CN114339417B - Video recommendation method, terminal equipment and readable storage medium - Google Patents
Video recommendation method, terminal equipment and readable storage medium Download PDFInfo
- Publication number
- CN114339417B CN114339417B CN202111652195.4A CN202111652195A CN114339417B CN 114339417 B CN114339417 B CN 114339417B CN 202111652195 A CN202111652195 A CN 202111652195A CN 114339417 B CN114339417 B CN 114339417B
- Authority
- CN
- China
- Prior art keywords
- video
- user
- historical
- recommended
- portrait
- 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
- 238000000034 method Methods 0.000 title claims abstract description 89
- 230000006399 behavior Effects 0.000 claims description 24
- 238000012549 training Methods 0.000 claims description 16
- 230000000873 masking effect Effects 0.000 claims description 12
- 238000012216 screening Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 5
- 230000008569 process Effects 0.000 description 23
- 238000004891 communication Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 8
- 230000009471 action Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 230000011664 signaling Effects 0.000 description 2
- 241000237858 Gastropoda Species 0.000 description 1
- 241000282320 Panthera leo Species 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The application provides a video recommendation method, terminal equipment and a readable storage medium, wherein the method comprises the steps of obtaining portrait features of a user watching a terminal, wherein the portrait features of the user comprise operation features and preference features of the user; the method comprises the steps of inputting a historical play sequence of a terminal, portrait characteristics of a user and a video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, wherein the video recommendation model is obtained according to the portrait characteristics of the user and historical play sequence samples. The method can achieve the effect of personalized video recommendation according to different family members.
Description
Technical Field
The present application relates to the field of personalized video recommendation for multimedia terminals, and in particular, to a video recommendation method, a terminal device, and a readable storage medium.
Background
With the development of the mobile internet, the small-screen mobile phone end occupies most of traffic users in the current market. The end users faced by the mobile internet are typically relatively solid, i.e., the preference dimension span of the faced users is not very large. In addition, the small screen end and the user depth interaction are relatively high. Both of these are natural defects of the large screen end.
Users who are targeted for large-screen-end video recommendation are usually home users. The interests of each person in the family members are quite different, so that the recommended content is not the content wanted by the user in the video recommendation process, and great trouble is brought to the user.
Therefore, how to individually recommend videos according to different family members is a technical problem to be solved.
Disclosure of Invention
The embodiment of the application aims to provide a video recommendation method, and the technical scheme of the embodiment of the application can achieve the effect of personalized video recommendation according to different family members.
In a first aspect, an embodiment of the present application provides a method for video recommendation, where the method includes obtaining portrait features of a user of a viewing terminal, where the portrait features of the user include operation features and preference features of the user; the method comprises the steps of inputting a historical play sequence of a terminal, portrait characteristics of a user and a video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, wherein the video recommendation model is obtained according to the portrait characteristics of the user and historical play sequence samples.
In the process, through the portrait features of the user of the viewing terminal, the historical playing sequence of the viewing terminal and the video to be recommended are input into the video recommendation model trained in advance, the recommendation degree of the video to be recommended can be calculated according to the preference of the user, the video recommended to the user is determined according to the recommendation degree, and the effect of recommending the video personalized for the user according to the preference of the user can be achieved.
Optionally, in acquiring the portrait features of the user of the viewing terminal, the method further includes:
Acquiring historical play sequence samples and portrait features of all users watching videos in the historical play sequence;
And training the basic model by utilizing the historical play sequence sample and the portrait characteristics of all users watching the videos in the historical play sequence to obtain a video recommendation model.
In the process, the video in the historical playing sequence in the terminal is taken as a sample, the sample and the portrait characteristics of the user watching the video in the historical playing sequence in the terminal are input into the basic model, the model can be trained according to the characteristics of different users, and the model can be used for recommending the video to the user in a personalized mode according to the hobbies of different users.
Alternatively, training the basic model by using the historical play sequence sample and portrait features of all users watching the video in the historical play sequence to obtain a video recommendation model, including:
randomly masking one historical video sample in the historical play sequence samples to obtain a masked historical play sequence sample;
Predicting a historical video sample according to the masked historical play sequence sample and the image characteristics of the user to obtain a prediction result, wherein the prediction result comprises a predicted video and the recommendation degree of the predicted video;
and adjusting the basic model according to the difference between the prediction result and the historical video sample to obtain a video recommendation model.
In the process, masking is carried out on one historical video in the historical playing sequence, prediction is carried out on the portrait characteristics of different users and the historical video in the masked historical playing sequence, the video and the recommendation degree of the video can be obtained, and a model is adjusted according to the difference between the predicted result and the masking video, so that the model is more accurate.
Optionally, obtaining the portrait features of the user of the viewing terminal includes:
and acquiring portrait characteristics of the user according to the operation behaviors of the user and/or according to the time of the user watching the terminal.
In the above process, by the operation of the user and/or the time of the user watching the terminal, it can be deduced which user is watching the terminal, and further the portrait features of the user are obtained according to the user.
Optionally, before the historical playing sequence of the terminal, the portrait characteristic of the user and the video set to be recommended are input into the pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, the method further comprises:
acquiring an initial video set to be screened in a system, wherein the initial video set to be screened comprises all videos in the system;
and screening the initial set to be screened according to the portrait characteristics of the user to obtain a set of video to be recommended.
In the process, before the video to be recommended is input into the model, the videos in the system are screened according to the portrait features of the user, so that disliked videos of some users can be filtered, the input of model data can be reduced, a large amount of time is saved, and the error rate is reduced.
The video in the system can be sourced from databases of various websites or from local video libraries, and the application is not limited to the above.
Optionally, after the historical playing sequence of the terminal, the portrait characteristic of the user and the video set to be recommended are input into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, the method further comprises:
according to the recommendation degree of each video, ordering each video in the video set to be recommended in a descending order to obtain an ordered recommended video set;
And recommending the ordered recommended video set to the user.
In the process, through descending order sequencing of the videos to be recommended, the videos which are more preferred by the user can be placed in front, so that the user can watch according to the preference, and better experience is brought.
Optionally, after the historical playing sequence of the terminal, the portrait characteristic of the user and the video set to be recommended are input into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, the method further comprises:
comparing the recommendation degree of each video in the video set to be recommended with a recommendation threshold value to obtain a comparison result;
and recommending the videos corresponding to the videos in the video set to be recommended, the recommendation degree of which is greater than the recommendation threshold value, to the user according to the comparison result.
In the above process, a recommendation threshold may be set, and only when the recommendation degree of the video to be recommended output by the model is greater than the recommendation threshold, the corresponding video may be recommended to the user.
In a second aspect, an embodiment of the present application provides a terminal device, including:
The acquisition module is used for acquiring portrait features of a user of the viewing terminal, wherein the portrait features of the user comprise operation features and preference features of the user;
The input module is used for inputting the historical play sequence of the terminal, the portrait characteristic of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, wherein the video recommendation model is obtained according to the portrait characteristic of the user and the historical play sequence sample.
Optionally, the terminal device further includes:
The training module is used for acquiring historical play sequence samples and portrait characteristics of all users watching videos in the historical play sequence before acquiring portrait characteristics of users watching the terminal;
And training the basic model by utilizing the historical play sequence sample and the portrait characteristics of all users watching the videos in the historical play sequence to obtain a video recommendation model.
Optionally, the training module is specifically configured to:
randomly masking one historical video sample in the historical play sequence samples to obtain a masked historical play sequence sample;
Predicting a historical video sample according to the masked historical play sequence sample and the image characteristics of the user to obtain a prediction result, wherein the prediction result comprises a predicted video and the recommendation degree of the predicted video;
and adjusting the basic model according to the difference between the prediction result and the historical video sample to obtain a video recommendation model.
Optionally, the acquiring module is specifically configured to:
and acquiring portrait characteristics of the user according to the operation behaviors of the user and/or according to the time of the user watching the terminal.
Optionally, the terminal device further includes:
The screening module is used for acquiring an initial video set to be screened in the system before the input module inputs the historical play sequence of the terminal, the portrait characteristics of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, wherein the initial video set to be screened comprises all videos in the system;
and screening the initial set to be screened according to the portrait characteristics of the user to obtain a set of video to be recommended.
Optionally, the terminal device further includes:
the first recommendation module is used for inputting the historical play sequence of the terminal, the portrait characteristics of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, and then ordering each video in the video set to be recommended in a descending order according to the recommendation degree of each video to obtain an ordered recommended video set;
And recommending the ordered recommended video set to the user.
Optionally, the terminal device further includes:
The second recommendation module is used for comparing the recommendation degree of each video in the video set to be recommended with a recommendation threshold value after the input module inputs the historical play sequence of the terminal, the portrait characteristic of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, so as to obtain a comparison result;
and recommending the videos corresponding to the videos in the video set to be recommended, the recommendation degree of which is greater than the recommendation threshold value, to the user according to the comparison result.
In a third aspect, an embodiment of the present application provides an electronic device comprising a processor and a memory storing computer readable instructions which, when executed by the processor, perform the steps of the method as provided in the first aspect above.
In a fourth aspect, embodiments of the present application provide a readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the method as provided in the first aspect above.
Additional features and advantages of the application will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the embodiments of the application. The objectives and other advantages of the application will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments of the present application will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and should not be considered as limiting the scope, and other related drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for video recommendation according to an embodiment of the present application;
FIG. 2 is a detailed flowchart of a method for video recommendation according to an embodiment of the present application;
Fig. 3 is a schematic block diagram of a terminal device according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of a terminal device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be described below with reference to the accompanying drawings in the embodiments of the present application.
It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. Meanwhile, in the description of the present application, the terms "first", "second", and the like are used only to distinguish the description, and are not to be construed as indicating or implying relative importance.
The method is applied to the scenes of video personalized recommendation of the intelligent internet digital television multi-terminal users, and the specific scenes are the scenes that the user can acquire the user preference according to the operation behaviors of the user at different times when the user uses the intelligent multimedia terminal, and the user can recommend some videos for the user in a personalized mode according to the user preference.
But users facing large-screen-end video recommendation are generally home users. Family members have great differences in their habits of age, sex, interests and viewing behavior. The user characteristics to be considered when recommending the view are more comprehensive. Such as: the household users watch different contents on holidays and the like due to different watching behaviors in the morning and evening. The weight of the recommended video is greatly affected.
In the field of mobile internet, a more excellent deep learning CTR (click through rate) model is already developed, and the effect is better than that of a traditional model, but the video recommendation scene of a home terminal is quite different from that of an internet video recommendation service mode, some deep learning ordering models applied to families at present are user features and commodity features, cross features, sequence features are only click play or search, but operations such as simple average value taking, maximum value taking and the like are not performed, a model with finer granularity is not provided, the characteristics of the home user are not obvious and clear, the viewing history of the home user is generally considered to be passive, low-frequency, but the interests of the user of the home terminal are also wide and dynamic, and the user needs to experience personalized recommendation experience, so the center of gravity should be placed on the sequence of behaviors.
In summary, two points are considered, firstly, a designed model constructs a modeling family portrait according to a historical sequence, and meanwhile, in the same historical viewing sequence, the weights for predicting the influence of the next program clicking of a user are different, so that when the predicted targets are different, different attention programs are given to videos watched by the user according to the targets when the family portrait model is designed. Based on the consideration, the application recommends a plurality of videos for the user in a personalized way according to the characteristic information such as the preference characteristics, the operation behaviors, the operation time and the like of the user by preferentially acquiring the portrait characteristics of the user watching the terminal. The recommendation of the self-attention video recommendation model can be responded to the real-time recommendation request service of the terminal, and finally, the user who is in front of the current television can be distinguished, and the user can make a judgment to conduct personalized recommendation according to the historical behaviors and the real-time behaviors of the user, so that the home-end multi-user personalized accurate real-time recommendation is achieved.
The user watching the terminal can be the user just opening the terminal, the user watching the video or watching the video program, and the like, and the terminal can be multimedia terminal equipment such as a television, a computer, a mobile phone, and the like, and the application is not limited to the above, and the application can be the recommended video personalized for the user according to some operation behaviors, operation time, preference characteristics, and the like of the user under different conditions, and the portrait characteristics can be some data of the user, including but not limited to age, sex, family member structure, whether children, regions, members, liveness, video primary-secondary classification preference, preference time period, preference week number or cold-day fake, video title, description, primary classification, director, actor guest, winner, year, language, play behavior, search behavior, watching duration, time, page, scene, and the like. Some manually constructed cross higher order features are also possible, such as: the collection of features such as online preference time period, day of week and summer holidays can also be used for constructing cross features according to some behaviors, for example: preference time period, play sequence, search sequence, etc., some derivative features and non-higher order features, etc., may also be provided according to the above features, and the application is not limited thereto.
The method of video recommendation according to an embodiment of the present application is described in detail below with reference to fig. 1.
Referring to fig. 1, fig. 1 is a flowchart of a video recommendation method provided in an embodiment of the present application, and the method applied to a terminal device, where the method for video recommendation shown in fig. 1 includes:
110: and obtaining the portrait features of the user of the viewing terminal.
Wherein the portrait features of the user include one or more of the following features, for example: the user's operating characteristics, preference characteristics, age, gender, family member structure, whether children are present, regions are members, liveness, video one-level classification preference, preference period, preference week number or cold and hot false, video title, description, one-level classification, director, actor guests, prize winning, age, language, play behavior, search behavior, viewing duration, time of occurrence, page, scene or a combination of the above features, and the like. The user viewing the terminal may be when the terminal is turned on or while the terminal is being used. When the portrait features of the user are acquired, the portrait features can be manually input or accumulated and stored by the user continuously watching the video. The user's portrait may be obtained by estimating the user based on the portrait characteristics of the user, some operations of the user, or the time of using the terminal, or by directly inputting the user in a search box of the terminal to confirm the user.
Optionally, obtaining the portrait features of the user of the viewing terminal includes:
and acquiring portrait characteristics of the user according to the operation behaviors of the user and/or according to the time of the user watching the terminal.
In the above process, by the operation of the user and/or the time of the user watching the terminal, it can be deduced which user is watching the terminal, and further the portrait features of the user are obtained according to the user.
The operation behavior of the user may be related behavior such as searching behavior, playing channel, and viewing time, and the time when the user views the terminal is, for example: the present application is not limited to this, and the weekends are weekends, weekdays, cold and summer holidays, school hours, and working hours.
Optionally, before obtaining the portrait features of the user of the viewing terminal, the method shown in fig. 1 may further include:
Acquiring historical play sequence samples and portrait features of all users watching videos in the historical play sequence;
And training the basic model by utilizing the historical play sequence sample and the portrait characteristics of all users watching the videos in the historical play sequence to obtain a video recommendation model.
In the process, the video in the historical playing sequence in the terminal is taken as a sample, the sample and the portrait characteristics of the user watching the video in the historical playing sequence in the terminal are input into the basic model, the model can be trained according to the characteristics of different users, and the model can be used for recommending the video to the user in a personalized mode according to the hobbies of different users.
The historical play sequence sample can be all videos in the historical play sequence; it may also be a video of a recent period of time of a historical play sequence, for example: video within two weeks from the current time; it may also be part of the video in the historical play sequence closest to the current time, for example: according to the 1000 most recent videos at the current time. The basic model can select the existing video recommendation model, and the model can be trained to realize personalized video recommendation aiming at different users according to the preference of the different users.
Alternatively, training the basic model by using the historical play sequence sample and portrait features of all users watching the video in the historical play sequence to obtain a video recommendation model, including:
randomly masking one historical video sample in the historical play sequence samples to obtain a masked historical play sequence sample;
Predicting a historical video sample according to the masked historical play sequence sample and the image characteristics of the user to obtain a prediction result, wherein the prediction result comprises a predicted video and the recommendation degree of the predicted video;
and adjusting the basic model according to the difference between the prediction result and the historical video sample to obtain a video recommendation model.
In the process, masking is carried out on one historical video in the historical playing sequence, prediction is carried out on the portrait characteristics of different users and the historical video in the masked historical playing sequence, the video and the recommendation degree of the video can be obtained, and a model is adjusted according to the difference between the predicted result and the masking video, so that the model is more accurate.
The method comprises the steps that the steps are needed to be carried out for a plurality of times, each video in a sequence is required to be masked, the historical video is predicted according to the portrait characteristics of different users and the historical play sequence samples after masking, the video in the historical play sequence has a corresponding relation with different users, when the model is trained, the video in the historical video sequence watched by the different users is preferentially determined, the video is assigned with larger weight, and then the masked historical video is predicted, so that accurate model training is realized. In the prediction process, portrait features of different users and historical videos watched by the users play a greater role on prediction results, for example: one terminal device historically plays three videos, namely limit challenge, ottman and snail, and to predict the probability of a user clicking on the life of the user, the limit challenge should be assigned more weight in the recommendation, and also ottman should play a larger role in predicting whether the user will click on one of the small lions, and so on.
120: And inputting the historical play sequence of the terminal, the portrait characteristic of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended.
In the above process, the portrait characteristic of the current user, the historical playing sequence in the terminal and the video to be recommended of the system are input into the video recommendation model, each video to be recommended can be scored, the corresponding recommendation degree is obtained, and the video can be recommended to the user according to the preference of the current user.
The video recommendation model is obtained by training the video recommendation model according to the portrait characteristics of the user and the historical play sequence samples.
The application mainly refers to a multi-head self-attention model which is different from the training process of other attention mechanisms and introduces mask masking ideas. The main construction composition of the network is firstly multi-head self-attention with a shade, then a dropout (taking off network neurons and not participating in training) layer, some neurons are randomly removed according to a certain self-defining method, generalization capability is increased, and then a skip connection layer and a layerNorm layer are used for preventing network gradient elimination and explosion. Then, the feed-forward neural network follows, with the dropout and skip connection layers and layerNorm layers being connected later similarly as above. The training stage adopts a mask idea, namely, a random mask is adopted in an input sequence, the position is replaced by an identifier of 'mask', the content id (Identity document identity identification number) index of the corresponding position is replaced, the same operation, indexing and embedding are carried out on the identifier of 'mask', and then the final loss soft layer is reached, namely, the replaced id is predicted, in addition, a plurality of attention models are designed, and the models are arranged in parallel, namely, the models are similar to a plurality of convolution kernels in the image field, each convolution kernel is responsible for attention of different aspects, and the network design can enable the models to fully learn the association relation and mutual influence of the user history sequence. When the model is used, the mask is only required to be added to the last position of the sequence, the multi-head self-attention model output only needs to take the last column to represent the content vector to be clicked by the user, and then the clicking probability, interest preference degree or score and the like of the user on the video to be recommended are output through mlp (Mobile Location Protocol mobile positioning protocol) and softmax (logistic regression).
Optionally, before the historical playing sequence of the terminal, the portrait characteristic of the user and the video set to be recommended are input into the pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, the method shown in fig. 1 may further include:
acquiring an initial video set to be screened in a system, wherein the initial video set to be screened comprises all videos in the system;
and screening the initial set to be screened according to the portrait characteristics of the user to obtain a set of video to be recommended.
In the process, before the video to be recommended is input into the model, the videos in the system are screened according to the portrait features of the user, so that disliked videos of some users can be filtered, the input of model data can be reduced, a large amount of time is saved, and the error rate is reduced.
Based on the representation of the user and the historical operational behavior of the user over a longer period of time, thousand-sided personalized recommendation systems typically include two steps of recall and sorting. The main functions of the recall phase are: from a large number of candidate articles, adopting a specified rule to quickly screen and reduce the initial video recommendation set to thousands or hundreds of levels according to the portrait characteristics of the user. Typically, a CTR (click through rate) model is used as a ranking model, and a click rate prediction model (input user portrait features, candidate content features, user content cross features, etc.) is used to rank recalled content. The sorting stage can be used for carrying out preference scoring and sorting on the video contents recalled in multiple ways and then recommending the video contents to the user, and can also be used for selecting a plurality of video contents which are most suitable for the hobbies of the user according to a threshold value set in advance and recommending the video contents to the user.
Optionally, when preference scoring and sorting are performed on the multiple recalled video content and then recommending the video content to the user, after the historical playing sequence of the terminal, the portrait characteristic of the user and the video set to be recommended are input into the pre-trained video recommendation model, the method shown in fig. 1 may further include:
according to the recommendation degree of each video, ordering each video in the video set to be recommended in a descending order to obtain an ordered recommended video set;
And recommending the ordered recommended video set to the user.
In the process, through descending order sequencing of the videos to be recommended, the videos which are more preferred by the user can be placed in front, so that the user can watch according to the preference, and better experience is brought.
Optionally, when a plurality of video contents which best meet the user preference are selected and recommended to the user according to the threshold value set in advance. After the historical playing sequence of the terminal, the portrait characteristic of the user and the video set to be recommended are input into the pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, the method shown in fig. 1 may further include:
comparing the recommendation degree of each video in the video set to be recommended with a recommendation threshold value to obtain a comparison result;
and recommending the videos corresponding to the videos in the video set to be recommended, the recommendation degree of which is greater than the recommendation threshold value, to the user according to the comparison result.
In the above process, a recommendation threshold may be set, and only when the recommendation degree of the video to be recommended output by the model is greater than the recommendation threshold, the corresponding video may be recommended to the user.
The method of video recommendation was described above by means of fig. 1, and the detailed steps of the method of video recommendation are described below in connection with fig. 2.
Referring to fig. 2, fig. 2 is a detailed flowchart of a video recommendation method according to an embodiment of the present application:
210: user data input.
The input may be operation data of some users, for example: the related behavior data such as searching behavior, playing channel and watching time length can also be data about time when some users use the terminal, for example: weekends are weekdays, cold and summer holidays are school hours or working hours, and the like.
220: And acquiring the user portrait.
According to the input of the user data, the portrait features of the user can be obtained from the memory.
230: The input model includes: user portraits, historical sequences, and system video.
And inputting the portrait characteristics of the user, the historical sequence played by the terminal and the video to be recommended in the system into a model to be recommended.
240: And outputting data.
And recommending the video preferred by the user for the user according to the portrait characteristic of the user.
The method of video recommendation is described above by means of fig. 1-2, and the terminal device of video recommendation is described below in connection with fig. 3-4.
Referring to fig. 3, a schematic block diagram of a terminal device 300 according to an embodiment of the present application is shown, where the terminal device 300 may be a module, a program segment, or a code on an electronic device. The terminal device 300 corresponds to the above-described embodiment of the method of fig. 1, and is capable of executing the steps involved in the embodiment of the method of fig. 1, and specific functions of the terminal device 300 may be referred to as the following description, and detailed descriptions are omitted herein as appropriate to avoid redundancy.
Optionally, the terminal device 300 includes:
An acquisition module 310, configured to acquire portrait features of a user of the viewing terminal, where the portrait features of the user include operation features and preference features of the user;
the input module 320 is configured to input the historical play sequence of the terminal, the portrait characteristic of the user, and the video set to be recommended into a pre-trained video recommendation model, to obtain the recommendation degree of each video in the video set to be recommended, where the video recommendation model is obtained from the video recommendation model according to the portrait characteristic of the user and the historical play sequence sample.
Optionally, the terminal device further includes:
The training module is used for acquiring historical play sequence samples and portrait characteristics of all users watching videos in the historical play sequence before acquiring portrait characteristics of users watching the terminal;
And training the basic model by utilizing the historical play sequence sample and the portrait characteristics of all users watching the videos in the historical play sequence to obtain a video recommendation model.
Optionally, the training module is specifically configured to:
randomly masking one historical video sample in the historical play sequence samples to obtain a masked historical play sequence sample;
Predicting a historical video sample according to the masked historical play sequence sample and the image characteristics of the user to obtain a prediction result, wherein the prediction result comprises a predicted video and the recommendation degree of the predicted video;
and adjusting the basic model according to the difference between the prediction result and the historical video sample to obtain a video recommendation model.
Optionally, the acquiring module is specifically configured to:
and acquiring portrait characteristics of the user according to the operation behaviors of the user and/or according to the time of the user watching the terminal.
Optionally, the terminal device further includes:
The screening module is used for acquiring an initial video set to be screened in the system before the input module inputs the historical play sequence of the terminal, the portrait characteristics of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, wherein the initial video set to be screened comprises all videos in the system;
and screening the initial set to be screened according to the portrait characteristics of the user to obtain a set of video to be recommended.
Optionally, the terminal device further includes:
the first recommendation module is used for inputting the historical play sequence of the terminal, the portrait characteristics of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, and then ordering each video in the video set to be recommended in a descending order according to the recommendation degree of each video to obtain an ordered recommended video set;
And recommending the ordered recommended video set to the user.
Optionally, the terminal device further includes:
The second recommendation module is used for comparing the recommendation degree of each video in the video set to be recommended with a recommendation threshold value after the input module inputs the historical play sequence of the terminal, the portrait characteristic of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, so as to obtain a comparison result;
and recommending the videos corresponding to the videos in the video set to be recommended, the recommendation degree of which is greater than the recommendation threshold value, to the user according to the comparison result.
Referring to fig. 4, a schematic block diagram of a terminal device 400 according to an embodiment of the present application may include a processor 420 and a memory 410. Optionally, the terminal device may further include: a communication interface 430 and a communication bus 440. The terminal device corresponds to the embodiment of the method of fig. 1, and can execute the steps involved in the embodiment of the method of fig. 1, and specific functions of the terminal device can be referred to in the following description.
In particular, the memory 410 is used to store computer readable instructions.
The processor 420, configured to process the readable instructions stored in the memory, is capable of performing the steps of the method embodiments 1 to 3 of fig. 2.
Communication interface 430 is used for signaling or data communication with other node devices. For example: for communication with a server or terminal, or with other device nodes, although embodiments of the application are not limited in this regard.
A communication bus 440 for enabling direct connection communication of the above-described components.
The communication interface 430 of the device in the embodiment of the present application is used for performing signaling or data communication with other node devices. The memory 410 may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one disk memory. Memory 410 may also optionally be at least one storage device located remotely from the aforementioned processor. The memory 410 has stored therein computer readable instructions which, when executed by the processor 420, perform the method process described above in fig. 1. Processor 420 may be used on terminal device 300 and for performing functions in the present application. By way of example, the Processor 420 described above may be a general purpose Processor, a digital signal Processor (DIGITAL SIGNAL Processor, DSP), an Application SPECIFIC INTEGRATED Circuit (ASIC), an off-the-shelf programmable gate array (Field Programmable GATE ARRAY, FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, to which embodiments of the application are not limited.
Embodiments of the present application also provide a readable storage medium, which when executed by a processor, performs a method process performed by an electronic device in the method embodiment shown in fig. 1.
It will be clear to those skilled in the art that, for convenience and brevity of description, reference may be made to the corresponding procedure in the foregoing method for the specific working procedure of the apparatus described above, and this will not be repeated here.
In summary, the embodiments of the present application provide a method, a terminal device, and a readable storage medium for video recommendation, where the method includes obtaining portrait characteristics of a user viewing a terminal, where the portrait characteristics of the user include operation characteristics and preference characteristics of the user; the method comprises the steps of inputting a historical play sequence of a terminal, portrait characteristics of a user and a video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, wherein the video recommendation model is obtained according to the portrait characteristics of the user and historical play sequence samples. The method can achieve the effect of personalized video recommendation according to different family members.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The apparatus embodiments described above are merely illustrative, for example, of the flowcharts and block diagrams in the figures that illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, functional modules in the embodiments of the present application may be integrated together to form a single part, or each module may exist alone, or two or more modules may be integrated to form a single part.
The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The above description is only an example of the present application and is not intended to limit the scope of the present application, and various modifications and variations will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application. It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.
The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
Claims (8)
1. A method for video recommendation, applied to a terminal, comprising:
Acquiring portrait features of a user watching the terminal, wherein the portrait features of the user comprise operation features and preference features of the user;
Inputting the historical play sequence of the terminal, the portrait characteristic of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, wherein the video recommendation model is obtained by training the video recommendation model according to the portrait characteristic of the user and the historical play sequence sample, and the historical play sequence comprises all historical videos of the terminal or videos in a period of time;
Before the obtaining of the portrait features of the user viewing the terminal, the method further comprises: acquiring the historical play sequence sample and portrait characteristics of all users watching videos in the historical play sequence; randomly masking one historical video sample in the historical play sequence samples to obtain masked historical play sequence samples; predicting the historical video sample according to the masked historical play sequence sample and the portrait characteristic of the user to obtain a prediction result, wherein the prediction result comprises a predicted video and the recommendation degree of the predicted video; and adjusting a basic model according to the difference between the prediction result and the historical video sample to obtain the video recommendation model.
2. The method of claim 1, wherein the obtaining the portrait features of the user viewing the terminal comprises:
and acquiring portrait features of the user according to the operation behaviors of the user and/or according to the time when the user views the terminal.
3. The method according to claim 1 or 2, wherein before the inputting the historical play sequence of the terminal, the portrait characteristics of the user, and the video to be recommended set into the pre-trained video recommendation model, the method further comprises:
Acquiring an initial video set to be screened in a system, wherein the initial video set to be screened comprises all videos in the system;
And screening the initial set to be screened according to the portrait characteristics of the user to obtain the set of video to be recommended.
4. The method according to claim 1 or 2, wherein after the inputting the historical play sequence of the terminal, the portrait characteristic of the user, and the video to be recommended set into a pre-trained video recommendation model, obtaining the recommendation degree of each video in the video to be recommended set, the method further comprises:
according to the recommendation degree of each video, ordering each video in the video set to be recommended in a descending order to obtain an ordered recommended video set;
And recommending the ordered recommended video set to the user.
5. The method according to claim 1 or 2, wherein after the inputting the historical play sequence of the terminal, the portrait characteristic of the user, and the video to be recommended set into a pre-trained video recommendation model, obtaining the recommendation degree of each video in the video to be recommended set, the method further comprises:
Comparing the recommendation degree of each video in the video set to be recommended with a recommendation threshold value to obtain a comparison result;
And recommending the video corresponding to the video with the recommendation degree of the video in the video set to be recommended being greater than the recommendation threshold to the user according to the comparison result.
6. A terminal device, comprising:
The acquisition module is used for acquiring portrait features of a user watching the terminal, wherein the portrait features of the user comprise operation features and preference features of the user;
The input module is used for inputting the historical play sequence of the terminal, the portrait characteristic of the user and the video set to be recommended into a pre-trained video recommendation model to obtain the recommendation degree of each video in the video set to be recommended, wherein the video recommendation model is obtained by the video recommendation model according to the portrait characteristic of the user and the historical play sequence sample, and the historical play sequence comprises all historical videos of the terminal or videos in a period of time;
Before the acquisition module acquires the portrait features of the user who views the terminal, the acquisition module is further configured to: acquiring the historical play sequence sample and portrait characteristics of all users watching videos in the historical play sequence; randomly masking one historical video sample in the historical play sequence samples to obtain masked historical play sequence samples; predicting the historical video sample according to the masked historical play sequence sample and the portrait characteristic of the user to obtain a prediction result, wherein the prediction result comprises a predicted video and the recommendation degree of the predicted video; and adjusting a basic model according to the difference between the prediction result and the historical video sample to obtain the video recommendation model.
7. A terminal device, comprising:
a memory and a processor, the memory storing computer readable instructions that when executed by the processor perform the steps in the method of any one of claims 1 to 5.
8. A computer-readable storage medium, comprising:
computer program which, when run on a computer, causes the computer to perform the method of any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111652195.4A CN114339417B (en) | 2021-12-30 | 2021-12-30 | Video recommendation method, terminal equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111652195.4A CN114339417B (en) | 2021-12-30 | 2021-12-30 | Video recommendation method, terminal equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114339417A CN114339417A (en) | 2022-04-12 |
| CN114339417B true CN114339417B (en) | 2024-05-10 |
Family
ID=81018871
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111652195.4A Active CN114339417B (en) | 2021-12-30 | 2021-12-30 | Video recommendation method, terminal equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114339417B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114722238B (en) * | 2022-05-13 | 2022-09-30 | 北京达佳互联信息技术有限公司 | Video recommendation method and device, electronic equipment, storage medium and program product |
| CN115695897B (en) * | 2022-09-29 | 2024-04-30 | 中国广播电视网络集团有限公司 | Multi-role-based multi-screen interactive content recommendation method, storage medium and equipment |
| CN116821475B (en) * | 2023-05-19 | 2024-02-02 | 广州蜜糖网络科技有限公司 | Video recommendation method and device based on client data and computer equipment |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105930425A (en) * | 2016-04-18 | 2016-09-07 | 乐视控股（北京）有限公司 | Personalized video recommendation method and apparatus |
| EP3416390A1 (en) * | 2017-06-12 | 2018-12-19 | Facebook, Inc. | Interactive spectating interface for live videos |
| WO2019051835A1 (en) * | 2017-09-18 | 2019-03-21 | 华为技术有限公司 | Recommendation method and terminal |
| CN110365994A (en) * | 2019-07-18 | 2019-10-22 | 广州虎牙科技有限公司 | Recommended method, device, server and readable storage medium storing program for executing is broadcast live |
| CN110704674A (en) * | 2019-09-05 | 2020-01-17 | 苏宁云计算有限公司 | Video playing integrity prediction method and device |
| CN111008332A (en) * | 2019-12-03 | 2020-04-14 | 腾讯科技（深圳）有限公司 | Content item recommendation method, device, server and storage medium |
| CN111050193A (en) * | 2019-11-12 | 2020-04-21 | 汉口北进出口服务有限公司 | User portrait construction method and device, computer equipment and storage medium |
| CN111339355A (en) * | 2020-05-21 | 2020-06-26 | 北京搜狐新媒体信息技术有限公司 | Video recommendation method and system |
| CN111669622A (en) * | 2020-06-10 | 2020-09-15 | 北京奇艺世纪科技有限公司 | Method and device for determining default play relationship of videos and electronic equipment |
| CN111966909A (en) * | 2020-08-26 | 2020-11-20 | 腾讯科技（深圳）有限公司 | Video recommendation method and device, electronic equipment and computer-readable storage medium |
| CN112328778A (en) * | 2020-11-03 | 2021-02-05 | 腾讯科技（深圳）有限公司 | Method, apparatus, device and medium for determining user characteristics and model training |
| CN112364203A (en) * | 2020-11-30 | 2021-02-12 | 未来电视有限公司 | Television video recommendation method, device, server and storage medium |
| CN112417207A (en) * | 2020-11-24 | 2021-02-26 | 未来电视有限公司 | Video recommendation method, device, equipment and storage medium |
| CN113139856A (en) * | 2021-05-14 | 2021-07-20 | 青岛聚看云科技有限公司 | Movie and television member package recommendation method and device |
| CN113268633A (en) * | 2021-06-25 | 2021-08-17 | 北京邮电大学 | Short video recommendation method |
| CN113568769A (en) * | 2020-04-28 | 2021-10-29 | 北京达佳互联信息技术有限公司 | Exception handling method, exception handling device, server and storage medium |
| CN113613075A (en) * | 2021-08-11 | 2021-11-05 | 苏州律点信息科技有限公司 | Video recommendation method and device and cloud server |
-
2021
- 2021-12-30 CN CN202111652195.4A patent/CN114339417B/en active Active
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105930425A (en) * | 2016-04-18 | 2016-09-07 | 乐视控股（北京）有限公司 | Personalized video recommendation method and apparatus |
| EP3416390A1 (en) * | 2017-06-12 | 2018-12-19 | Facebook, Inc. | Interactive spectating interface for live videos |
| WO2019051835A1 (en) * | 2017-09-18 | 2019-03-21 | 华为技术有限公司 | Recommendation method and terminal |
| CN110365994A (en) * | 2019-07-18 | 2019-10-22 | 广州虎牙科技有限公司 | Recommended method, device, server and readable storage medium storing program for executing is broadcast live |
| CN110704674A (en) * | 2019-09-05 | 2020-01-17 | 苏宁云计算有限公司 | Video playing integrity prediction method and device |
| CN111050193A (en) * | 2019-11-12 | 2020-04-21 | 汉口北进出口服务有限公司 | User portrait construction method and device, computer equipment and storage medium |
| CN111008332A (en) * | 2019-12-03 | 2020-04-14 | 腾讯科技（深圳）有限公司 | Content item recommendation method, device, server and storage medium |
| CN113568769A (en) * | 2020-04-28 | 2021-10-29 | 北京达佳互联信息技术有限公司 | Exception handling method, exception handling device, server and storage medium |
| CN111339355A (en) * | 2020-05-21 | 2020-06-26 | 北京搜狐新媒体信息技术有限公司 | Video recommendation method and system |
| CN111669622A (en) * | 2020-06-10 | 2020-09-15 | 北京奇艺世纪科技有限公司 | Method and device for determining default play relationship of videos and electronic equipment |
| CN111966909A (en) * | 2020-08-26 | 2020-11-20 | 腾讯科技（深圳）有限公司 | Video recommendation method and device, electronic equipment and computer-readable storage medium |
| CN112328778A (en) * | 2020-11-03 | 2021-02-05 | 腾讯科技（深圳）有限公司 | Method, apparatus, device and medium for determining user characteristics and model training |
| CN112417207A (en) * | 2020-11-24 | 2021-02-26 | 未来电视有限公司 | Video recommendation method, device, equipment and storage medium |
| CN112364203A (en) * | 2020-11-30 | 2021-02-12 | 未来电视有限公司 | Television video recommendation method, device, server and storage medium |
| CN113139856A (en) * | 2021-05-14 | 2021-07-20 | 青岛聚看云科技有限公司 | Movie and television member package recommendation method and device |
| CN113268633A (en) * | 2021-06-25 | 2021-08-17 | 北京邮电大学 | Short video recommendation method |
| CN113613075A (en) * | 2021-08-11 | 2021-11-05 | 苏州律点信息科技有限公司 | Video recommendation method and device and cloud server |
Non-Patent Citations (3)
| Title |
|---|
| Learning to Recommend Frame for Interactive Video Object Segmentation in the Wild;Zhaoyuan Yin;《2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)》;20211102;全文 * |
| 基于用户行为数据分析的短视频移动产品设计;严家卓;《中国优秀硕士学位论文全文数据库》;20210615;全文 * |
| 大数据时代基于用户画像的视频精准推荐;王正友;《电子商务》;20191030(第10期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114339417A (en) | 2022-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114339417B (en) | Video recommendation method, terminal equipment and readable storage medium | |
| CN108875022B (en) | Video recommendation method and device | |
| Gomez-Uribe et al. | The netflix recommender system: Algorithms, business value, and innovation | |
| CN104462292B (en) | Socially collaborative filtering | |
| JP2005521144A (en) | Recommendation system using multiple recommendation scores | |
| CN111274440B (en) | Video recommendation method based on visual and audio content relevancy mining | |
| CN103747343A (en) | Method and device for resource day parting recommendation | |
| CN108521586B (en) | IPTV television program personalized recommendation method giving consideration to time context and implicit feedback | |
| CN112507163B (en) | Duration prediction model training method, recommendation method, device, equipment and medium | |
| KR101372928B1 (en) | Apparatus, System, Method and Computer Readable Recording Media Storing the Program for Automatic Recommendation of TV Program Contents based on User-Preferred Topic | |
| CN111062527B (en) | Video traffic collection prediction method and device | |
| CN111858969B (en) | Multimedia data recommendation method, device, computer equipment and storage medium | |
| Khadse et al. | Recommendation engine for predicting best rated movies | |
| Yuyan et al. | A novel movie recommendation system based on deep reinforcement learning with prioritized experience replay | |
| CN115455280A (en) | Recommendation list determining method and server | |
| US9020863B2 (en) | Information processing device, information processing method, and program | |
| CN110781377A (en) | Article recommendation method and device | |
| CN113468413B (en) | Multi-user sharing-oriented multimedia network video recommendation method | |
| CN113852867B (en) | Method and device for recommending programs based on kernel density estimation | |
| Li et al. | Personalized TV recommendation with mixture probabilistic matrix factorization | |
| Singh et al. | A Recommender System for YouTube Video based on deep neural network | |
| CN113761364B (en) | Multimedia data pushing method and device | |
| CN117651165B (en) | Video recommendation method and device based on client data | |
| CN117633371B (en) | Recommendation method, device and readable storage medium based on multi-attention mechanism | |
| CN112650884B (en) | Video pushing method based on deep learning |
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 |