Server, display device and media asset recommendation method

Abstract

Embodiments of the invention show a server, a display device and a media asset recommendation method. The method comprises the steps of obtaining media asset exposure times and media asset click times; converting the media asset exposure times into label exposure times of media asset labels corresponding to the media assets, and converting the media asset click times into label click times of the media asset labels corresponding to the media assets; calculating label income according to the label exposure times and the label click times; determining the media asset label corresponding to the label revenue of the previous preset rank with the highest label revenue as a target label; and sending the target media asset corresponding to the target tag to a display device, so that the display device displays the target media asset on a media asset recommendation page. According to the scheme provided by the embodiment of the invention, the media assets corresponding to the media asset labels with the highest label income are recommended to the user, and the media assets are replaced by the tentative objects by the labels, so that the calculation amount of the tentative object income is reduced, and the media assets with cross-category interests and hobbies can be quickly recommended to the user.

Description

Technical Field

[0001] The present application relates to the technical field of display devices, and in particular to a server, a display device, and a media resource recommendation method. Background Art

[0002] The content of large-screen movies and TV shows is becoming increasingly rich, and personalized recommendation strategies are emerging one after another. When user interests are relatively fixed, each recommendation position displayed on the first screen is in line with user interests, but there are still problems such as a single recommended media category and users repeatedly consuming the same type of films, which will have a negative impact on the utilization rate of media resources and the breadth of user consumption, and thus affect the user's viewing experience.

[0003] In order to avoid recommending a single category of media assets, the recommendation algorithm can be used to conduct cross-category interest exploration. The cross-category interest exploration method can take each media asset as the object of rotation, calculate the revenue of each media asset through the algorithm of the upper bound of the confidence interval, and then select the media asset with the highest revenue to recommend to the user.

[0004] However, due to the wide variety of media resources for video applications and the corresponding number of media resources reaching tens of thousands, the calculation of media resource revenue is extremely complex, making it impossible to accurately and quickly obtain users' current cross-category interests and hobbies. Summary of the invention

[0005] Some embodiments of the present application provide a server, a display device, and a media resource recommendation method, which recommends media resources corresponding to media resources tags with the highest tag revenue to users, and can recommend media resources that are more in line with users' category interests and hobbies. Moreover, replacing media resources with tags as trial objects can reduce the amount of calculation of trial object revenue, which helps to quickly recommend media resources with cross-category interests and hobbies to users.

[0006] In a first aspect, some embodiments of the present application provide a server, including:

[0007] A communication module, configured to establish a communication connection with a display device;

[0008] The processing module is configured as follows:

[0009] Get the number of media exposures and media clicks;

[0010] Convert the number of media asset exposures to the number of label exposures of the media asset's corresponding media asset label, and convert the number of media asset clicks to the number of label clicks of the media asset's corresponding media asset label. The number of label exposures is the same as the number of media asset exposures, and the number of media asset clicks is the same as the number of label clicks.

[0011] The tag revenue is calculated based on the number of tag exposures and tag clicks. The tag revenue is used to represent the predicted probability of being clicked by users after the corresponding media asset tag is exposed.

[0012] The media asset tag corresponding to the tag revenue of the first preset ranking with the highest tag revenue is determined as the target tag;

[0013] The target media asset corresponding to the target tag is sent to the display device, so that the display device displays the target media asset on the media asset recommendation page.

[0014] The above technical solution has the following advantages or beneficial effects: the number of media asset clicks and exposures is converted into the number of tag clicks and exposures, and the tag revenue is calculated, and the media asset corresponding to the media asset tag with the highest tag revenue is recommended to the user. The media asset that is more in line with the user's category interests and hobbies can be recommended to the user, and the trial object is replaced by the tag instead of the media asset, which can reduce the amount of calculation of the trial object revenue, and help to quickly recommend media assets with cross-category interests and hobbies to users.

[0015] In some embodiments, the processing module is further configured to obtain the number of media asset exposures and the number of media asset clicks:

[0016] Receive user behavior logs sent by display devices;

[0017] According to the media assets and behaviors in the user behavior log, the number of media asset exposures and media asset clicks are counted. Behaviors include exposures and clicks.

[0018] The above technical solution has the following advantages or beneficial effects: according to the user behavior log uploaded in real time by the display device, the number of media exposures and the number of media clicks can be obtained in time, so that the media to be recommended can be quickly determined and sent to the display device for display.

[0019] In some embodiments, the processing module is further configured to calculate the tag revenue according to the number of tag exposures and the number of tag clicks:

[0020] The ratio of the number of tag clicks to the number of tag exposures is determined as the average revenue of the media asset tag;

[0021] The basic revenue of the media asset tag is calculated based on the average revenue, the number of recommendations, and the total number of recommendations. The number of recommendations is the number of times the target media asset is sent to the display device when the target media asset corresponds to the media asset tag. The total number of recommendations is the total number of times the target media asset is sent to the display device. The basic revenue is used to represent the basic probability of the media asset corresponding to the media tag being clicked by the user after being exposed.

[0022] The weighted benefit is calculated based on the media asset tags, similar tags, and heterogeneous tags. Similar tags are media asset tags with similar attributes to the media asset tags in the positive feedback tags. Heterogeneous tags are media asset tags with heterogeneous attributes to the media asset tags in the positive feedback tags. The positive feedback tags are the media asset tags corresponding to the media asset identifiers that are clicked after exposure. The weighted benefit is used to represent the probability of adding or reducing the weight of the media asset corresponding to the media tag being clicked by the user after exposure.

[0023] The sum of the basic income and the weighted income is determined as the label income of the media asset label.

[0024] The above technical solution has the following advantages or beneficial effects: by introducing weighted benefits into the benefit calculation of media asset tags, the accuracy of media asset tag benefit calculation can be improved, thereby improving the accuracy of media asset recommendation.

[0025] In some embodiments, the processing module is further configured to calculate the weighted benefit according to the media asset tags, the same tags and the different tags:

[0026] Obtain a first media asset vector of a media asset group corresponding to a media asset tag, a second media asset vector of a media asset group corresponding to a similar label, and a third media asset vector of a media asset group corresponding to a different label;

[0027] Determine the average vector of the first media asset vector as the media asset label vector, determine the average vector of the second media asset vector as the same type label vector, and determine the average vector of the third media asset vector as the different type label vector;

[0028] Calculate a first similarity between a media asset label vector and a label vector of the same type, and a second similarity between a media asset label vector and a label vector of a different type;

[0029] The weighted benefit is calculated according to the first similarity and the second similarity.

[0030] The above technical solution has the following advantages or beneficial effects: by weighting the benefits by using similar attribute labels in the positive feedback labels, and by downweighting the benefits by using heterogeneous attribute labels in the positive feedback labels, the accuracy of media resource label benefit calculation can be improved, thereby improving the accuracy of media resource recommendation.

[0031] In some embodiments, the processing module executes sending the target media asset corresponding to the target tag to the display device, and is further configured to:

[0032] Send the target media asset with the largest number of matches to the target tag to the display device; or

[0033] The target media assets whose number of matches with the target tags reaches the preset number are sent to the display device.

[0034] The above technical solution has the following advantages or beneficial effects: the media assets that match the target tag the most or reach a preset number are sent to the display device as target media assets, so that the recommended media assets can better meet the user's interests.

[0035] In some embodiments, the processing module executes sending the target media asset corresponding to the target tag to the display device, and is further configured to:

[0036] Calculate the target media asset vector of the target media asset, and obtain the cached media asset vector of the delivered media asset;

[0037] Input the target media asset vector into the Gaussian clustering model to obtain the number of first cluster centers;

[0038] Inputting the target media asset vector and the cached media asset vector into the Gaussian clustering model to obtain the number of second cluster centers;

[0039] Calculate the difference between the number of the second cluster centers and the number of the first cluster centers;

[0040] If the difference is greater than or equal to 1, the target media asset corresponding to the target tag is sent to the display device.

[0041] The above technical solution has the following advantages or beneficial effects: a Gaussian clustering model is introduced to determine whether the target media needs to be sent to the display device, and the server can update the media recommendation page in time according to user behavior without the user actively pressing a button to expose more recommendation positions to update.

[0042] In a second aspect, some embodiments of the present application provide a display device, including:

[0043] A display; configured to display a media resource recommendation page;

[0044] A communication device configured to establish a communication connection with a server

[0045] A controller is coupled to the display and is configured to:

[0046] The target media asset is sent by the receiving server. The target media asset is the media asset corresponding to the target tag. The target tag is the media asset tag corresponding to the tag revenue of the preset ranking with the highest tag revenue. The tag revenue is used to represent the predicted probability of the media asset corresponding to the media asset tag being clicked by the user after being exposed. The tag revenue is calculated based on the tag exposure times and the tag click times of the media asset tag. The tag exposure times are converted from the media asset exposure times corresponding to the media asset tag, and the tag click times are converted from the media click times corresponding to the media asset tag. The tag exposure times are the same as the media asset exposure times, and the media click times are the same as the tag click times.

[0047] The control display displays the target media asset on the media asset recommendation page.

[0048] The above technical solution has the following advantages or beneficial effects: the number of media asset clicks and exposures is converted into the number of tag clicks and exposures, and the tag revenue is calculated, and the media asset corresponding to the media asset tag with the highest tag revenue is recommended to the user. The media asset that is more in line with the user's category interests and hobbies can be recommended to the user, and the trial object is replaced by the tag instead of the media asset, which can reduce the amount of calculation of the trial object revenue, and help to quickly recommend media assets with cross-category interests and hobbies to users.

[0049] In a third aspect, some embodiments of the present application provide a media resource recommendation method, which is applied to a server and includes:

[0050] Get the number of media exposures and media clicks;

[0051] Convert the number of media asset exposures to the number of label exposures of the media asset's corresponding media asset label, and convert the number of media asset clicks to the number of label clicks of the media asset's corresponding media asset label. The number of label exposures is the same as the number of media asset exposures, and the number of media asset clicks is the same as the number of label clicks.

[0052] The tag revenue is calculated based on the number of tag exposures and tag clicks. The tag revenue is used to represent the predicted probability of being clicked by users after the corresponding media asset tag is exposed.

[0053] The media asset tag corresponding to the tag revenue of the first preset ranking with the highest tag revenue is determined as the target tag;

[0054] The target media asset corresponding to the target tag is sent to the display device, so that the display device displays the target media asset on the media asset recommendation page.

[0055] The above technical solution has the following advantages or beneficial effects: the number of media asset clicks and exposures is converted into the number of tag clicks and exposures, and the tag revenue is calculated, and the media asset corresponding to the media asset tag with the highest tag revenue is recommended to the user. The media asset that is more in line with the user's category interests and hobbies can be recommended to the user, and the trial object is replaced by the tag instead of the media asset, which can reduce the amount of calculation of the trial object revenue, and help to quickly recommend media assets with cross-category interests and hobbies to users.

[0056] In some embodiments, the step of obtaining the number of media asset exposures and the number of media asset clicks includes:

[0057] Receive user behavior logs sent by display devices;

[0058] According to the media assets and behaviors in the user behavior log, the number of media asset exposures and media asset clicks are counted. Behaviors include exposures and clicks.

[0059] The above technical solution has the following advantages or beneficial effects: according to the user behavior log uploaded in real time by the display device, the number of media exposures and the number of media clicks can be obtained in time, so that the media to be recommended can be quickly determined and sent to the display device for display.

[0060] In a fourth aspect, some embodiments of the present application provide a media resource recommendation method, which is applied to a display device, including:

[0061] The target media asset is sent by the receiving server. The target media asset is the media asset corresponding to the target tag. The target tag is the media asset tag corresponding to the tag revenue of the preset ranking with the highest tag revenue. The tag revenue is used to represent the predicted probability of the media asset corresponding to the media asset tag being clicked by the user after being exposed. The tag revenue is calculated based on the tag exposure times and the tag click times of the media asset tag. The tag exposure times are converted from the media asset exposure times corresponding to the media asset tag, and the tag click times are converted from the media click times corresponding to the media asset tag. The tag exposure times are the same as the media asset exposure times, and the media click times are the same as the tag click times.

[0062] The control display displays the target media asset on the media asset recommendation page.

[0063] The above technical solution has the following advantages or beneficial effects: the number of media asset clicks and exposures is converted into the number of tag clicks and exposures, and the tag revenue is calculated, and the media asset corresponding to the media asset tag with the highest tag revenue is recommended to the user. The media asset that is more in line with the user's category interests and hobbies can be recommended to the user, and the trial object is replaced by the tag instead of the media asset, which can reduce the amount of calculation of the trial object revenue, and help to quickly recommend media assets with cross-category interests and hobbies to users.

[0064] The technical solution provided in the embodiment of the present application obtains the number of media asset exposures and the number of media asset clicks, and converts the number of media asset exposures and the number of media asset clicks into the number of label exposures and the number of label clicks of the media asset labels corresponding to the media assets, and then calculates the label revenue according to the number of label exposures and the number of label clicks, and pushes the media assets corresponding to the media asset labels with the largest label revenue to the display device for display. The solution in the embodiment of the present application recommends the media assets corresponding to the media asset labels with the highest label revenue to the user, and can recommend to the user media assets that are more in line with their category interests and hobbies, and replaces the media assets with labels as the test objects, which can reduce the amount of calculation of the test object revenue, and helps to quickly recommend media assets with cross-category interests and hobbies to the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0066] Figure 1 A schematic diagram of an operation scenario between a display device and a control device provided in some embodiments of the present application;

[0067] Figure 2 A schematic diagram of the hardware configuration of a display device provided in some embodiments of the present application;

[0068] Figure 3 A schematic diagram of software configuration of a display device provided in some embodiments of the present application;

[0069] Figure 4 A flowchart of a media resource recommendation method provided in some embodiments of the present application;

[0070] Figure 5 A schematic diagram of a user's movie-watching behavior within a week provided in some embodiments of the present application;

[0071] Figure 6 A flowchart of a method for determining the number of exposures and clicks of a media asset tag provided in some embodiments of the present application;

[0072] Figure 7 A flowchart of a method for calculating the tag revenue of a media asset tag provided in some embodiments of the present application;

[0073] Figure 8 A flowchart of a method for determining whether to conduct a trial is provided for some embodiments of the present application;

[0074] Fig. 9 A flowchart of a media resource recommendation method provided in some embodiments of the present application;

[0075] Fig.10 A timing diagram of a media resource recommendation method provided in some embodiments of the present application. DETAILED DESCRIPTION

[0076] The following embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following embodiments do not represent all implementations consistent with the present application. They are only examples of systems and methods consistent with some aspects of the present application as detailed in the claims.

[0077] It should be noted that the brief description of terms in this application is only for the convenience of understanding the embodiments described below, and is not intended to limit the embodiments of this application. Unless otherwise specified, these terms should be understood according to their common and usual meanings.

[0078] The terms "first", "second", "third", etc. in the specification and claims of this application and the above drawings are used to distinguish similar or similar objects or entities, and do not necessarily mean to limit a specific order or sequence, unless otherwise noted. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances.

[0079] The terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover but not exclude inclusion, for example, a product or device comprising a list of components is not necessarily limited to all the components expressly listed but may include other components not expressly listed or inherent to such product or device.

[0080] The term "module" refers to any known or later developed hardware, software, firmware, artificial intelligence, fuzzy logic, or combination of hardware and / or software code that is capable of performing the functions associated with that element.

[0081] In the embodiment of the present application, the display device 200 generally refers to a device with image display and data processing capabilities. For example, the display device 200 includes but is not limited to a smart TV, a mobile terminal, a computer, a monitor, an advertising screen, a wearable device, a virtual reality device, an augmented reality device, etc.

[0082] Figure 1 This is a schematic diagram of an operation scenario between a display device and a control device provided in some embodiments of the present application. Figure 1 As shown in FIG. 1 , a user can operate the display device 200 through touch operation, the mobile terminal 300 and the control device 100. The control device 100 is used to receive operation instructions input by the user and convert the operation instructions into control instructions that can be recognized and responded to by the display device 200. For example, the control device 100 can be a remote controller, a stylus pen, a handle, etc.

[0083] The mobile terminal 300 can be used as a control device for performing human-computer interaction between the user and the display device 200. The mobile terminal 300 can also be used as a communication device for establishing a communication connection with the display device 200 and performing data interaction. In some embodiments, the mobile terminal 300 can install software applications with the display device 200, and achieve connection and communication through a network communication protocol to achieve the purpose of one-to-one control operation and data communication. The audio and video content displayed on the mobile terminal 300 can also be transmitted to the display device 200 to achieve a synchronous display function.

[0084] In some embodiments, the mobile terminal 300 or other electronic devices may also simulate the functions of the control device 100 by running an application program for controlling the display device 200 .

[0085] like Figure 1 As also shown in FIG. 4 , the display device 200 also communicates data with the server 400 through various communication methods. The display device 200 may be allowed to communicate and connect through a local area network (LAN), a wireless local area network (WLAN), and other networks.

[0086] After receiving the input instruction from the user on the media asset recommendation page, the display device 200 performs the operation corresponding to the input instruction, generates a user behavior log corresponding to the input instruction, and then sends the user behavior log to the server 400. The server determines the exposure times and click times of the media asset tag according to the user behavior log, calculates the tag revenue of the media asset tag, and then determines the target media asset according to the media asset tag with the largest tag revenue and the top preset ranking, and finally sends the target media asset to the display device 200. The display device 200 displays the target media asset on the media asset recommendation page.

[0087] The display device 200 may provide a broadcast receiving television function, and may also additionally provide an intelligent network television function with a computer support function, including but not limited to network television, smart television, Internet Protocol television (IPTV), and the like.

[0088] Figure 2 Some embodiments of the present application provide Figure 1 2 is a block diagram of the hardware configuration of the display device 200.

[0089] In some embodiments, the display device 200 may include at least one of a tuner 210, a communication device 220, a detector 230, a device interface 240, a controller 250, a display 260, an audio output device 270, a memory, a power supply, and a user input interface.

[0090] In some embodiments, the detector 230 is used to collect signals of the external environment or external interaction. For example, the detector 230 includes a light receiver, a sensor for collecting the intensity of ambient light; or, the detector 230 includes an image collector, such as a camera, which can be used to collect external environment scenes, user attributes or user interaction gestures; or, the detector 230 includes a sound collector, such as a microphone, etc., for receiving external sounds.

[0091] In some embodiments, the display 260 includes a display function component for presenting a picture, and a driving component for driving an image display. The display 260 is used to receive an image signal output from the controller 250 for display. For example, the display 260 can be used to display video content, image content, and components of a menu control interface and a user control UI interface.

[0092] In some embodiments, the communication device 220 is a component for communicating with an external device or server 400 according to various communication protocol types. The display device 200 may be provided with a plurality of communication devices 220 according to different supported communication modes. For example, when the display device 200 supports wireless network communication, the display device 200 may be provided with a communication device 220 including a WiFi function. When the display device 200 supports Bluetooth connection communication, the display device 200 needs to be provided with a communication device 220 including a Bluetooth function.

[0093] The communication device 220 can enable the display device 200 to communicate with the external device or server 400 by wireless or wired connection. Among them, the wired connection can connect the display device 200 with the external device through components such as data cables and interfaces. The wireless connection can connect the display device 200 with the external device through wireless signals or wireless networks. The display device 200 can establish a connection relationship with the external device directly, or indirectly establish a connection relationship through a gateway, a router, a connection device, etc.

[0094] In some embodiments, the controller 250 may include at least one of a central processing unit, a video processor, an audio processor, a graphics processor, and a power processor, and a first interface to an nth interface for input / output. The controller 250 controls the operation of the display device and responds to the user's operation through various software control programs stored in the memory. The controller 250 controls the overall operation of the display device 200.

[0095] In some embodiments, the controller 250 and the tuner-demodulator 210 may be located in different separate devices, that is, the tuner-demodulator 210 may also be located in an external device of the main device where the controller 250 is located, such as an external set-top box.

[0096] In some embodiments, the user may input a user command through a graphical user interface (GUI) displayed on the display 260 , and the user input interface receives the user input command through the graphical user interface (GUI).

[0097] In some embodiments, the audio output device 270 may be a local speaker of the display device 200, or may be an external audio output device of the display device 200. In particular, for the external audio output device of the display device 200, the display device 200 may also be provided with an external audio output terminal, and the audio output device may be connected to the display device 200 through the external audio output terminal to output the sound of the display device 200.

[0098] In some embodiments, the user input interface 280 can be used to receive instructions from the user. For example, the user input interface 280 can receive instructions from the user to select a control or move the focus in the media resource recommendation page. The user input interface 280 can also receive voice instructions from the user.

[0099] In order to perform user interaction, in some embodiments, the display device 200 may run an operating system. The operating system is a computer program for managing and controlling hardware resources and software resources in the display device 200. The operating system can control the display device to provide a user interface. For example, the operating system can directly control the display device to provide a user interface, or provide a user interface by running an application program. The operating system also allows the user to interact with the display device 200.

[0100] It should be noted that the operating system can be a native operating system based on a specific operating platform, a third-party operating system deeply customized based on a specific operating platform, or an independent operating system specially developed for the display device.

[0101] The operating system can be divided into different modules or layers according to the functions implemented, such as Figure 3 As shown, in some embodiments, the system is divided into four layers, from top to bottom, namely, the application layer (Applications) layer (referred to as "application layer"), the application framework layer (Application Framework) layer (referred to as "framework layer"), the system library layer and the kernel layer.

[0102] In some embodiments, the application layer is used to provide services and interfaces for applications so that the display device 200 can run applications and interact with users based on the applications. At least one application can be run in the application layer, and these applications can be window programs, system settings programs, clock programs, etc. that come with the operating system; they can also be applications developed by third-party developers. In specific implementations, the application packages in the application layer are not limited to the above examples.

[0103] In the embodiment of the present application, the application program layer may include a media resource application, and the media resource application is used to provide users with massive media resources for users to choose to watch.

[0104] The framework layer provides application programming interfaces (APIs) and programming frameworks for applications. The application framework layer includes some predefined functions. The application framework layer is equivalent to a processing center that determines the actions that applications in the application layer take. Through the API interface, applications can access system resources and obtain system services during execution.

[0105] like Figure 3 As shown, in the embodiment of the present application, the application framework layer includes a view system, managers, content providers, etc., wherein the view system can design and implement the interface and interaction of the application, and the view system includes lists, grids, text boxes, buttons, etc. The manager includes at least one of the following modules: an activity manager for interacting with all activities running in the system; a location manager for providing system services or applications with access to system location services; a package manager for retrieving various information related to the application package currently installed on the device; a notification manager for controlling the display and clearing of notification messages; and a window manager for managing icons, windows, toolbars, wallpapers, and desktop widgets on the user interface.

[0106] In some embodiments, the activity manager is used to manage the life cycle of each application and the usual navigation back function, such as controlling the exit, opening, and back of the application. The window manager is used to manage all window programs, such as obtaining the display screen size, determining whether there is a status bar, locking the screen, capturing the screen, and controlling the display window changes, for example, reducing the display window, shaking the display, distorting the display, etc.

[0107] In some embodiments, the system runtime layer can provide support for the framework layer. When the framework layer is used, the operating system will run the instruction library contained in the system runtime layer, such as the C / C++ instruction library, to implement the functions to be implemented by the framework layer.

[0108] In some embodiments, the kernel layer is a functional layer between the hardware and software of the display device 200. The kernel layer can implement functions such as hardware abstraction, multitasking, and memory management. Figure 3 As shown, the kernel layer may be configured with hardware drivers, and the drivers included in the kernel layer may be at least one of the following drivers: audio driver, display driver, Bluetooth driver, camera driver, WIFI driver, USB driver, HDMI driver, sensor driver (such as fingerprint sensor, temperature sensor, pressure sensor, etc.), and power driver, etc.

[0109] It should be noted that the above example is only a simple division of the operating system functions and does not constitute a limitation on the specific operating system form of the display device 200 in the embodiment of the present application. Depending on factors such as the function of the display device and the type of operating system, the number of levels and specific level types contained in the operating system may be expressed in other forms.

[0110] The content of large-screen movies and TV shows is becoming increasingly rich, and personalized recommendation strategies are emerging one after another. Accurately obtaining user interests and completing video recommendations based on user preferences and interests has become a relatively mature direction. With the deepening of personalized recommendations, the personalized recommendation results and categories that the same user can obtain are gradually fixed, and the number of recommended positions on large-screen terminals is relatively fixed and precious. It is impossible to continuously scroll down and refresh to complete the recommendation like short video apps. From the perspective of the recommendation model alone, it is impossible to accurately and quickly obtain the user's current cross-category interests and hobbies.

[0111] In video applications on large-screen TV terminals, personalized video recommendation categories include TV series, movies, variety shows, documentaries, animations, etc. However, when user interests are relatively fixed, in the limited number of recommendation positions exposed on the first screen, manual configuration and personalized recommendation results often recommend only 1-2 categories of corresponding media assets. Even if each recommendation position fits the user's interests, there are still problems such as a single recommended media asset category and users repeatedly consuming the same type of films, which will have a negative impact on the utilization rate of media assets and the breadth of user consumption, and thus affect the user's viewing experience.

[0112] In order to avoid recommending a single category of media assets, recommendation algorithms can be used to conduct cross-category interest exploration. Typical interest exploration algorithms in the field of recommendation algorithms include: ε-greedy algorithm, Thompson sampling, confidence interval upper bound, and deep model. The core idea is to calculate the next action to maximize the benefit or minimize the regret based on historical behavior. Different directions have their own advantages and disadvantages according to different application scenarios.

[0113] In the large-screen terminal application scenario, since it is necessary to calculate the average return for different categories, it is a good choice to achieve a better balance in the exploration and utilization process, and to use the upper confidence bound (UCB) as the basic algorithm for improvement. For example, user A often watches the fairy-tale costume drama starring Xiao X in the 1 to N-1 viewing history. Conventional recommendation systems or models often infer more similar fairy-tale and costume TV series based on the viewing behavior characteristics of user A through the model. This results in the user receiving more TV drama category media when the user turns on the TV for the Nth time, although the user may have other preferences for the Nth time. In the current recommendation algorithm, the user cannot intuitively feel other types of media, such as variety shows, documentaries, etc.

[0114] In some embodiments, in the large-screen media asset video recommendation scenario, the classic confidence interval upper bound algorithm is a fair experiment for each media asset. The specific method may include: calculating the profit probability of each media asset based on historical user-media asset pairs, and by calculating the subsequent observation probability, there is always a difference between the observation probability and the profit probability. The confidence interval upper bound algorithm optimistically takes the sum of the observation probability and the difference as the expected profit, thereby returning the top-N media assets with the highest profit. The calculation formula of the confidence interval upper bound is as follows:

[0115]

[0116] Among them, i∈k, a total of k user-media pairs; X i is the average revenue of the i user-media resource pairs so far, N is the total number of trials so far, and n i is the number of attempts for the ith user-media pair.

[0117] The above embodiment uses each media asset identifier as a rotation object, calculates the revenue of each media asset through the upper bound of the confidence interval algorithm, and selects the media asset with the highest revenue to recommend to the user. However, due to the wide variety of media assets in video applications, the corresponding number of media assets also reaches tens of thousands, and the amount of calculation of media asset revenue is extremely large, resulting in the inability to accurately and quickly obtain the user's current cross-category interests and hobbies.

[0118] In order to accurately and quickly obtain the user's current cross-category interests and hobbies, the embodiment of the present application provides a server 400, which includes a communication module and a processing module. The communication module is configured to establish a communication connection with the display device 200. The embodiment of the present application further improves some functions of the server 400. Figure 4 As shown, the processing module enables the server 400 to perform the following steps by running the application:

[0119] Step S401: Obtain the number of media asset exposures and the number of media asset clicks.

[0120] When receiving a user input command to start the display device 200, start a media asset application, or switch a media asset page, the display device 200 sends a page acquisition request to the server 400. The server 400 determines the page data based on the page acquisition request and sends the page data to the display device 200. The display device 200 displays the page corresponding to the page data, that is, the media asset recommendation page.

[0121] The page acquisition request may include a page identifier. After the server 400 parses the page acquisition request and obtains the page identifier, it can determine the page requested by the display device 200 according to the page identifier, and determine the page data corresponding to the page identifier, and the page data includes first page data and second page data. The first page data includes media data compiled by the operator and the user's historical viewing media data. The second page data includes media data compiled by the operator and personalized recommended media data. The media recommendation page includes a recommended position and a non-recommended position. The recommended position is used to display personalized recommended media data, and the non-recommended position is used for media data compiled by the operator.

[0122] The page acquisition request may include a personalized recommendation identifier. The server 400 parses the page acquisition request, and after obtaining the personalized recommendation identifier, determines whether the personalized recommendation identifier is a preset value. If the personalized recommendation identifier is a preset value, for example, the personalized recommendation identifier is 1, the second page data corresponding to the page identifier may be sent to the display device 200. If the personalized recommendation identifier is not a preset value, for example, the personalized recommendation identifier is 0, the first page data corresponding to the page identifier may be sent to the display device 200. When the personalized recommendation switch is turned on, the personalized recommendation identifier is set to 1, and when the personalized recommendation switch is turned off, the personalized recommendation identifier is set to 0.

[0123] The second page data is not static. In some embodiments, the page acquisition request includes historical media records. The historical media records include user historical browsing, exposure, click behavior logs and user historical media viewing records. The server 400 parses the page acquisition request to obtain the historical media records. When the personalized recommendation flag is a preset value, the historical media records are input into the personalized recommendation model to obtain recommended media. The second page data is generated based on the recommended media.

[0124] In other embodiments, the page acquisition request includes a user account identifier or a device identifier of the display device 200. The server 400 parses the page acquisition request, obtains the user account identifier or the device identifier of the display device 200, and can obtain the historical media asset record corresponding to the user account identifier or the device identifier of the display device 200, thereby generating the second page data according to the historical media asset record.

[0125] like Figure 5As shown in , the media resources displayed to the user for the first time after the display device 200 is turned on each time may come from the media resources manually arranged by the operation engineer, or from the personalized media resources provided by the system-side recommendation system. Among them, the manually arranged media resources are the focus of recent operation media resources and are non-personalized. The media resources provided by the recommendation system, although personalized, are all based on the user's historical behavior and interests. The results given by conventional intelligent recommendation algorithms such as recall-sort-rearrangement can better fit the user's historical interests. However, when the user turns on the device and browses the video platform this time, the current recommendation algorithm cannot capture in real time whether the user's interests have changed and where the focus is after the change. This results in that once the user's viewing interest changes, the personalized media resources displayed to the user are still based on past experience, which will lead to a poor user experience.

[0126] In order to update personalized media assets in a timely manner, the server needs to obtain the number of media asset exposures and media asset clicks in a timely manner. The specific implementation method may include: receiving the user behavior log sent by the display device, and then counting the media asset exposure times and media asset click times based on the media assets and behaviors in the user behavior log. Behaviors include exposures and clicks.

[0127] When displaying the media asset recommendation page, the display device 200 receives the user's confirmation operation on the media asset control in the media asset recommendation page, and sends a page request for obtaining the media asset details page corresponding to the media asset control to the server 400. After receiving the page request, the server 400 sends the page data of the media asset details page, so that the display device 200 displays the media asset details page. When sending the page request, the display device 200 generates a user behavior log of clicking the media asset control, and sends the user behavior log to the server 400.

[0128] When displaying the media asset recommendation page, the display device 200 generates a user behavior log of browsing or exposure of all media asset controls in the current media asset recommendation page, and sends the user behavior log to the server 400. When receiving the position of the focus of the media asset recommendation page moved by the user and the media asset recommendation page is turned or updated, a user behavior log of browsing or exposure of the media asset controls in the current media asset recommendation page or the newly displayed media asset controls can be generated and sent to the server 400.

[0129] The difference between media asset control browsing and exposure is that media asset control browsing means that the media asset control is displayed on the current media asset recommendation page, while media asset control exposure means that the duration of the media asset control displaying on the current media asset recommendation page reaches a preset time. In some embodiments, media asset control browsing can be equivalent to media asset control exposure, that is, only one of media asset exposure and browsing can be obtained.

[0130] User behavior logs include user ID, media asset ID, behavior ID, and session ID. User ID is used to represent the user identity. Media asset ID is used to represent the media asset corresponding to the exposure or click of the media asset control. Behavior ID is used to represent the user's behavior, such as browsing, exposure, and clicking. Session ID is used to determine whether at least two behavior logs belong to the same startup behavior log.

[0131] In some embodiments, after generating the user behavior log, the display device 200 performs data cleansing on the user behavior log and sends the cleaned user behavior log to the server 400 . The server 400 counts the number of media asset exposures and the number of media asset clicks based on the user behavior log.

[0132] In other embodiments, after generating the user behavior log, the display device 200 directly sends the user behavior log to the server 400. The server 400 first cleans the user behavior log, and then counts the number of media asset exposures and the number of media asset clicks based on the user behavior log.

[0133] In some embodiments, the media asset recommendation page may include non-media asset controls. If the user selects a non-media asset control, the user behavior log has no effect on the media asset recommendation. Therefore, the user behavior log needs to be deleted through data cleaning. A specific implementation of data cleaning may include: determining whether the control selected by the user in the user behavior log is a media asset control. If the control selected by the user in the user behavior log is a media asset control, retaining the user behavior log. If the control selected by the user in the user behavior log is not a media asset control, deleting the user behavior log.

[0134] In other embodiments, the media asset controls in the media asset recommendation page may correspond to media asset categories that do not need to be recommended. For example, media assets related to education do not need to be recommended based on the user behavior log. Therefore, it is necessary to delete the user behavior log through data cleaning. Another specific implementation method of data cleaning may include: determining whether the media asset category corresponding to the media asset identifier is the target category. If the media asset category corresponding to the media asset identifier is the target category, deleting the user behavior log. If the media asset category corresponding to the media asset identifier is not the target category, retaining the user behavior log. Exemplarily, the target category is education.

[0135] In some embodiments, after receiving the user behavior log sent by the display device 200 , the number of media asset exposures and the number of media asset clicks can be directly counted according to the user behavior log.

[0136] In other embodiments, the session identifier can be used to determine whether the user behavior log is generated at the same startup. Each time the display device 200 is powered on, a new session identifier is generated to distinguish it from the session identifier of the last startup. The session identifiers of all user behavior logs uploaded at this startup are the same.

[0137] After receiving the user behavior log sent by the display device 200, determine whether the session identifier of this user behavior log is the same as the session identifier of the previous user behavior log. If the session identifier of this user behavior log is the same as the session identifier of the previous user behavior log, then the number of media asset exposures and the number of media asset clicks are counted according to the user behavior log; if the session identifier of this user behavior log is different from the session identifier of the previous user behavior log, then delete the historical data saved before this user behavior log, that is, the number of media asset exposures and the number of media asset clicks are reset to zero, and the number of media asset exposures and the number of media asset clicks are re-counted according to this user behavior log. The embodiment of the present application can ensure that only the user behavior log generated by this startup is used as the basic data for the current media asset recommendation, and will not be interfered by the user's usage behavior of the previous time.

[0138] The number of exposures and clicks of the media assets can be counted by reading all the media assets and their corresponding behaviors in the saved user behavior logs each time the user behavior log uploaded by the display device 200 is obtained, and then counting the number of exposures and clicks of each media asset. The number of exposures and clicks of the media assets can also be counted by obtaining a user behavior log once, saving the number of exposures and clicks of each media asset corresponding to the current user behavior log, and when obtaining a new user behavior log, adding the number of media asset exposures and clicks of the media asset and its behavior in the new user behavior log to the number of media asset exposures and clicks that have been saved.

[0139] Step S402: converting the number of media asset exposures into the number of label exposures of the media asset corresponding to the media asset label, and converting the number of media asset clicks into the number of label clicks of the media asset corresponding to the media asset label.

[0140] like Figure 6 As shown, a specific implementation method for determining the number of label exposures and the number of label clicks of a media asset label may include the following steps:

[0141] Step S601: Obtain the media asset identifier in the user behavior log.

[0142] Among them, the media asset identifier is used to represent the media asset that has been browsed, exposed or clicked. According to the media asset identifiers of exposure and click, the media assets can be divided into positive feedback media assets and negative feedback media assets, and distinguished by feedback identifiers. The media asset corresponding to the media asset identifier that is clicked after exposure is the positive feedback media asset, and the media asset corresponding to the media asset identifier that is not clicked after exposure is the negative feedback media asset.

[0143] Step S602: Obtain at least one media asset tag corresponding to the media asset identifier in the media asset tag dimension table.

[0144] The media asset label dimension table includes the mapping relationships between multiple media asset identifiers and media asset labels. One media asset identifier can correspond to one or more media asset labels. In the media asset label dimension table, one or more corresponding labels can be obtained according to the media asset identifier.

[0145] Exemplarily, the media asset label dimension table is shown in Table 1. In Table 1, the labels corresponding to the media asset identifier 04 are Movie - Workplace and Inspirational respectively.

[0146] Table 1

[0147] Media Name Media asset identification Tag 1 Tag 2 Tag N Media A 01 TV series-Romance Palace Fight …… Media Resources B 02 Movie-Romance Costume …… Media Resources C 03 Animation-Sports basketball …… Media Resources 04 Movie-Workplace Inspirational …… …… …… …… …… ……

[0148] Step S603: Convert the media asset exposure times into the label exposure times of the media asset corresponding media asset labels, and convert the media asset click times into the label click times of the media asset corresponding media asset labels. Among them, the label exposure times are the same as the media asset exposure times, and the media asset click times are the same as the label click times.

[0149] Exemplarily, if the exposure times of media asset A are 10 times and the click times are 2 times, then the exposure times of the media asset label "TV Drama - Love" are 10 times and the click times are 2 times, and the exposure times of the media asset label "Palace Drama" are 10 times and the click times are 2 times.

[0150] In some embodiments, the server can receive the user behavior logs uploaded by multiple users. In order to distinguish users, the media asset labels can be represented in the form of user - label pairs, that is, the user identifier in the user behavior log and at least one media asset label can be combined to form at least one user - label pair. Hereinafter, the media asset labels can be equivalent to the user - label pairs. For example, calculating the revenue of the media asset label can be equivalent to calculating the revenue of the user - label pair.

[0151] If the label corresponding to the media asset identifier is one, one user - label pair can be formed. If the label corresponding to the media asset identifier is three, three user - label pairs can be formed.

[0152] Exemplarily, the user identifier is user1, and the labels corresponding to the media asset identifier 04 are Movie - Workplace and Inspirational respectively, then the user - label pairs that can be formed are <user1 - Movie - Workplace> and <user1 - Inspirational>.

[0153] Step S403: Calculate the label revenue according to the label exposure times and the label click times.

[0154] As Figure 7 shown, the steps of calculating the label revenue according to the label exposure times and the label click times can include:

[0155] Step S701: Determine the ratio of the label click times to the label exposure times as the average revenue of the media asset label.

[0156] The ratio of the number of clicks on the media asset tag to the number of exposures of the media asset tag is determined as the average revenue of the media asset tag, that is, the average revenue X i = Number of clicks a / Number of exposures b.

[0157] It should be noted that if multiple exposure media assets correspond to the same media asset tag, then the exposure count of the media asset tag is the sum of the exposure counts of the multiple exposure media assets. If multiple click media assets correspond to the same media asset tag, then the click count of the media asset tag is the sum of the click counts of the multiple click media assets.

[0158] For example, the exposure times of media asset A1 are 10 times, the click times are 3 times, and the media asset tags are tag 1 and tag 2. The exposure times of media asset A2 are 3 times, the click times are 1 time, and the media asset tags are tag 1 and tag 3. The exposure times of tag 1 are 13 times, and the click times are 4 times.

[0159] Step S702: Calculate the basic revenue of the media asset tag according to the average revenue, the number of recommendations and the total number of recommendations.

[0160] Among them, the number of recommendations is the number of times the target media asset is sent to the display device when the target media asset corresponds to the tag, that is, the target tag is a media asset tag; the total number of recommendations is the total number of times the target media asset is sent to the display device; the basic revenue is used to represent the basic probability of being clicked by the user after the media asset corresponding to the media asset tag is exposed.

[0161] After the target media asset is sent to the display device 200, the media asset tag corresponding to the target media asset is obtained, and the number of attempts corresponding to the media asset tag, that is, the number of recommendations, is increased by 1. When calculating the basic income of the media asset tag, the number of recommendations of the media asset tag n can be directly obtained. i Among them, in the same round of recommendation data, if there are multiple target data corresponding to the same media asset label, the number of attempts for the media asset label is only increased by 1.

[0162] After a round of target media assets is sent to the display device 200, the total number of attempts up to now, that is, the total number of recommendations, is increased by 1. The total number of recommendations N can be directly obtained when calculating the basic income of the media asset tag.

[0163] Basic income UCB-tag i The calculation formula of ' is as follows:

[0164]

[0165] Step S703: Calculate weighted benefits according to media asset tags, similar tags, and heterogeneous tags.

[0166] Among them, similar tags are media asset tags with similar attributes as media asset tags in positive feedback tags, different tags are media asset tags with different attributes as media asset tags in positive feedback tags, positive feedback tags are media asset tags corresponding to media asset identifiers that are clicked after exposure, and weighted benefits are used to represent the probability of adding or reducing weights of media asset tags corresponding to users after exposure.

[0167] In some embodiments, the media asset tag includes a combination tag of a primary classification tag and a secondary classification tag. The primary classification tag can be used to characterize the type or category of the media asset. For example, the primary classification tags include TV series, movies, variety shows, documentaries, and animations. The secondary classification tags can be used to characterize the subject matter or content of the media asset. For example, the secondary classification tags include romance, science fiction, sports, and talk shows. The combination tag is a primary classification tag-secondary classification tag. For example, movies-romance, variety shows-talk shows, TV series-science fiction, etc.

[0168] For the combined tags, a specific implementation of determining the same type of tags and different type of tags of the media asset tags may include:

[0169] Determine whether the positive feedback tag is the same as the primary classification tag or the secondary classification tag of the media asset tag. If the positive feedback tag is the same as the primary classification tag or the secondary classification tag of the media asset tag, the positive feedback tag is determined to be a similar tag. If the positive feedback tag is different from both the primary classification tag and the secondary classification tag of the media asset tag, the positive feedback tag is determined to be a different tag.

[0170] For example, if the media asset tag is TV series-romance, then movie-romance and TV series-suspense belong to the same category, and variety show-talk show belongs to a different category.

[0171] Since the amount of media assets on the media asset platform is in the tens of thousands, it is extremely computationally intensive to continuously display each media asset identifier as a rotating object to the user. The embodiment of the present application uses the first and second level classification labels of media assets as explorable objects, takes advantage of the rich media asset labels in the media asset platform, and divides the same type of media assets into the same category, for example: movies-romance, variety shows-talk shows, TV series-science fiction (first level classification-second level classification), etc. After the media asset label division, the five major categories of TV series, movies, variety shows, documentaries, and animation can eventually be divided into 135 categories of labels, that is, the exploratory objects are reduced from tens of thousands to hundreds of objects, reducing the computational complexity of the explorable object benefits, and helping to quickly obtain media assets of cross-category interests and push them to users.

[0172] In some embodiments, the media asset tag includes a single tag, for example, the media asset tag is palace fighting.

[0173] For a single tag, a specific implementation of determining the same type of tags and different types of tags may include:

[0174] Obtain a list of similar tags corresponding to the media asset tag, the list of similar tags including at least one similar tag. Determine whether the positive feedback tag is the same as the similar tag. If the positive feedback tag is the same as the similar tag, determine the positive feedback tag as the similar tag. If the positive feedback tag is not the same as the similar tag, determine the positive feedback tag as a different tag.

[0175] For example, the media asset tag is palace fighting, and the list of similar tags corresponding to the palace fighting tag includes ancient costumes and history, etc. Ancient costumes and history tags belong to the same category, and modern tags belong to different categories.

[0176] The specific implementation method of calculating the weighted benefit based on the media asset tags, similar tags and heterogeneous tags may include: first, obtaining the first media asset vector of the media asset group corresponding to the media asset tag, the second media asset vector of the media asset group corresponding to the similar tag and the third media asset vector of the media asset group corresponding to the heterogeneous tag. Secondly, determining the average vector of the first media asset vector as the media asset tag vector, determining the average vector of the second media vector as the similar tag vector, and determining the average vector of the third media vector as the heterogeneous tag vector. Thirdly, calculating the first similarity between the media asset tag vector and the similar tag vector, and the second similarity between the media asset tag vector and the heterogeneous tag vector. Finally, calculating the weighted benefit based on the first similarity and the second similarity.

[0177] The media asset group corresponding to the media asset label includes all the first media assets in the media asset library that contain the media asset label. The first media asset is input into the deep learning model to obtain a first media asset vector. The media asset group corresponding to the same type of label includes all the second media assets in the media asset library that contain the same type of label. The second media asset is input into the deep learning model to obtain a second media asset vector. The media asset group corresponding to the heterogeneous label includes all the third media assets in the media asset library that contain heterogeneous labels. The third media asset is input into the deep learning model to obtain a third media asset vector. The deep learning model can capture the complex features in the media asset content and represent it as a vector of fixed length, namely a media asset vector. There are many ways to obtain media asset vectors, which are not required by this application.

[0178] The average vectors of the first media asset vector, the second media asset vector, and the third media asset vector are calculated respectively, and the average vectors are used as label vectors of the media asset labels, the same type labels, and the different type labels.

[0179] It should be noted that when the media assets in the media asset library are not updated, the label vector of each label remains unchanged. Therefore, the label vector of each label can be calculated in advance and directly obtained when needed. When the media assets in the media asset library are updated, the label vector of each label is updated.

[0180] By calculating the vector distance between the same or different labels, that is, the L2-norm distance (Euclidean distance), the similarity between the same or different labels is obtained, and the weighted benefit is calculated based on the similarity.

[0181] Weighted income UCB-tag i The calculation formula of ” is as follows:

[0182]

[0183] Among them, i is the media resource label, E i is the label vector of media resource label i, E i+ is the same label vector, E i- is the heterogeneous label vector, E g is the label vector of the positive feedback label, g∈[1,G] and g≠i, d is the L2-norm distance between label vectors.

[0184] Step S704: the sum of the basic revenue and the weighted revenue is determined as the label revenue of the media asset label.

[0185] Tag income UCB-tag i =UCB-tag i '+UCB-tag i ”.

[0186] It should be noted that when the total number of recommendations is 1, the weighted benefit is the initialized random number δ.

[0187] Tag income UCB-tag i The calculation formula is as follows:

[0188]

[0189] In the present application, when calculating the upper bound of the confidence interval, the similar and different attributes between different categories are taken into consideration. When calculating the revenue of media asset tags, the similarities and differences between tag attributes are weighted differently, which helps to improve the accuracy of calculating the revenue of media asset tags.

[0190] Step S404: determine the media asset tag corresponding to the tag revenue of the first preset ranking with the highest tag revenue as the target tag.

[0191] After calculating the label revenue of each media asset label, the label revenues are sorted from large to small, and the media asset labels ranked in the top preset positions, that is, the media asset labels with label revenue TOP-N, are determined as target labels.

[0192] Exemplarily, the preset ranking is the top five, and the media asset tags ranked in the top five in terms of tag revenue are determined as target tags.

[0193] Step S405: sending the target media asset corresponding to the target tag to the display device, so that the display device displays the target media asset on the media asset recommendation page.

[0194] After the target tag is determined, the media asset corresponding to the target tag may be added to the media asset candidate set as the target media asset.

[0195] In some embodiments, a specific implementation of determining the media asset corresponding to the target tag as the target media asset includes: determining the media asset with the largest number of matches with the target tag as the target media asset.

[0196] If there are multiple target media assets in the media asset candidate set, the matching numbers may be sorted from large to small, and the media assets ranked in a preset position may be taken as target media assets and added to the media asset candidate set.

[0197] For example, the target tags are <TV series-romance>, <movie-romance>, <ancient costume>, <palace fighting>, and <funny>. The number of tags matching the target tags of media asset 1 is 4, the number of tags matching the target tags of media asset 2 is 2, and the number of tags matching the target tags of media asset 3 is 3. If there are two positions in the media asset candidate set, media asset 1 and media asset 3 are added to the media asset candidate set.

[0198] In some other embodiments, another specific implementation of determining the media asset corresponding to the target tag as the target media asset includes: determining the media asset whose number of matches with the target tag reaches a preset number as the target media asset.

[0199] For example, the preset number is 3, and the target tags are <TV series-love>, <movie-love>, <costume>, <palace fighting>, and <funny>. The number of tags matching the target tags of media asset 1 is 4, the number of tags matching the target tags of media asset 2 is 2, and the number of tags matching the target tags of media asset 3 is 1. Then media asset 1 is added as the target media asset to the media asset candidate set.

[0200] After receiving the target media asset, the display device 200 may randomly replace the content displayed in one or more recommendation positions in the media asset recommendation page.

[0201] The solution provided in the embodiment of the present application can be used to solve the problem of cross-category interest exploration by combining the user's real-time interactive behavior sequence, introducing an interest exploration method, and obtaining a user's real-time interest matrix. The real-time interest matrix of each user is further decoupled, and based on the UCB algorithm, the user's cumulative revenue in different categories is determined according to the upper bound of the confidence interval of different categories in the large-screen cross-category video recommendation, thereby determining the categories that the user can currently explore, and achieving the purpose of exploring the user's cross-category interests.

[0202] In some embodiments, the calculation of cross-category interests and the trial of returning media assets are realized by receiving user behavior logs. However, from the perspective of how the user of the display device 200 perceives the changes, it is necessary to update the recommendation engine results and refresh the terminal cache to achieve this. Since the number of recommendation positions displayed to the user by the display device 200 each time is in the range of 6-10, if the media assets in the recommendation positions do not change, it is necessary to rely on the user's active button behavior to expose more recommendation positions to achieve the purpose of display and exposure. If the user does not actively press the button to expose more recommendation positions, the media asset recommendation page cannot be updated in time, that is, the purpose of rotating the media assets in the recommendation positions cannot be achieved.

[0203] In order to achieve the effect of timely updating the media recommendation page, the server 400 can determine whether the media in the recommendation position needs to be rotated according to the user's interest in the displayed media, and agree with the display device 200 that the server 400 will inform the display device 200 whether the cache needs to be refreshed, thereby achieving the purpose of controlling the rotation of the media in the recommendation position of the display device 200.

[0204] Whether a trial is needed is determined based on the similarity between the result of the current media asset trial and the result of the last time sent and displayed to the user, that is, whether the recommended media asset in the media asset candidate set is sent to the display device 200. The Gaussian clustering algorithm can be used to calculate the similarity.

[0205] like Figure 8 As shown, the specific implementation of determining whether to try may include:

[0206] Step S801: Calculate the target media asset vector of the target media asset, and obtain the cached media asset vector of the delivered media asset.

[0207] The target media asset vector can be obtained by inputting the target media asset into the deep learning model. The target media asset vector is also calculated and saved in advance, and can be directly obtained when needed. The target media asset vector is the media asset vector of this trial result, denoted by I e =[I1, I2, …, I m ]. The delivered media asset refers to the target media asset delivered last time, which needs to be cached by the recommendation engine in advance. The media asset vector of the cached result is recorded as I o =[I1,I2,…,In].

[0208] Step S802: input the target media asset vector into the Gaussian clustering model to obtain the first number of cluster centers.

[0209] Step S803: input the target media asset vector and the cached media asset vector into the Gaussian clustering model to obtain the second number of cluster centers.

[0210] Step S804: Calculate the difference between the number of the second cluster centers and the number of the first cluster centers.

[0211] Step S805: Determine whether the difference is greater than or equal to 1.

[0212] If the difference is greater than or equal to 1, step S806 is executed: sending the target media asset to the display device.

[0213] If the difference is less than 1, step S807 is executed: the target media asset is not sent to the display device.

[0214] For example, consider cross-category interests of 5 categories, that is, the maximum number of cluster centers is 5. o The media vector is introduced into the Gaussian clustering model to obtain X cluster centers, and then I e and I o Directly or randomly scatter them into the Gaussian clustering model to obtain Y cluster centers, and calculate the difference of YX. When the difference is ≥ 1, it is determined that there is a new cross-category interest that needs to be explored, so the target media asset and the refresh cache identifier are transparently transmitted to the display device 200. After receiving the refresh cache identifier, the display device 200 replaces the displayed media asset according to the target media asset, thereby achieving the purpose of cross-category interest exploration.

[0215] In some embodiments, the flowchart of the media resource recommendation method can be as follows: Fig. 9 As shown. The server 400 continuously receives the user's real-time behavior log, processes the data and constructs samples by associating media asset tags, and inputs the samples into the cross-category interest trial model to obtain the cross-category trial media asset vector. The Gaussian clustering model is introduced to cluster the cross-category trial media asset vector with the issued media asset vector displayed by the display device 200, and determine whether a trial is needed, that is, a cache refresh, based on the number of cluster centers. If no trial is made, the user behavior log continues to be received. If a trial is made, the trial media asset is sent to the display device 200 through the recommendation engine, and at the same time, the issued media asset vector is cached, ultimately achieving the purpose of real-time control of the user's cross-category interest trial. The solution provided in the embodiment of the present application reasonably controls the display device 200 to refresh the media asset cache mechanism, so that when the media asset browsed by the user does not meet the current interest during use, the cross-category media asset is promptly responded to and refreshed to timely reach the user's interest, thereby improving the user's viewing experience.

[0216] In some embodiments, the timing diagram of the media resource recommendation method may be as follows: Fig.10As shown. The display device 200 receives the power-on command input by the user, and the display device 200 starts. The display device 200 sends a page acquisition request to the server 400. The page acquisition request includes a page identifier and a personalized recommendation identifier. The server 400 determines whether the personalized recommendation is turned on based on the personalized recommendation identifier. When the personalized recommendation is not turned on, the server 400 sends a backup result to the display device 200, that is, only the user's viewing history + the operator's manually arranged media resources are sent. When the personalized recommendation is turned on, the recommendation engine of the server 400 determines the personalized recommended media resources based on the user's historical viewing records, etc., and sends the personalized recommended media resources to the display device 200. The display device 200 displays the media resource recommendation page based on the page data sent by the server 400.

[0217] After receiving the user operation, the display device 200 generates a user behavior log and sends the user behavior log to the server 400. The server 400 constructs a label sample based on the user behavior log associated label, that is, the number of media asset clicks and media asset exposures are counted according to the user behavior log, so as to determine the number of exposures and clicks of the media asset label or user-label pair corresponding to the media asset. The cross-category interest trial model is used to calculate the media asset vector for this trial. Among them, the cross-category interest trial model first calculates the basic income of the media asset label or user-label pair, and then uses the similarities and differences between the label attributes to add and reduce weights to calculate the weighted income, and finally calculates the label income. Among them, the label income is the sum of the basic income and the weighted income. Sort the label income from large to small, and determine the media asset corresponding to the media asset label or user-label pair ranked in TOP-N as the recommended media asset, that is, the media asset for this trial. Calculate the media asset vector for this trial media asset. Input this trial media asset vector into the Gaussian clustering model to obtain the first number of cluster centers. The current trial media asset vector and the cached media asset vector of the last recommended media asset in the recommendation engine are input into the Gaussian clustering model to obtain the second number of cluster centers. If the difference between the second number of cluster centers and the first number of cluster centers is greater than or equal to 1, the current trial media asset is sent to the display device 200. The display device updates the media asset recommendation page based on the current trial media asset.

[0218] The solution provided by the embodiment of the present application can utilize the browsing, exposure and click behavior logs reported by the display device 200 to obtain the behavior trajectory of the user who is turned on this time. With respect to the consumption of media assets that have been displayed to the user, the probability of the user consuming in different categories is quickly calculated through the UCB-tag algorithm, and the TOP-N cross-category consumer interests with the largest cumulative revenue are passed to the sorting model through label recall, so as to realize the issuance of a new round of media assets by the recommendation system. Finally, through the issuance of media asset updates by the recommendation system and the cache refresh mechanism of the display device 200, the original media assets that have been displayed to the user and in which the user is not interested are quickly replaced with the latest recommendation results, and the media assets recalled by cross-category consumer interests are used for display during the user's pull-down process, so as to capture the interests of this user more quickly and accurately, and tentatively issue cross-category media assets.

[0219] Some embodiments of the present application also provide a computer-readable storage medium, which may store a program. When the computer storage medium is configured in a display device or a server, the program may include the program steps involved in the media resource recommendation method in the above embodiment when the program is executed. Among them, the computer storage medium may be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0220] The embodiment of the present application provides an electronic device, which includes: a processor and a memory for storing instructions executable by the processor. The processor is used to read the executable instructions from the memory and execute the instructions to implement the media resource recommendation method in the above embodiment.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0222] For ease of explanation, the above description has been made in conjunction with specific embodiments. However, the above exemplary discussion is not intended to be exhaustive or limit the embodiments to the specific forms disclosed above. Based on the above teachings, various modifications and variations can be obtained. The selection and description of the above embodiments are intended to better explain the principles and practical applications, so that those skilled in the art can better use the embodiments and various different variations of the embodiments suitable for specific use considerations.

Claims

1. A server, characterized in that: include: A communication module, configured to establish a communication connection with a display device; The processing module is configured as follows: Get the number of media exposures and media clicks; Convert the media asset exposure times into the label exposure times of the media asset label corresponding to the media asset, and convert the media asset click times into the label click times of the media asset label corresponding to the media asset, wherein the label exposure times are the same as the media asset exposure times, and the media asset click times are the same as the label click times; Calculate the label revenue according to the number of times the label is exposed and the number of times the label is clicked, where the label revenue is used to represent the predicted probability of the media asset corresponding to the media asset label being clicked by the user after being exposed; The media asset tag corresponding to the tag revenue of the first preset ranking with the highest tag revenue is determined as the target tag; The target media asset corresponding to the target tag is sent to the display device, so that the display device displays the target media asset on a media asset recommendation page.

2. The server according to claim 1, characterized in that: The processing module is further configured to obtain the number of media asset exposures and the number of media asset clicks: Receiving a user behavior log sent by the display device; According to the media assets and behaviors in the user behavior log, the number of media asset exposures and the number of media asset clicks of the media assets are counted, and the behaviors include exposures and clicks.

3. The server according to claim 1, characterized in that: The processing module is further configured to calculate the tag revenue according to the number of tag exposures and the number of tag clicks: Determine the ratio of the number of times the tag is clicked to the number of times the tag is exposed as the average revenue of the media asset tag; The basic income of the media asset tag is calculated according to the average income, the number of recommendations and the total number of recommendations, the number of recommendations being the number of times the target media asset is sent to the display device when the target media asset corresponding label is the media asset tag, the total number of recommendations being the total number of times the target media asset is sent to the display device, and the basic income is used to represent the basic probability of the media asset corresponding to the media asset tag being clicked by the user after being exposed; Calculate the weighted benefit according to the media asset tag, similar tags and heterogeneous tags, wherein the similar tags are media asset tags with similar attributes to the media asset tag in the positive feedback tags, and the heterogeneous tags are media asset tags with heterogeneous attributes to the media asset tag in the positive feedback tags. The positive feedback tags are media asset tags corresponding to the media asset identifiers clicked after exposure, and the weighted benefit is used to characterize the probability of adding or reducing weights of the media asset corresponding to the media tag being clicked by the user after exposure; The sum of the basic income and the weighted income is determined as the label income of the media asset label.

4. The server according to claim 3, characterized in that: The processing module is further configured to calculate the weighted benefit according to the media asset tags, the same tags and the different tags: Obtaining a first media asset vector of a media asset group corresponding to the media asset tag, a second media asset vector of a media asset group corresponding to the same type of tag, and a third media asset vector of a media asset group corresponding to the different type of tag; Determine the average vector of the first media asset vector as a media asset label vector, determine the average vector of the second media asset vector as a homogeneous label vector, and determine the average vector of the third media asset vector as a heterogeneous label vector; Calculating a first similarity between the media asset label vector and the same type label vector, and a second similarity between the media asset label vector and the different type label vector; The weighted benefit is calculated according to the first similarity and the second similarity.

5. The server according to claim 1, characterized in that: The processing module executes sending the target media asset corresponding to the target tag to the display device, and is further configured to: Sending the target media asset with the largest number of matches to the target tag to the display device; or, The target media assets whose matching quantity with the target tag reaches a preset quantity are sent to the display device.

6. The server according to claim 1, characterized in that: The processing module executes sending the target media asset corresponding to the target tag to the display device, and is further configured to: Calculating a target media asset vector of the target media asset, and obtaining a cached media asset vector of the delivered media asset; Inputting the target media asset vector into a Gaussian clustering model to obtain the number of first cluster centers; Inputting the target media asset vector and the cached media asset vector into the Gaussian clustering model to obtain a second number of cluster centers; Calculating the difference between the number of the second cluster centers and the number of the first cluster centers; If the difference is greater than or equal to 1, the target media asset corresponding to the target tag is sent to the display device.

7. A display device, characterized in that: include: monitor; is configured to display a media asset recommendation page; A communication device configured to establish a communication connection with a server A controller is coupled to the display and is configured to: Receive a target media asset sent by the server, wherein the target media asset is the media asset corresponding to a target tag, and the target tag is the media asset tag corresponding to the tag revenue of the top preset ranking with the highest tag revenue, and the tag revenue is used to characterize the predicted probability of the media asset corresponding to the media asset tag being clicked by the user after being exposed, and the tag revenue is calculated based on the tag exposure times and the tag click times of the media asset tag; the tag exposure times are converted from the media asset exposure times corresponding to the media asset tag, and the tag click times are converted from the media asset click times corresponding to the media asset tag, and the tag exposure times are the same as the media asset exposure times, and the media asset click times are the same as the tag click times; The display is controlled to display the target media asset in the media asset recommendation page.

8. A media resource recommendation method, characterized in that: Applicable to servers, including: Get the number of media exposures and media clicks; Convert the media asset exposure times into the label exposure times of the media asset label corresponding to the media asset, and convert the media asset click times into the label click times of the media asset label corresponding to the media asset, wherein the label exposure times are the same as the media asset exposure times, and the media asset click times are the same as the label click times; Calculate the label revenue according to the number of times the label is exposed and the number of times the label is clicked, where the label revenue is used to represent the predicted probability of the media asset corresponding to the media asset label being clicked by the user after being exposed; The media asset tag corresponding to the tag revenue of the first preset ranking with the highest tag revenue is determined as the target tag; The target media asset corresponding to the target tag is sent to a display device, so that the display device displays the target media asset on a media asset recommendation page.

9. The method according to claim 8, characterized in that The step of obtaining the number of media asset exposures and the number of media asset clicks includes: Receiving a user behavior log sent by the display device; According to the media assets and behaviors in the user behavior log, the number of media asset exposures and the number of media asset clicks of the media assets are counted, and the behaviors include exposures and clicks.

10. A media resource recommendation method, characterized in that: Applied to display devices, including: Receive target media assets sent by a server, the target media assets are media assets corresponding to target tags, the target tags are media asset tags corresponding to the tag revenue of the top preset ranking with the highest tag revenue, the tag revenue is used to characterize the predicted probability of the media asset tag corresponding to the media asset tag being clicked by the user after exposure, and the tag revenue is calculated based on the tag exposure times and tag click times of the media asset tag; the tag exposure times are converted from the media asset exposure times corresponding to the media asset tag, the tag click times are converted from the media asset click times corresponding to the media asset tag, the tag exposure times are the same as the media asset exposure times, and the media asset click times are the same as the tag click times; The display is controlled to display the target media asset in the media asset recommendation page.

Images