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

Abstract

The invention relates to a video recommendation method and device, electronic equipment and a readable storage medium, and belongs to the technical field of computers. The method comprises the following steps: acquiring user feature data of a target user, video feature data of a first video and video feature data of each second video; optimizing a pre-trained video recommendation model based on the user feature data, the video feature data of the first video and the video feature data of each second video, and outputting a user behavior score corresponding to each second video based on the optimized video recommendation model; and according to the user behavior score corresponding to each second video, determining a recommended video corresponding to the second-level page in each second video. By adopting the method and the device, the video page view in the second-level page can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a video recommendation method, device, electronic device and readable storage medium. Background Art

[0002] At present, the architecture of short video platforms usually includes a first-level page and a second-level page. The first-level page is the homepage of the short video platform. The short video platform displays video content that can quickly attract users' attention on the first-level page. The second-level page is the video details page, which is entered by jumping from the video link on the first-level page. The short video platform displays the complete video content on the second-level page, and further displays other video content based on user operations (such as sliding down).

[0003] However, since the video recommendation model used by short video platforms on secondary pages mainly focuses on users' general interests (i.e., what types of videos users may be interested in), it does not take into account users' prior interests represented by the video content clicked by users on the primary page (i.e., the types of videos users currently want to watch). This may result in users finding that the video content does not meet their expectations after clicking on the video content on the secondary page. For example, a user enters a secondary page from the entertainment primary page, but the secondary page recommends military videos. The user will quickly exit the video content, which in turn leads to lower video views on the secondary page. Summary of the invention

[0004] Based on this, it is necessary to provide a video recommendation method, device, electronic device and readable storage medium to address the above technical issues.

[0005] In a first aspect, a video recommendation method is provided, the method comprising:

[0006] Obtain user feature data of a target user, video feature data of a first video, and video feature data of each second video; wherein the first video is a video that the target user browses on a primary page, and the second video is a video to be recommended corresponding to a secondary page, and the secondary page is a page that is triggered by the first video;

[0007] Optimizing a pre-trained video recommendation model based on the user feature data, the video feature data of the first video, and the video feature data of each of the second videos, and outputting a user behavior score corresponding to each of the second videos based on the optimized video recommendation model;

[0008] According to the user behavior score corresponding to each of the second videos, a recommended video corresponding to the secondary page is determined in each of the second videos.

[0009] As an optional implementation manner, optimizing a pre-trained video recommendation model based on the user feature data, the video feature data of the first video, and the video feature data of each of the second videos, and outputting a user behavior score corresponding to each of the second videos based on the optimized video recommendation model includes:

[0010] Inputting the user feature data and the video feature data of the first video into a pre-trained network parameter optimization model, and outputting network parameter weights of the pre-trained video recommendation model;

[0011] The video recommendation model is subjected to network parameter optimization processing based on the network parameter weights to obtain an optimized video recommendation model, and the user feature data and the video feature data of each second video are input into the optimized video recommendation model to output a user behavior score corresponding to each second video.

[0012] As an optional implementation manner, the network parameter weights include network parameter weights corresponding to each rectified linear unit in the video recommendation model;

[0013] The performing network parameter optimization processing on the video recommendation model based on the network parameter weight to obtain an optimized video recommendation model includes:

[0014] For each corrected linear unit in the video recommendation model, the product of the current network parameter corresponding to the corrected linear unit and the network parameter weight corresponding to the corrected linear unit is determined as the optimized network parameter corresponding to the corrected linear unit to obtain the optimized video recommendation model.

[0015] As an optional implementation manner, before inputting the user feature data and the video feature data of each second video into the optimized video recommendation model, the method further includes:

[0016] Acquire first characteristic data of the target user; wherein the first characteristic data is characteristic data obtained through function interaction;

[0017] The step of inputting the user feature data and the video feature data of each second video into the optimized video recommendation model comprises:

[0018] The user feature data, the first feature data, and the video feature data of each of the second videos are input into the optimized video recommendation model.

[0019] As an optional implementation manner, determining, in each second video, a recommended video corresponding to the secondary page according to the user behavior score corresponding to each second video includes:

[0020] Among the second videos, a preset number of the second videos are selected in descending order of the user behavior scores and determined as the recommended videos corresponding to the secondary page.

[0021] As an optional implementation, the method further includes:

[0022] Acquire training sample data and label data; wherein the training sample data includes user feature data of a sample user, video feature data of a third video, and video feature data of a fourth video, and the label data includes user behavior data of the fourth video, the third video is a video that the sample user browses on the primary page, and the fourth video is a video that the sample user continues to browse on the secondary page after jumping to the secondary page based on the third video;

[0023] The initial network parameter optimization model and the initial video recommendation model are jointly trained according to the training sample data and the label data to obtain a trained network parameter optimization model and a trained video recommendation model.

[0024] As an optional implementation, the training sample data also includes second feature data; wherein the second feature data is feature data obtained through functional interaction when the sample user browses the fourth video in the secondary page.

[0025] As an optional implementation manner, the obtaining of training sample data and label data includes:

[0026] In response to a trigger operation of the sample user on the fourth video in the secondary page, obtaining user feature data of the sample user, the second feature data, and video feature data and user behavior data of the fourth video;

[0027] According to the user ID of the sample user and the video ID of the fourth video, the third video belonging to the first-level page and associated with the user ID and the video ID of the fourth video is searched in the video browsing log of the sample user.

[0028] In a second aspect, a video recommendation device is provided, the device comprising:

[0029] A first acquisition module is used to acquire user feature data of a target user, video feature data of a first video, and video feature data of each second video; wherein the first video is a video browsed by the target user on a primary page, and the second video is a video to be recommended corresponding to a secondary page, and the secondary page is a page to which the first video triggers a jump;

[0030] a processing unit, configured to optimize a pre-trained video recommendation model based on the user feature data, the video feature data of the first video, and the video feature data of each of the second videos, and output a user behavior score corresponding to each of the second videos based on the optimized video recommendation model;

[0031] A determination module is used to determine, in each of the second videos, a recommended video corresponding to the secondary page according to the user behavior score corresponding to each of the second videos.

[0032] As an optional implementation, the processing unit includes a first output unit and a second output unit;

[0033] A first output module, configured to input the user feature data and the video feature data of the first video into a pre-trained network parameter optimization model, and output network parameter weights of the pre-trained video recommendation model;

[0034] The second output module is used to perform network parameter optimization processing on the video recommendation model based on the network parameter weights to obtain an optimized video recommendation model, and input the user feature data and the video feature data of each second video into the optimized video recommendation model, and output the user behavior score corresponding to each second video.

[0035] As an optional implementation manner, the network parameter weights include network parameter weights corresponding to each rectified linear unit in the video recommendation model; the first output module is specifically used to:

[0036] For each corrected linear unit in the video recommendation model, the product of the current network parameter corresponding to the corrected linear unit and the network parameter weight corresponding to the corrected linear unit is determined as the optimized network parameter corresponding to the corrected linear unit to obtain the optimized video recommendation model.

[0037] As an optional implementation, the device further includes:

[0038] A second acquisition module is used to acquire first characteristic data of the target user; wherein the first characteristic data is characteristic data obtained through functional interaction;

[0039] The second output module is specifically used for:

[0040] The user feature data, the first feature data, and the video feature data of each of the second videos are input into the optimized video recommendation model.

[0041] As an optional implementation manner, the determining module is specifically configured to:

[0042] Among the second videos, a preset number of the second videos are selected in descending order of the user behavior scores and determined as the recommended videos corresponding to the secondary page.

[0043] As an optional implementation, the device further includes:

[0044] A third acquisition module is used to acquire training sample data and label data; wherein the training sample data includes user feature data of the sample user, video feature data of the third video and video feature data of the fourth video, the label data includes user behavior data of the fourth video, the third video is a video browsed by the sample user in the first-level page, and the fourth video is a video that the sample user continues to browse in the second-level page after jumping to the second-level page based on the third video;

[0045] The training module is used to jointly train the initial network parameter optimization model and the initial video recommendation model according to the training sample data and the label data to obtain a trained network parameter optimization model and a trained video recommendation model.

[0046] As an optional implementation, the training sample data also includes second feature data; wherein the second feature data is feature data obtained through functional interaction when the sample user browses the fourth video in the secondary page.

[0047] As an optional implementation manner, the third acquisition module is specifically used to:

[0048] In response to a trigger operation of the sample user on the fourth video in the secondary page, obtaining user feature data of the sample user, the second feature data, and video feature data and user behavior data of the fourth video;

[0049] According to the user ID of the sample user and the video ID of the fourth video, the third video belonging to the first-level page and associated with the user ID and the video ID of the fourth video is searched in the video browsing log of the sample user.

[0050] According to a third aspect, an electronic device is provided, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; the memory is used to store computer programs; and the processor is used to implement any method described in the first aspect when executing the program stored in the memory.

[0051] According to a fourth aspect, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the method described in any one of the first aspects is implemented.

[0052] The present application provides a video recommendation method, device, electronic device and readable storage medium. The technical solution provided by the embodiments of the present application brings at least the following beneficial effects: the computer device obtains the user feature data of the target user, the video feature data of the first video and the video feature data of each second video. Among them, the first video is the video browsed by the target user on the first-level page, the second video is the video to be recommended corresponding to the second-level page, and the second-level page is the page triggered by the first video. Then, the computer device optimizes the pre-trained video recommendation model based on the user feature data, the video feature data of the first video and the video feature data of each second video, and outputs the user behavior score corresponding to each second video based on the optimized video recommendation model. Afterwards, the computer device determines the recommended video corresponding to the second-level page in each second video according to the user behavior score corresponding to each second video. In this way, when recommending videos on the secondary page, after different users enter the secondary page through videos on different primary pages, the network parameters used by the video recommendation model of the secondary page are different, so that the video recommendation model used by the computer device on the secondary page takes into account the user's prior interests represented by the video content clicked by the user on the primary page, so that the videos recommended on the secondary page are related to the user's prior interests and are more in line with the type of videos the user currently wants to watch, thereby increasing the video views on the secondary page.

[0053] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention 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 invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 A flowchart of a video recommendation method provided in an embodiment of the present application;

[0056] Figure 2 A schematic diagram of a network parameter optimization model combined with a video recommendation model provided in an embodiment of the present application;

[0057] Figure 3A flowchart of a joint training method for a network parameter optimization model and a video recommendation model provided in an embodiment of the present application;

[0058] Figure 4 A schematic diagram of joint training of a network parameter optimization model and a video recommendation model provided in an embodiment of the present application;

[0059] Figure 5 A schematic diagram of the structure of a video recommendation device provided in an embodiment of the present application;

[0060] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] At present, the architecture of short video platforms usually includes a first-level page and a second-level page. The first-level page is the homepage of the short video platform. The short video platform displays video content that can quickly attract the user's attention on the first-level page. The second-level page is the video details page, which is jumped into by the video link on the first-level page. The short video platform displays the complete video content on the second-level page, and further displays other video content based on user operations (such as sliding down). However, since the video recommendation model used by the short video platform on the second-level page mainly focuses on the user's general interests (that is, what types of videos the user may be interested in), it does not take into account the user's pre-interest represented by the video content clicked by the user on the first-level page (that is, the type of video the user currently wants to watch). This may cause the user to find that the video content does not meet expectations after clicking on the video content on the second-level page, and quickly exit the video content, which in turn leads to a low number of video views on the second-level page.

[0063] The following will describe in detail a video recommendation method provided in an embodiment of the present application in conjunction with a specific implementation method. Figure 1 This is a flowchart of a video recommendation method provided in an embodiment of the present application. This embodiment of the present application takes the method as an example of applying to a computer device, which may be a terminal or a server. It is understandable that the method may also be applied to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. Figure 1 As shown, the specific steps are as follows:

[0064] Step 101, obtaining user feature data of a target user, video feature data of a first video, and video feature data of each second video, wherein the first video is a video that the target user browses on a primary page, and the second video is a video to be recommended corresponding to a secondary page, and the secondary page is a page that is triggered by the first video.

[0065] In implementation, when the target user triggers a video (i.e., the first video) displayed on the primary page, the computer device can obtain the video link of the first video, jump to the secondary page, and play the complete first video on the secondary page based on the video link of the first video. At this time, in order to recommend videos to the target user on the secondary page, the computer device can obtain the user feature data of the target user, the video feature data of the first video, and the video feature data of each second video. Among them, the first video is the video that the target user browses on the primary page, the second video is the video to be recommended corresponding to the secondary page, and the secondary page is the page to which the first video triggers the jump.

[0066] Optionally, user feature data may include one or more of user ID, user age, user gender, membership attributes (such as whether a member is a member, membership level, etc.), login attributes (such as whether logged in, type of login device, etc.), browsing history behavior, click history behavior, and user long-term and short-term portraits. Video feature data may include one or more of video ID, video attributes, video creator ID, and video creator main attributes.

[0067] Step 102, based on the user feature data, the video feature data of the first video and the video feature data of each second video, optimize the pre-trained video recommendation model, and output the user behavior score corresponding to each second video based on the optimized video recommendation model.

[0068] During implementation, the computer device optimizes a pre-trained video recommendation model based on user feature data, video feature data of the first video, and video feature data of each second video, and then outputs a user behavior score corresponding to each second video based on the optimized video recommendation model, so that the optimized video recommendation model uses the video feature data of the first video as the basis for video recommendation on the secondary page, that is, the optimized video recommendation model takes into account the user's prior interest represented by the video content clicked by the user on the primary page.

[0069] Step 103, according to the user behavior score corresponding to each second video, determine the recommended video corresponding to the secondary page in each second video.

[0070] During implementation, the computer device obtains the user behavior score corresponding to each second video, and can determine the recommended video corresponding to the secondary page in each second video.

[0071] It can be seen that in this embodiment, the computer device optimizes the pre-trained video recommendation model based on the user feature data, the video feature data of the first video and the video feature data of each second video, and outputs the user behavior score corresponding to each second video based on the optimized video recommendation model, so that when recommending videos on the secondary page, after different users enter the secondary page through videos on different primary pages, the network parameters used by the video recommendation model of the secondary page are different, so that the video recommendation model used by the computer device on the secondary page takes into account the user's prior interest represented by the video content clicked by the user on the primary page, so that the videos recommended on the secondary page are related to the user's prior interest and are more in line with the type of video that the user currently wants to watch, thereby increasing the video views on the secondary page.

[0072] In some embodiments, in step 102, based on the user feature data, the video feature data of the first video and the video feature data of each second video, a pre-trained video recommendation model is optimized, and based on the optimized video recommendation model, a user behavior score corresponding to each second video is output, including Figure 2 The following steps 1021 and 1022 are not shown:

[0073] Step 1021: input the user feature data and the video feature data of the first video into a pre-trained network parameter optimization model, and output the network parameter weights of the pre-trained video recommendation model.

[0074] In implementation, Figure 2 Schematic diagram of the network parameter optimization model combined with the video recommendation model provided in the embodiment of the present application. Figure 2 As shown, the computer device implements video recommendation for the secondary page through the network parameter optimization model and the video recommendation model. The network parameter optimization model can use the PPNet (Parameter Personalized Network) model, or other network models, which are not limited in the embodiments of the present application; the video recommendation model can use the DNN (Deep Neural Networks) model, or other types of network models, which are not limited in the embodiments of the present application.

[0075] When the computer device recommends a video to a target user in the secondary page, the user feature data of the target user and the video feature data of the first video can be input into a pre-trained network parameter optimization model. Accordingly, the network parameter optimization model can output the network parameter weights of the pre-trained video recommendation model.

[0076] Step 1022, performing network parameter optimization processing on the video recommendation model based on the network parameter weights to obtain an optimized video recommendation model, and inputting the user feature data and the video feature data of each second video into the optimized video recommendation model, and outputting the user behavior score corresponding to each second video.

[0077] In implementation, Figure 2 As shown, after the computer device obtains the network parameter weights of the video recommendation model output by the network parameter optimization model, it can further perform network parameter optimization processing on the video recommendation model based on the network parameter weights to obtain the optimized video recommendation model. Then, the computer device can input the user feature data of the target user and the video feature data of each second video into the optimized video recommendation model. Correspondingly, the video recommendation model can output the user behavior score corresponding to each second video. Among them, the user behavior score is used to characterize the predicted probability of the user performing various operation behaviors on the second video. The operation behavior can include multiple dimensions of behavior used to reflect the depth of interaction between the user and the second video. For example, the operation behavior may include but is not limited to one or more of clicking on the video, browsing for a certain length of time, watching to the end, liking, commenting, collecting, and forwarding. Correspondingly, the user behavior score may include but is not limited to one or more of the predicted click probability, predicted viewing time, completion probability, liking probability, comment probability, collection probability, and forwarding probability of the user for the second video, which is not limited in the embodiment of the present application.

[0078] It can be seen that in this embodiment, the computer device inputs the video feature data and user feature data of the first video in the first-level page into the network parameter optimization model, and outputs the network parameter weights of the video recommendation model, so that when recommending videos on the second-level page, after different users enter the second-level page through videos on different first-level pages, the network parameters used by the video recommendation model of the second-level page are different, so that the video recommendation model used by the computer device on the second-level page takes into account the user's prior interests represented by the video content clicked by the user on the first-level page, so that the videos recommended on the second-level page are related to the user's prior interests and are more in line with the type of video that the user currently wants to watch, thereby increasing the video views on the second-level page.

[0079] In some embodiments, in step 1022, before inputting the user feature data and the video feature data of each second video into the optimized video recommendation model, the method further includes: obtaining first feature data of the target user.

[0080] In this embodiment, the computer device further recommends videos to the target user in the secondary page based on the feature data based on the user feature data and the video feature data. The specific processing process is: obtaining the first feature data of the target user. The first feature data is feature data obtained through functional interaction. For example, the first feature data is a feature obtained through functional interaction when the target user currently uses a short video platform. Correspondingly, the processing process of inputting the user feature data and the video feature data of each second video into the optimized video recommendation model is: inputting the user feature data, the first feature data and the video feature data of each second video into the optimized video recommendation model.

[0081] In implementation, Figure 2 As shown, the computer device can obtain the first feature data through functional interaction, and the first feature data includes one or more of holidays, time, place, season and weather. For example, holidays are obtained through the calendar function, and the location is obtained through the positioning function. Then, the computer device can input the user feature data of the target user, the first feature data and the video feature data of each second video into the optimized video recommendation model. Correspondingly, the video recommendation model can output the user behavior score corresponding to each second video. The user behavior data includes one or more of clicks, viewing time, completion rate, comments and likes.

[0082] In some embodiments, in step 1023, the network parameter weights include the network parameter weights corresponding to each corrected linear unit in the video recommendation model. The computer device performs network parameter optimization processing on the video recommendation model based on the network parameter weights to obtain the optimized video recommendation model in the following process: for each corrected linear unit in the video recommendation model, the product of the current network parameter corresponding to the corrected linear unit and the network parameter weight corresponding to the corrected linear unit is determined as the optimized network parameter corresponding to the corrected linear unit to obtain the optimized video recommendation model.

[0083] In implementation, Figure 2 As shown, the network parameter weights include the network parameter weights corresponding to each rectified linear unit (ReLu) in the video recommendation model. After the computer device obtains the network parameter weights corresponding to each ReLu in the video recommendation model, it can further determine the product of the current network parameter corresponding to the rectified linear unit and the network parameter weight corresponding to the rectified linear unit for each rectified linear unit in the video recommendation model as the optimized network parameter corresponding to the rectified linear unit, thereby obtaining the optimized video recommendation model.

[0084] In some embodiments, in step 103, based on the user behavior score corresponding to each second video, the processing process of determining the recommended video corresponding to the secondary page in each second video is as follows: in each second video, in descending order of the user behavior score, a preset number of second videos are selected and determined as the recommended videos corresponding to the secondary page.

[0085] During implementation, after the computer device obtains the user behavior score corresponding to each second video, it can select a preset number of second videos from each second video in descending order of the user behavior score and determine them as recommended videos corresponding to the secondary page.

[0086] The computer device may sort the second videos in descending order of the user behavior score to obtain a second video sequence, and then determine the first preset number of second videos as recommended videos corresponding to the secondary page. Alternatively, the computer device may also determine the second videos with user behavior scores greater than a preset threshold as recommended videos corresponding to the secondary page. In this way, second videos with higher user behavior scores may be selected as recommended videos, thereby improving the accuracy of video recommendations.

[0087] The following will describe in detail a joint training method for a network parameter optimization model and a video recommendation model provided in an embodiment of the present application in combination with a specific implementation method. Figure 3 A flowchart of a joint training method of a network parameter optimization model and a video recommendation model provided in an embodiment of the present application. Figure 3 As shown, the specific steps are as follows:

[0088] Step 301, obtaining training sample data and label data. The training sample data includes user feature data of the sample user, video feature data of the third video, and video feature data of the fourth video. The label data includes user behavior data of the fourth video. The third video is a video that the sample user browses on the primary page, and the fourth video is a video that the sample user continues to browse on the secondary page after jumping to the secondary page based on the third video.

[0089] In implementation, the computer device may obtain training sample data and label data, wherein the training sample data includes user feature data of the sample user, video feature data of the third video, and video feature data of the fourth video, and the label data includes user behavior data of the fourth video, the third video is a video browsed by the sample user on the primary page, and the fourth video is a video that the sample user continues to browse on the secondary page after jumping to the secondary page based on the third video.

[0090] Step 302: jointly train the initial network parameter optimization model and the initial video recommendation model according to the training sample data and the label data to obtain a trained network parameter optimization model and a trained video recommendation model.

[0091] In implementation, Figure 4 A schematic diagram of the joint training of a network parameter optimization model and a video recommendation model provided in an embodiment of the present application. Figure 4 As shown, after the computer device obtains the training sample data and the label data, it can jointly train the initial network parameter optimization model and the initial video recommendation model according to the training sample data and the label data to obtain the trained network parameter optimization model and the trained video recommendation model.

[0092] In some embodiments, in an embodiment of the present application, for a situation in which a computer device can recommend a video to a target user in a secondary page based on user feature data and video feature data, further based on feature data, in step 301, the training sample data also includes second feature data; wherein the second feature data is feature data obtained through functional interaction when the sample user browses the fourth video in the secondary page.

[0093] As an optional implementation, the processing process of the computer device acquiring the training sample data and the label data is as follows:

[0094] Step one, in response to a trigger operation of a sample user on a fourth video in a secondary page, obtaining user feature data, second feature data of the sample user, and video feature data and user behavior data of the fourth video.

[0095] In implementation, when a sample user triggers a recommended video (i.e., the fourth video) in the secondary page, the computer device can obtain the user feature data, the second feature data of the sample user, and the video feature data and user behavior data of the fourth video.

[0096] Step 2: According to the user ID of the sample user and the video ID of the fourth video, in the video browsing log of the sample user, search for a third video that belongs to a first-level page and is associated with the user ID and the video ID of the fourth video.

[0097] In implementation, the computer device can store the videos browsed by the user in the primary page, the videos browsed in the secondary page, and the association between the videos in the primary page and the videos in the secondary page (i.e., which video in the primary page jumps to the secondary page and further browses the video in the secondary page) in the video browsing log. After the computer device obtains the user ID of the sample user and the video ID of the fourth video, it can query the third video belonging to the primary page and associated with the user ID and the video ID of the fourth video in the video browsing log of the sample user. Subsequently, the computer device can splice the obtained user feature data of the sample user, the video feature data of the third video, and the video feature data of the fourth video into training sample data, and use the user behavior data of the fourth video as label data.

[0098] An embodiment of the present application provides a method for recommending videos. A computer device obtains user feature data of a target user, video feature data of a first video, and video feature data of each second video. Among them, the first video is a video browsed by the target user on a first-level page, and the second video is a video to be recommended corresponding to a second-level page, and the second-level page is a page triggered by the first video. Then, the computer device inputs the user feature data and the video feature data of the first video into a pre-trained network parameter optimization model, and outputs the network parameter weights of the pre-trained video recommendation model; performs network parameter optimization processing on the video recommendation model based on the network parameter weights to obtain an optimized video recommendation model, and inputs the user feature data and the video feature data of each second video into the optimized video recommendation model, and outputs the user behavior score corresponding to each second video. Afterwards, the computer device determines the recommended video corresponding to the second-level page in each second video according to the user behavior score corresponding to each second video. In this way, the computer device inputs the video feature data and user feature data of the first video in the first-level page into the network parameter optimization model, and outputs the network parameter weights of the video recommendation model. Therefore, when recommending videos on the second-level page, after different users enter the second-level page through videos on different first-level pages, the network parameters used by the video recommendation model of the second-level page are different, so that the video recommendation model used by the computer device on the second-level page takes into account the user's prior interests represented by the video content clicked by the user on the first-level page, so that the videos recommended on the second-level page are related to the user's prior interests and are more in line with the type of videos the user currently wants to watch, thereby increasing the video views on the second-level page.

[0099] It should be understood that although Figure 1 and Figure 3The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 and Figure 3 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0100] It can be understood that the same / similar parts between the various embodiments of the above method in this specification can refer to each other, and each embodiment focuses on the differences from other embodiments. For related points, please refer to the description of other method embodiments.

[0101] The present application also provides a video recommendation device, such as Figure 5 As shown, the device comprises:

[0102] The first acquisition module 510 is used to acquire user feature data of the target user, video feature data of the first video, and video feature data of each second video; wherein the first video is a video browsed by the target user on a primary page, and the second video is a video to be recommended corresponding to a secondary page, and the secondary page is a page to which the first video triggers a jump;

[0103] The processing unit 520 is used to optimize the pre-trained video recommendation model based on the user feature data, the video feature data of the first video, and the video feature data of each second video, and output the user behavior score corresponding to each second video based on the optimized video recommendation model;

[0104] The determination module 530 is used to determine the recommended video corresponding to the secondary page in each second video according to the user behavior score corresponding to each second video.

[0105] As an optional implementation, the processing unit 520 includes a first output module and a second output module.

[0106] A first output module, used to input the user feature data and the video feature data of the first video into a pre-trained network parameter optimization model, and output the network parameter weights of the pre-trained video recommendation model;

[0107] The second output module is used to optimize the network parameters of the video recommendation model based on the network parameter weights to obtain an optimized video recommendation model, and input the user feature data and the video feature data of each second video into the optimized video recommendation model, and output the user behavior score corresponding to each second video.

[0108] As an optional implementation, the network parameter weights include network parameter weights corresponding to each rectified linear unit in the video recommendation model; the first output module is specifically used to:

[0109] For each corrected linear unit in the video recommendation model, the product of the current network parameter corresponding to the corrected linear unit and the network parameter weight corresponding to the corrected linear unit is determined as the optimized network parameter corresponding to the corrected linear unit, thereby obtaining the optimized video recommendation model.

[0110] As an optional implementation, the device further includes:

[0111] A second acquisition module is used to acquire first characteristic data of a target user; wherein the first characteristic data is characteristic data obtained through functional interaction;

[0112] The second output module is specifically used for:

[0113] The user feature data, the first feature data, and the video feature data of each second video are input into the optimized video recommendation model.

[0114] As an optional implementation, the determination module 530 is specifically configured to:

[0115] Among the second videos, a preset number of second videos are selected in descending order of the user behavior scores and determined as the recommended videos corresponding to the secondary page.

[0116] As an optional implementation, the device further includes:

[0117] A third acquisition module is used to acquire training sample data and label data; wherein the training sample data includes user feature data of the sample user, video feature data of the third video and video feature data of the fourth video, and the label data includes user behavior data of the fourth video, the third video is a video browsed by the sample user on the primary page, and the fourth video is a video that the sample user continues to browse on the secondary page after jumping to the secondary page based on the third video;

[0118] The training module is used to jointly train the initial network parameter optimization model and the initial video recommendation model according to the training sample data and the label data to obtain the trained network parameter optimization model and the trained video recommendation model.

[0119] As an optional implementation, the training sample data also includes second feature data; wherein the second feature data is feature data obtained through functional interaction when the sample user browses the fourth video in the secondary page.

[0120] As an optional implementation manner, the third acquisition module is specifically used to:

[0121] In response to a trigger operation of the sample user on the fourth video in the secondary page, obtaining user feature data, second feature data of the sample user, and video feature data and user behavior data of the fourth video;

[0122] According to the user ID of the sample user and the video ID of the fourth video, a third video belonging to a first-level page and associated with the user ID and the video ID of the fourth video is searched in the video browsing log of the sample user.

[0123] For the specific definition of the video recommendation device, please refer to the definition of the video recommendation method above, which will not be repeated here. Each module in the above video recommendation device can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the electronic device in the form of hardware, or can be stored in the memory of the electronic device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0124] In one embodiment, an electronic device is provided, such as Figure 6 As shown, it includes a processor 601, a communication interface 602, a memory 603 and a communication bus 604, wherein the processor 601, the communication interface 602, and the memory 603 communicate with each other through the communication bus 604; the memory 603 is used to store computer programs; the processor 601 is used to implement the recommended method steps of the above video when executing the program stored in the memory 603.

[0125] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned video recommendation method are implemented.

[0126] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0127] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0128] It should also be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0129] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0130] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0131] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A video recommendation method, characterized in that: The method comprises: Obtain user feature data of a target user, video feature data of a first video, and video feature data of each second video; wherein the first video is a video that the target user browses on a primary page, and the second video is a video to be recommended corresponding to a secondary page, and the secondary page is a page that is triggered by the first video; Optimizing a pre-trained video recommendation model based on the user feature data, the video feature data of the first video, and the video feature data of each of the second videos, and outputting a user behavior score corresponding to each of the second videos based on the optimized video recommendation model; According to the user behavior score corresponding to each of the second videos, a recommended video corresponding to the secondary page is determined in each of the second videos.

2. The method according to claim 1, characterized in that The optimizing a pre-trained video recommendation model based on the user feature data, the video feature data of the first video, and the video feature data of each of the second videos, and outputting a user behavior score corresponding to each of the second videos based on the optimized video recommendation model, includes: Inputting the user feature data and the video feature data of the first video into a pre-trained network parameter optimization model, and outputting network parameter weights of the pre-trained video recommendation model; The video recommendation model is subjected to network parameter optimization processing based on the network parameter weights to obtain an optimized video recommendation model, and the user feature data and the video feature data of each second video are input into the optimized video recommendation model to output a user behavior score corresponding to each second video.

3. The method according to claim 2, characterized in that The network parameter weights include network parameter weights corresponding to each rectified linear unit in the video recommendation model; The performing network parameter optimization processing on the video recommendation model based on the network parameter weight to obtain an optimized video recommendation model includes: For each corrected linear unit in the video recommendation model, the product of the current network parameter corresponding to the corrected linear unit and the network parameter weight corresponding to the corrected linear unit is determined as the optimized network parameter corresponding to the corrected linear unit to obtain the optimized video recommendation model.

4. The method according to claim 2, characterized in that: Before inputting the user feature data and the video feature data of each second video into the optimized video recommendation model, the method further includes: Acquire first characteristic data of the target user; wherein the first characteristic data is characteristic data obtained through function interaction; The step of inputting the user feature data and the video feature data of each second video into the optimized video recommendation model comprises: The user feature data, the first feature data, and the video feature data of each of the second videos are input into the optimized video recommendation model.

5. The method according to claim 1, characterized in that The method further comprises: Acquire training sample data and label data; wherein the training sample data includes user feature data of a sample user, video feature data of a third video, and video feature data of a fourth video, and the label data includes user behavior data of the fourth video, the third video is a video that the sample user browses on the primary page, and the fourth video is a video that the sample user continues to browse on the secondary page after jumping to the secondary page based on the third video; The initial network parameter optimization model and the initial video recommendation model are jointly trained according to the training sample data and the label data to obtain a trained network parameter optimization model and a trained video recommendation model.

6. The method according to claim 5, characterized in that The training sample data also includes second feature data; wherein the second feature data is feature data obtained through functional interaction when the sample user browses the fourth video in the secondary page.

7. The method according to claim 6, characterized in that The obtaining of training sample data and label data includes: In response to a trigger operation of the sample user on the fourth video in the secondary page, obtaining user feature data of the sample user, the second feature data, and video feature data and user behavior data of the fourth video; According to the user ID of the sample user and the video ID of the fourth video, the third video belonging to the first-level page and associated with the user ID and the video ID of the fourth video is searched in the video browsing log of the sample user.

8. A video recommendation device, characterized in that: The device comprises: A first acquisition module is used to acquire user feature data of a target user, video feature data of a first video, and video feature data of each second video; wherein the first video is a video browsed by the target user on a primary page, and the second video is a video to be recommended corresponding to a secondary page, and the secondary page is a page to which the first video triggers a jump; a processing unit, configured to optimize a pre-trained video recommendation model based on the user feature data, the video feature data of the first video, and the video feature data of each of the second videos, and output a user behavior score corresponding to each of the second videos based on the optimized video recommendation model; A determination module is used to determine, in each of the second videos, a recommended video corresponding to the secondary page according to the user behavior score corresponding to each of the second videos.

9. An electronic device, characterized in that: The invention comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; the memory is used to store computer programs; and the processor is used to implement the method described in any one of claims 1 to 7 when executing the program stored in the memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images