CN110730369B - Video recommendation method and server - Google Patents
Video recommendation method and server Download PDFInfo
- Publication number
- CN110730369B CN110730369B CN201910980081.9A CN201910980081A CN110730369B CN 110730369 B CN110730369 B CN 110730369B CN 201910980081 A CN201910980081 A CN 201910980081A CN 110730369 B CN110730369 B CN 110730369B
- Authority
- CN
- China
- Prior art keywords
- video
- recommended
- historical
- evaluation value
- exposure
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/251—Learning process for intelligent management, e.g. learning user preferences for recommending movies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/262—Content or additional data distribution scheduling, e.g. sending additional data at off-peak times, updating software modules, calculating the carousel transmission frequency, delaying a video stream transmission, generating play-lists
- H04N21/26258—Content or additional data distribution scheduling, e.g. sending additional data at off-peak times, updating software modules, calculating the carousel transmission frequency, delaying a video stream transmission, generating play-lists for generating a list of items to be played back in a given order, e.g. playlist, or scheduling item distribution according to such list
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4667—Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Computing Systems (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The application relates to the field of computers, and discloses a video recommendation method and a server, which are used for solving the problem that the video playing conversion rate is reduced due to repeated recommendation of videos with high exposure rates. The method comprises the steps of firstly, acquiring a set of videos to be recommended, and calculating a first video recommendation evaluation value of each video to be recommended; secondly, determining historical exposed videos and unexposed videos in a video set to be recommended; thirdly, adjusting the first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by adopting an attenuation model; and finally, generating and outputting a target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video. In this way, the first video recommendation evaluation value of the historical exposure video is subjected to weight reduction processing through the attenuation model so as to improve the video playing conversion rate.
Description
Technical Field
The present application relates to the field of computers, and in particular, to a video recommendation method and a server.
Background
With the development of science and technology, a large amount of video content is contained on the existing network, and a user can browse and play the video content in a mode of actively searching or receiving push information.
In order to improve the video playing conversion rate, the existing strategy is to adopt a recommendation system to recommend videos which may be liked by the user and have not been played before to the user. However, when the user only browses but does not play a certain video, the recommendation system still pushes the video to the user for many times, which causes the user to feel the dislike due to repeated recommendation of videos with high exposure rate, thereby reducing the video playing conversion rate.
In view of the above, the present application provides a new video recommendation method to overcome the above-mentioned drawbacks.
Disclosure of Invention
The embodiment of the application provides a video recommendation method and a server, which aim to solve the problem that the video playing conversion rate is reduced due to repeated recommendation of videos with high exposure rates.
The video recommendation method provided by the embodiment of the application comprises the following steps:
acquiring a video set to be recommended, and calculating a first video recommendation evaluation value of each video to be recommended in the video set to be recommended;
determining historical exposed videos and unexposed videos in the video set to be recommended;
adjusting a first video recommendation evaluation value of the historical exposure video to a second video recommendation evaluation value of the historical exposure video by using a preset attenuation model, wherein the second video recommendation evaluation value of the historical exposure video is smaller than the first video recommendation evaluation value of the historical exposure video;
and generating a corresponding target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video, and outputting the target recommended video set.
Optionally, determining the historical exposed video and the unexposed video in the video set to be recommended includes:
the following operations are respectively executed for each video to be recommended:
judging whether a video to be recommended carries exposure identification information or not;
if so, judging that the video to be recommended is the historical exposure video;
otherwise, judging that the video to be recommended is the unexposed video.
Optionally, adjusting the first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by using a preset attenuation model, where the method includes:
obtaining an attenuation adjustment value of the historical exposure video from the attenuation model, wherein the value of the attenuation adjustment value of the historical exposure video is in negative correlation with the recommended times of the historical exposure video;
and generating a second video recommendation evaluation value of the historical exposure video based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video.
Optionally, obtaining an attenuation adjustment value of the historical exposure video in the attenuation model includes:
acquiring video identification information of the historical exposure video;
and determining an attenuation adjustment value corresponding to the video identification information in the attenuation model.
Optionally, generating a second video recommendation evaluation value of the historical exposure video based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video includes:
and determining the product of the first video recommendation evaluation value of the historical exposure video and the attenuation adjustment value as a second video recommendation evaluation value of the historical exposure video.
Optionally, generating a corresponding target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video, where the generating includes:
sequencing the unexposed videos and the historical exposed videos according to the sequence from high to low of the first video recommendation evaluation value and the second video recommendation evaluation value;
and determining the reordered video set to be recommended as the target recommended video set.
Correspondingly, an embodiment of the present application further provides a video recommendation server, including:
the device comprises an acquisition unit, a recommendation unit and a recommendation unit, wherein the acquisition unit is used for acquiring a video set to be recommended and calculating a first video recommendation evaluation value of each video to be recommended in the video set to be recommended;
the processing unit is used for determining historical exposed videos and unexposed videos in the video set to be recommended;
adjusting a first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by adopting a preset attenuation model, wherein the second video recommended evaluation value of the historical exposure video is smaller than the first video recommended evaluation value of the historical exposure video;
and the recommending unit is used for generating a corresponding target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video, and outputting the target recommended video set.
Optionally, the historical exposed video and the unexposed video in the video set to be recommended are determined, and the processing unit is configured to:
the following operations are respectively executed for each video to be recommended:
judging whether a video to be recommended carries exposure identification information or not;
if so, judging that the video to be recommended is the historical exposure video;
otherwise, judging that the video to be recommended is the unexposed video.
Optionally, adjusting the first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by using a preset attenuation model, where the method includes:
obtaining an attenuation adjustment value of the historical exposure video from the attenuation model, wherein the value of the attenuation adjustment value of the historical exposure video is in negative correlation with the recommended times of the historical exposure video;
and generating a second video recommendation evaluation value of the historical exposure video based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video.
Optionally, the processing unit is configured to obtain an attenuation adjustment value of the historical exposure video in the attenuation model, and:
acquiring video identification information of the historical exposure video;
and determining an attenuation adjustment value corresponding to the video identification information in the attenuation model.
Optionally, based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video, a second video recommendation evaluation value of the historical exposure video is generated, and the processing unit is configured to:
and determining the product of the first video recommendation evaluation value of the historical exposure video and the attenuation adjustment value as a second video recommendation evaluation value of the historical exposure video.
Optionally, a corresponding target recommended video set is generated based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video, and the recommending unit is specifically configured to:
sequencing the unexposed videos and the historical exposed videos according to the sequence from high to low of the first video recommendation evaluation value and the second video recommendation evaluation value;
and determining the reordered video set to be recommended as the target recommended video set.
Correspondingly, an embodiment of the present application further provides a computing device, including:
a memory for storing program instructions;
and the processor is used for calling the program instructions stored in the memory and executing the video recommendation method according to the obtained program.
Accordingly, an embodiment of the present application further provides a computer-readable non-volatile storage medium, which includes computer-readable instructions, and when the computer-readable instructions are read and executed by a computer, the computer-readable instructions cause the computer to execute the above video recommendation method.
The beneficial effect of this application is as follows:
in the embodiment of the application, a video set to be recommended is obtained first, and a first video recommendation evaluation value of each video to be recommended is calculated; secondly, determining historical exposed videos and unexposed videos in a video set to be recommended; thirdly, adjusting the first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by adopting an attenuation model; and finally, generating and outputting a target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video. And carrying out reduction processing on the first video recommendation evaluation value of the historical exposure video through an attenuation model to generate a second video recommendation evaluation value of the historical exposure video, so that the videos with less repeated recommendation times and/or the videos which are not recommended can be ranked in front of the videos with more repeated recommendation times as much as possible, and the video playing conversion rate is improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic flowchart of a process for training an attenuation model according to an embodiment of the present application;
fig. 2 is a schematic flowchart of a video recommendation method according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of a video recommendation server according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of a computing device according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail with reference to the accompanying drawings, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to solve the problem that the video playing conversion rate is reduced due to the fact that a video with high exposure rate is recommended repeatedly, a new technical scheme is provided in the embodiment of the application. The scheme comprises the following steps: firstly, acquiring a set of videos to be recommended and calculating a first video recommendation evaluation value of each video to be recommended; secondly, determining historical exposed videos and unexposed videos in a video set to be recommended; thirdly, adjusting the first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by adopting an attenuation model; and finally, generating and outputting a target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video.
The following detailed description of preferred embodiments of the present application refers to the accompanying drawings.
The embodiment of the application provides a recommendation engine comprising a recommendation model and an attenuation model, and the recommendation engine is used for generating and sending a target recommendation video set to a user. Referring to fig. 1, the process of training the attenuation model is as follows:
s101: all combinations of attenuation functions are determined using an initial attenuation model.
Specifically, the process of determining all combinations of attenuation functions using the initial attenuation model is as follows:
and A1, adopting an initial attenuation model, and associating corresponding attenuation functions for each preset conversion influence factor respectively based on a preset attenuation function set.
In the embodiment of the present application, a factor that affects the transition of the exposed video from the non-playing state to the playing state is referred to as a conversion factor. And the preset conversion influence factor at least comprises three elements of the number of days of the video from the last exposure, the total number of times of the current video exposure and the position of the video in the target recommended video set.
The following four attenuation functions are included in the initial attenuation model:
f(x)＝α1*x+α2an attenuation function (1);
f(x)＝α1(x-α2)2+α3an attenuation function (4);
wherein x represents a certain conversion influence factor, α1、α2And alpha3All represent attenuation function coefficients, and f (x) represents a certainA decay function of the conversion impact factor.
And the initial attenuation model is used for selecting one of the four attenuation functions for each conversion influence factor as a corresponding attenuation function.
And A2, adopting the attenuation functions related to the conversion influence factors to form an attenuation function combination.
A3, judging whether all attenuation function combinations are output, wherein attenuation functions contained in each attenuation function combination are not completely the same, if yes, executing step A4; otherwise, step a1 is performed.
A4, the initial decay model determines all combinations of decay functions.
For ease of understanding, the description is made with reference to three specific examples.
For example, the initial attenuation function is associated with an attenuation function (1) for a conversion factor 1, an attenuation function (2) for a conversion factor 2, and an attenuation function (4) for a conversion factor 3, and thus the combination 1 of the attenuation functions is [ attenuation function (1), attenuation function (2), and attenuation function (4) ].
For example, the initial attenuation function is associated with an attenuation function (1) for a conversion factor 1, an attenuation function (2) for a conversion factor 2, and an attenuation function (3) for a conversion factor 3, and thus the attenuation function combination 2 is formed as [ attenuation function (1), attenuation function (2), and attenuation function (3) ].
For another example, the initial attenuation function is a combination of attenuation functions 3 [ attenuation function (1), and attenuation function (1) ] in which the conversion factor 1 is associated with the attenuation function (1), the conversion factor 2 is associated with the attenuation function (1), and the conversion factor 3 is associated with the attenuation function (1).
As can be seen from the above three examples, the same attenuation function may be included in the same attenuation function combination, but the attenuation functions included in a plurality of attenuation function combinations are not completely the same.
S102: and obtaining the predicted video playing conversion rate variation of the historical playing sample set based on a decay function combination.
Before step 102 is executed, a historical play sample set needs to be constructed, and the actual video play conversion rate variation of the historical play sample set needs to be calculated.
The historical playing sample set refers to a training sample data set formed by historical playing records of a plurality of users.
The actual video playing conversion rate refers to the efficiency of converting an exposed video from an unplayed state to a playing state, so that the actual video playing conversion rate of a historical playing video can reflect the actual favorite degree of a user on the exposed video; the actual video playing conversion rate variation refers to the overall variation degree of the actual video playing conversion rates of all the historical playing records.
The predicted video playing conversion rate refers to the efficiency of predicting that an exposed video is converted from an unplayed state to a playing state, so that the user's preference degree on the exposed video can be predicted by the predicted video playing conversion rate of a historical playing video; the predicted video playing conversion rate variation refers to the overall variation degree of the predicted video playing conversion rates of all historical playing records.
Specifically, in the embodiment of the application, the historical video sets of a plurality of users are acquired by collecting historical video logs of the plurality of users. The historical video collection of a user comprises historical exposure records of videos which are not played and historical playing records of videos which are browsed and played by the user.
Wherein one historical exposure record contains the following elements: the video recommendation system comprises a user Identity Identification (ID), a video ID, the number of days of last exposure of the video, the total number of times of current video exposure, the position of the video in a target recommended video set, a video recommendation evaluation value, a video exposure time point and non-conversion identification information, wherein a historical play record comprises all elements from the user ID to the video exposure time point and the conversion identification information.
Specifically, the process of calculating the actual video playing conversion rate variation of the historical playing sample set is as follows:
and B1, calculating the actual video playing conversion rate of each historical playing sample.
The method for calculating the actual video playing conversion rate of a historical playing sample is to determine the ratio of the number of times of playing the video to the total number of times of exposure of the current video as the actual video playing conversion rate. Generally, the recommendation model does not recommend repeatedly played videos to the user, and therefore, when the user plays the videos, the historical video logs will stop updating, that is, the videos are played only once.
And B2, drawing a curve 1 in a two-dimensional coordinate system with the vertical axis of the actual video playing conversion rate and the horizontal axis of the nth historical playing sample, wherein the curve 1 represents the variation of the actual video playing conversion rate.
Specifically, based on a combination of decay functions, the process of obtaining the predicted video playing conversion rate variation of the historical playing sample set is as follows:
and C1, combining through a decay function, and generating a predicted video playing conversion rate corresponding to each historical playing sample in the historical playing sample set.
In the embodiment of the present application, formula (1) is used
And generating a predicted video playing conversion rate. Wherein m represents the total number of transformation-affecting factors, f (x)i) A decay function, ω, representing the ith conversion influence factoriIs the weight of the ith conversion influence factor, and
indicating the predicted video playback conversion rate.
And C2, determining the variable quantity of the predicted video playing conversion rate based on the corresponding predicted video playing conversion rate.
And drawing a curve 2 in a two-dimensional coordinate system taking the predicted video playing conversion rate as a vertical axis and taking the m conversion influence factors of the nth historical playing sample as a horizontal axis, wherein the curve 2 represents the variation of the actual video playing conversion rate.
S103: judging whether the predicted video playing conversion rate variation accords with a preset error threshold value, if so, executing step 104; otherwise, step 105 is performed.
And calculating the difference between the predicted video playing conversion rate variation and the actual video playing conversion rate variation, wherein whether the difference does not exceed an error threshold value. If the difference value does not exceed the error threshold value, the output predicted video playing conversion rate is close to the actual video playing conversion rate based on the attenuation function combination; otherwise, it indicates that the parameter value in the attenuation function combination is not optimal, the parameter value of the attenuation function combination should be adjusted, and the process returns to step 102, and calculates a new predicted video playing conversion rate variation according to the new parameter value of the attenuation function combination. Wherein the parameter values comprise attenuation function coefficients and weights of the ith conversion impact factor.
S104: a decay function combination is determined as a candidate decay function combination output.
S105: the parameter values of a combination of attenuation functions are adjusted and the process returns to step 102.
S106: judging whether all attenuation function combinations are processed, if so, executing step 107; otherwise, step 102 is performed.
S107: and combining the candidate attenuation functions with the minimum error of the corresponding predicted video playing conversion rate in the obtained candidate attenuation combinations to serve as the target attenuation function combination of the attenuation model.
S108: and outputting the attenuation model containing the target attenuation function combination as a target attenuation model.
The training of the attenuation model is completed through the steps 101-108.
Specifically, using the trained target attenuation model, the process of determining the attenuation adjustment value of each historical exposure video is as follows:
d1, acquiring the latest historical exposure record of each historical exposure video in a log collection mode;
the historical exposure video refers to a video which is recommended to the user at least once and is only browsed but not played by the user;
d2, sequentially inputting the latest historical exposure records into an attenuation model to obtain the predicted video playing conversion rate of each historical exposure video;
d3, classifying the historical exposure videos according to the user ID, namely dividing the historical exposure videos pushed to the same user into a group;
d4, adopting formula (2)
And calculating attenuation coefficients of the historical exposure videos in the same group.
Wherein d represents an attenuation adjustment value,
predictive video playback conversion, max, representing a historical exposure video
Indicating the maximum predicted video playback conversion rate in the same group. d has a value range of [0,1 ]]As the number of times the historical exposure video is recommended increases, d of the historical exposure video decreases.
B5, storing d of each historical exposure video in the attenuation model according to the corresponding relation between the video ID and d.
In addition, a plurality of new historical play record sets of the users can be obtained periodically according to a set period, and the attenuation model is retrained according to the steps 101-111 so as to improve the prediction accuracy. And calculating new attenuation adjustment values for each historical exposure video using the updated attenuation model.
Referring to fig. 2, a specific process of generating a target recommended video set in the embodiment of the present application is as follows:
s201: the method comprises the steps of obtaining a video set to be recommended and calculating a first video recommendation evaluation value of each video to be recommended in the video set to be recommended.
The method comprises the steps of deducing videos which are probably liked by a user through a pre-trained recommendation model, generating a set of videos to be recommended, and calculating a first video recommendation evaluation value of each video to be recommended through the recommendation model. The first video recommendation evaluation value is a value for predicting the preference degree of the user to the video to be recommended, so that if the first video recommendation evaluation value is higher, the user is indicated to be more likely to be interested in the corresponding video to be recommended; otherwise, it indicates that the user may be less interested in the corresponding video to be recommended.
S202: and determining historical exposed videos and unexposed videos in the video set to be recommended.
Specifically, the process of determining whether each video to be recommended is a history exposure video is as follows:
judging whether a video to be recommended carries exposure identification information or not, if so, judging that the video to be recommended is a historical exposure video; otherwise, judging that the video to be recommended is an unexposed video.
The historical exposure video refers to a video which is recommended to the user at least once and is only browsed but not played by the user; and unexposed video refers to video that has never been recommended to the user. The more the number of the historical exposure videos contained in the target recommended video set sent to the user is, and/or the more the times that each historical exposure video is repeatedly recommended are, the more the video playing conversion rate of the target recommended video set is reduced. The video playing conversion rate refers to the percentage of videos browsed and played by the user in the target recommended video set. Therefore, in the embodiment of the application, the decay model is adopted to perform the de-emphasis processing on the video recommendation evaluation value of the historical exposure video, so that the videos which are recommended repeatedly less and/or not recommended are arranged in front of the videos which are recommended repeatedly more as much as possible, and the video playing conversion rate is improved.
Before this, it is most important to screen out all historical exposure videos in the video set to be recommended. In the embodiment of the application, when the historical exposure video is determined as the video to be recommended again by the recommendation model, the exposure identification information is given to the video to be recommended, so that the recommendation engine can screen out the corresponding historical exposure video through the exposure identification information and send the corresponding historical exposure video to the attenuation model for weight reduction processing.
S203: and adjusting the first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by adopting a preset attenuation model, wherein the second video recommended evaluation value of the historical exposure video is smaller than the first video recommended evaluation value of the historical exposure video.
The process of adjusting the video recommendation evaluation value of the historical exposure video is specifically described as follows:
e1, acquiring a historical exposure video.
And E2, obtaining an attenuation adjustment value of a historical exposure video from the attenuation model, wherein the value of the attenuation adjustment value of the historical exposure video is in negative correlation with the recommended times of the historical exposure video.
By obtaining a video ID for a historical exposure video, the corresponding attenuation adjustment value is matched in the attenuation model. Wherein the attenuation adjustment value of one historical exposure video decreases as the recommended number of times of the one historical exposure video increases.
E3, generating a second video recommendation evaluation value of the historical exposure video based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video.
Specifically, the product of the first video recommendation evaluation value of one historical exposure video and the attenuation adjustment value is determined as the second video recommendation evaluation value of one historical exposure video.
E4, judging whether all the historical exposure videos are processed completely, if so, outputting second video recommended evaluation values of all the historical exposure videos, and executing the step 204; otherwise, return to step E1.
S204: and generating a corresponding target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video, and outputting the target recommended video set.
Specifically, sequencing each unexposed video and each historical exposed video according to a first video recommended evaluation value and a second video recommended evaluation value from high to low;
and determining the reordered video set to be recommended as the target recommended video set.
Through the steps 201-204, the recommendation engine reorders the video sets to be recommended, so that videos with few repeated recommendation times and/or videos which are not recommended in the finally generated target recommended video set can be ranked in front of videos with many repeated recommendation times as much as possible, and the video playing conversion rate is improved.
Based on the same inventive concept, fig. 3 exemplarily shows a schematic structural diagram of a video recommendation server provided in an embodiment of the present application, and includes at least an obtaining unit 301, a processing unit 302, and a recommending unit 303, where,
the acquiring unit 301 is configured to acquire a video set to be recommended and calculate a first video recommendation evaluation value of each video to be recommended in the video set to be recommended;
a processing unit 302, configured to determine a historical exposed video and an unexposed video in the video set to be recommended;
adjusting a first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by adopting a preset attenuation model, wherein the second video recommended evaluation value of the historical exposure video is smaller than the first video recommended evaluation value of the historical exposure video;
a recommending unit 303, configured to generate a corresponding target recommended video set based on the first video recommendation evaluation value of each unexposed video and the second video recommendation evaluation value of each historical exposed video, and output the target recommended video set.
Optionally, the historical exposed video and the unexposed video in the video set to be recommended are determined, and the processing unit 302 is configured to:
the following operations are respectively executed for each video to be recommended:
judging whether a video to be recommended carries exposure identification information or not;
if so, judging that the video to be recommended is the historical exposure video;
otherwise, judging that the video to be recommended is the unexposed video.
Optionally, a preset attenuation model is adopted, the first video recommended evaluation value of the historical exposure video is adjusted to be the second video recommended evaluation value of the historical exposure video, and the processing unit 302 is configured to:
obtaining an attenuation adjustment value of the historical exposure video from the attenuation model, wherein the value of the attenuation adjustment value of the historical exposure video is in negative correlation with the recommended times of the historical exposure video;
and generating a second video recommendation evaluation value of the historical exposure video based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video.
Optionally, the attenuation adjustment value of the historical exposure video is obtained in the attenuation model, and the processing unit 302 is configured to:
acquiring video identification information of the historical exposure video;
and determining an attenuation adjustment value corresponding to the video identification information in the attenuation model.
Optionally, based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video, a second video recommendation evaluation value of the historical exposure video is generated, and the processing unit 302 is configured to:
and determining the product of the first video recommendation evaluation value of the historical exposure video and the attenuation adjustment value as a second video recommendation evaluation value of the historical exposure video.
Optionally, a corresponding target recommended video set is generated based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video, and the recommending unit 303 is specifically configured to:
sequencing the unexposed videos and the historical exposed videos according to the sequence from high to low of the first video recommendation evaluation value and the second video recommendation evaluation value;
and determining the reordered video set to be recommended as the target recommended video set.
Based on the same inventive concept, fig. 4 exemplarily illustrates a schematic structural diagram of a computing device provided in an embodiment of the present application, and includes at least a memory 401 and a processor 402;
a memory 401 for storing program instructions;
and the processor 402 is used for calling the program instructions stored in the memory and executing the video recommendation method according to the obtained program.
Based on the same inventive concept, the embodiment of the present invention further provides a computer-readable non-volatile storage medium, which includes computer-readable instructions, and when the computer reads and executes the computer-readable instructions, the computer is caused to execute the above video recommendation method.
In conclusion, a set of videos to be recommended is obtained, and a first video recommendation evaluation value of each video to be recommended is calculated; secondly, determining historical exposed videos and unexposed videos in a video set to be recommended; thirdly, adjusting the first video recommended evaluation value of the historical exposure video to a second video recommended evaluation value of the historical exposure video by adopting an attenuation model; and finally, generating and outputting a target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video.
Obviously, an attenuation model is introduced into the recommendation engine of the embodiment of the application, and the attenuation model is used for performing weight reduction processing on the first video recommendation evaluation value of the historical exposure video to generate the second video recommendation evaluation value of the historical exposure video, so that videos with few repeated recommendation times and/or videos which are not recommended can be ranked in front of videos with many repeated recommendation times as much as possible, and the video playing conversion rate is improved.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (10)
1. A method for video recommendation, comprising:
acquiring a video set to be recommended, and acquiring a first video recommendation evaluation value of each video to be recommended in the video set to be recommended based on a pre-trained recommendation model;
determining historical exposed videos and unexposed videos in the video set to be recommended, wherein the historical exposed videos are videos which are recommended to a target object at least once, but the target object only browses unplayed videos;
the preset attenuation model adjusts the first video recommendation evaluation value of the historical exposure video to be a second video recommendation evaluation value of the historical exposure video based on the conversion influence factor of the historical exposure video, wherein the conversion influence factor at least comprises: the historical exposure video is far away from the last exposure day, the total exposure times of the historical exposure video and the position of the historical exposure video in the video set to be recommended; the second video recommendation evaluation value of the historical exposure video is smaller than the first video recommendation evaluation value of the historical exposure video;
and generating a corresponding target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video, and outputting the target recommended video set.
2. The method of claim 1, wherein determining historical exposed videos and unexposed videos in the set of videos to be recommended comprises:
the following operations are respectively executed for each video to be recommended:
judging whether a video to be recommended carries exposure identification information or not;
if so, judging that the video to be recommended is the historical exposure video;
otherwise, judging that the video to be recommended is the unexposed video.
3. The method of claim 1, wherein adjusting the first video recommendation evaluation value of the historical exposure video to the second video recommendation evaluation value of the historical exposure video using a preset attenuation model comprises:
obtaining an attenuation adjustment value of the historical exposure video from the attenuation model, wherein the value of the attenuation adjustment value of the historical exposure video is in negative correlation with the recommended times of the historical exposure video;
and generating a second video recommendation evaluation value of the historical exposure video based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video.
4. The method of claim 3, wherein obtaining attenuation adjustment values for the historical exposure video in the attenuation model comprises:
acquiring video identification information of the historical exposure video;
and determining an attenuation adjustment value corresponding to the video identification information in the attenuation model.
5. The method of claim 4, wherein generating a second video recommendation evaluation value for the historical exposure video based on the attenuation adjustment value and the first video recommendation evaluation value for the historical exposure video comprises:
and determining the product of the first video recommendation evaluation value of the historical exposure video and the attenuation adjustment value as a second video recommendation evaluation value of the historical exposure video.
6. The method of claim 1, wherein generating a respective set of target recommended videos based on the first video recommendation evaluation value for each unexposed video and the second video recommendation evaluation value for each historical exposed video comprises:
sequencing the unexposed videos and the historical exposed videos according to the sequence from high to low of the first video recommendation evaluation value and the second video recommendation evaluation value;
and determining the reordered video set to be recommended as the target recommended video set.
7. A video recommendation server, comprising:
the device comprises an acquisition unit, a recommendation unit and a recommendation unit, wherein the acquisition unit is used for acquiring a video set to be recommended and acquiring a first video recommendation evaluation value of each video to be recommended in the video set to be recommended based on a pre-trained recommendation model;
the processing unit is used for determining historical exposed videos and unexposed videos in the video set to be recommended, wherein the historical exposed videos are videos which are recommended to a target object at least once but are only browsed by the target object and are not played;
the preset attenuation model adjusts the first video recommendation evaluation value of the historical exposure video to be a second video recommendation evaluation value of the historical exposure video based on the conversion influence factor of the historical exposure video, wherein the conversion influence factor at least comprises: the historical exposure video is far away from the last exposure day, the total exposure times of the historical exposure video and the position of the historical exposure video in the video set to be recommended; the second video recommendation evaluation value of the historical exposure video is smaller than the first video recommendation evaluation value of the historical exposure video;
and the recommending unit is used for generating a corresponding target recommended video set based on the first video recommended evaluation value of each unexposed video and the second video recommended evaluation value of each historical exposed video, and outputting the target recommended video set.
8. The server according to claim 7, wherein historical exposed videos and unexposed videos in the video set to be recommended are determined, and the processing unit is configured to:
the following operations are respectively executed for each video to be recommended:
judging whether a video to be recommended carries exposure identification information or not;
if so, judging that the video to be recommended is the historical exposure video;
otherwise, judging that the video to be recommended is the unexposed video.
9. The server according to claim 7, wherein the first video recommendation evaluation value of the historical exposure video is adjusted to the second video recommendation evaluation value of the historical exposure video by using a preset attenuation model, and the processing unit is configured to:
obtaining an attenuation adjustment value of the historical exposure video from the attenuation model, wherein the value of the attenuation adjustment value of the historical exposure video is in negative correlation with the recommended times of the historical exposure video;
and generating a second video recommendation evaluation value of the historical exposure video based on the attenuation adjustment value and the first video recommendation evaluation value of the historical exposure video.
10. The server according to claim 7, wherein a corresponding target recommended video set is generated based on the first video recommendation evaluation value of each unexposed video and the second video recommendation evaluation value of each historical exposed video, and the recommending unit is specifically configured to:
sequencing the unexposed videos and the historical exposed videos according to the sequence from high to low of the first video recommendation evaluation value and the second video recommendation evaluation value;
and determining the reordered video set to be recommended as the target recommended video set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910980081.9A CN110730369B (en) | 2019-10-15 | 2019-10-15 | Video recommendation method and server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910980081.9A CN110730369B (en) | 2019-10-15 | 2019-10-15 | Video recommendation method and server |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110730369A CN110730369A (en) | 2020-01-24 |
| CN110730369B true CN110730369B (en) | 2022-01-04 |
Family
ID=69221298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910980081.9A Active CN110730369B (en) | 2019-10-15 | 2019-10-15 | Video recommendation method and server |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110730369B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112423134B (en) * | 2020-07-06 | 2022-03-29 | 上海哔哩哔哩科技有限公司 | Video content recommendation method and device, computer equipment and storage medium |
| CN112291297B (en) * | 2020-09-04 | 2022-04-26 | 腾讯科技（深圳）有限公司 | Information data processing method, device, storage medium and electronic equipment |
| CN113781086B (en) * | 2021-01-21 | 2024-08-20 | 北京沃东天骏信息技术有限公司 | Article recommendation method, device, medium and electronic equipment |
| CN114268815B (en) * | 2021-12-15 | 2024-08-13 | 北京达佳互联信息技术有限公司 | Video quality determining method, device, electronic equipment and storage medium |
| CN114387056A (en) * | 2021-12-29 | 2022-04-22 | 中国电信股份有限公司 | Information recommendation method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170105036A1 (en) * | 2015-10-08 | 2017-04-13 | Cisco Technology, Inc. | Method and system for segment based recording |
| CN108259939A (en) * | 2017-12-25 | 2018-07-06 | 广州华多网络科技有限公司 | new video push control method, device and server |
| CN109274987A (en) * | 2018-08-30 | 2019-01-25 | 武汉斗鱼网络科技有限公司 | A kind of video collection sort method, server and readable storage medium storing program for executing |
| CN109729395A (en) * | 2018-12-14 | 2019-05-07 | 广州市百果园信息技术有限公司 | Video quality evaluation method, device, storage medium and computer equipment |
| CN109800324A (en) * | 2018-12-18 | 2019-05-24 | 北京达佳互联信息技术有限公司 | Video data recommended method, device, server and storage medium |
-
2019
- 2019-10-15 CN CN201910980081.9A patent/CN110730369B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170105036A1 (en) * | 2015-10-08 | 2017-04-13 | Cisco Technology, Inc. | Method and system for segment based recording |
| CN108259939A (en) * | 2017-12-25 | 2018-07-06 | 广州华多网络科技有限公司 | new video push control method, device and server |
| CN109274987A (en) * | 2018-08-30 | 2019-01-25 | 武汉斗鱼网络科技有限公司 | A kind of video collection sort method, server and readable storage medium storing program for executing |
| CN109729395A (en) * | 2018-12-14 | 2019-05-07 | 广州市百果园信息技术有限公司 | Video quality evaluation method, device, storage medium and computer equipment |
| CN109800324A (en) * | 2018-12-18 | 2019-05-24 | 北京达佳互联信息技术有限公司 | Video data recommended method, device, server and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110730369A (en) | 2020-01-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110730369B (en) | Video recommendation method and server | |
| CN110149540B (en) | Recommendation processing method and device for multimedia resources, terminal and readable medium | |
| CN110929052B (en) | Multimedia resource recommendation method and device, electronic equipment and storage medium | |
| CN109558514B (en) | Video recommendation method, device thereof, information processing equipment and storage medium | |
| US10353944B2 (en) | User-specific media playlists | |
| CN101281540B (en) | Apparatus, method and computer program for processing information | |
| CN110197258A (en) | Neural network searching method, image processing method and device, equipment and medium | |
| CN114896492A (en) | Recommending live streaming content using machine learning | |
| US20070219994A1 (en) | Methods and systems for displaying media utilizing user-generated data | |
| CN110489574B (en) | Multimedia information recommendation method and device and related equipment | |
| CN106649647B (en) | Search result ordering method and device based on artificial intelligence | |
| CN107766360A (en) | A kind of video temperature Forecasting Methodology and device | |
| CN112860937B (en) | KNN and word embedding based mixed music recommendation method, system and equipment | |
| JP2007317068A (en) | Recommending device and recommending system | |
| CN111858969B (en) | Multimedia data recommendation method, device, computer equipment and storage medium | |
| CN112163149B (en) | Method and device for recommending message | |
| CN112836081A (en) | Neural network model training method, information recommendation method and storage medium | |
| US20240107125A1 (en) | System and method for modelling access requests to multi-channel content sharing platforms | |
| CN113592589A (en) | Textile raw material recommendation method and device and processor | |
| CN110569447B (en) | Network resource recommendation method and device and storage medium | |
| CN111143695B (en) | Searching method, searching device, server and storage medium | |
| CN116757783A (en) | Product recommendation method and device | |
| KR20210090706A (en) | Sort | |
| CN110708361A (en) | System, method and device for determining grade of digital content publishing user and server | |
| AU2020335019B2 (en) | Evaluation method based on mobile news client and system thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |