Information recommendation method and device, electronic equipment and computer readable storage medium

Abstract

The application provides an information recommendation method and device, electronic equipment and a computer readable storage medium. The method comprises: obtaining information features of information to be recommended, object features of a target object, and respective style features of a plurality of display styles; performing fusion processing on the information features of the information, the object features of the target object, and the style features of each display style respectively to obtain fusion features corresponding to each display style; determining recommendation parameters of each display style based on the fusion features corresponding to each display style; wherein the recommendation parameters represent the preference degree of the target object when the information is displayed through the display style; taking the display style with the highest recommendation parameter as the target display style of the information, and responding to a recommendation request for the target object based on the information and the target display style. Through the application, the display style of the recommended information can be enriched, thereby meeting the personalized needs of different objects.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, and particularly relates to an information recommendation method and device, electronic equipment and computer readable storage medium. BACKGROUND

[0002] The popularity of the Internet, especially mobile Internet, makes information recommendation an important way to obtain information. The recommended information stream is calculated offline or online when the user requests based on the user's interests (such as food, games, movies, sports, etc.) and behaviors (such as likes, collections, comments, and forwards) to provide multiple matching information to the user for browsing.

[0003] However, in the related art, the recommended information is usually displayed in a fixed display style, that is, the same display style is used to display the same recommended information on different user terminal devices. That is, the display style of the recommended information in the related art is relatively single and cannot meet the personalized needs of different users. SUMMARY

[0004] The embodiments of the present application provide an information recommendation method, device, electronic equipment, computer readable storage medium and computer program product, which can enrich the display style of the recommended information and meet the personalized needs of different objects.

[0005] The technical solutions of the embodiments of the present application are as follows:

[0006] The embodiments of the present application provide an information recommendation method, comprising:

[0007] Obtaining the information features of the information to be recommended, the object features of the target object, and the respective style features of a plurality of display styles;

[0008] Respectively fusing the information features of the information, the object features of the target object, and the style features of each of the display styles to obtain the fusion features corresponding to each of the display styles;

[0009] Based on the fusion features corresponding to each of the display styles, determining the recommendation parameters of each of the display styles; wherein the recommendation parameters represent the preference degree of the target object when the information is displayed through the display style;

[0010] Taking the display style with the highest recommendation parameter as the target display style of the information, and responding to the recommendation request for the target object based on the information and the target display style.

[0011] The embodiments of the present application provide an information recommendation device, comprising:

[0012] The acquisition module is configured to acquire information features of information to be recommended, object features of a target object, and respective style features of a plurality of display styles;

[0013] The fusion module is configured to perform fusion processing on the information features of the information, the object features of the target object, and the style features of each of the display styles, respectively, to obtain fusion features corresponding to each of the display styles.

[0014] The determination module is configured to determine a recommendation parameter for each of the display styles based on the fusion features corresponding to each of the display styles, wherein the recommendation parameter represents a preference degree of the target object when the information is displayed through the display style.

[0015] The determination module is further configured to take the display style with the highest recommendation parameter as a target display style of the information, and respond to a recommendation request for the target object based on the information and the target display style.

[0016] In the above solution, the acquisition module is further configured to perform embedding processing on first category features included in the information to be recommended to obtain vector representations of the first category features, perform normalization processing on first numerical features included in the information to obtain vector representations of the first numerical features, perform splicing processing on the vector representations of the first category features and the vector representations of the first numerical features to obtain the information features, perform embedding processing on second category features included in the target object to obtain vector representations of the second category features, perform normalization processing on second numerical features included in the target object to obtain vector representations of the second numerical features, and perform splicing processing on the vector representations of the second category features and the vector representations of the second numerical features to obtain the object features of the target object. For each of the plurality of display styles, the following processing is performed: perform embedding processing on third category features included in the display style to obtain vector representations of the third category features, perform normalization processing on third numerical features included in the display style to obtain vector representations of the third numerical features, and perform splicing processing on the vector representations of the third category features and the vector representations of the third numerical features to obtain the style features of the display style.

[0017] In the above solution, the recommendation parameter is determined by calling a neural network model. The determination module is further configured to perform the following processing on the fusion features of each of the display styles: perform multiple times of nonlinear transformation processing on the fusion features of the display style through a plurality of activation layers included in the neural network model to obtain a corresponding recommendation parameter when the information is displayed through the display style.

[0018] In the above scheme, each object type corresponds to a pre-trained neural network model; the device further includes an object type identification module configured to perform object type identification processing on the target object to obtain an object type to which the target object belongs; and the determination module is further configured to obtain a target neural network model corresponding to the object type to which the target object belongs, and perform cascaded multiple times of nonlinear transformation processing on the fusion feature of the display style based on multiple activation layers included in the target neural network model, to obtain the recommended parameter corresponding to displaying the information by using the display style.

[0019] In the above scheme, each information type corresponds to a pre-trained neural network model; the device further includes an information type identification module configured to perform information type identification processing on the information to be recommended to obtain an information type to which the information belongs; and the determination module is further configured to obtain a target neural network model corresponding to the information type to which the information belongs, and perform cascaded multiple times of nonlinear transformation processing on the fusion feature of the display style based on multiple activation layers included in the target neural network model, to obtain the recommended parameter corresponding to displaying the information by using the display style.

[0020] In the above scheme, each display style corresponds to a pre-trained neural network model; and the determination module is further configured to obtain a target neural network model corresponding to the display style, and perform cascaded multiple times of nonlinear transformation processing on the fusion feature of the display style based on multiple activation layers included in the target neural network model, to obtain the recommended parameter corresponding to displaying the information by using the display style.

[0021] In the above scheme, the device further includes a training module configured to train the neural network model by: obtaining information features of sample information, object features of sample objects, and respective style features of multiple sample display styles; performing fusion processing on the information features of the sample information and the object features of the sample objects respectively with the style features of each of the sample display styles, to obtain a fusion feature corresponding to each of the sample display styles; performing forward propagation based on the fusion feature corresponding to each of the sample display styles, to obtain a prediction result; determining a difference between the prediction result and labeled data of the sample information, and performing backward propagation based on the difference, and updating parameters of the neural network model layer by layer in the backward propagation process; wherein the labeled data includes feedback data corresponding to the sample objects when the sample information is displayed by using each of the sample display styles.

[0022] In the scheme, the obtaining module is further configured to obtain an identifier of the sample information, associate the identifier of the sample information with an information feature library, and obtain an information feature of the sample information; obtain an identifier of the sample object, associate the identifier of the sample object with an object feature library, and obtain an object feature of the sample object; and for each sample display style in the plurality of sample display styles, perform the following processing: obtain an identifier of the sample display style, associate the identifier of the sample display style with a display style feature library, and obtain a display style feature of the sample display style.

[0023] In the scheme, the apparatus further includes a receiving module configured to receive a recommendation request for the target object sent by a terminal device; and the determining module is further configured to, when the target object is a cold start object, obtain a plurality of candidate information ranked in front from an information library, and take the plurality of candidate information ranked in front as the information to be recommended; and when the target object is a non-cold start object, obtain historical behavior data of the target object, determine portrait information of the target object according to the historical behavior data, obtain a plurality of candidate information matching the portrait information of the target object from an information library, and take the plurality of candidate information matching the target object as the information to be recommended.

[0024] In the scheme, the determining module is further configured to, when receiving a recommendation request for the target object sent by a terminal device, perform at least one of the following processing: obtain a candidate object with a similarity between an object feature of the target object and an object feature of the candidate object greater than a similarity threshold value from an object library, obtain information preferred by the candidate object, and take the preferred information as the information to be recommended; obtain historical preference information of the target object, determine an information feature of the historical preference information, obtain a plurality of candidate information with a similarity between the information feature of the historical preference information and an information feature of each candidate information greater than a similarity threshold value from an information library, and take the plurality of candidate information as the information to be recommended.

[0025] Embodiments of the present application provide an information recommendation method, comprising:

[0026] receiving at least one information to be recommended for a target object, and a target display style of each of the information;

[0027] displaying the at least one information in a human-computer interaction interface based on the target display style of each of the information;

[0028] The target display style of each of the information is selected from a plurality of candidate display styles based on an information feature of the information, an object feature of the target object, and a display style feature of each of the plurality of candidate display styles.

[0029] The embodiment of the present application provides an information recommendation device, comprising:

[0030] A receiving module is used for receiving at least one information to be recommended for a target object and a target display style of each information;

[0031] A display module is used for displaying the at least one information in a man-machine interactive interface based on the target display style of each information;

[0032] The target display style of each information is selected from a plurality of candidate display styles based on information features of the information, object features of the target object and style features of the plurality of candidate display styles.

[0033] The embodiment of the present application provides an electronic device, comprising:

[0034] A memory is used for storing executable instructions;

[0035] A processor is used for executing the executable instructions stored in the memory, and the information recommendation method provided by the embodiment of the present application is realized.

[0036] The embodiment of the present application provides a computer readable storage medium, which stores executable instructions, and is used for causing a processor to execute the information recommendation method provided by the embodiment of the present application.

[0037] The embodiment of the present application provides a computer program product, comprising a computer program or instructions, and the computer program or instructions are executed by a processor to realize the information recommendation method provided by the embodiment of the present application.

[0038] The embodiment of the present application has the following beneficial effects:

[0039] Based on the information features of the information to be recommended, the object features of the target object and the style features of the plurality of candidate display styles, the target display style of the information to be recommended is selected from the plurality of candidate display styles, so that the information to be recommended is displayed through the target display style. Therefore, for the same information, the target display styles corresponding to different objects are different, the display styles of the recommended information are enriched, and the personalized needs of different objects are met. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 FIG. 1 is an architecture schematic diagram of an information recommendation system 100 provided by the embodiment of the present application;

[0041] Figure 2 FIG. 2 is a structure schematic diagram of a server 200 provided by the embodiment of the present application;

[0042] Figure 3 is a structural schematic diagram of a terminal device 400 provided by an embodiment of the present application;

[0043] Figure 4 is a flowchart of an information recommendation method provided by an embodiment of the present application;

[0044] Figure 5 is a flowchart of an information recommendation method provided by an embodiment of the present application;

[0045] Figure 6 is an advertisement style schematic diagram provided by the related art;

[0046] Figure 7 is an advertisement style schematic diagram provided by the related art;

[0047] Figure 8 is an item display schematic diagram of different types provided by the related art;

[0048] Figure 9 is an advertisement style schematic diagram provided by an embodiment of the present application;

[0049] Figure 10 is an advertisement style schematic diagram provided by an embodiment of the present application;

[0050] Figure 11 is an architecture schematic diagram of an advertisement recommendation system provided by an embodiment of the present application;

[0051] Figure 12 is a model construction and training schematic diagram provided by an embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings, and the described embodiments should not be regarded as limiting the present application, and all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0053] In the following description, "some embodiments" are described, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0054] In the following description, the terms "first\second\third" are merely to distinguish similar objects, and do not represent a specific order of the objects, and it can be understood that "first\second\third" can be interchanged in a specific order or sequence as allowed, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to be limiting of this application.

[0056] Before the embodiments of the present application are explained in further detail, the nomenclature and terminology involved in the embodiments of the present application are explained, which are applicable to the following explanations.

[0057] 1) Embedding: essentially a kind of mapping, i.e. from semantic space to vector space, while trying to keep the relationship of the original samples in the semantic space as much as possible in the vector space, for example, two words with semantic proximity are also close in the vector space.

[0058] 2) Neural Network (NN): a network organized by many logistics units according to different levels, and the output variable of each layer is used as the input variable of the next layer. Neural network has a wide application prospect in system identification, pattern recognition, intelligent control and other fields.

[0059] 3) Numerical Feature: i.e. the feature that can be actually measured, such as the exposure of the advertisement, the click volume, the access times of the commodity, the number of adding to the shopping cart, the final sales volume, etc.

[0060] 4) Categorical Feature: refers to the feature that only takes values within a limited option, such as gender (male, female), grade (e.g. divided into A, B, C), etc.

[0061] 5) Cold Start Object: i.e. the object with no historical behavior data or very little historical behavior data, such as a new user.

[0062] 6) Cascade: in the process of nonlinear transformation, cascade refers to the dependency relationship between the nonlinear transformation of the previous layer and the nonlinear transformation of the next layer, i.e. the nonlinear transformation of the previous layer depends on the nonlinear transformation of the next layer.

[0063] The embodiment of the present application provides an information recommendation method and device, electronic equipment, computer readable storage medium and computer program product, which can enrich the display style of recommended information, and further meet the personalized needs of different objects. The following describes the exemplary application of the electronic equipment provided by the embodiment of the present application. The electronic equipment provided by the embodiment of the present application can be implemented as notebook computers, tablet computers, desktop computers, set-top boxes, mobile devices (for example, mobile phones, portable music players, personal digital assistants, dedicated messaging devices, portable game devices), vehicle-mounted terminals and various types of terminal devices. It can also be implemented as a server, or cooperatively implemented by a terminal device and a server. In the following, the information recommendation method provided by the embodiment of the present application will be described by taking the terminal device and the server cooperatively implementing the information recommendation method as an example.

[0064] Referring to Figure 1 , Figure 1 is the architecture diagram of the information recommendation system 100 provided by the embodiment of the present application. In order to support an application of a rich display style of recommended information, the terminal device 400 is connected to the server 200 through the network 300. The network 300 can be a wide area network or a local area network, or a combination of the two.

[0065] As Figure 1 indicated, the terminal device 400 runs a client 410. The client 410 can be various types of clients, such as instant messaging clients, social media clients, information reading clients, e-commerce shopping clients, browsers, etc. In response to a refresh (for example, the client 410 receives a user-triggered pull-to-refresh operation, or receives a click operation on a refresh button displayed in a human-computer interaction interface, or the client 410 is started, for example, the user clicks the icon corresponding to the client 410 displayed in the human-computer interaction interface of the terminal device 400) of the client 410, the server 200 sends a recommendation request for a target object to the server 200 through the network 300.

[0066] After the server 200 receives the recommendation request sent by the client 410, the server 200 obtains candidate information matched with the object feature of the target object from the database 500 as information to be recommended; then the server 200 obtains the information feature of the information to be recommended, the object feature of the target object (for example, the user feature of the user A) and the style feature of each of the plurality of candidate display styles, and respectively fuses the information feature of the information and the object feature of the target object with the style feature of each display style to obtain the fusion feature corresponding to each display style; subsequently, the server 200 determines the recommendation parameter (for example, the click rate, the score, etc.) of each display style based on the fusion feature corresponding to each display style, wherein the recommendation parameter represents the preference degree of the target object when the information is displayed through the display style; finally, the server 200 takes the display style with the highest recommendation parameter as the target display style of the information, and sends at least one information to be recommended for the target object and the target display style of each information to the terminal device 400, so that the terminal device 400 can call the human-computer interaction interface of the client 410, and display at least one information in the human-computer interaction interface based on the target display style of each information, so that the display style of the recommended information is enriched, the personalized needs of different users are met, and the visual experience of the user is improved.

[0067] In some embodiments, the information recommendation method provided by the embodiments of the present application can also be implemented by the terminal device alone, for example, the terminal device 400 shown in FIG. 4. Figure 1 For example, the terminal device 400 receives at least one information to be recommended for the target object and the target display style of each information; then the terminal device 400 calls the human-computer interaction interface of the client 410, and displays at least one information in the human-computer interaction interface based on the target display style of each information.

[0068] In some embodiments, the embodiments of the present application can be implemented by means of cloud technology. Cloud technology refers to a kind of hosting technology that unifies a series of resources such as hardware, software and network in a wide area network or a local area network to realize the calculation, storage, processing and sharing of data.

[0069] Cloud technology is a general term of network technology, information technology, integration technology, management platform technology and application technology applied based on cloud computing business model, can form a resource pool, and can be used on demand, flexibly and conveniently. Cloud computing technology will become an important support. The background service of the technical network system needs a large amount of calculation and storage resources.

[0070] For example, Figure 1The server 200 shown in FIG. 1 can be a standalone physical server, a server cluster composed of multiple physical servers, or a distributed system, and can also be a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, and basic cloud computing services such as big data and artificial intelligence platforms. The terminal device 400 can be a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, and the like, but is not limited thereto. The terminal device 400 and the server 200 can be connected directly or indirectly through wired or wireless communication, and the present application is not limited in this embodiment.

[0071] In some other embodiments, the terminal device or the server can also implement the information recommendation method provided in the present application by running a computer program. For example, the computer program can be a native program or a software module in the operating system, for example, an information recommendation module embedded in any program; or a native application program (APP), that is, a program that needs to be installed in the operating system to run, for example, a native application program (APP) such as WeChat, QQ, and Alipay. Figure 1 The above computer program can be any form of application program, module, or plug-in.

[0072] The structure of the server 200 shown in FIG. 1 will be described below. Figure 1 The structure of the server 200 shown in FIG. 1 will be described below. Figure 2 , Figure 2 is a structural schematic diagram of the server 200 provided in the present application, Figure 2 The server 200 shown in FIG. 1 includes at least one processor 210, a memory 240, and at least one network interface 220. The various components in the server 200 are coupled together through a bus system 230. It can be understood that the bus system 230 is used to realize the connection and communication between the components. The bus system 230 includes a data bus, a power supply bus, a control bus, and a state signal bus. However, for the purpose of clear illustration, all the buses are marked as the bus system 230 in Figure 2 .

[0073] The processor 210 can be an integrated circuit chip with signal processing capability, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0074] The memory 240 can be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard drives, optical drives, and the like. The memory 240 optionally includes one or more storage devices remotely located from the processor 210.

[0075] The memory 240 includes volatile memory or nonvolatile memory, and can also include both volatile and nonvolatile memory. Nonvolatile memory can be read only memory (ROM), volatile memory can be random access memory (RAM). The memory 240 described in the embodiments of the present application is intended to include any suitable type of memory.

[0076] In some embodiments, the memory 240 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, which are exemplarily illustrated below.

[0077] The operating system 241 includes system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, and the like, for implementing various basic services and processing hardware-based tasks;

[0078] The network communication module 242 is used to reach other computing devices via one or more (wired or wireless) network interfaces 220, exemplary network interfaces 220 including Bluetooth, wireless compatibility certification (WiFi), and universal serial bus (USB), and the like;

[0079] In some embodiments, the information recommendation apparatus provided by the embodiments of the present application can be realized in a software manner, Figure 2 The information recommendation apparatus 243 stored in the memory 240 is shown, which can be software in the form of programs and plug-ins, including the following software modules: an acquisition module 2431, a fusion module 2432, a determination module 2433, an object type identification module 2434, an information type identification module 2435, a training module 2436, and a receiving module 2437. These modules are logical, and thus can be combined or further split according to the implemented functions. It should be noted that, in order to facilitate expression, all the above modules are shown at one time, but it should not be considered that the information recommendation apparatus 243 excludes the implementation that only includes the acquisition module 2431, the fusion module 2432, and the determination module 2433, the functions of each module will be described below. Figure 2

[0080] The structure of the terminal device 400 shown in Figure 1 will be described below. Referring to​Figure 3 , Figure 3 is a structural schematic diagram of the terminal device 400 provided in an embodiment of the present application. As shown in Figure 3 , the terminal device 400 includes a network interface 430, a user interface 440, a system bus 450, a memory 460, and a processor 420. The user interface 440 includes one or more output devices 441 that enable presentation of media content, including one or more speakers and / or one or more visual display screens. The user interface 440 also includes one or more input devices 442, including user interface components that facilitate user input, such as a keyboard, a mouse, a microphone, a touch screen display, a camera, other input buttons and controls. The memory 460 includes an operating system 461, a network communication module 462, a presentation module 463 for enabling display of information (e.g., a user interface for operating peripheral devices and displaying content and information) via one or more output devices 441 associated with the user interface 440 (e.g., a display screen, speakers, etc.), an input processing module 464 for detecting and interpreting one or more user inputs or interactions from one or more of the input devices 442, and an information recommendation device 465. Further, the information recommendation device 465 stored in the memory 460 includes a receiving module 4651 and a display module 4652, which are logical and thus can be combined or further split according to the implemented functions, and the functions of the modules will be described below.

[0081] In some embodiments, the information recommendation device provided in the embodiments of the present application can be implemented in a hardware manner. As an example, the information recommendation device provided in the embodiments of the present application can be a processor in the form of a hardware decoding processor, which is programmed to perform the information recommendation method provided in the embodiments of the present application. For example, the processor in the form of a hardware decoding processor can use one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs), or other electronic elements.

[0082] The information recommendation method provided by the embodiments of the present application will be described in detail from the perspective of the interaction between the terminal device and the server. It should be noted that the steps performed by the terminal device are performed by various forms of computer programs running on the terminal device, and are not limited to the client 410 running on the terminal device 400, for example, the operating system 461, software modules, scripts, and applets described above, and therefore, the following examples of the client should not be regarded as a limitation of the embodiments of the present application.

[0083] Referring to Figure 4 , Figure 4 is a flowchart of the information recommendation method provided by the embodiments of the present application, which will be described in combination with the steps shown in Figure 4 .

[0084] In step 101, the server obtains the information features of the information to be recommended, the object features of the target object, and the respective style features of the plurality of display styles.

[0085] In some embodiments, referring to Figure 5 , Figure 5 is a flowchart of the information recommendation method provided by the embodiments of the present application, as shown in Figure 5 , before performing step 101 shown in Figure 4 , steps 107 and 108 shown in Figure 5 may also be performed, which will be described in combination with the steps shown in Figure 5 .

[0086] In step 107, the terminal device sends a recommendation request for the target object to the server.

[0087] In some embodiments, the timing of sending the recommendation request can be that the client running in the terminal device is refreshed, for example, when the client is started (when the client is started, the corresponding information needs to be obtained to be displayed to the user, therefore, the start of the client can be regarded as a special refresh), for example, when the user first opens the client installed on the terminal device or the client is switched back from the background to the foreground, the recommendation request for the target object is automatically sent to the server; of course, the timing of sending the recommendation request can also be that the client receives a refresh operation triggered by the user, for example, when the client receives a pull-to-refresh operation triggered by the user or a click operation on the refresh button displayed in the human-computer interaction interface, the recommendation request for the target object is sent to the server.

[0088] In step 108, the server obtains the information to be recommended.

[0089] In some embodiments, the server can obtain the information to be recommended in the following manner: when the target object is a cold start object, obtaining a plurality of candidate information ranked in the front (i.e., popular) from the information library, and taking the plurality of information ranked in the front as the information to be recommended; when the target object is a non-cold start object, obtaining historical behavior data of the target object; determining portrait information of the target object according to the historical behavior data; obtaining a plurality of candidate information matching the portrait information of the target object from the information library, and taking the plurality of candidate information matching as the information to be recommended.

[0090] For example, for a cold start object (e.g., a new user or a user with less rich historical behavior data), since the historical behavior data (e.g., browsing, clicking, ordering or downloading) related to the cold start object (e.g., user A) cannot be obtained, the portrait information of user A cannot be determined, and therefore a plurality of candidate information ranked in the front in the information library can be taken as the information to be recommended for user A. For example, in an e-commerce scenario, items with sales exceeding a sales threshold within a certain time period can be taken as the information to be recommended for user A; or items with scores greater than a score threshold in evaluations generated by a plurality of old users can be taken as the information to be recommended for user A.

[0091] For example, for a non-cold start object (e.g., an old user or a user with rich historical behavior data), the historical behavior data (e.g., clicking, ordering, purchasing, payment, collection, refund, evaluation, etc. of user B in the past month) of the non-cold start object (e.g., user B) can be obtained first, then the portrait information (i.e., the label of user B, e.g., user B is a sports enthusiast) of user B can be determined according to the historical behavior data, and subsequently a plurality of candidate information (e.g., basketball related information) matching the portrait information of user B can be obtained from the information library, and the plurality of candidate information matching can be taken as the information to be recommended for user B.

[0092] In other embodiments, the server can also obtain the information to be recommended in the following manner: performing at least one of the following processing: obtaining a candidate object from the object library with a similarity between the object features of the target object and the candidate object greater than a similarity threshold, obtaining information preferred by the candidate object, and taking the information preferred by the candidate object as the information to be recommended; obtaining historical preference information of the target object, determining information features of the historical preference information, obtaining a plurality of candidate information from the information library with a similarity between the information features of the historical preference information and the plurality of candidate information greater than a similarity threshold, and taking the plurality of candidate information as the information to be recommended.

[0093] For example, when a target object (e.g., user A) needs personalized recommendation, other users similar to user A can be found first, and information preferred by the other users but unknown to user A (i.e., information not exposed to the terminal device of user A) can be taken as information to be recommended to user A. The specific process is as follows: first, the k nearest neighbors of user A are calculated using a selected similarity calculation method (e.g., cosine similarity, Jaccard similarity, etc.), and the k nearest neighbors are the users closest to user A, i.e., the k users most similar to user A; then, after selecting the k users, the predicted interest of information not purchased or known by user A (i.e., information not exposed to the terminal device of user A) in the k users is calculated; and then, the N information with the highest predicted interest is selected and taken as information to be recommended to user A.

[0094] For example, when a target object (e.g., user A) needs personalized recommendation, the historical preference information of user A (e.g., items ordered by user A in the past week) can be obtained first, and information similar to the historical preference information in the information library can be taken as information to be recommended to user A. The specific process is as follows: first, the similarity between the information features of the historical preference information of user A and the information features corresponding to a plurality of information stored in the information library is calculated using a selected similarity calculation method (e.g., cosine similarity, Jaccard similarity, etc.); then, a plurality of candidate information with a similarity greater than a similarity threshold is selected from the plurality of information, and the plurality of candidate information is taken as information to be recommended to user A.

[0095] It should be noted that, in actual applications, the server can also obtain information to be recommended in the following manner: for example, one of the biggest differences between a mobile device and a PC is that the location of the mobile device changes frequently. Different geographic locations reflect different user scenarios, and therefore, when a target object needs personalized recommendation, a corresponding recall strategy can be triggered according to the real-time geographic location, work location, residence location, etc., of the user. According to the granularity of the geographic location, the recall strategy can be divided into nearby recall, current business district recall, current city recall, etc.

[0096] For example, when the server receives a recommendation request for a target object sent by a terminal device, the current location of the target object can be obtained, and a plurality of candidate information adapted to the current location of the target object can be obtained from the information library, and the plurality of candidate information is taken as information to be recommended to the target object. For example, when the server detects that the target object is currently located in city A, the information of stores, food, hotels, etc., in city A can be obtained from the information library as information to be recommended to the target object.

[0097] In some embodiments, after recalling a plurality of candidate information from the information library based on the portrait information of the target object or based on collaborative filtering, since the number of the recalled candidate information is large (e.g., can reach the order of ten thousand), after recalling the plurality of candidate information, the following processing can also be performed: uniformly scoring and ranking the plurality of recalled candidate information (e.g., the plurality of recalled candidate information can be ranked using a click rate estimation model, wherein the click rate estimation model is trained based on the user features of the sample users and the information features of the sample information), and selecting the top N information in the ranking result as the information to be recommended, so that the number of the information to be recommended can be reduced, and accurate personalized recommendation can be realized.

[0098] In some embodiments, after recalling a plurality of candidate information from the information library based on the portrait information of the target object or based on collaborative filtering, since the number of the recalled candidate information is large (e.g., can reach the order of ten thousand), after recalling the plurality of candidate information, the following processing can also be performed: uniformly scoring and ranking the plurality of recalled candidate information (e.g., the plurality of recalled candidate information can be ranked using a click rate estimation model, wherein the click rate estimation model is trained based on the user features of the sample users and the information features of the sample information), and selecting the top N information in the ranking result as the information to be recommended, so that the number of the information to be recommended can be reduced, and accurate personalized recommendation can be realized.

[0099] For example, taking information A to be recommended as an example, the category feature of information A can be the information type to which information A belongs (for example, when information A is a computer, the information type to which information A belongs is a digital product), and for the category feature of information A, an embedding manner can be used to obtain the vector representation of the category feature of information A. For example, natural number coding or one-hot coding can be used to convert the category feature of information A into a corresponding vector representation. The natural number coding refers to assigning a number to each category. For example, assuming that the information type is divided into “digital products, fresh food, clothing, books”, a corresponding number is assigned to each category. For example, “1” represents “digital products”, “2” represents “fresh food”, “3” represents “clothing”, and “4” represents “books”. One-hot coding, also known as one-bit effective coding, mainly uses an N-bit state register to encode N categories. Each category has an independent register bit, and only one bit is effective at any time. For example, assuming that the information type is divided into “digital products, fresh food, clothing, books”, after conversion to one-hot vector, (1, 0, 0, 0) represents “digital products”, (0, 1, 0, 0) represents “fresh food”, (0, 0, 1, 0) represents “clothing”, and (0, 0, 0, 1) represents “books”. Compared with natural number coding, using one-hot coding can significantly increase the dimension of the feature space.

[0100] For example, taking information A to be recommended as an example, the category feature of information A can be the information type to which information A belongs (for example, when information A is a computer, the information type to which information A belongs is a digital product), and for the category feature of information A, an embedding manner can be used to obtain the vector representation of the category feature of information A. For example, natural number coding or one-hot coding can be used to convert the category feature of information A into a corresponding vector representation. The natural number coding refers to assigning a number to each category. For example, assuming that the information type is divided into “digital products, fresh food, clothing, books”, a corresponding number is assigned to each category. For example, “1” represents “digital products”, “2” represents “fresh food”, “3” represents “clothing”, and “4” represents “books”. One-hot coding, also known as one-bit effective coding, mainly uses an N-bit state register to encode N categories. Each category has an independent register bit, and only one bit is effective at any time. For example, assuming that the information type is divided into “digital products, fresh food, clothing, books”, after conversion to one-hot vector, (1, 0, 0, 0) represents “digital products”, (0, 1, 0, 0) represents “fresh food”, (0, 0, 1, 0) represents “clothing”, and (0, 0, 0, 1) represents “books”. Compared with natural number coding, using one-hot coding can significantly increase the dimension of the feature space.

[0101] For example, taking the target object as user B, the category features of user B can be the gender, marital status, hometown, or favorite movie type of user B, and the like. For the category features of user B, an embedding manner (for example, a one-hot encoding manner or a natural number encoding manner) can be used to convert the category features of user B into corresponding vector representations. In addition, the numerical features of user B can be the number of orders, the number of comments, the number of collections, and the like placed by user B in the past week, and the like. For the numerical features of user B, a normalization processing (for example, a linear function normalization processing or a zero-mean normalization processing) can be used to convert the numerical features of user B into corresponding vector representations. After obtaining the vector representations of the category features of user B and the vector representations of the numerical features of user B, the vector representations of the category features of user B and the vector representations of the numerical features of user B can be spliced to obtain the user features of user B.

[0102] For example, taking the display style as display style C, the category features of display style C can be the style type to which display style C belongs (for example, the style type of the display style can be divided into: a picture type, that is, focusing on picture performance and reducing the script to one line display; a text type, that is, focusing on the content introduction of the displayed information, and ignoring the picture performance), and the like. For the category features of display style C, an embedding manner (for example, a one-hot encoding manner or a natural number encoding manner) can be used to convert the category features of display style C into corresponding vector representations. In addition, the numerical features of display style C can be the number of clicks, the number of orders, and the like of the user when displaying information through display style C, and the like. For the numerical features of display style C, a normalization processing (for example, a linear function normalization processing or a zero-mean normalization processing) can be used to obtain the vector representations corresponding to the numerical features of display style C. After obtaining the vector representations of the category features of display style C and the vector representations of the numerical features of display style C, the vector representations of the category features of display style C and the vector representations of the numerical features of display style C can be spliced to obtain the style features of display style C.

[0103] It should be noted that in actual application, for category features, a word set model (for example, an N-Gram sequence) or a bag-of-words model can also be used to convert them into corresponding vector representations; for numerical features, in addition to normalization processing, binarization, bucketing, feature intersection, etc. can also be used for processing, wherein bucketing refers to dividing numerical variables into a bucket and assigning a bucket number, common bucketing methods include fixed-width bucketing, for fixed-width bucketing, the value range of each bucket is fixed, if the size of each bucket is the same, it is also called uniform bucketing, for example, the age of a user is divided into 0-9 years old, 10-19 years old, etc.; feature intersection can represent the interaction between numerical features, for example, two numerical features can be added, subtracted, multiplied, and divided, etc., useful intersection combinations can be selected through feature selection methods (for example, statistical tests or model feature importance), so as to enrich numerical features.

[0104] Continuing to refer to Figure 4 In step 102, the server fuses the information features of the information, the object features of the target object, and the style features of each display style respectively to obtain the fusion features corresponding to each display style.

[0105] In some embodiments, after obtaining the information features of the information to be recommended, the object features of the target object, and the respective style features of the plurality of candidate display styles, the server can fuse the information features of the information, the object features of the target object, and the style features of each display style respectively (for example, perform addition processing) to obtain the fusion features corresponding to each display style.

[0106] For example, taking the information to be recommended as information A, after obtaining the information features M A of information A, the object features (i.e., the user features U A of the target object (for example, user A), and the respective style features of the plurality of candidate display styles (for example, three candidate display styles, namely display style A, display style B, and display style C, and their corresponding style features are S1, S2, and S3, respectively), the server can fuse the information features M A of information A, the user features U A of user A, and the style features of each display style respectively to obtain the fusion features corresponding to each display style, for example, for display style A, its corresponding fusion features are: M A + U A + S1; for display style B, its corresponding fusion features are: M A + U A + S2; and for display style C, its corresponding fusion features are: M A + U A + S3.

[0107] In step 103, the server determines the recommendation parameter through each display style based on the fusion feature corresponding to each display style.

[0108] Here, the recommendation parameter represents the preference degree of the target object when the information is displayed through the display style, for example, the recommendation parameter can be the click rate, the score, etc.

[0109] In some embodiments, the recommendation parameter can be determined by calling a neural network model, and then the server can implement the above-mentioned step 103 in the following manner: for the fusion feature of each display style, perform the following processing: perform multiple times of nonlinear transformation processing on the fusion feature of the display style through multiple activation layers included in the neural network model to obtain the corresponding recommendation parameter when the information is displayed through the display style.

[0110] For example, taking display style A as an example, after obtaining the fusion feature of display style A, the fusion feature of display style A can be subjected to 3 times of nonlinear transformation processing through multiple activation layers (for example, 3 activation layers) included in the neural network model (for example, the nonlinear transformation result output by the first activation layer is input into the second activation layer, so that the second activation layer performs nonlinear transformation processing on the output of the first activation layer, and then the nonlinear transformation result output by the second activation layer is input into the third activation layer, so that the third activation layer performs nonlinear transformation processing on the output of the second activation layer) to obtain the corresponding recommendation parameter when the information is displayed through display style A, for example, the click rate of the target object when the information is displayed through display style A.

[0111] It should be noted that in actual application, the activation function (Activation Function) included in the activation layer can be a Sigmoid function, a linear rectification (Relu, Rectified Linear Unit) function, or a Tanh function, etc., wherein the Sigmoid function is a commonly used S-shaped function in biology, also known as S-shaped growth curve, in information science, due to its single-increasing and inverse function single-increasing characteristics, the Sigmoid function is often used as a threshold function of the neural network model, for mapping variables between 0 and 1; the Relu function is a commonly used activation function in neural network model, usually refers to a nonlinear function represented by a ramp function and its variants; the Tanh function is one of the hyperbolic functions, Tanh() is the hyperbolic tangent derived from the basic hyperbolic functions hyperbolic sine and hyperbolic cosine.

[0112] In addition, it also needs to be explained that the neural network model in the embodiments of the present application can be a deep neural network model, a convolutional neural network model, a long short-term memory network model, etc. The embodiments of the present application do not specifically limit the type of neural network model.

[0113] In some embodiments, each object type can correspond to a pre-trained neural network model, and then the server can implement the above-mentioned multiple times of nonlinear transformation processing of the fusion features of the display style by the multiple activation layers included in the neural network model to obtain the corresponding recommendation parameters when the information is displayed by the display style in the following manner: performing object type identification processing on the target object to obtain the object type to which the target object belongs; obtaining a target neural network model corresponding to the object type to which the target object belongs, and performing multiple times of nonlinear transformation processing of the fusion features of the display style by the multiple activation layers included in the target neural network model to obtain the corresponding recommendation parameters when the information is displayed by the display style.

[0114] For example, a corresponding neural network model can be trained for each object type. For example, assuming that the object types can be divided into: price-sensitive type, brand-preference type and active type, then for the three object types, three corresponding neural network models can be pre-trained, for example, neural network model 1 corresponds to price-sensitive users, neural network model 2 corresponds to brand-preference users, and neural network model 3 corresponds to active users. Subsequently, for a target object (for example, user A), the type identification processing can be performed on user A first (for example, the historical behavior data of user A can be obtained, and the historical behavior data of user A is input into a pre-trained user type identification model to obtain the user type to which user A belongs). Assuming that user A is identified as a price-sensitive user, then neural network model 1 can be called to determine the corresponding recommendation parameters of each display style. In this way, by training a corresponding neural network model for each object type, on the one hand, the training amount of the model is reduced, and on the other hand, the accuracy of the predicted recommendation parameters is also improved.

[0115] In other embodiments, each information type can correspond to a pre-trained neural network model, and then the server can implement the above-mentioned multiple times of nonlinear transformation processing of the fusion features of the display style by the multiple activation layers included in the neural network model to obtain the corresponding recommendation parameters when the information is displayed by the display style in the following manner: performing information type identification processing on the information to be recommended to obtain the information type to which the information belongs; obtaining a target neural network model corresponding to the information type to which the information belongs, and performing multiple times of nonlinear transformation processing of the fusion features of the display style by the multiple activation layers included in the target neural network model to obtain the corresponding recommendation parameters when the information is displayed by the display style.

[0116] For example, one corresponding neural network model can be trained for each information type. For example, it is assumed that the information types can be classified as: digital products, clothing, fresh food, and books. Four corresponding neural network models can be pre-trained for the four information types. For example, neural network model 1 corresponds to digital products, neural network model 2 corresponds to clothing, neural network model 3 corresponds to fresh food, and neural network model 4 corresponds to books. Subsequently, for the information to be recommended (for example, information A), the information type identification processing can be performed on the information A first (for example, the text and pictures included in the information A can be obtained, and the obtained text and pictures can be input into the pre-trained information type identification model to obtain the information type to which the information A belongs). It is assumed that the information type to which the information A belongs is identified as digital products. The neural network model 1 can be called to determine the recommendation parameters corresponding to the plurality of candidate display styles configured for the information A. In this way, by training one corresponding neural network model for each information type, when the information type changes, the entire neural network model does not need to be retrained, thereby reducing the calculation scale.

[0117] In some embodiments, each display style can correspond to a pre-trained neural network model. The server can obtain the target neural network model corresponding to the display style, and perform the cascaded multiple times of nonlinear transformation processing on the fusion features of the display style based on the multiple activation layers included in the target neural network model, to obtain the recommendation parameters corresponding to the display of the information.

[0118] For example, one corresponding neural network model can be trained for each display style in the plurality of candidate display styles. For example, it is assumed that the plurality of candidate display styles include: display style A, display style B, and display style C. Three corresponding neural network models can be pre-trained for the three display styles. For example, neural network model 1 corresponds to display style A, neural network model 2 corresponds to display style B, and neural network model 3 corresponds to display style C. Subsequently, when determining the recommendation parameters corresponding to the display of the information by the display style A, the neural network model 1 can be called to determine the recommendation parameters. When determining the recommendation parameters corresponding to the display of the information by the display style B, the neural network model 2 can be called to determine the recommendation parameters. When determining the recommendation parameters corresponding to the display of the information by the display style C, the neural network model 3 can be called to determine the recommendation parameters. In this way, by training one corresponding neural network model for each display style, the efficiency of determining the recommendation parameters can be improved. When a new display style is added, only the new display style needs to be trained, and the entire neural network model does not need to be retrained, thereby reducing the calculation scale.

[0119] The training process of the neural network model is described below.

[0120] In some embodiments, the neural network model described above can be trained in the following manner: obtaining information features of sample information, object features of sample objects, and respective style features of a plurality of sample display styles; performing fusion processing on the information features of the sample information, the object features of the sample objects, and the style features of each sample display style, respectively, to obtain fusion features corresponding to each sample display style; performing forward propagation based on the fusion features corresponding to each sample display style to obtain a prediction result; determining the difference between the prediction result and the labeled data of the sample information, and performing backward propagation based on the difference, and updating the parameters of the neural network model layer by layer in the backward propagation process (for example, the gradient descent method can be used to update the parameters); wherein the labeled data includes feedback data (such as whether to click, whether to be interested, etc.) corresponding to the sample objects when the sample information is displayed through each sample display style.

[0121] For example, the information features of the sample information, the object features of the sample objects, and the respective style features of the plurality of sample display styles can be obtained in the following manner: obtaining the identity (Identity Document) of the sample information, associating the identity of the sample information with the information feature library to obtain the information features of the sample information (for example, the information feature library stores a plurality of information identities and information features corresponding to each information identity, so that the information identity of the sample information can be queried in the information feature library, and the information features corresponding to the queried identity are taken as the information features of the sample information); obtaining the identity of the sample object (for example, the social network account used by the sample object), associating the identity of the sample object with the object feature library to obtain the object features of the sample object (for example, the object feature library stores a plurality of object identities and object features corresponding to each object identity, so that the identity of the sample object can be queried in the object feature library, and the object features corresponding to the queried identity are taken as the object features of the sample object); for each sample display style in the plurality of sample display styles, the following processing is performed: obtaining the identity of the sample display style, associating the identity of the sample display style with the style feature library to obtain the style features of the sample display style (for example, the style feature library stores a plurality of display style identities and style features corresponding to each display style identity, so that for each sample display style in the plurality of sample display styles, such as sample display style A, the identity of the sample display style A can be queried in the style feature library, and the style features corresponding to the queried identity are taken as the style features of the sample display style A).

[0122] In step 104, the server takes the display style of the highest recommended parameter as the target display style of the information.

[0123] In some embodiments, after obtaining the recommended parameters corresponding to the plurality of candidate display styles respectively, the server can take the display style with the highest recommended parameter as the target display style of the information.

[0124] For example, taking information A to be recommended as an example, assuming that the plurality of candidate display styles for information A are display style A, display style B, and display style C, after obtaining the recommended parameters corresponding to display style A, display style B, and display style C respectively (for example, the click rate corresponding to display style A is 80%, the click rate corresponding to display style B is 50%, and the click rate corresponding to display style C is 40%), the display style with the highest recommended parameter (i.e., display style A) can be taken as the target display style of information A.

[0125] In step 105, the server sends at least one information to be recommended for the target object and the target display style of each information to the terminal device.

[0126] In some embodiments, after determining the target display style corresponding to each information to be recommended for the target object, the server can send at least one information to be recommended for the target object and the target display style of each information to the terminal device.

[0127] For example, assuming that the information to be recommended for the target object includes information 1 and information 2, and the server determines that the target display style corresponding to information 1 is display style A and the target display style corresponding to information 2 is display style B, the server can send information 1 and information 2 and display style A and display style B to the terminal device, wherein information 1 is bound to display style A (i.e., informing the terminal device that display style A is used to display information 1), and information 2 is bound to display style B (i.e., informing the terminal device that display style B is used to display information 2).

[0128] In step 106, the terminal device displays at least one information in the human-computer interaction interface based on the target display style of each information.

[0129] In some embodiments, after receiving at least one information to be recommended for the target object and the target display style of each information sent by the server, the terminal device can call the human-computer interaction interface of the client running on the terminal device, and display at least one information in the human-computer interaction interface based on the target display style of each information, for example, display information 1 through display style A and display information 2 through display style B.

[0130] The embodiment of the application selects a target display style of the information to be recommended from the plurality of candidate display styles based on the information features of the information to be recommended, the object features of the target object, and the style features of the plurality of candidate display styles, so as to display the information to be recommended through the target display style. Thus, since the target display style of the information to be recommended is related to the object features of the target object, the target display style corresponding to different users is different for the same information, the display style of the recommended information is enriched, and the personalized needs of different users are met.

[0131] In the following, an example application of the embodiment of the application in an actual application scenario is described by taking an advertisement recommendation scenario as an example.

[0132] The reasons why the user clicks the advertisement are various, for example, the user trusts the brand a or wants to purchase the goods of the category b, and the like. Therefore, different advertisement contents are pushed to each user according to the label of the user, which can greatly hit the needs of the user and also can improve the advertisement consumption (i.e., the amount of money paid by the advertiser to the traffic platform when the advertiser puts the advertisement) of the traffic platform (also referred to as a traffic channel).

[0133] However, the advertisement style (corresponding to the display style described above) as a medium for carrying the advertisement creative does not achieve the thousand faces of a thousand people, and the preference of the user for different advertisement styles is not recorded in the related art, so that the advertisement materials of different industries / brands are usually displayed in several fixed styles provided by the traffic platform and are distributed to different people in the same way, that is, the advertisement styles seen by different users for the same advertisement are the same.

[0134] For example, referring to Figure 6 , Figure 6 is a schematic diagram of an advertisement style provided by the related art, as shown in Figure 6 , the advertiser can preview the advertisement styles provided by the traffic platform when putting the advertisement, but cannot select.

[0135] For example, referring to Figure 7 , Figure 7 is a schematic diagram of different advertisement styles provided by the related art, as shown in Figure 7 , the traffic platform can regularly perform A / B testing, and determine which advertisement style has better effect according to the total results of the A / B testing, for example, the traffic platform can simultaneously provide Figure 7The four different types of advertisement styles 1.1, 1.2, 1.3 and 1.n are shown, and the four different types of advertisement styles are put to users to test which advertisement style is the best (for example, to test which advertisement style corresponds to the highest user click rate), and then the best advertisement style is used to distribute to all users, however, the above method does not take into account the differences between different users, in addition, there is no call to action or additional information (such as timeliness, geographic information, etc.).

[0136] For example, referring to Figure 8 , Figure 8 The different types of item display diagrams provided by the related art are shown, for example Figure 8 Based on user behavior and commercial industry, different user interfaces (UI, User Interface) can be displayed, for example, when a user searches for different items, different information flows are automatically used, for example, when a user searches for a mobile phone, configuration information that the user will focus on can be displayed in a list flow, and when a user searches for clothing (for example, a men's T-shirt), a picture flow can be used, focusing on picture performance and reducing the title to one line, however, the above scheme still uses a fixed style to display the same item.

[0137] Therefore, the embodiments of the present application provide an information recommendation method, based on individual preferences for advertisement styles, by analyzing the relationship between advertisement styles, advertisement materials and users, to achieve a thousand faces for a thousand people, thereby achieving more refined distribution of advertisements, and also improving the consumption of traffic platform advertisements.

[0138] The information recommendation method provided by the embodiments of the present application can be applied to all advertisements, for example, when advertisement a2 is sent to user c1 in advertisement style b1, the click-through rate (CTR, Click-Through-Rate) of user c1 is the best, which may be due to the fact that user c1 agrees with the brand of advertisement a2, and the advertisement style b1 highlights the advertiser, so the advertisement style is displayed to user c1; similarly, assuming that user c2 is a price-sensitive user, then when the advertisement a1 has a special price in the advertisement word, the advertisement style b3 that can highlight the text can be used for user c2.

[0139] For example, referring to Figure 9 , Figure 9 The multiple candidate advertisement style diagrams provided by the embodiments of the present application are shown, for example Figure 9As shown, for the same advertising material (e.g. lipstick advertisement), the traffic platform can provide multiple different types of advertising styles in advance, for example, including advertising style 1 to advertising style 5, so that when it is analyzed that the user is a price-sensitive user, the advertising style 5 highlighting the text can be used as the target advertising style of the lipstick advertisement, for example, the ID of the advertising style 5 and the advertising material can be sent to the terminal device of the user, so that the terminal device displays the advertising material through the advertising style 5.

[0140] For example, referring to Figure 10 , Figure 10 is a plurality of candidate advertising style diagrams provided by the embodiment of the application, as Figure 10 shown, for the same advertising material (e.g. car advertisement), the traffic platform can provide multiple candidate advertising styles in advance, for example, including advertising style 1 to advertising style 7, when the user c1 is viewing the advertisement, it is analyzed that the brand a cannot promote the decision of the user c1, then when displaying the advertisement of the brand a, some advertising styles highlighting the brand logo can be excluded.

[0141] For example, referring to Figure 11 , Figure 11 is a schematic diagram of the architecture of the advertising recommendation system provided by the embodiment of the application, as Figure 11 shown, when a user (e.g. user A) is browsing information, the point terminal can initiate an advertising request to the background, the background requests to recall the advertisement from the advertising delivery system (e.g. Guangdiantong) after receiving the advertising request sent by the point terminal, so that Guangdiantong returns the corresponding advertising material to the background; then the background can call the advertising style matching algorithm to match the advertising material returned by Guangdiantong and the multiple selectable advertising styles, calculate the optimal advertising style for the user A for the advertisement, and finally, the background returns the advertising material and the optimal advertising style to the point terminal, so that the point terminal displays the advertising material through the optimal advertising style in the human-computer interaction interface.

[0142] The advertising style matching algorithm shown in Figure 11 will be described in detail below.

[0143] In some embodiments, the advertising style matching algorithm mainly matches the advertising material with multiple candidate advertising styles to find the advertising style most liked by the current user for display on the terminal, thereby improving the advertising click rate and user experience, and ultimately improving the consumption of the advertisement.

[0144] The advertising style matching algorithm provided by the embodiment of the application mainly includes four parts: data preparation, feature processing, model construction and training, and online service of the model, which will be described below.

[0145] (1) Data preparation

[0146] In the data preparation stage, the historical advertisement exposure and click data of a user in the past week is mainly pulled from the highlight platform, wherein the data fields can include user identification (such as a social network account of the user, such as QQ), advertisement ID, advertisement style ID, and whether the advertisement is clicked.

[0147] (2) Feature processing

[0148] In the feature processing stage, the user QQ in the advertisement exposure and click data is associated with the user feature library of the highlight to obtain user-related features; the advertisement ID is associated with the advertisement feature library of the highlight to obtain advertisement-related features; the advertisement style ID is associated with the advertisement style feature library to obtain advertisement style-related features, and finally the three types of feature data are associated to obtain the feature data table as shown in Table 1.

[0149] Table 1 User feature data table

[0150]

[0151] (3) Model construction and training

[0152] In some embodiments, referring to Figure 12 , Figure 12 is a model construction and training schematic diagram provided by the embodiments of the present application, as shown in Figure 12 , the category features (Categorical Feature) in the advertisement style are subjected to embedding (Embedding) processing to obtain vector representation of each category feature, and the numerical features (Numerical Feature) in the advertisement style are subjected to normalization (Normalization) processing to obtain vector representation of each numerical feature; then the two types of vector features are spliced to obtain the style feature Ui_vector of the advertisement style. Similarly, the user feature User_vector of the user and the advertisement feature Ad_vector of the advertisement can also be obtained.

[0153] For example, for the user feature User_vector, the following formula can be used to calculate:

[0154] u_cate1=x cate1 *W cate1

[0155] u_num1=norm(x num1 )

[0156] User_vector=concat(u_cate1,…,u_num1,…)

[0157] wherein, W cate1 represents an embedding matrix, x cate1 represents a value of a category feature, u_cate1 represents a vector representation of the category feature, x num1 represents a value of a numerical feature, u_num1 represents a vector representation of the numerical feature, which is obtained through normalization processing, used to ensure the stability of the numerical value, and User_vector represents a user feature, which is obtained by splicing a plurality of vector representations of category features and a plurality of vector representations of numerical features.

[0158] In some other embodiments, based on the above examples, after obtaining the user feature User_vector, the advertisement feature Ad_vector and the advertisement style feature Ui_vector, the obtained Ui_vector, User_vector and Ad_vector can be spliced to obtain a vector h (corresponding to the fusion feature described above), and the vector h is input into a multi-layer neural network model, and finally the probability y of the user clicking the advertisement when the advertisement is displayed through the advertisement style is output, and the specific formula is as follows:

[0159] h = concat(Ui_vector, User_vector, Ad_vector)

[0160] y = f(f(f(h)))

[0161] wherein, h represents the fusion vector obtained by splicing the above three features, f represents Figure 12 the activation function ReLU shown in formula (3).

[0162] In addition, the loss function of the model can adopt a cross-entropy loss function, and the specific formula is as follows:

[0163] Loss = -(y*log(y') + (1-y)*log(1-y'))

[0164] wherein, Loss represents the loss function, y' represents the result of the advertisement label, and of course, in actual application, the loss function of the model can also adopt other types of loss functions, such as square loss function, exponential loss function, perception loss function, etc.

[0165] In the process of training the obtained data according to the above process, the optimizer of the whole model can adopt adam (a first-order optimization algorithm that can replace the traditional stochastic gradient descent process, which can iteratively update the weight parameters of the neural network model based on the training data), and the learning rate can be set to 0.01. After the model training is completed, the model can be stored and put online for prediction.

[0166] (4) Online service of the model

[0167] When the trained model is called in the background to obtain the optimal advertisement style, the user features and the advertisement features need to be input, and at the same time, all the advertisement styles currently available in the advertisement style library and the style features of each advertisement style can be obtained, and then the style features of each advertisement style, the user features and the advertisement features are input into the trained model at the same time to obtain the matching score of the advertisement style, and finally the advertisement style with the highest matching score is taken as the optimal advertisement style for the user, and the advertisement and the advertisement style ID are returned to the look point terminal for advertisement display. For example, as shown in Figure 9 the same lipstick advertisement, five different types of advertisement styles are stored in the advertisement style library in advance, including advertisement style 1 to advertisement style 5, and then the style features of each of the five advertisement styles, the user features (such as the user features of user A) and the advertisement features are input into the model at the same time to predict the click-through rates corresponding to the five advertisement styles, and assuming that the click-through rate corresponding to advertisement style 5 is the highest for user A, the background can return the advertisement material and advertisement style 5 to the look point terminal of user A, so that the look point terminal displays the advertisement material through advertisement style 5, and in addition, assuming that the click-through rate corresponding to advertisement style 3 is the highest for user B, the background can return the advertisement material and advertisement style 3 to the look point terminal of user B, so that the display style of the advertisement is enriched, and at the same time, the personalized needs of different users can be met.

[0168] The information recommendation method provided by the embodiments of the present application is based on the preference of individuals for advertisement styles, and through analyzing the relationship among advertisement styles, advertisement materials and users, the presentation of thousands of people with thousands of faces is realized, so that the distribution of advertisements is more refined, and the consumption of traffic platform advertisements can also be improved.

[0169] The following continues to illustrate an exemplary structure of the implementation of the information recommendation device 243 provided by the embodiments of the present application as a software module, and in some embodiments, as shown in Figure 2 the software module stored in the information recommendation device 243 in the storage 240 can include an acquisition module 2431, a fusion module 2432 and a determination module 2433.

[0170] The acquisition module 2431 is configured to acquire information features of information to be recommended, object features of a target object, and respective style features of a plurality of display styles; the fusion module 2432 is configured to perform fusion processing on the information features of the information, the object features of the target object, and the style features of each display style respectively, to obtain fusion features corresponding to each display style; the determination module 2433 is configured to determine a recommendation parameter of each display style based on the fusion features corresponding to each display style; the recommendation parameter represents a preference degree of the target object when the information is displayed through the display style; the determination module 2433 is further configured to take a display style with the highest recommendation parameter as a target display style of the information, and respond to a recommendation request for the target object based on the information and the target display style.

[0171] In some embodiments, the acquisition module 2431 is further configured to perform embedding processing on first category features included in the information to be recommended, to obtain vector representations of the first category features, perform normalization processing on first numerical value features included in the information, to obtain vector representations of the first numerical value features, and perform splicing processing on the vector representations of the first category features and the vector representations of the first numerical value features, to obtain the information features of the information; perform embedding processing on second category features included in the target object, to obtain vector representations of the second category features, perform normalization processing on second numerical value features included in the target object, to obtain vector representations of the second numerical value features, and perform splicing processing on the vector representations of the second category features and the vector representations of the second numerical value features, to obtain the object features of the target object; and perform the following processing on each display style in the plurality of display styles: perform embedding processing on third category features included in the display style, to obtain vector representations of the third category features, perform normalization processing on third numerical value features included in the display style, to obtain vector representations of the third numerical value features, and perform splicing processing on the vector representations of the third category features and the vector representations of the third numerical value features, to obtain the style features of the display style.

[0172] In some embodiments, the recommendation parameter is determined by calling a neural network model; the determination module 2433 is further configured to perform the following processing on the fusion features of each display style: perform a plurality of times of nonlinear transformation processing on the fusion features of the display style in a cascaded manner through a plurality of activation layers included in the neural network model, to obtain a corresponding recommendation parameter when the information is displayed through the display style.

[0173] In some embodiments, each object type corresponds to a pre-trained neural network model; the information recommendation apparatus 243 further comprises an object type identification module 2434 configured to perform object type identification processing on the target object to obtain an object type to which the target object belongs; and the determination module 2433 is further configured to obtain a target neural network model corresponding to the object type to which the target object belongs, and perform multiple times of cascaded nonlinear transformation processing on the fusion feature of the display style based on multiple activation layers included in the target neural network model, to obtain the recommendation parameter corresponding to the display of the information by the display style.

[0174] In some embodiments, each information type corresponds to a pre-trained neural network model; the information recommendation apparatus 243 further comprises an information type identification module 2435 configured to perform information type identification processing on the information to be recommended to obtain an information type to which the information belongs; and the determination module 2433 is further configured to obtain a target neural network model corresponding to the information type to which the information belongs, and perform multiple times of cascaded nonlinear transformation processing on the fusion feature of the display style based on multiple activation layers included in the target neural network model, to obtain the recommendation parameter corresponding to the display of the information by the display style.

[0175] In some embodiments, each display style corresponds to a pre-trained neural network model; and the determination module 2433 is further configured to obtain a target neural network model corresponding to the display style, and perform multiple times of cascaded nonlinear transformation processing on the fusion feature of the display style based on multiple activation layers included in the target neural network model, to obtain the recommendation parameter corresponding to the display of the information by the display style.

[0176] In some embodiments, the information recommendation apparatus 243 further comprises a training module 2436 configured to train the neural network model by: obtaining information features of sample information, object features of sample objects, and respective style features of multiple sample display styles; performing fusion processing on the information features of the sample information, the object features of the sample objects, and the style features of each sample display style, respectively, to obtain a fusion feature corresponding to each sample display style; performing forward propagation based on the fusion feature corresponding to each sample display style to obtain a prediction result; determining a difference between the prediction result and labeled data of the sample information, and performing backward propagation based on the difference to update parameters of the neural network model layer by layer in the process of backward propagation; wherein the labeled data comprises feedback data corresponding to the sample object when the sample information is displayed by each sample display style.

[0177] In some embodiments, the obtaining module 2431 is further configured to obtain an identifier of the sample information, associate the identifier of the sample information with the information feature library to obtain an information feature of the sample information; obtain an identifier of the sample object, associate the identifier of the sample object with the object feature library to obtain an object feature of the sample object; and for each sample display style in the plurality of sample display styles, perform the following processing: obtain an identifier of the sample display style, associate the identifier of the sample display style with the style feature library to obtain a style feature of the sample display style.

[0178] In some embodiments, the information recommendation apparatus 243 further includes a receiving module 2437 configured to receive a recommendation request for a target object sent by a terminal device; the determining module 2433 is further configured to, when the target object is a cold start object, obtain a plurality of candidate information ranked in front from the information library, and take the plurality of candidate information ranked in front as information to be recommended; and when the target object is a non-cold start object, obtain historical behavior data of the target object, determine portrait information of the target object according to the historical behavior data, obtain a plurality of candidate information matching the portrait information of the target object from the information library, and take the plurality of candidate information matching as information to be recommended.

[0179] In some embodiments, the determining module 2433 is further configured to, when receiving a recommendation request for a target object sent by a terminal device, perform at least one of the following processing: obtain a candidate object with a similarity between the object feature of the target object and the object feature of the candidate object greater than a similarity threshold from the object library, obtain information preferred by the candidate object, and take the preferred information as information to be recommended; obtain historical preference information of the target object, determine an information feature of the historical preference information, obtain a plurality of candidate information with a similarity between the information feature of the historical preference information and the information feature of the candidate information greater than a similarity threshold from the information library, and take the plurality of candidate information as information to be recommended.

[0180] The following continues to illustrate an exemplary structure of the information recommendation apparatus 465 provided by the embodiments of the present application, which is implemented as a software module. In some embodiments, as shown in FIG. 4B, the software modules in the information recommendation apparatus 465 stored in the memory 460 can include a receiving module 4651 and a display module 4652. Figure 3

[0181] The receiving module 4651 is configured to receive at least one information to be recommended for a target object and a target display style of each information; and the display module 4652 is configured to display the at least one information in a human-computer interaction interface based on the target display style of each information; wherein the target display style of each information is selected from a plurality of candidate display styles based on an information feature of the information, an object feature of the target object, and a style feature of each of the plurality of candidate display styles.

[0182] ​It should be noted that the description of the apparatus in the embodiments of the present application is similar to the implementation of the information recommendation method in the foregoing, and has similar beneficial effects, and thus will not be described herein. For technical details of the information recommendation apparatus provided in the embodiments of the present application, the description can be understood according to the description of the information recommendation method in the foregoing. Figures 4-5 The description of any figure.

[0183] The embodiments of the present application provide a computer program product or a computer program, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the information recommendation method provided in the embodiments of the present application.

[0184] The embodiments of the present application provide a computer readable storage medium storing executable instructions, wherein the executable instructions, when executed by a processor, cause the processor to execute the information recommendation method provided in the embodiments of the present application, for example, the information recommendation method shown in Figure 4 , or Figure 5 .

[0185] In some embodiments, the computer readable storage medium can be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disc, or CD-ROM; or various devices including one or any combination of the above memories.

[0186] In some embodiments, the executable instructions can be in the form of a program, software, software module, script or code, written in any form of programming language (including a compiled or interpreted language, or a declarative or procedural language), and can be deployed in any form, including being deployed as a standalone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0187] As an example, the executable instructions can but not necessarily correspond to a file in a file system, can be stored in a part of a file storing other programs or data, for example, stored in one or more scripts in a Hyper Text Markup Language (HTML) document, stored in a single file dedicated to the program in question, or stored in multiple cooperating files (for example, files storing one or more modules, subroutines or code portions).

[0188] As an example, the executable instructions can be deployed to be executed on one computing device, or executed on multiple computing devices located in one place, or executed on multiple computing devices distributed in multiple places and interconnected through a communication network.

[0189] To sum up, the embodiment of the present application selects the target display style of the information to be recommended from the plurality of candidate display styles based on the information features of the information to be recommended, the object features of the target object, and the style features of the plurality of candidate display styles, so as to display the information to be recommended through the target display style. Thus, since the target display style of the information to be recommended is related to the object features of the target object, the target display style corresponding to different users is different for the same information, the display style of the recommended information is enriched, and the personalized needs of different users are met.

[0190] The above merely describes the embodiments of the present application, but is not used to limit the protection scope of the present application. Any modification, equivalent replacement, and improvement within the spirit and scope of the present application shall be included in the protection scope of the present application.

Claims

1. An information recommendation method characterized by comprising: The method comprises: in response to receiving a recommendation request for a target object, obtaining information to be recommended; performing embedding processing on first category features included in the information to obtain vector representations of the first category features, performing normalization processing on first numerical features included in the information to obtain vector representations of the first numerical features, and performing concatenation processing on the vector representations of the first category features and the vector representations of the first numerical features to obtain information features of the information; performing embedding processing on second category features included in the target object to obtain vector representations of the second category features, performing normalization processing on second numerical features included in the target object to obtain vector representations of the second numerical features, and performing concatenation processing on the vector representations of the second category features and the vector representations of the second numerical features to obtain object features of the target object; for each display style in a plurality of display styles, performing embedding processing on third category features included in the display style to obtain vector representations of the third category features, performing normalization processing on third numerical features included in the display style to obtain vector representations of the third numerical features, and performing concatenation processing on the vector representations of the third category features and the vector representations of the third numerical features to obtain style features of the display style; performing fusion processing on the information features of the information and the object features of the target object and the style features of each of the display styles to obtain fusion features corresponding to each of the display styles; for the fusion features of each of the display styles, performing cascaded multiple times of nonlinear transformation processing on the fusion features of the display style through a plurality of activation layers included in a neural network model to obtain recommendation parameters corresponding to displaying the information through the display style; wherein the recommendation parameters represent a preference degree of the target object when the information is displayed through the display style; taking a display style with the highest recommendation parameter as a target display style of the information, and responding to the recommendation request for the target object based on the information and the target display style.

2. The method of claim 1, wherein: each object type corresponds to a pre-trained neural network model; the performing, by the neural network model including a plurality of activation layers, cascaded multiple times of nonlinear transformation processing on the fusion features of the display style to obtain recommendation parameters corresponding to displaying the information through the display style comprises: performing object type identification processing on the target object to obtain an object type to which the target object belongs; obtaining a target neural network model corresponding to the object type to which the target object belongs, and performing, based on a plurality of activation layers included in the target neural network model, cascaded multiple times of nonlinear transformation processing on the fusion features of the display style to obtain recommendation parameters corresponding to displaying the information through the display style.

3. The method of claim 1, wherein: each information type corresponds to a pre-trained neural network model; The multiple activation layers included in the neural network model perform multiple times of nonlinear transformation processing on the fusion features of the display style in cascade, to obtain the recommended parameters corresponding to the display of the information by the display style. information type identification processing on the information to be recommended is performed to obtain the information type to which the information belongs; a target neural network model corresponding to the information type to which the information belongs is obtained, and multiple times of nonlinear transformation processing on the fusion features of the display style in cascade is performed based on the multiple activation layers included in the target neural network model, to obtain the recommended parameters corresponding to the display of the information by the display style.

4. The method of claim 1, wherein each display style corresponds to a pre-trained neural network model; the multiple activation layers included in the neural network model perform multiple times of nonlinear transformation processing on the fusion features of the display style in cascade, to obtain the recommended parameters corresponding to the display of the information by the display style. a target neural network model corresponding to the display style is obtained, and multiple times of nonlinear transformation processing on the fusion features of the display style in cascade is performed based on the multiple activation layers included in the target neural network model, to obtain the recommended parameters corresponding to the display of the information by the display style.

5. The method of claim 1, wherein, Before the multiple activation layers included in the neural network model perform multiple times of nonlinear transformation processing on the fusion features of the display style in cascade, the method further comprises: The neural network model is trained by: obtaining information features of sample information, object features of a sample object, and respective style features of multiple sample display styles; performing fusion processing on the information features of the sample information, the object features of the sample object, and the style features of each of the sample display styles, respectively, to obtain fusion features corresponding to each of the sample display styles; performing forward propagation based on the fusion features corresponding to each of the sample display styles to obtain a prediction result; determining a difference between the prediction result and labeled data of the sample information, and performing backward propagation based on the difference, and updating parameters of the neural network model layer by layer in the backward propagation process; wherein the labeled data includes feedback data corresponding to the sample object when the sample information is displayed by each of the sample display styles.

6. The method of claim 5, wherein, The obtaining of the information features of sample information, the object features of a sample object, and the respective style features of multiple sample display styles comprises: obtaining an identifier of the sample information, and associating the identifier of the sample information with an information feature library to obtain the information features of the sample information; obtaining an identifier of the sample object, and associating the identifier of the sample object with an object feature library to obtain the object features of the sample object; for each of the multiple sample display styles, the following processing is performed: obtaining an identifier of the sample display style, and associating the identifier of the sample display style with a style feature library to obtain the style features of the sample display style.

7. The method of claim 1, wherein, The obtaining of the information to be recommended in response to receiving a recommendation request for a target object comprises: receiving a recommendation request for a target object sent by a terminal device; when the target object is a cold start object, obtaining a plurality of candidate information ranked in front from an information library, and taking the plurality of candidate information ranked in front as information to be recommended; when the target object is a non-cold start object, obtaining historical behavior data of the target object, determining portrait information of the target object according to the historical behavior data, obtaining a plurality of candidate information matched with the portrait information of the target object from the information library, and taking the plurality of matched candidate information as information to be recommended.

8. The method of claim 1, wherein, The method comprises: when receiving a recommendation request for a target object sent by a terminal device, at least one of the following processes is performed: obtaining a candidate object with a similarity between an object feature of the candidate object and an object feature of the target object greater than a similarity threshold from an object library, obtaining information preferred by the candidate object, and taking the preferred information as information to be recommended; obtaining historical preference information of the target object, determining an information feature of the historical preference information, obtaining a plurality of candidate information with a similarity between the information feature of the historical preference information and an information feature greater than a similarity threshold from an information library, and taking the plurality of candidate information as information to be recommended.

9. An information recommendation method characterized by comprising: The method comprises: receiving at least one information to be recommended for a target object and a target display style of each of the information; displaying the at least one information based on the target display style of each of the information in a human-computer interaction interface. The target display style of each information is determined by: performing embedding processing on first category features included in the information to obtain vector representations of the first category features, performing normalization processing on first numerical value features included in the information to obtain vector representations of the first numerical value features, and performing splicing processing on the vector representations of the first category features and the vector representations of the first numerical value features to obtain information features of the information; performing embedding processing on second category features included in the target object to obtain vector representations of the second category features, performing normalization processing on second numerical value features included in the target object to obtain vector representations of the second numerical value features, and performing splicing processing on the vector representations of the second category features and the vector representations of the second numerical value features to obtain object features of the target object; for each display style in a plurality of display styles, performing embedding processing on third category features included in the display style to obtain vector representations of the third category features, performing normalization processing on third numerical value features included in the display style to obtain vector representations of the third numerical value features, and performing splicing processing on the vector representations of the third category features and the vector representations of the third numerical value features to obtain style features of the display style; performing fusion processing on the information features of the information, the object features of the target object, and the style features of each display style to obtain fusion features corresponding to each display style; and performing, by a plurality of activation layers included in a neural network model, multiple times of nonlinear transformation processing in cascade on the fusion features of each display style to obtain a recommended parameter corresponding to a case where the information is displayed by the display style; wherein the recommended parameter represents a preference degree of the target object in the case where the information is displayed by the display style; and the display style with the highest recommended parameter is taken as the target display style of the information.

10. An information recommendation device characterized by comprising: The apparatus comprises: an acquisition module configured to, in response to receiving a recommendation request for a target object, acquire information to be recommended; the acquisition module is further configured to perform embedding processing on first category features included in the information to obtain vector representations of the first category features, perform normalization processing on first numerical value features included in the information to obtain vector representations of the first numerical value features, and perform splicing processing on the vector representations of the first category features and the vector representations of the first numerical value features to obtain information features of the information; the acquisition module is further configured to perform embedding processing on second category features included in the target object to obtain vector representations of the second category features, perform normalization processing on second numerical value features included in the target object to obtain vector representations of the second numerical value features, and perform splicing processing on the vector representations of the second category features and the vector representations of the second numerical value features to obtain object features of the target object; The acquisition module is further configured to, for each of the plurality of display styles, perform embedding processing on a third category feature included in the display style to obtain a vector representation of the third category feature, perform normalization processing on a third numerical feature included in the display style to obtain a vector representation of the third numerical feature, and perform concatenation processing on the vector representation of the third category feature and the vector representation of the third numerical feature to obtain a style feature of the display style; The fusion module is configured to fuse the information feature of the information, the object feature of the target object, and the style feature of each of the display styles respectively to obtain a fusion feature corresponding to each of the display styles; The determination module is configured to, for the fusion feature of each of the display styles, perform cascaded multiple times of nonlinear transformation processing on the fusion feature of the display style by using a plurality of activation layers included in a neural network model to obtain a recommendation parameter corresponding to displaying the information by using the display style; the recommendation parameter represents a preference degree of the target object when the information is displayed by using the display style; The determination module is further configured to take the display style with the highest recommendation parameter as a target display style of the information, and respond to a recommendation request for the target object based on the information and the target display style.

11. An information recommendation device characterized by comprising: The apparatus comprises: a receiving module configured to receive at least one information to be recommended for a target object and a target display style of each of the information; a display module configured to display the at least one information based on the target display style of each of the information in a human-computer interaction interface. The target display style of each information is determined by: performing embedding processing on first category features included in the information to obtain vector representations of the first category features, performing normalization processing on first numerical features included in the information to obtain vector representations of the first numerical features, and performing splicing processing on the vector representations of the first category features and the vector representations of the first numerical features to obtain information features of the information; performing embedding processing on second category features included in the target object to obtain vector representations of the second category features, performing normalization processing on second numerical features included in the target object to obtain vector representations of the second numerical features, and performing splicing processing on the vector representations of the second category features and the vector representations of the second numerical features to obtain object features of the target object; for each display style in a plurality of display styles, performing embedding processing on third category features included in the display style to obtain vector representations of the third category features, performing normalization processing on third numerical features included in the display style to obtain vector representations of the third numerical features, and performing splicing processing on the vector representations of the third category features and the vector representations of the third numerical features to obtain style features of the display style; performing fusion processing on the information features of the information, the object features of the target object, and the style features of each display style to obtain fusion features corresponding to each display style; for the fusion features of each display style, performing multiple times of nonlinear transformation processing on the fusion features of the display style through multiple activation layers included in a neural network model to obtain a recommended parameter corresponding to the display style for displaying the information through the display style; wherein the recommended parameter represents a preference degree of the target object when the information is displayed through the display style; and taking a display style with the highest recommended parameter as the target display style of the information.

12. An electronic device, comprising: The electronic device comprises: a memory for storing executable instructions; a processor for executing the executable instructions stored in the memory to implement the information recommendation method of any one of claims 1 to 8 or claim 9.

13. A computer-readable storage medium storing executable instructions, wherein the instructions, when executed by a processor, cause the processor to perform operations comprising: The executable instructions are executed by the processor to implement the information recommendation method of any one of claims 1 to 8 or claim 9.

14. A computer program product comprising computer programs or instructions, characterized in that, The computer program or instructions are executed by the processor to implement the information recommendation method of any one of claims 1 to 8 or claim 9.

Images