recommendation score threshold is reached, the object to be recommended is determined as the target object data to be displayed to the user, and the target object data can be sent to the target terminal device. The target object data is used to be displayed on the current graphical user interface of the target terminal device. By inputting the attribute vector corresponding to the user data, the historical session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended into the trained object recommendation model and processing them using the attention mechanism, the importance of each historical session to the current recommendation task can be dynamically evaluated, thereby improving the accuracy of the recommendation results. The application of the attention mechanism allows the object recommendation model to assign different weights when processing different users and historical sessions, so that it can adapt to different recommendation scenarios and user behavior patterns, thereby enhancing the flexibility of the object recommendation model. By aligning the weighted session vectors corresponding to the user data to obtain the aligned session vector, the subsequent calculation process can be simplified, the calculation efficiency and response speed can be improved, and the understanding of user interests can be enhanced.

[0053] In some embodiments, before obtaining at least one historical session vector corresponding to the user data, the method further includes: obtaining at least one historical behavior vector corresponding to the user data; and concatenating the historical behavior vectors corresponding to the user data to obtain the historical session vectors corresponding to the user data.

[0054] Specifically, at least one historical behavior vector corresponding to the user data can be extracted from the user's historical behavior records. The historical behavior vector corresponding to the user data can include various behavioral characteristics of the user in the past period of time, and the behavioral characteristics include but are not limited to clicks, browsing, and / or purchases. The historical behavior vectors corresponding to the user data can be spliced ​​and combined according to time sequence or logical rules to obtain various historical session vectors corresponding to the user data.

[0055] Among them, the historical behavior vector corresponding to the user data can be a data vector generated by various behaviors performed by the user in the past period of time, and the historical behavior vector corresponding to the user data can include information such as behavior type, behavior time, and / or behavior object; the historical session vector corresponding to the user data can be obtained by splicing multiple historical behavior vectors, and the historical session vector corresponding to the user data can be used to characterize the user's overall behavior pattern and preferences in a specific session or time period.

[0056] The historical session vector corresponding to the user data may be obtained by concatenating multiple historical behavior vectors, and may be used to characterize the overall behavior pattern and preference within a specified time period.

[0057] For example, in a recommendation system of an online shopping platform, a user's interactive behavior may be browsing a product page. The object recommendation model can extract multiple historical behavior vectors corresponding to the current user data from the user's historical behavior records, such as the user's clicks, browsing, and adding to shopping cart behaviors in the past week. The historical behavior vectors corresponding to the user data can be concatenated to obtain the user's historical session vector in the current session.

[0058] According to the technical solution provided in the embodiments of the present disclosure, by extracting at least one historical behavior vector corresponding to the user data from the user's historical behavior records, the historical behavior vectors corresponding to the user data can be spliced ​​to obtain each historical session vector corresponding to the user data, thereby enhancing the data integration capability, improving the accuracy and comprehensiveness of the user data, and improving the personalization and precision of the object recommendation model.

[0059] In some embodiments, before determining the target object data to be displayed to the user based on a preset recommendation score threshold and the recommendation score corresponding to the object to be recommended, and sending the target object data to the target terminal device, it also includes: obtaining at least one object to be recommended and the recommendation score corresponding to each object to be recommended; sorting the recommendation scores corresponding to each object to be recommended to obtain a target object recommendation list corresponding to the user data; and sending the target object recommendation list corresponding to the user data to the target terminal device for display.

[0060] Specifically, at least one object to be recommended and the recommendation score corresponding to each object to be recommended can be obtained through the processing and calculation of the object recommendation model. The object to be recommended can be a type of commodity, multimedia, or service, etc., which is not limited here. Then, according to the recommendation score corresponding to each object to be recommended, each object to be recommended can be sorted from high to low or from low to high to obtain a target object recommendation list corresponding to the current user data. The sorted target object recommendation list can be sent to the user's target terminal device for display.

[0061] Among them, the target object recommendation list corresponding to the user data can refer to a list containing multiple target objects, and the recommendation score corresponding to each target object can be used to characterize the degree of match or relevance between the target object and the current user. The target object recommendation list can be obtained by calculating and processing the object recommendation model.

[0062] For example, in the recommendation system of an online music playback platform, users can use mobile phone applications to listen to songs, and the object recommendation model can calculate multiple songs to be recommended and the recommendation scores corresponding to each song to be recommended based on the user's listening history, likes record, search behavior and other data. The recommended songs can then be sorted according to the recommendation scores corresponding to each song to obtain a target song recommendation list corresponding to the current user data, and the target song recommendation list corresponding to the current user data can be sent to the user's mobile phone application for display.

[0063] According to the technical solution provided by the embodiment of the present disclosure, at least one object to be recommended and the recommendation score corresponding to each object to be recommended are obtained through processing and calculation of the object recommendation model, and then each object to be recommended can be sorted from high to low or from low to high according to the recommendation score corresponding to each object to be recommended, to obtain a target object recommendation list corresponding to the current user data, and the sorted target object recommendation list can be sent to the user's target terminal device for display, thereby enhancing the intelligence and personalization of the object recommendation model, improving the accuracy and relevance of the recommendation results, and enhancing the user experience and satisfaction.

[0064] In some embodiments, a recommendation score corresponding to an object to be recommended is determined based on an attribute vector corresponding to user data, each aligned session vector corresponding to the user data, and an attribute vector corresponding to the object to be recommended, including: concatenating the attribute vector corresponding to the user data, each aligned session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended to obtain a user recommendation vector corresponding to the object to be recommended; mapping the user recommendation vector corresponding to the object to be recommended through a multi-layer perceptron to obtain a space mapping vector corresponding to the object to be recommended; and calculating the recommendation score of the space mapping vector corresponding to the object to be recommended to obtain a recommendation score corresponding to the object to be recommended.

[0065] Specifically, the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended can be concatenated to obtain a user recommendation vector corresponding to the object to be recommended. The user recommendation vector can then be nonlinearly mapped using a multi-layer perceptron to obtain a mapping vector of the object to be recommended in space. The spatial mapping vector can then be used to calculate a recommendation score to obtain a recommendation score corresponding to the object to be recommended.

[0066] Among them, the user recommendation vector corresponding to the object to be recommended can be obtained by concatenating the attribute vector corresponding to the user data, the aligned session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended. The user recommendation vector corresponding to the object to be recommended can be used to characterize the relationship between the user and the object to be recommended.

[0067] The multilayer perceptron can be a feedforward artificial neural network model. The multilayer perceptron can include multiple hidden layers. The neurons in each hidden layer are connected to all the neurons in the previous layer, and can perform nonlinear transformation and mapping on the input data.

[0068] The spatial mapping vector corresponding to the object to be recommended can be a vector obtained by performing nonlinear mapping processing on the user recommendation vector through a multi-layer perceptron. The position of the spatial mapping vector corresponding to the object to be recommended in space can be used to characterize the potential relationship between the object to be recommended and the user. The spatial mapping vector corresponding to the object to be recommended can be used for subsequent recommendation score calculation and ranking of recommendation results.

[0069] For example, in an e-commerce platform, the user's historical behavior data, such as browsing and purchase records, is converted into attribute vectors and aligned session vectors, and the product information is converted into corresponding attribute vectors. The attribute vector corresponding to the user data, the aligned session vector corresponding to the user data, and the attribute vector corresponding to the product can be concatenated to obtain a user recommendation vector, which can then be mapped through a multi-layer perceptron to obtain a spatial mapping vector corresponding to the product. The recommendation score of the spatial mapping vector corresponding to the product is calculated, and the products are sorted according to the recommendation score, so that the product that best meets the user's needs can be recommended to the user.

[0070] According to the technical solution provided by the embodiments of the present disclosure, the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended are concatenated to obtain a user recommendation vector corresponding to the object to be recommended. The user recommendation vector can then be nonlinearly mapped by a multi-layer perceptron to obtain a mapping vector of the object to be recommended in space. The spatial mapping vector can be calculated to obtain a recommendation score corresponding to the object to be recommended, thereby enhancing the personalization of the object recommendation model, improving the accuracy of the recommendation results, and improving the user experience and satisfaction.

[0071] In some embodiments, before obtaining the attribute vector corresponding to the user, at least one historical session vector corresponding to the user, and the attribute vector corresponding to the object to be recommended, it also includes: obtaining a user attribute training vector, multiple training session vectors corresponding to the user data, attribute vectors corresponding to multiple training objects, and real labels corresponding to each training object; inputting the multiple training session vectors corresponding to the user data and the attribute vectors corresponding to each training object into the object recommendation model for attention mechanism processing to obtain the weight value corresponding to each training session vector; determining each weighted training session vector corresponding to the user data according to the weight value corresponding to each training session vector and each training session vector corresponding to the user data; aligning each weighted training session vector corresponding to the user data to obtain each aligned training session vector corresponding to the user data; determining the recommendation score corresponding to each training object according to the user attribute training vector, each aligned training session vector corresponding to the user data, and the attribute vector corresponding to each training object; determining the loss of the object recommendation model according to the recommendation score corresponding to each training object and the real label corresponding to each training object; updating the parameters in the object recommendation model according to the loss corresponding to the object recommendation model in a loop iteration manner.

[0072] Specifically, a user attribute training vector, multiple training session vectors corresponding to user data, attribute vectors corresponding to multiple training objects, and a true label corresponding to each training object can be obtained. The true label can be used to characterize the actual interaction behavior between the user and the candidate, including but not limited to purchase or click, etc. The multiple training session vectors corresponding to the user data and the training object attribute vector can be input into the object recommendation model, and a weight value can be assigned to each training session vector through an attention mechanism. The weight value corresponding to each training session vector can be used to characterize the importance of the corresponding session to the user preference. The weighted weight training session vector can be calculated according to the weight value corresponding to the training session vector and the training session vector corresponding to the user data. The weighted training session vector can be aligned to obtain the aligned training session vector corresponding to the user data. The recommendation score corresponding to each training object can be calculated based on the user attribute training vector, the aligned training session vector, and the attribute vector corresponding to the training object. Then, the recommendation score can be compared with the true label to calculate the loss value of the object recommendation model. The loss value can be used to measure the deviation between the prediction result of the object recommendation model and the actual behavior of the user. The parameters of the object recommendation model can be continuously adjusted according to the loss value through loop iteration until the model performance is stable.

[0073] Among them, the user attribute training vector can be a vector representation of the user's basic features after numerical processing, wherein the user's basic features can be age, gender, or region, etc., which are not limited here; the training session vector corresponding to the user data can be a vector representation of the session level obtained by processing the user's historical behavior data; the attribute vector corresponding to the training object can be a feature vector of the candidate object used for training; the real label corresponding to the training object can be the interaction behavior label actually generated by the user with each training object; the weight value corresponding to the training session vector can be calculated by the attention mechanism, and the weight value corresponding to the training session vector can be used to characterize the difference in the contribution of different sessions to user preferences; the weight training session vector corresponding to the user data can be a weighted session vector calculated based on the weight value and the training session vector; the aligned training session vector corresponding to the user data can be a session vector processed by a specific alignment method, and the aligned training session vector corresponding to the user data can be used for comparison in the same dimensional space; the recommendation score corresponding to the training object can be used to characterize the object recommendation model predicting the user's preference for a specific training object based on the user attributes and session information; the loss of the object recommendation model can be the error between the model prediction result and the actual user behavior, and the loss of the object recommendation model can be used to guide the update of the parameters of the object recommendation model.

[0074] According to the technical solution provided by the embodiment of the present disclosure, by obtaining a user attribute training vector, multiple training session vectors corresponding to user data, attribute vectors corresponding to multiple training objects, and a true label corresponding to each training object, the multiple training session vectors corresponding to the user data and the training object attribute vector can be input into the object recommendation model, a weight value can be assigned to each training session vector through an attention mechanism, a weighted weight training session vector can be calculated according to the weight value corresponding to the training session vector and the training session vector corresponding to the user data, the weighted training session vector can be aligned to obtain the aligned training session vector corresponding to the user data, the recommendation score corresponding to each training object can be calculated based on the user attribute training vector, the aligned training session vector, and the attribute vector corresponding to the training object, and then the recommendation score can be compared with the true label to calculate the loss value of the object recommendation model, and the parameters of the object recommendation model can be continuously adjusted according to the loss value through loop iteration until the model performance is stable, thereby enhancing the generalization ability of the object recommendation model, improving the accuracy of the recommendation, and improving the satisfaction of the user experience.

[0075] In some embodiments, each historical conversation vector corresponding to the user data and the attribute vector corresponding to the object to be recommended are input into the trained object recommendation model for attention mechanism processing to obtain a weight value corresponding to each historical conversation vector, including: performing nonlinear transformation processing on each historical conversation vector corresponding to the user data and the attribute vector corresponding to the object to be recommended to obtain each conversation attention score corresponding to the object to be recommended; performing weight allocation processing on each historical conversation vector corresponding to the user data and each conversation attention score corresponding to the object to be recommended to obtain a weight value corresponding to each historical conversation vector.

[0076] Specifically, each historical conversation vector corresponding to the user data and the attribute vector corresponding to the object to be recommended can be input into the trained object recommendation model, and the input vector can be transformed through the nonlinear transformation layer inside the object recommendation model to calculate the session attention scores corresponding to the object to be recommended. The session attention scores corresponding to the object to be recommended can be used to characterize the importance of each historical conversation vector in determining the recommendation degree of the object to be recommended. The historical conversation vectors corresponding to the user data can be weighted according to the session attention scores corresponding to the object to be recommended, that is, each historical conversation vector is assigned a weight value according to the corresponding session attention score.

[0077] Among them, the conversation attention score corresponding to the object to be recommended can be a numerical representation obtained through nonlinear transformation processing in the object recommendation model, and the conversation attention score corresponding to the object to be recommended can be used to represent the numerical contribution of each historical conversation vector to the recommendation degree of the object to be recommended.

[0078] For example, in a video recommendation system, a user's viewing history can be represented as multiple historical conversation vectors. The video objects to be recommended can have attribute vectors corresponding to each video object to be recommended, such as video type or cast, etc. The conversation attention score of each historical conversation vector for the video to be recommended can be calculated through the nonlinear transformation layer in the trained object recommendation model. If the user has recently watched science fiction movies frequently, the historical conversation vectors related to science fiction movies can obtain a higher conversation attention score. Then, according to the conversation attention score, the object recommendation model can assign weights to each historical conversation vector to obtain the weight value corresponding to the historical conversation vector.

[0079] According to the technical solution provided by the embodiments of the present disclosure, by inputting each historical conversation vector corresponding to the user data and the attribute vector corresponding to the object to be recommended into the trained object recommendation model, the input vector can be transformed by the nonlinear transformation layer inside the object recommendation model to calculate the each conversation attention score corresponding to the object to be recommended, and the each historical conversation vector corresponding to the user data can be weighted according to the each conversation attention score corresponding to the object to be recommended, that is, each historical conversation vector is assigned a weight value according to the corresponding conversation attention score, thereby enhancing the object recommendation model's ability to understand and use the user's historical behavior data, improving the personalization and accuracy of the recommendation results, and improving the user experience satisfaction.

[0080] In some embodiments, each weighted session vector corresponding to the user data is aligned to obtain an aligned session vector corresponding to the user data, including: performing weighted summation processing on each weighted session vector corresponding to the user data to obtain a high-dimensional session vector corresponding to the user data; and performing alignment processing on the high-dimensional session vector corresponding to the user data to obtain each aligned session vector corresponding to the user data.

[0081] Specifically, a weighted summation process can be performed on each weighted session vector corresponding to the user data, that is, a weighted average can be performed according to the weight value of each weighted session vector to obtain a high-dimensional session vector corresponding to the user data. The high-dimensional session vector can be a single vector representation that integrates multiple historical session information of the user. The high-dimensional session vector can be aligned, and the alignment process includes but is not limited to dimensionality reduction or standardization, to obtain each aligned session vector corresponding to the user data.

[0082] Among them, the high-dimensional session vector corresponding to the user data can be a single vector representation obtained by weighted summing up the weighted session vectors corresponding to the user data, the high-dimensional session vector corresponding to the user data can be used to characterize multiple historical session information of the user, and the high-dimensional session vector corresponding to the user data can be obtained by weighted averaging multiple session vectors in the user's historical behavior according to the corresponding weight values.

[0083] For example, in a news recommendation system, a user's reading history can be represented as multiple weighted session vectors, each of which can respectively characterize the user's news reading behavior in different time periods. Then, each weighted session vector can be weighted summed to obtain a high-dimensional session vector corresponding to the user data. The high-dimensional session vectors corresponding to the user data can be aligned to obtain each aligned session vector corresponding to the user data.

[0084] According to the technical solution provided by the embodiment of the present disclosure, by performing weighted summation processing on each weighted session vector corresponding to the user data, a high-dimensional session vector corresponding to the user data is obtained, and the high-dimensional session vector can be aligned to obtain each aligned session vector corresponding to the user data, thereby enhancing the object recommendation model's ability to integrate user historical behavior information and improving the accuracy and efficiency of recommendation decisions.

[0085] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present disclosure, and will not be described in detail here.

[0086] The following are embodiments of the device disclosed herein, which can be used to execute the method embodiments disclosed herein. For details not disclosed in the device embodiments disclosed herein, please refer to the method embodiments disclosed herein.

[0087] Figure 3 is a schematic diagram of an object recommendation device provided by an embodiment of the present disclosure. Figure 3 As shown, the object recommendation device includes:

[0088] An acquisition module 301 is used to acquire an attribute vector corresponding to user data, at least one historical session vector corresponding to the user data, and an attribute vector corresponding to an object to be recommended;

[0089] The first processing module 302 is used to input each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended into the trained object recommendation model for attention mechanism processing to obtain the weight value corresponding to each historical session vector;

[0090] A first determination module 303, configured to determine each weighted session vector corresponding to the user data according to each session weight value corresponding to the to-be-recommended object and each historical session vector corresponding to the user data;

[0091] The second processing module 304 is used to align each weighted session vector corresponding to the user data to obtain an aligned session vector corresponding to the user data;

[0092] The second determination module 305 is used to determine the recommendation score corresponding to the object to be recommended according to the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended;

[0093] The third processing module 306 is used to determine the target object data to be displayed to the user based on the preset recommendation score threshold and the recommendation score corresponding to the object to be recommended, and send the target object data to the target terminal device for displaying the target object data in the current graphical user interface of the target terminal device.

[0094] According to the technical solution provided by the embodiment of the present disclosure, an attribute vector corresponding to user data, at least one historical session vector corresponding to the user data, and an attribute vector corresponding to the object to be recommended are obtained, and through the trained object recommendation model, attention mechanism processing is performed on each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended to obtain a weight value corresponding to each historical session vector, and then, based on the weight value and each historical session vector corresponding to the user data, each weighted session vector corresponding to the user data is determined, and dimension alignment processing is performed on each weighted session vector corresponding to the user data to obtain an aligned session vector corresponding to the user data, and a recommendation score corresponding to the object to be recommended is obtained by calculating the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended, and the recommendation score corresponding to the object to be recommended is numerically compared with a preset recommendation score threshold, and when the recommendation score corresponding to the object to be recommended is greater than or equal to the value of the threshold, the recommendation score is calculated. When the preset recommendation score threshold is reached, the object to be recommended is determined as the target object data to be displayed to the user, and the target object data can be sent to the target terminal device. The target object data is used to be displayed on the current graphical user interface of the target terminal device. By inputting the attribute vector corresponding to the user data, the historical session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended into the trained object recommendation model and processing them using the attention mechanism, the importance of each historical session to the current recommendation task can be dynamically evaluated, thereby improving the accuracy of the recommendation results. The application of the attention mechanism allows the object recommendation model to assign different weights when processing different users and historical sessions, so that it can adapt to different recommendation scenarios and user behavior patterns, thereby enhancing the flexibility of the object recommendation model. By aligning the weighted session vectors corresponding to the user data to obtain the aligned session vector, the subsequent calculation process can be simplified, the calculation efficiency and response speed can be improved, and the understanding of user interests can be enhanced.

[0095] In some embodiments, the object recommendation device is further used to obtain at least one historical behavior vector corresponding to the user data; and concatenate the historical behavior vectors corresponding to the user data to obtain the historical session vectors corresponding to the user data.

[0096] In some embodiments, the above-mentioned object recommendation device is also used to obtain at least one object to be recommended and the recommendation score corresponding to each object to be recommended; sort the recommendation scores corresponding to each object to be recommended to obtain a target object recommendation list corresponding to the user data; and send the target object recommendation list corresponding to the user data to the target terminal device for display.

[0097] In some embodiments, the second determination module 305 is specifically used to concatenate the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended to obtain a user recommendation vector corresponding to the object to be recommended; map the user recommendation vector corresponding to the object to be recommended through a multi-layer perceptron to obtain a space mapping vector corresponding to the object to be recommended; and calculate the recommendation score of the space mapping vector corresponding to the object to be recommended to obtain a recommendation score corresponding to the object to be recommended.

[0098] In some embodiments, the object recommendation device is further used to obtain a user attribute training vector, multiple training session vectors corresponding to user data, attribute vectors corresponding to multiple training objects, and a true label corresponding to each training object; input multiple training session vectors corresponding to user data and attribute vectors corresponding to each training object into an object recommendation model for attention mechanism processing to obtain a weight value corresponding to each training session vector; determine each weighted training session vector corresponding to the user data according to the weight value corresponding to each training session vector and each training session vector corresponding to the user data; align each weighted training session vector corresponding to the user data to obtain each aligned training session vector corresponding to the user data; determine a recommendation score corresponding to each training object according to the user attribute training vector, each aligned training session vector corresponding to the user data, and an attribute vector corresponding to each training object; determine the loss of the object recommendation model according to the recommendation score corresponding to each training object and the true label corresponding to each training object; and update the parameters in the object recommendation model according to the loss corresponding to the object recommendation model in a loop iteration manner.

[0099] In some embodiments, the first processing module 302 is specifically used to perform nonlinear transformation processing on each historical conversation vector corresponding to the user data and the attribute vector corresponding to the object to be recommended to obtain each conversation attention score corresponding to the object to be recommended; perform weight allocation processing on each historical conversation vector corresponding to the user data and the attribute vector corresponding to the object to be recommended to obtain the weight value corresponding to each historical conversation vector.

[0100] In some embodiments, the second processing module 304 is specifically used to perform weighted summation processing on each weighted session vector corresponding to the user data to obtain a high-dimensional session vector corresponding to the user data; and perform alignment processing on the high-dimensional session vector corresponding to the user data to obtain each aligned session vector corresponding to the user data.

[0101] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure.

[0102] Figure 4 Schematic diagram of an electronic device 4 provided in an embodiment of the present disclosure. Figure 4 As shown, the electronic device 4 of this embodiment includes: a processor 401, a memory 402, and a computer program 403 stored in the memory 402 and executable on the processor 401. When the processor 401 executes the computer program 403, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor 401 executes the computer program 403, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0103] The electronic device 4 may be a desktop computer, a notebook, a PDA, a cloud server, or other electronic device. The electronic device 4 may include, but is not limited to, a processor 401 and a memory 402. Those skilled in the art will appreciate that Figure 4 The electronic device 4 is merely an example and does not limit the electronic device 4 , and may include more or less components than those shown in the figure, or different components.

[0104] The processor 401 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0105] The memory 402 may be an internal storage unit of the electronic device 4, for example, a hard disk or memory of the electronic device 4. The memory 402 may also be an external storage device of the electronic device 4, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 4. The memory 402 may also include both an internal storage unit and an external storage device of the electronic device 4. The memory 402 is used to store computer programs and other programs and data required by the electronic device.

[0106] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0107] If the integrated module / 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 readable storage medium (e.g., a computer-readable storage medium). Based on this understanding, the present disclosure implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. The computer program may include computer program code, and the computer program code may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable storage medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0108] The above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should all be included in the protection scope of the present disclosure.

Claims

1. An object recommendation method, characterized in that: include: Acquire an attribute vector corresponding to the user data, at least one historical session vector corresponding to the user data, and an attribute vector corresponding to the object to be recommended; Input each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended into the trained object recommendation model for attention mechanism processing to obtain the weight value corresponding to each historical session vector; Determine each weighted session vector corresponding to the user data according to each session weight value corresponding to the to-be-recommended object and each historical session vector corresponding to the user data; Performing alignment processing on each weighted session vector corresponding to the user data to obtain an aligned session vector corresponding to the user data; Determine a recommendation score corresponding to the object to be recommended according to the attribute vector corresponding to the user data, each aligned session vector corresponding to the user data, and the attribute vector corresponding to the object to be recommended; Based on a preset recommendation score threshold and the recommendation score corresponding to the object to be recommended, the target object data to be displayed to the user is determined, and the target object data is sent to a target terminal device for displaying the target object data on a current graphical user interface of the target terminal device.

2. The object recommendation method according to claim 1, characterized in that: Before acquiring at least one historical session vector corresponding to the user data, the method further includes: Obtaining at least one historical behavior vector corresponding to the user data; The historical behavior vectors corresponding to the user data are concatenated to obtain the historical session vectors corresponding to the user data.

3. The object recommendation method according to claim 1, characterized in that: Before determining the target object data to be displayed to the user based on the preset recommendation score threshold and the recommendation score corresponding to the to-be-recommended object, and sending the target object data to the target terminal device, the method further includes: Obtaining at least one of the objects to be recommended and a recommendation score corresponding to each of the objects to be recommended; Sorting the recommendation scores corresponding to the objects to be recommended to obtain a target object recommendation list corresponding to the user data; The target object recommendation list corresponding to the user data is sent to the target terminal device for display.

4. The object recommendation method according to claim 1, characterized in that: The determining, according to the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended, a recommendation score corresponding to the object to be recommended, comprises: Concatenate the attribute vector corresponding to the user data, the aligned session vectors corresponding to the user data, and the attribute vector corresponding to the object to be recommended, to obtain a user recommendation vector corresponding to the object to be recommended; Mapping the user recommendation vector corresponding to the object to be recommended by a multi-layer perceptron to obtain a spatial mapping vector corresponding to the object to be recommended; A recommendation score calculation process is performed on the space mapping vector corresponding to the object to be recommended to obtain a recommendation score corresponding to the object to be recommended.

5. The object recommendation method according to claim 1, characterized in that: Before acquiring the attribute vector corresponding to the user, at least one historical conversation vector corresponding to the user, and the attribute vector corresponding to the object to be recommended, the method further includes: Obtaining a user attribute training vector, a plurality of training session vectors corresponding to the user data, attribute vectors corresponding to a plurality of training objects, and a true label corresponding to each of the training objects; Inputting a plurality of training session vectors corresponding to the user data and an attribute vector corresponding to each training object into the object recommendation model for attention mechanism processing to obtain a weight value corresponding to each training session vector; Determine each weighted training session vector corresponding to the user data according to the weight values ​​corresponding to each training session vector and each training session vector corresponding to the user data; Performing alignment processing on each weight training session vector corresponding to the user data to obtain each aligned training session vector corresponding to the user data; Determine a recommendation score corresponding to each of the training objects according to the user attribute training vector, each aligned training session vector corresponding to the user data, and an attribute vector corresponding to each of the training objects; Determining the loss of the object recommendation model according to the recommendation score corresponding to each of the training objects and the true label corresponding to each of the training objects; In a loop iterative manner, the parameters in the object recommendation model are updated according to the loss corresponding to the object recommendation model.

6. The object recommendation method according to claim 1, characterized in that: The inputting each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended into the trained object recommendation model for attention mechanism processing to obtain the weight value corresponding to each historical session vector includes: Performing nonlinear transformation processing on each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended, to obtain each session attention score corresponding to the object to be recommended; A weight distribution process is performed on each historical conversation vector corresponding to the user data and each conversation attention score corresponding to the object to be recommended, so as to obtain a weight value corresponding to each historical conversation vector.

7. The object recommendation method according to claim 1, characterized in that: The aligning of the weighted session vectors corresponding to the user data to obtain the aligned session vector corresponding to the user data includes: Performing weighted sum processing on each weighted session vector corresponding to the user data to obtain a high-dimensional session vector corresponding to the user data; Alignment processing is performed on the high-dimensional session vectors corresponding to the user data to obtain each aligned session vector corresponding to the user data.

8. An object recommendation device, characterized in that: include: An acquisition module, used to acquire an attribute vector corresponding to user data, at least one historical session vector corresponding to the user data, and an attribute vector corresponding to an object to be recommended; A first processing module is used to input each historical session vector corresponding to the user data and the attribute vector corresponding to the object to be recommended into the trained object recommendation model for attention mechanism processing to obtain a weight value corresponding to each historical session vector; A first determination module, configured to determine each weighted session vector corresponding to the user data according to each session weight value corresponding to the to-be-recommended object and each historical session vector corresponding to the user data; A second processing module is used to align each weighted session vector corresponding to the user data to obtain an aligned session vector corresponding to the user data; A second determination module, configured to determine a recommendation score corresponding to the object to be recommended according to an attribute vector corresponding to the user data, each aligned session vector corresponding to the user data, and an attribute vector corresponding to the object to be recommended; The third processing module is used to determine the target object data to be displayed to the user based on a preset recommendation score threshold and the recommendation score corresponding to the object to be recommended, and send the target object data to the target terminal device, so as to display the target object data on the current graphical user interface of the target terminal device.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images