CN118035542A

CN118035542A - Object list recommendation method and device, electronic equipment and readable storage medium

Info

Publication number: CN118035542A
Application number: CN202410163045.4A
Authority: CN
Inventors: 徐琳
Original assignee: Shenzhen Xumi Yuntu Space Technology Co Ltd
Current assignee: Shenzhen Xumi Yuntu Space Technology Co Ltd
Priority date: 2024-02-05
Filing date: 2024-02-05
Publication date: 2024-05-14

Abstract

The disclosure relates to the technical field of list recommendation, and provides an object list recommendation method, an object list recommendation device, electronic equipment and a readable storage medium. The method comprises the following steps: acquiring user historical behavior information and current graphical user interface information of terminal equipment, wherein the user historical behavior information is used for representing historical interaction behaviors of an object to be recommended; performing text generation processing on the user history behavior information and the current graphical user interface information through the trained large language model to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface; inputting a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector into the deep interest network model to obtain a target object recommendation sequence; according to the target object recommendation sequence, a target object recommendation list displayed for a user is determined, and the target object recommendation list is sent to the terminal equipment, so that the accuracy of object list recommendation is improved by displaying the target object recommendation list on the current graphical user interface of the terminal equipment, the contact of cross-scene recommendation is enhanced, the richness of multi-behavior characteristics of the user is improved, the uniformity of interest characterization of the user is improved, and the problem that in the prior art, the accuracy of target object recommendation is low due to the fact that the cross-scene information characteristics are different is solved.

Description

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

Technical Field

The disclosure relates to the technical field of list recommendation, and in particular relates to an object list recommendation method, an object list recommendation device, electronic equipment and a readable storage medium.

Background

The recommendation system is a personalized information recommendation system which recommends information, products and the like which are interested by a user to the user according to the information demand, the interest and the like of the user through a recommendation model, the recommendation system performs personalized calculation by researching the interest preference of the user, and the system discovers the interest points of the user, so that the user is guided to discover the information demand of the user, the recommendation system is widely applied to a plurality of fields, an independent subject is gradually formed, but in the current development, the recommendation model has low accuracy of recommendation results under the condition that the history behavior information of the user is insufficiently learned or the history behavior information characteristics of the user are sparse, and conventionally, the history behavior information characteristics of the user in a plurality of scenes are singly analyzed for recommendation through modeling of different scenes, but the method increases a large number of calculation processes and slows down the processing speed.

Therefore, the problems of low accuracy, large processing capacity and low processing speed of the recommendation result caused by independent operation of the recommendation model on different scenes are solved.

Disclosure of Invention

In view of the above, the embodiments of the present disclosure provide a method, an apparatus, an electronic device, and a readable storage medium for recommending an object list, so as to solve the problems in the prior art that, due to isolated modeling and separate operation of a recommendation model on different scenes, the recommendation result is low in accuracy, large in processing amount, and slow in processing speed.

In a first aspect of an embodiment of the present disclosure, there is provided an object list recommendation method, including: acquiring user historical behavior information and current graphical user interface information of terminal equipment, wherein the user historical behavior information is used for representing historical interaction behaviors of an object to be recommended; performing text generation processing on the user history behavior information and the current graphical user interface information through the trained large language model to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface; inputting a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector into the deep interest network model to obtain a target object recommendation sequence; and determining a target object recommendation list displayed for the user according to the target object recommendation sequence, and sending the target object recommendation list to the terminal equipment, wherein the target object recommendation list is used for displaying the target object recommendation list on the current graphical user interface of the terminal equipment.

In a second aspect of the embodiments of the present disclosure, there is provided an object list recommendation apparatus, including: the acquisition module is used for acquiring user historical behavior information and current graphical user interface information of the terminal equipment, wherein the user historical behavior information is used for representing historical interaction behaviors of the object to be recommended; the first processing module is used for generating and processing texts of the historical behavior information of the user and the current graphical user interface information through the trained large language model to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface; the second processing module is used for inputting the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector into the deep interest network model to obtain a target object recommendation sequence; the determining module is used for determining a target object recommendation list displayed for the user according to the target object recommendation sequence, and sending the target object recommendation list to the terminal equipment, and displaying the target object recommendation list on the current graphical user interface of the terminal equipment.

In a third aspect of the disclosed embodiments, an electronic device is provided, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the above method when executing the computer program.

In a fourth aspect of the disclosed embodiments, there is provided a readable storage medium storing a computer program which, when executed by a processor, implements the steps of the above method.

Compared with the prior art, the embodiment of the disclosure has the beneficial effects that: the method comprises the steps of acquiring interaction history data of a user and information of a current graphical user interface displayed on terminal equipment, analyzing the user history behavior information and the current graphical user interface information based on a trained large language model, converting the user history behavior information and the current graphical user interface information into a user preference vector and a current graphical user interface feature vector through a text generation technology, inputting the two vectors into a deep interest network model, recommending proper target objects for the user by the deep interest network model according to the preference of the user and the characteristics of the current graphical user interface, arranging the target objects according to the sequence of the recommended objects, generating a target object recommendation list according to the sequence of the recommended objects, sending the target object recommendation list to the terminal equipment for displaying on the current graphical user interface, and therefore improving the recommendation accuracy of the target object list, enhancing the cross-scene recommendation relativity, improving the richness of multi-behavior characteristics of the user and improving the uniformity of user interest characterization.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings that are required for the embodiments or the description of the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and other drawings may be obtained according to these drawings without inventive effort for a person of ordinary skill in the art.

Fig. 1 is a scene schematic diagram of an application scene of an embodiment of the present disclosure;

Fig. 2 is a flowchart of an object list recommendation method according to an embodiment of the present disclosure;

FIG. 3 is a schematic diagram of an object list recommendation model provided by an embodiment of the present disclosure;

Fig. 4 is a schematic structural diagram of an object list recommending apparatus according to an embodiment of the present disclosure;

fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system configurations, techniques, etc. in order to provide a thorough understanding of the disclosed embodiments. However, it will be apparent to one skilled in the art that the present disclosure may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present disclosure with unnecessary detail.

It should be noted that, the user information (including, but not limited to, terminal device information, user personal information, etc.) and the data (including, but not limited to, data for presentation, analyzed data, etc.) related to the present disclosure are information and data authorized by the user or sufficiently authorized by each party.

An object list recommendation method and apparatus according to embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

Fig. 1 is a scene diagram of an application scene 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.

The terminal devices 1,2 and 3 may be hardware or software. When the terminal devices 1,2 and 3 are hardware, they may be various electronic devices having a display screen and supporting communication with the server 4, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like; when the terminal apparatuses 1,2, and 3 are software, they can be installed in the electronic apparatus as above. The terminal devices 1,2 and 3 may be implemented as a plurality of software or software modules, or as a single software or software module, to which the embodiments of the present disclosure are not limited. Further, various applications, such as a data processing application, an instant messaging tool, social platform software, a search class application, a shopping class application, and the like, may be installed on the terminal devices 1,2, and 3.

The server 4 may be a server that provides various services, for example, a background server that receives a request transmitted from a terminal device with which communication connection is established, and the background server may perform processing such as receiving and analyzing the request transmitted from the terminal device and generate a processing result. The server 4 may be a server, a server cluster formed by a plurality of servers, or a cloud computing service center, which is not limited in the embodiment of the present disclosure.

The server 4 may be hardware or software. When the server 4 is hardware, it may be various electronic devices that provide various services to the terminal devices 1, 2, and 3. When the server 4 is software, it may be a plurality of software or software modules providing various services to the terminal devices 1, 2, and 3, or may be a single software or software module providing various services to the terminal devices 1, 2, and 3, which is not limited by the embodiments of the present disclosure.

The network 5 may be a wired network using coaxial cable, twisted pair wire, and optical fiber connection, or may be a wireless network that can implement interconnection of various Communication devices without wiring, for example, bluetooth (Bluetooth), near Field Communication (NFC), infrared (Infrared), etc., which are not limited by the embodiments of the present disclosure.

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 transmit information or the like. Specifically, the server 4 may obtain, via the terminal devices 1, 2, and 3, the interaction history data of the user and the information of the current graphical user interface displayed on the terminal device, analyze, based on the trained large language model, the user history behavior information and the current graphical user interface information, convert, through a text generation technique, the user history behavior information and the current graphical user interface information into a user preference vector and a current graphical user interface feature vector, input the two vectors into the deep interest network model, and the deep interest network model may recommend a suitable target object for the user according to the preference of the user and the feature of the current graphical user interface, and arrange in order, generate a target object recommendation list according to the order of the recommended objects, and send the target object recommendation list to the terminal device for displaying on the current graphical user interface.

It should be noted that the specific types, numbers and combinations of the terminal devices 1,2 and 3, the server 4 and the network 5 may be adjusted according to the actual requirements of the application scenario, which is not limited by the embodiment of the present disclosure.

Fig. 2 is a flowchart of an object list recommendation method according to an embodiment of the present disclosure. The object list recommendation method of fig. 2 may be performed by the server of fig. 1. As shown in fig. 2, the object list recommendation method includes:

step 201, obtaining user historical behavior information and current graphical user interface information of a terminal device, wherein the user historical behavior information is used for representing historical interaction behaviors of an object to be recommended.

Specifically, user history behavior information authorized by the user and current graphical user interface information can be obtained, wherein the user history behavior information can be used for representing the user's history interaction behavior in a recommendation system and an object to be recommended can be an object which is used for carrying out interaction behavior with the user in the recommendation system, and the object to be recommended can be a commodity, a video or a picture, and the like, the interaction behavior can be collection, forwarding or purchasing, and the like, the current graphical user interface can be different scenes in the recommendation system, including but not limited to shopping cart pages, comment pages, homepages or repurchase pages, and the like, the current graphical user interface information is used for representing a specific scene of the user at the current moment, and the utilization rate of information resources is improved by obtaining the user history behavior information authorized by the user and the current graphical user interface information, the processing efficiency of a large language model is improved, and the recommendation accuracy of the user is enhanced.

For example, the historical behavior information of the user a can be obtained, wherein the historical behavior information comprises 10 commodities which are clicked by the user a on a first page displayed by the terminal equipment, and the commodities comprise mother and infant products, electronic products, food and the like; the current graphical user interface information characterizes that the current graphical user interface may be an order page of the commodity.

And 202, performing text generation processing on the user history behavior information and the current graphical user interface information through the trained large language model to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface.

Specifically, the user historical behavior information and the current graphical user interface information may be input to a trained large language model for processing, the large language model may be a deep learning model trained by using a large amount of text data, natural language text may be generated or meaning of the language text may be understood, the large language model may be generally based on neural network technology, training and optimizing may be performed by using large-scale training data, and various natural language tasks may be processed, including but not limited to text classification, question and answer, or dialogue, etc., the large language model may also generate semantically relevant output according to the input text, by learning a large amount of text data, the large language model may obtain understanding of aspects including but not limited to language structure, grammar, or semantics, etc., and may be applied in various natural language processing tasks, such as machine translation, text abstract, question and answer systems, etc., where no limitation is made, the large language model may include generating pre-training converter 3 (GENERATIVE PRE-trained Transformer, gpt-3), deep bi-directional pre-training transducer (Bidirectional Encoder Representation from Transformers, BERT) and XLNet, etc., where no limitation is made.

In some embodiments, the user history behavior information and the current graphical user interface information which are characterized by natural language can be input into a trained large language model, text generation is performed according to a target and constraint conditions of a generated text, the constraint conditions can be used for generating style, language, length and the like of the generated text, and further, in the process of generating the text, the large language model can gradually generate the next word or character according to language rules and context information in combination with an initial context, continuous text output is formed, wherein the initial context can be the user history behavior information and the current graphical user interface information, a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector are obtained through text generation processing, the user preference vector corresponding to the current graphical user interface can be used for representing preference feature information of the user on the current graphical user interface, and the current graphical user interface feature vector can be used for representing fusion features of the current graphical user interface, including but not limited to activity information of the current graphical user interface, product types provided by the current graphical user interface, or skip to the current graphical user interface, and the like, so that the application range of the text is improved, the recommendation context is improved, the application range is expanded, the recommendation range is improved, the application range is enlarged, and the application range is generated by a specific text-oriented context is enlarged, and the application range is improved.

For example, the historical behavior information of the user a can include 10 commodities which are clicked by the user a on a first page displayed by the terminal device, wherein the commodities include mother and infant products, electronic products, food and the like; the current graphical user interface information characterizes that the current graphical user interface can be a list page, the list page is input into a large language model, text generation processing is carried out, two 16-dimensional dense vectors are obtained, the dense vectors comprise user preference vectors corresponding to the current graphical user interface and current graphical user interface feature vectors, wherein the user preference vectors corresponding to the current graphical user interface can be used for characterizing that a user A clicks and browses electronic products in the list page preference displayed by terminal equipment, and the current graphical user interface feature vectors can be used for characterizing activity information of the list page, provided product types, or skip conditions of the user and the like, and are not limited herein.

And 203, inputting the user preference vector corresponding to the current graphical user interface and the feature vector of the current graphical user interface into the deep interest network model to obtain the recommendation sequence of the target object.

Specifically, the user preference vector and the current graphical user interface feature vector corresponding to the current graphical user interface can be input to a deep interest network model for processing, wherein the deep interest network model can be a deep learning model for a recommendation system and can be used for acquiring and predicting interest information of a user and providing more accurate and personalized content for the user in the recommendation system, the deep interest network model can acquire long-term and short-term interest information of the user and preference degrees of the user for different interests so as to generate interest representations of the user, the representation mode can acquire interest changes and dynamics of the user, so that understanding degree of demands and intentions of the user is improved, the deep interest network model can be applied to various scenes including but not limited to video recommendation, music recommendation, news recommendation and the like, and accuracy of the recommendation system is improved by combining the deep interest network model with recommendation algorithms.

In some embodiments, the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector can be received through the input layer, the input vector is processed through the middle layer, the click prediction probability of each target object is output through the output layer, the click prediction probability can be used for representing the possibility that the corresponding target object is clicked, further, the target objects can be ordered according to the click prediction probability of each target object to obtain a target object recommendation sequence, the target objects can be objects to be recommended in the current graphical user interface, including but not limited to videos, pictures, commodities or the like, so that the object list recommendation model can learn the internal rule of data through the deep learning method of the deep interest network model, prediction accuracy is improved, the capability of processing large-scale data is improved, and the adaptability and the expandability of the object list recommendation model are improved.

For example, a user preference vector and a current graphical user interface feature vector corresponding to a current graphical user interface of the user a may be input into the deep interest network model for processing, a predicted click probability of a plurality of electronic products of the user a with the preference of the next page displayed by the terminal device may be obtained through middle layer processing, and the corresponding plurality of electronic products may be ranked according to the predicted click probability, so as to obtain a recommendation sequence of the plurality of electronic products of the preference of the user a with the preference of the next page displayed by the terminal device.

And 204, determining a target object recommendation list displayed for the user according to the target object recommendation sequence, and sending the target object recommendation list to the terminal equipment for displaying the target object recommendation list on the current graphical user interface of the terminal equipment.

Specifically, the target object recommendation list displayed for the user can be determined according to the target object recommendation sequence, the target object recommendation list can be a personalized recommendation list generated by a recommendation system and used for displaying contents or services for the user according to the behavior and interest information of the user, the recommendation system is not limited in this place, the recommendation system can process historical behavior data and interest preference information of the user, calculate recommendation scores of target objects through corresponding algorithms and models, order the target objects according to the scores, the generation and display process of the target object recommendation list can be adjusted and optimized according to actual requirements and scenes, the target object recommendation list is further sent to terminal equipment of the user, the display mode can be text, pictures, videos or the like through the current graphic user interface, the display mode is not limited in this place, user experience is improved, resource utilization efficiency is improved, safety is improved, and recommendation accuracy is improved by combining with the current graphic user interface.

For example, the target object recommendation sequence of the order page displayed by the user a at the terminal device may be a product a, a product B, a product C, and a product D, the products are arranged into a list according to the existing sequence, the list is sent to the terminal device currently used by the user a, and the user a may view the target object recommendation list at the order page displayed by the terminal device.

According to the technical scheme provided by the embodiment of the disclosure, through acquiring the interaction history data of the user and the information of the current graphical user interface displayed on the terminal equipment, based on the trained large language model, the user history behavior information and the current graphical user interface information are analyzed, the user history behavior information and the current graphical user interface information are converted into the user preference vector and the current graphical user interface feature vector through a text generation technology, the two vectors are input into the deep interest network model, the deep interest network model can recommend proper target objects for the user according to the preference of the user and the features of the current graphical user interface, and are arranged according to the sequence of the recommended objects, a target object recommendation list is generated according to the sequence of the recommended objects, and the target object recommendation list is sent to the terminal equipment for displaying on the current graphical user interface, so that the recommendation accuracy of the target object list is improved, the cross-scene recommendation contact is enhanced, the richness of the multi-behavior features of the user is improved, and the uniformity of the user interest characterization is improved.

In some embodiments, prior to obtaining the user historical behavior information and the current graphical user interface information, further comprising: acquiring a training behavior information set, wherein the training behavior information set comprises training behavior information corresponding to a plurality of different training graphical user interfaces and labels corresponding to the training behavior information, and the labels corresponding to the training behavior information are used for representing to-be-recommended objects of a user on a preset training graphical user interface; fusion processing is carried out on the user training behavior information corresponding to a plurality of different training graphical user interfaces through generating a pre-training transducer model, so that training behavior fusion information corresponding to the user is obtained; according to a preset prompt instruction, inputting training behavior fusion information corresponding to a user into a large language model for information classification processing to obtain user training behavior classification information corresponding to a plurality of different training graphical user interfaces; determining the loss corresponding to the large language model according to the user training behavior classification information corresponding to the plurality of different training graphical user interfaces and the labels corresponding to the training behavior information; and updating parameters in the large language model according to the loss corresponding to the large language model in a cyclic iteration mode.

Specifically, a training behavior information set may be obtained, where the training behavior information set may include training behavior information corresponding to a plurality of different training graphical user interfaces and labels corresponding to the training behavior information, where the training behavior information corresponding to the plurality of different training graphical user interfaces may be historical behavior information corresponding to the user through different training graphical user interfaces, the training graphical user interfaces may be graphical user interfaces that may be included in the training process of the large language model, may include a menu page, a shopping cart page, or a homepage, etc., where the historical behavior information corresponding to the training graphical user interfaces may represent historical interaction behaviors of the user in the recommendation system and an object to be recommended, including but not limited to collection, forwarding, or purchase, etc., and the labels corresponding to the training behavior information may be used to represent the object to be recommended by the user in a preset training graphical user interface, including but not limited to video, pictures, products, etc.

In some embodiments, the user training behavior information corresponding to the plurality of different training graphical user interfaces may be fused into training behavior fusion information corresponding to the user by generating a pre-training transducer model, where the generating pre-training transducer model may be an internet-based, data-enabled deep learning model for text generation that may be fine-tuned to perform a variety of natural language processing tasks, such as text generation, code generation, video generation, text question-and-answer, image generation, paper authoring, movie authoring, or scientific experimental design, and the like, without limitation. The model can adopt a multi-layer transducer structure for predicting probability distribution of the next word, natural language text can be generated through language modes learned in a large text corpus, and training behavior fusion information corresponding to a user can be used for representing fusion information of historical behavior information of the user under a plurality of different training graphical user interfaces.

In some embodiments, information classification may be performed on training behavior fusion information corresponding to a user through a large language model according to a preset prompting instruction, so as to obtain user behavior classification information corresponding to a plurality of different training graphical user interfaces, where the preset prompting instruction may be used to instruct the large language model to perform classification processing on input information, and the user behavior classification information corresponding to the plurality of different training graphical user interfaces may include collection, purchase, forwarding, and the like, and is not limited herein.

In some embodiments, the loss corresponding to the large language model may be determined according to the user training behavior classification information corresponding to the multiple different training graphical user interfaces and the labels corresponding to the training behavior information, where the greater the difference between the user training behavior classification information corresponding to the multiple different training graphical user interfaces and the label results corresponding to the training behavior information is, the greater the loss is, and the loss corresponding to the large language model may be cross entropy loss, mean square error loss, average absolute error loss, or the like, which is not limited herein, and further, by a loop iteration mode, parameters in the large language model are updated according to the loss corresponding to the large language model.

According to the technical scheme provided by the embodiment of the disclosure, the training behavior information set can comprise training behavior information corresponding to a plurality of different training graphical user interfaces and labels corresponding to the training behavior information through acquiring the training behavior information set, the user training behavior information corresponding to the plurality of different training graphical user interfaces can be fused into training behavior fusion information corresponding to the user through generating the pre-training transducer model, further, according to a preset prompting instruction, the training behavior fusion information corresponding to the user can be classified through the large language model, the user behavior classification information corresponding to the plurality of different training graphical user interfaces is obtained, the loss corresponding to the large language model is determined according to the user training behavior classification information corresponding to the plurality of different training graphical user interfaces and labels corresponding to the training behavior information, the larger the difference between the label results corresponding to the user training behavior classification information corresponding to the plurality of different training graphical user interfaces is, further, the parameters in the large language model are updated according to the loss corresponding to the large language model through a cyclic iteration mode, the accuracy of the large language model is improved, the processing efficiency is improved, the multi-scene behavior information is improved, and the language information integration capability of the large language model is improved.

In some embodiments, performing text generation processing on the historical behavior information of the user and the current graphical user interface information through the trained large language model to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface, including: word embedding processing is respectively carried out on the user historical behavior information and the current graphical user interface information, so that an embedded vector corresponding to the user historical behavior information and an embedded vector corresponding to the current graphical user interface information are obtained; respectively carrying out self-attention mechanism processing on the embedded vector corresponding to the user history behavior information and the embedded vector corresponding to the current graphical user interface information to obtain a weighted vector corresponding to the user history behavior information and a weighted vector corresponding to the current graphical user interface information; carrying out nonlinear transformation on the weighted vector corresponding to the user history behavior information and the weighted vector corresponding to the current graphical user interface information to obtain a global integrated vector corresponding to the user; and processing the global integration vector corresponding to the user according to a preset prompting instruction to obtain a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector.

Specifically, feature extraction can be performed on the historical behavior information of the user and the current graphical user interface information through a large language model, word embedding can be achieved through a word embedding mode, word embedding can be a technology of mapping words or phrases from a vocabulary to real number domain vectors, each word or phrase can be mapped to a vector on a real number domain, through a training model, the model can predict the occurrence probability of a word in a given context, the vector representation of the word can be obtained, for example, the vector of a near-meaning word is closer in space, the vector of an anti-meaning word is farther away, some other characteristics of the words can be obtained, including but not limited to gender, complex number, state and/or the like, the embedded vector corresponding to the historical behavior information of the user and the embedded vector corresponding to the current graphical user interface information can be obtained through word embedding processing, the embedded vector corresponding to the historical behavior information of the user can be used for representing the historical behavior information of the user in a recommendation system, the embedded vector corresponding to the current graphical user interface information can be used for representing the feature information of the current graphical user interface, and the feature information of the current graphical user interface can include but is not limited to pages, shopping pages, shopping pages or multiple pages or the like.

In some embodiments, the embedded vector corresponding to the user historical behavior information and the embedded vector corresponding to the current graphical user interface information may be processed according to a self-attention mechanism to obtain the weighted vector corresponding to the user historical behavior information and the weighted vector corresponding to the current graphical user interface information, where the self-attention mechanism may be a special attention mechanism, and when the object list recommendation model is processing the sequence data, a relationship between each element in the sequence and all other elements may be obtained, so as to improve the understanding degree of the object list recommendation model on the context information in the sequence, and further process the sequence data more accurately. The self-attention mechanism can determine the importance of each position by calculating the relevance score between different positions in the sequence, so that the processing efficiency and accuracy are improved, the application range is expanded, the interpretability of the object list recommendation model is improved, the weighting vector corresponding to the user history behavior information can be used for representing preference information of the user history behavior, history behavior mode information of the user, importance degree of the history behavior information on the user, context information in the user history behavior information and the like, the limitation is omitted, and the weighting vector corresponding to the current graphical user interface information can be used for representing the attention degree of the user on interface elements, interaction history, task completion degree, user preference, habit and the like, and the limitation is omitted. Therefore, the user experience and interaction efficiency are improved.

In some embodiments, the weighted vector corresponding to the historical behavior information of the user and the weighted vector corresponding to the current graphical user interface information may be subjected to nonlinear transformation through an activation function to obtain a global integration vector corresponding to the user, where the activation function may be used to provide nonlinear modeling capability of the network, so that the neural network may express and process nonlinear problems, and the types of the global integration vector corresponding to the user obtained through nonlinear transformation include, but are not limited to, sigmoid function, tanh function, reLU function, and the like, and the global integration vector corresponding to the user obtained through nonlinear transformation may be used to characterize the preference of the user on the current graphical user interface, the historical behavior information, the association degree of the user and the object to be recommended, and the like, and is not limited herein, and the global integration vector corresponding to the user is processed according to a preset prompt instruction of a large language model, so as to obtain the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector.

For example, the historical behavior information and the ordering page information of the user a can be respectively processed through a word embedding technology to obtain an embedded vector corresponding to the historical behavior information of the user a and an embedded vector corresponding to the ordering page information, weight distribution is carried out on the embedded vectors through a self-attention mechanism to obtain a weighted vector corresponding to the historical behavior information of the user a and a weighted vector corresponding to the ordering page information, nonlinear transformation is carried out on the obtained two weighted vectors through a Sigmoid function to obtain a global integration vector corresponding to the user a, and the global integration vector corresponding to the user a is processed according to a preset prompt fusion instruction to obtain a preference vector of the user a corresponding to the ordering page and a feature vector of the ordering page.

According to the technical scheme provided by the embodiment of the disclosure, word embedding processing is performed on the user history behavior information and the current graphical user interface information through a word embedding technology to obtain an embedded vector corresponding to the user history behavior information and an embedded vector corresponding to the current graphical user interface information, weight distribution is further performed on the embedded vector corresponding to the user history behavior information and the embedded vector corresponding to the current graphical user interface information through a self-attention mechanism to obtain a weighted vector corresponding to the user history behavior information and a weighted vector corresponding to the current graphical user interface information, nonlinear transformation is performed on the two weighted vectors through an activation function to obtain a global integration vector corresponding to a user, and the global integration vector corresponding to the user is processed through a large language model according to a preset prompt instruction to obtain a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector, so that accuracy of determining user preferences is improved, processing efficiency of an object list recommendation model is improved, and application range of the object list recommendation model is expanded.

In some embodiments, inputting the user preference vector and the current graphical user interface feature vector corresponding to the current graphical user interface into the deep interest network model to obtain the target object recommendation order, including: splicing and fusing the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector to obtain a fusion vector corresponding to the current graphical user interface; performing logistic regression processing on the fusion vector corresponding to the current graphical user interface to obtain the weight value of each target object; and determining the recommendation sequence of the target objects according to the weight value of each target object.

Specifically, the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector can be spliced and fused through the deep interest network model to obtain a fusion vector corresponding to the current graphical user interface, the fusion vector corresponding to the current graphical user interface is used for representing the joint information of the user preference and the object feature to be recommended of the current graphical user interface, the preference degree of the user on a specific product or service can be obtained through analyzing the fusion vector, and then the fusion vector corresponding to the current graphical user interface is subjected to logistic regression processing to obtain the weight value of each target object, and the weight value of each target object can be calculated through a logistic regression function. The weight value can be used for representing the contribution degree of each feature in the fusion vector corresponding to the current graphical user interface to the target object, the fusion vector corresponding to the current graphical user interface can be input through a logistic regression function, the corresponding weight value is output, the recommendation sequence of the target object is further determined according to the weight value of each target object, the target objects are ordered from large to small according to the weight value of each target object, and the target object sequence corresponding to each weight value is obtained, wherein the target object can represent the current graphical user interface, and the target object possibly preferred by the user is to be recommended.

For example, through a deep interest network model, a preference vector of a user A corresponding to a lower page and a feature vector of the lower page displayed by a terminal device can be spliced and fused to obtain a fusion vector corresponding to the lower page, and logistic regression processing is performed on the fusion vector corresponding to the lower page to obtain a weight value 0.1 of a product A, a weight value 0.5 of a product B, a weight value 0.25 of a product C and a weight value 0.15 of a product D, and sorting the product A, the product B, the product C and the product D according to the weight values to obtain products B, C, D and A in sequence.

According to the technical scheme provided by the embodiment of the disclosure, the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector are spliced and fused through the deep interest network model to obtain the fusion vector corresponding to the current graphical user interface, the fusion vector corresponding to the current graphical user interface is subjected to logistic regression processing to obtain the weight value of each target object, and the recommendation sequence of the target objects is determined according to the weight value of each target object, so that the accuracy of object list recommendation is improved, the processing efficiency of object list recommendation is improved, and the fusion capability of multi-graphical user interface information is enhanced.

In some embodiments, before determining the target object recommendation list presented to the user according to the target object recommendation order, the method further comprises: determining a user recommendation text corresponding to a current graphical user interface according to the user history behavior information and the current graphical user interface information of the terminal equipment; and sending a user recommendation text corresponding to the current graphical user interface to the terminal equipment, wherein the user recommendation text corresponding to the current graphical user interface is used for representing the adaptability degree of the target object recommendation list.

Specifically, text generation can be performed through a large language model according to user historical behavior information and current graphical user interface information of terminal equipment to obtain user recommendation texts corresponding to the current graphical user interface, the user recommendation texts corresponding to the current graphical user interface are used for representing the adaptation degree of a target object recommendation list, specific contents can be selection reasons of a target object, recommendation reasons of the target recommendation list or preference prediction results of the user on the current graphical user interface and the like, the limitation is not made here, the user recommendation texts corresponding to the current graphical user interface are sent to terminal equipment used by the user, and the terminal equipment can be a mobile phone, a tablet, a personal computer or the like, and the limitation is not made here.

For example, according to the historical behavior information of the user a and the information of the order page displayed by the terminal device, generating a recommended text corresponding to the order page target object recommendation list displayed by the terminal device, wherein the content of the recommended text can be "displaying a possibly preferred product list to the user a according to prediction", and sending the recommended text to the terminal device of the user a for displaying the order page displayed by the terminal device of the user a.

According to the technical scheme provided by the embodiment of the disclosure, text generation is performed according to the historical behavior information of the user and the current graphical user interface information of the terminal equipment through the large language model, the user recommendation text corresponding to the current graphical user interface is obtained, and the user recommendation text corresponding to the current graphical user interface is sent to the terminal equipment used by the user, so that the matching degree of the current graphical user interface and the user is improved, the accuracy of object list recommendation is improved, and the information richness of object list recommendation is increased.

In some embodiments, fusion processing is performed on user training behavior information corresponding to a training graphical user interface by generating a pre-training transducer model, so as to obtain training behavior fusion information corresponding to a user, including: determining training prompt instructions according to user training behavior information corresponding to a plurality of different training graphical user interfaces; and according to the training prompt instruction, carrying out fusion processing on the training behavior information of the user corresponding to the training graphical user interface to obtain the training behavior fusion information corresponding to the user.

Specifically, a training prompt instruction can be determined by generating a pre-training transducer model according to user training behavior information corresponding to a plurality of different training graphical user interfaces, wherein the training prompt instruction is used for indicating the generating pre-training transducer model to fuse the user training behavior information corresponding to the plurality of different training graphical user interfaces, and further, according to the training prompt instruction, the generating pre-training transducer model fuses the user training behavior information corresponding to the plurality of different training graphical user interfaces, wherein the plurality of different training graphical user interfaces can comprise shopping cart pages, repurchase pages, order pages, personal homepages and the like, and the user training behavior fusion information corresponding to the user can be used for representing fusion information of historical behavior information of the user under the plurality of different training graphical user interfaces.

For example, the training prompt instruction can be determined to be expressed as 'fusion input information' by natural language according to the shopping cart page, the repurchase page, the order page and the user training behavior information corresponding to the personal homepage displayed by the terminal equipment by generating a pre-training transducer model, and fusion processing is performed on the shopping cart page, the repurchase page, the order page and the user training behavior information corresponding to the personal homepage displayed by the terminal equipment according to the training prompt instruction, so as to obtain training behavior fusion information corresponding to the user A.

According to the technical scheme provided by the embodiment of the disclosure, the training prompt instruction is determined according to the user training behavior information corresponding to the different training graphical user interfaces by the generation type pre-training transducer model, and then the user training behavior information corresponding to the different training graphical user interfaces is fused by the generation type pre-training transducer model according to the training prompt instruction, so that the understanding degree of the generation type pre-training transducer model on the input information is improved, and the accuracy of object list recommendation is improved.

In some embodiments, performing logistic regression processing on the fusion vector corresponding to the current graphical user interface to obtain a weight value of each target object, including: carrying out classification mapping processing on the fusion vector corresponding to the current graphical user interface to obtain category information corresponding to each target object; and carrying out linear regression processing on the fusion vector corresponding to the current graphical user interface and the category information corresponding to each target object to obtain the weight value of each target object.

Specifically, the fusion vector corresponding to the current graphical user interface can be subjected to classification mapping processing through a logistic regression function to obtain category information corresponding to each target object, the category information corresponding to the target object can be used for representing the category of the target object, including but not limited to the type of a product, the type of a service or the theme of content, the fusion vector corresponding to the current graphical user interface and the category information corresponding to each target object are subjected to linear regression processing to obtain the weight value of each target object, and the weight value of each target object can be 0.1, 0.25, 0.5 or the like, and is not limited herein, so that the interpretability of the object list recommendation system is improved, the flexibility of the object list recommendation model is enhanced, and the high efficiency of the object list recommendation model is enhanced.

For example, the fusion vector corresponding to the next page displayed by the terminal device can be subjected to classification mapping processing through a logistic regression function to obtain category information corresponding to the product A, the product B, the product C and the product D respectively, and further, the category information corresponding to the product A, the product B, the product C and the product D respectively is subjected to linear regression processing to obtain a weight value of 0.1 of the product A, a weight value of 0.5 of the product B, a weight value of 0.25 of the product C and a weight value of 0.15 of the product D.

According to the technical scheme provided by the embodiment of the disclosure, the fusion vector corresponding to the current graphical user interface is subjected to classification mapping processing through the logistic regression function to obtain the category information corresponding to each target object, and the linear regression processing is performed on the fusion vector corresponding to the current graphical user interface and the category information corresponding to each target object to obtain the weight value of each target object, so that the interpretability of the object list recommendation system is improved, the flexibility of the object list recommendation model is enhanced, and the high efficiency of the object list recommendation model is enhanced.

Fig. 3 is a schematic structural diagram of an object list recommendation model provided in an embodiment of the present disclosure. As shown in fig. 3, the structure of the object list recommendation model includes:

The large language model 301 is configured to perform text generation processing on the user history behavior information and the current graphical user interface information through the trained large language model, so as to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface;

The deep interest network model 302 is configured to input a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector into the deep interest network model to obtain a recommendation sequence of the target object.

According to the technical scheme provided by the embodiment of the disclosure, through acquiring the interaction history data of the user and the information of the current graphical user interface displayed on the terminal equipment, based on the large language model 301, the user history behavior information and the current graphical user interface information are analyzed, the user history behavior information and the current graphical user interface information are converted into the user preference vector and the current graphical user interface feature vector through the text generation technology, the two vectors are input into the deep interest network model 302, the deep interest network model can recommend proper target objects for the user according to the preference of the user and the features of the current graphical user interface, and the target object recommendation list is generated according to the sequence of the recommended objects and sent to the terminal equipment for displaying on the current graphical user interface, so that the recommendation accuracy of the target object list is improved, the cross-scene recommendation contact is enhanced, the richness of the multi-behavior features of the user is improved, and the uniformity of the user interest characterization is improved.

Any combination of the above-mentioned optional solutions may be adopted to form an optional embodiment of the present disclosure, which is not described herein in detail.

The following are device embodiments of the present disclosure that may be used to perform method embodiments of the present disclosure. For details not disclosed in the embodiments of the apparatus of the present disclosure, please refer to the embodiments of the method of the present disclosure.

Fig. 4 is a schematic diagram of an object list recommendation apparatus provided in an embodiment of the present disclosure. As shown in fig. 4, the object list recommending apparatus includes:

The acquiring module 401 is configured to acquire user historical behavior information and current graphical user interface information of the terminal device, where the user historical behavior information is used to characterize a historical interaction behavior of an object to be recommended;

The first processing module 402 is configured to perform text generation processing on the user history behavior information and the current graphical user interface information through the trained large language model, so as to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface;

the second processing module 403 is configured to input a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector to the deep interest network model, so as to obtain a recommendation sequence of the target object;

The determining module 404 is configured to determine a target object recommendation list displayed to the user according to the target object recommendation sequence, and send the target object recommendation list to the terminal device, where the target object recommendation list is displayed on a current graphical user interface of the terminal device.

In some embodiments, the object list recommending apparatus is further configured to obtain a training behavior information set, where the training behavior information set includes training behavior information corresponding to a plurality of different training graphical user interfaces and labels corresponding to the training behavior information, and the labels corresponding to the training behavior information are used to characterize an object to be recommended by a user on a preset training graphical user interface; fusing the user training behavior information corresponding to the training graphical user interface by generating a pre-training transducer model to obtain training behavior fusion information corresponding to the user; according to a preset prompt instruction, inputting training behavior fusion information corresponding to a user into a large language model for information classification processing to obtain user training behavior classification information corresponding to a plurality of different training graphical user interfaces; determining the loss corresponding to the large language model according to the user training behavior classification information corresponding to the plurality of different training graphical user interfaces and the labels corresponding to the training behavior information; and updating parameters in the large language model according to the loss corresponding to the large language model in a cyclic iteration mode.

In some embodiments, the first processing module 402 is specifically configured to perform word embedding processing on the user history behavior information and the current graphical user interface information, so as to obtain an embedded vector corresponding to the user history behavior information and an embedded vector corresponding to the current graphical user interface information; respectively carrying out self-attention mechanism processing on the embedded vector corresponding to the user history behavior information and the embedded vector corresponding to the current graphical user interface information to obtain a weighted vector corresponding to the user history behavior information and a weighted vector corresponding to the current graphical user interface information; carrying out nonlinear transformation on the weighted vector corresponding to the user history behavior information and the weighted vector corresponding to the current graphical user interface information to obtain a global integrated vector corresponding to the user; and processing the global integration vector corresponding to the user according to a preset prompting instruction to obtain a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector.

In some embodiments, the second processing module 403 is specifically configured to splice and fuse the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector to obtain a fusion vector corresponding to the current graphical user interface; performing logistic regression processing on the fusion vector corresponding to the current graphical user interface to obtain the weight value of each target object; and determining the recommendation sequence of the target objects according to the weight value of each target object.

In some embodiments, the object list recommending apparatus is further configured to determine, according to the user history behavior information and current graphical user interface information of the terminal device, a user recommendation text corresponding to the current graphical user interface; and sending a user recommendation text corresponding to the current graphical user interface to the terminal equipment, wherein the user recommendation text corresponding to the current graphical user interface is used for representing the adaptability degree of the target object recommendation list.

In some embodiments, fusion processing is performed on user training behavior information corresponding to a training graphical user interface by generating a pre-training transducer model, so that the obtained training behavior fusion information corresponding to the user is specifically used for determining training prompt instructions according to the user training behavior information corresponding to a plurality of different training graphical user interfaces; and according to the training prompt instruction, carrying out fusion processing on the training behavior information of the user corresponding to the training graphical user interface to obtain the training behavior fusion information corresponding to the user.

In some embodiments, performing logistic regression processing on the fusion vector corresponding to the current graphical user interface to obtain a weight value of each target object, wherein the weight value is specifically used for performing classification mapping processing on the fusion vector corresponding to the current graphical user interface to obtain class information corresponding to each target object; and carrying out linear regression processing on the fusion vector corresponding to the current graphical user interface and the category information corresponding to each target object to obtain the weight value of each target object.

It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic of each process, and should not constitute any limitation on the implementation process of the embodiments of the disclosure.

Fig. 5 is a schematic diagram of an electronic device 5 provided by an embodiment of the present disclosure. As shown in fig. 5, the electronic apparatus 5 of this embodiment includes: a processor 501, a memory 502 and a computer program 503 stored in the memory 502 and executable on the processor 501. The steps of the various method embodiments described above are implemented by processor 501 when executing computer program 503. Or the processor 501 when executing the computer program 503 performs the functions of the modules/units in the above-described apparatus embodiments.

The electronic device 5 may be a desktop computer, a notebook computer, a palm computer, a cloud server, or the like. The electronic device 5 may include, but is not limited to, a processor 501 and a memory 502. It will be appreciated by those skilled in the art that fig. 5 is merely an example of the electronic device 5 and is not limiting of the electronic device 5 and may include more or fewer components than shown, or different components.

The Processor 501 may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processors, digital signal processors (DIGITAL SIGNAL Processor, DSP), application SPECIFIC INTEGRATED Circuit (ASIC), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc.

The memory 502 may be an internal storage unit of the electronic device 5, for example, a hard disk or a memory of the electronic device 5. The memory 502 may also be an external storage device of the electronic device 5, such as a plug-in hard disk, a smart memory card (SMART MEDIA CARD, SMC), a Secure Digital (SD) card, a flash memory card (FLASH CARD) or the like, which are provided on the electronic device 5. Memory 502 may also include both internal storage units and external storage devices of electronic device 5. The memory 502 is used to store computer programs and other programs and data required by the electronic device.

It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions. The functional units and modules in the embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit, where the integrated units may be implemented in a form of hardware or a form of a software functional unit.

The integrated modules/units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a readable storage medium (e.g., a computer readable storage medium). Based on such understanding, the present disclosure may implement all or part of the flow of the method of the above-described embodiments, or may be implemented by a computer program to instruct related hardware, and the computer program may be stored in a computer readable storage medium, where the computer program, when executed by a processor, may implement the steps of the method embodiments described above. The computer program may comprise computer program code, which may be in source code form, object code form, executable file or in some intermediate form, etc. The computer readable storage medium may include: any entity or device capable of carrying computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth.

The above embodiments are merely for illustrating the technical solution of the present disclosure, and are not limiting thereof; although the present disclosure has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the disclosure, and are intended to be included in the scope of the present disclosure.

Claims

1. An object list recommendation method, comprising:

Acquiring user historical behavior information and current graphical user interface information of terminal equipment, wherein the user historical behavior information is used for representing historical interaction behaviors of an object to be recommended;

Performing text generation processing on the user history behavior information and the current graphical user interface information through the trained large language model to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface;

Inputting the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector into a deep interest network model to obtain a target object recommendation sequence;

and determining a target object recommendation list displayed for the user according to the target object recommendation sequence, and sending the target object recommendation list to the terminal equipment, wherein the target object recommendation list is used for displaying the target object recommendation list on a current graphical user interface of the terminal equipment.

2. The object list recommendation method of claim 1, further comprising, prior to said obtaining user historical behavior information and current graphical user interface information:

acquiring a training behavior information set, wherein the training behavior information set comprises training behavior information corresponding to a plurality of different training graphical user interfaces and labels corresponding to the training behavior information, and the labels corresponding to the training behavior information are used for representing to-be-recommended objects of the user on a preset training graphical user interface;

Fusing the user training behavior information corresponding to the plurality of different training graphical user interfaces by generating a pre-training transducer model to obtain training behavior fusion information corresponding to the user;

Inputting training behavior fusion information corresponding to the user into the large language model for information classification processing according to a preset prompt instruction, and obtaining user training behavior classification information corresponding to the plurality of different training graphical user interfaces;

determining the loss corresponding to the large language model according to the user training behavior classification information corresponding to the plurality of different training graphical user interfaces and the labels corresponding to the training behavior information;

And updating parameters in the large language model according to the loss corresponding to the large language model in a cyclic iteration mode.

3. The method for recommending an object list according to claim 1, wherein the performing text generation processing on the user history behavior information and the current gui information by using the trained large language model to obtain a user preference vector and a current gui feature vector corresponding to the current gui includes:

Word embedding processing is respectively carried out on the user historical behavior information and the current graphical user interface information, so that an embedded vector corresponding to the user historical behavior information and an embedded vector corresponding to the current graphical user interface information are obtained;

Respectively carrying out self-attention mechanism processing on the embedded vector corresponding to the user historical behavior information and the embedded vector corresponding to the current graphical user interface information to obtain a weighted vector corresponding to the user historical behavior information and a weighted vector corresponding to the current graphical user interface information;

performing nonlinear transformation on the weighted vector corresponding to the user historical behavior information and the weighted vector corresponding to the current graphical user interface information to obtain a global integration vector corresponding to the user;

And processing the global integration vector corresponding to the user according to a preset prompting instruction to obtain a user preference vector corresponding to the current graphical user interface and a current graphical user interface feature vector.

4. The method for recommending object list according to claim 1, wherein the step of inputting the user preference vector corresponding to the current gui and the current gui feature vector into a deep interest network model to obtain the target object recommendation order comprises:

Splicing and fusing the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector to obtain a fusion vector corresponding to the current graphical user interface;

Performing logistic regression processing on the fusion vector corresponding to the current graphical user interface to obtain weight values of all target objects;

And determining the recommendation sequence of the target objects according to the weight value of each target object.

5. The object list recommendation method according to claim 1, further comprising, before said determining a target object recommendation list to be presented to said user according to said target object recommendation order:

Determining a user recommendation text corresponding to a current graphical user interface according to the user history behavior information and the current graphical user interface information of the terminal equipment;

and sending a user recommendation text corresponding to the current graphical user interface to the terminal equipment, wherein the user recommendation text corresponding to the current graphical user interface is used for representing the adaptability degree of the target object recommendation list.

6. The method for recommending an object list according to claim 2, wherein the fusing processing is performed on the training behavior information of the user corresponding to the training graphical user interface by generating a pre-training transducer model, so as to obtain the training behavior fusion information corresponding to the user, and the method comprises the following steps:

Determining training prompt instructions according to user training behavior information corresponding to the plurality of different training graphical user interfaces;

and according to the training prompt instruction, carrying out fusion processing on the training behavior information of the user corresponding to the training graphical user interface to obtain the training behavior fusion information corresponding to the user.

7. The method for recommending object list according to claim 4, wherein the performing logistic regression on the fusion vector corresponding to the current gui to obtain the weight value of each target object comprises:

carrying out classification mapping processing on the fusion vector corresponding to the current graphical user interface to obtain category information corresponding to each target object;

and carrying out linear regression processing on the fusion vector corresponding to the current graphical user interface and the category information corresponding to each target object to obtain the weight value of each target object.

8. An object list recommendation apparatus, comprising:

The acquisition module is used for acquiring user historical behavior information and current graphical user interface information of the terminal equipment, wherein the user historical behavior information is used for representing historical interaction behaviors of the object to be recommended;

the first processing module is used for generating texts of the user history behavior information and the current graphical user interface information through the trained large language model to obtain a user preference vector and a current graphical user interface feature vector corresponding to the current graphical user interface;

The second processing module is used for inputting the user preference vector corresponding to the current graphical user interface and the current graphical user interface feature vector into the deep interest network model to obtain a target object recommendation sequence;

the determining module is used for determining a target object recommendation list displayed for the user according to the target object recommendation sequence, sending the target object recommendation list to the terminal equipment and displaying the target object recommendation list on a current graphical user interface of the terminal equipment.

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 the processor implements the steps of the method according to any of claims 1 to 7 when the computer program is executed.

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