CN112507163B - Duration prediction model training method, recommendation method, device, equipment and medium - Google Patents
Duration prediction model training method, recommendation method, device, equipment and medium Download PDFInfo
- Publication number
- CN112507163B CN112507163B CN202011405977.3A CN202011405977A CN112507163B CN 112507163 B CN112507163 B CN 112507163B CN 202011405977 A CN202011405977 A CN 202011405977A CN 112507163 B CN112507163 B CN 112507163B
- Authority
- CN
- China
- Prior art keywords
- multimedia
- multimedia data
- data
- sample
- historical
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to a training method, a recommending method, a device, equipment and a medium for a duration prediction model, wherein the method comprises the following steps: constructing a duration prediction model; acquiring a multimedia data sample library, wherein the multimedia data sample library comprises: historical multimedia sample data, and tag-associated multimedia sample data based on the historical multimedia sample data; extracting a preset amount of multimedia data samples from a multimedia data sample library to generate a plurality of multimedia data sample combinations; the following training procedure is performed separately for each multimedia data sample combination: inputting the multimedia data sample combination into a duration prediction model, and outputting the predicted playing duration of the multimedia data sample combination through the duration prediction model; calculating the consistency ratio of the predicted playing time length and the verification value; and if the consistency ratio is greater than a preset threshold, finishing training of the duration prediction model. The method and the device are used for solving the problems that the prior recommended video content is single, so that the user feels tired and the watching time length is reduced.
Description
Technical Field
The present disclosure relates to the field of computer technologies, and in particular, to a method, a device, equipment, and a medium for training a duration prediction model.
Background
The home page recommendation bit of the existing video application program is the maximum entry for receiving the personalized recommendation flow of the user, and the video application program recommends the video which the user is interested in through the home page recommendation bit and is used for improving the watching duration of the user.
The existing video displayed by the recommendation position is characterized in that videos possibly interested by a user are scored through a scoring model (a pointwise model), and video recommendation is carried out according to the score. However, in this process, videos that may be of interest to the user are mostly videos that the user has historically viewed, and the score of the videos that the user has historically viewed is high, thus causing the recommendation bit to display mostly videos that the user has historically viewed.
Under the recommendation condition, as most of recommended videos are videos which are historically watched by users, the recommendation inner barrel is single, so that the users feel tired, and the recommended videos in the recommendation position do not improve the watching time of the users.
Disclosure of Invention
The application provides a time length prediction model training method, a recommendation method, a device, equipment and a medium, which are used for solving the problems of fatigue of a user and reduced watching time length caused by single content of a recommended video.
In a first aspect, the present application provides a training method of a duration prediction model, including:
constructing a duration prediction model;
obtaining a multimedia data sample library, the multimedia data sample library comprising: historical multimedia sample data, and tag-associated multimedia sample data based on the historical multimedia sample data;
extracting a preset amount of multimedia data samples from the multimedia data sample library to generate a plurality of multimedia data sample combinations;
the following training procedure is performed for each of the multimedia data sample combinations separately: inputting the multimedia data sample combination into the duration prediction model, and outputting the predicted playing duration of the multimedia data sample combination through the duration prediction model;
calculating the consistency ratio of the predicted playing time length and the verification value;
and if the consistency ratio is greater than a preset threshold, finishing training of the duration prediction model.
Optionally, in the multimedia data sample combination, the number of the historical multimedia sample data is sequentially increased in proportion.
In a second aspect, the present application provides a recommendation method, applied to a server, the method including:
acquiring a multimedia database of a target account, wherein the multimedia database comprises: historical multimedia data associated with the target account, and tag-associated multimedia data based on the historical multimedia data;
Extracting a preset amount of multimedia data from the multimedia database to generate a plurality of multimedia data combinations;
respectively inputting a plurality of multimedia data combinations into a duration prediction model to obtain predicted playing duration of each multimedia data combination;
determining a target multimedia data content combination according to the predicted playing time length;
and generating recommendation information of the target multimedia data content combination, and sending the recommendation information to the client equipment corresponding to the target account so that the client equipment displays the recommendation information.
Optionally, extracting a preset amount of multimedia data from the multimedia database to generate a plurality of multimedia data combinations, including:
constructing a first data pool and a second data pool, wherein the first data pool is used for storing the historical multimedia data, and the second data pool is used for storing the tag-associated multimedia data;
extracting a first sub-number of the historical multimedia data from the first data pool, and extracting a second sub-number of the tag-associated multimedia data from the second data pool to generate a plurality of multimedia data combinations;
Wherein the sum of the first sub-quantity and the second sub-quantity is equal to the preset quantity.
Optionally, in the multimedia data combination, the ratio of the first sub-number of the extracted historical multimedia data is sequentially increased.
Optionally, after constructing the first data pool and the second data pool, the method further includes:
inputting the multimedia database into a scoring model, and respectively outputting scores corresponding to the historical multimedia data and the tag-associated multimedia data through the scoring model;
determining the playing probability of each historical multimedia data and each tag-associated multimedia data according to the scores;
extracting a first amount of the historical multimedia data from the multimedia database according to the play probability, storing the historical multimedia data in the first data pool, extracting a second amount of the tag-related multimedia data from the multimedia database, and storing the second amount of the tag-related multimedia data in the second data pool;
wherein an upper limit of the first number is greater than or equal to the preset amount, and an upper limit of the second number of the tag-associated multimedia data is greater than or equal to the preset amount.
Optionally, before inputting the multimedia database into the scoring model, the method further comprises:
Filtering the historical multimedia data with the playing duty ratio smaller than a first preset ratio from the multimedia database to obtain first filtered multimedia data; the playing duty ratio is as follows: a ratio of a played duration to an unplayed duration of the historical multimedia data;
and filtering the historical multimedia data which has a continuous relation and has a playing duty ratio larger than a second preset ratio in the first filtered multimedia data within a second preset time period to obtain second filtered multimedia data, and taking the second filtered multimedia data as the multimedia database.
Optionally, determining the target multimedia data content combination according to the predicted playing duration includes:
acquiring the multimedia data combination with the longest predicted playing duration;
and determining the target multimedia data content combination according to the multimedia data combination with the longest predicted playing duration.
In a third aspect, the present application provides a training device for a duration prediction model, including:
the construction module is used for constructing a duration prediction model;
an acquisition module, configured to acquire a multimedia data sample library, where the multimedia data sample library includes: historical multimedia sample data, and tag-associated multimedia sample data based on the historical multimedia sample data;
The generation module is used for extracting a preset amount of multimedia data samples from the multimedia data sample library and generating a plurality of multimedia data sample combinations;
the training module is used for respectively executing the following training process for each multimedia data sample combination: inputting the multimedia data sample combination into the duration prediction model, and outputting the predicted playing duration of the multimedia data sample combination through the duration prediction model;
the calculation module is used for calculating the consistency ratio of the predicted playing time length and the verification value;
and the adjusting module is used for finishing training the duration prediction model if the consistency ratio is greater than a preset threshold value.
In a fourth aspect, the present application provides a recommendation device, including:
the acquisition module is used for acquiring a multimedia database of the target account, and the multimedia database comprises: historical multimedia data associated with the target account, and tag-associated multimedia data based on the historical multimedia data;
the generation module is used for extracting a preset amount of multimedia data from the multimedia database and generating a plurality of multimedia data combinations;
the prediction module is used for respectively inputting a plurality of multimedia data combinations into a duration prediction model to obtain a predicted playing duration of each multimedia data combination;
The determining module is used for determining a target multimedia data content combination according to the predicted playing time length;
and the recommending module is used for generating recommending information of the target multimedia data content combination, and sending the recommending information to the client equipment corresponding to the target account so as to enable the client equipment to display the recommending information.
In a fifth aspect, the present application provides an electronic device, including: the device comprises a processor, a communication assembly, a memory and a communication bus, wherein the processor, the communication assembly and the memory are communicated with each other through the communication bus; the memory is used for storing a computer program; the processor is configured to execute the program stored in the memory, implement the training method of the duration prediction model according to the first aspect, and/or implement the recommendation method according to the second aspect.
In a sixth aspect, the present application provides a computer readable storage medium storing a computer program, which when executed by a processor implements a training method for implementing the duration prediction model according to the first aspect, and/or the recommendation method according to the second aspect.
Compared with the prior art, the technical scheme provided by the embodiment of the application has the following advantages: according to the method provided by the embodiment of the application, a duration prediction model is constructed; obtaining a multimedia data sample library, the multimedia data sample library comprising: the historical multimedia sample data and the tag-associated multimedia sample data based on the historical multimedia sample data are obtained through a multimedia data sample library, and the duration prediction model can meet the follow-up appeal of the user and the interest discovery of the user.
Further, extracting a preset amount of multimedia data samples from a multimedia data sample library to generate a plurality of multimedia data sample combinations; inputting each multimedia data sample combination into a duration prediction model, and outputting the predicted playing duration of each multimedia data sample combination through the duration prediction model; and calculating the consistency ratio of the predicted playing duration and the verification value, and finishing training of the duration prediction model when the consistency ratio is greater than a preset threshold value.
According to the method and the device, the combined data of the historical multimedia sample data and the tag-associated multimedia sample data are input through the training duration prediction model, and the predicted playing duration of the multimedia data is output, so that the method and the device are not limited to the traditional mode of recommending according to the historical viewing data of the user, and the problems that recommended content is occupied by single type of multimedia data and the recommended content is single are effectively avoided. In addition, the multimedia data recommendation content is various, the problems that users feel tired and the playing time length is reduced due to single recommendation content are avoided, and further, the retention rate of the users is improved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.
In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, and it will be obvious to a person skilled in the art that other drawings can be obtained from these drawings without inventive effort.
FIG. 1 is a schematic flow chart of a training method of a duration prediction model in an embodiment of the present application;
FIG. 2 is a schematic diagram of a machine translation implementation process according to an embodiment of the present application;
FIG. 3 is a flow chart of a recommendation method according to an embodiment of the present application;
fig. 4 is a schematic diagram of an implementation process of filtering a multimedia database according to an embodiment of the present application;
FIG. 5 is a process diagram of a specific implementation of a combination of multimedia data according to an embodiment of the present application;
fig. 6 is a flowchart of a method for acquiring historical multimedia data and tag-related multimedia data according to an embodiment of the present application;
FIG. 7 is a schematic diagram illustrating a specific implementation process of the recommendation method in the embodiment of the present application;
FIG. 8 is a schematic structural diagram of a training device of a duration prediction model in an embodiment of the present application;
FIG. 9 is a schematic structural diagram of a recommendation device according to an embodiment of the present application;
Fig. 10 is a schematic structural diagram of an electronic device in an embodiment of the present application.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present application based on the embodiments herein.
The first embodiment of the present application provides a training method of a duration prediction model, which can be applied to a client device, for example, a mobile phone, a computer, a tablet, a television, etc., and can also be applied to an application program installed on the client device, and can also be applied to a server.
The specific implementation of the method is as shown in fig. 1:
and step 101, constructing a duration prediction model.
Step 102, obtaining a multimedia data sample library, wherein the multimedia data sample library comprises: historical multimedia sample data, and tags based on the historical multimedia sample data.
Specifically, the multimedia data sample library includes: s historical multimedia sample data and C label associated multimedia sample data, wherein S is an integer greater than 1, and C is an integer greater than 1.
Step 103, extracting a preset amount of multimedia data samples from the multimedia data sample library to generate a plurality of multimedia data sample combinations.
Step 104, the following training procedure is performed for each multimedia data sample combination: and inputting the multimedia data sample combination into a duration prediction model, and outputting the predicted playing duration of the multimedia data sample combination through the duration prediction model.
Step 105, calculating the consistency ratio of the predicted playing time length and the verification value.
The verification value is the actual playing duration of the multimedia data sample combination.
And 106, if the consistency ratio is greater than a preset threshold, finishing training of the duration prediction model.
Specifically, if the consistency ratio is smaller than or equal to a preset threshold, after parameters in the duration prediction model are adjusted, the training process is repeatedly executed until the duration prediction model training is completed when the consistency ratio is larger than the preset threshold.
According to the method and the device, the duration prediction model is updated continuously, so that the output result of the duration prediction model is more and more accurate, and the retention rate of a user is further improved.
Specifically, the duration prediction model may be a list recommendation model (listwise model).
Specifically, each time a user enters a recommended video list through an application program recommendation bit, the data is recorded and stored in a multimedia data sample library, the data format is [ request identification, video list ], wherein the video list comprises: the identity of each video and the positional combination relationship of each video.
When the user exits the application program recommendation bit, the related behavior of the user is recorded and stored in a multimedia data sample library, and the data format is [ request identification, user unique identification and playing duration ]. Combining the two data through the request identification to form a multimedia data sample combination, and training a listwise model according to the multimedia data sample combination.
The method can be obtained through training an encoder (encoder) part in a machine translation model (converter model), and the interaction relation of different videos in a group of video lists can be learned by using the converter model, so that the method is used for calculating the playing time of the group of video lists:
fig. 2 is a standard transducer model comprising a vector layer 201, a processing layer 202 and an output layer 203, wherein the processing layer 202 comprises a multi-head attention layer, a first residual and normalization, a full-join layer and a second residual and normalization. The candidate multimedia data sample list is input to the vector layer 201, and G times of processing of the processing layer 202 are performed to obtain the playing duration of each candidate video, and the playing duration is output through the output layer 203, where G is an integer greater than or equal to 1.
In one embodiment, the number of historical multimedia sample data is sequentially incremented in the combination of multimedia data samples.
According to the method provided by the embodiment of the application, a duration prediction model is constructed; obtaining a multimedia data sample library, the multimedia data sample library comprising: the historical multimedia sample data and the tag-associated multimedia sample data based on the historical multimedia sample data are obtained through a multimedia data sample library, and the duration prediction model can meet the follow-up appeal of the user and the interest discovery of the user.
Further, extracting a preset amount of multimedia data samples from a multimedia data sample library to generate a plurality of multimedia data sample combinations; inputting each multimedia data sample combination into a duration prediction model, and outputting the predicted playing duration of each multimedia data sample combination through the duration prediction model; and calculating the consistency ratio of the predicted playing duration and the verification value, and finishing training of the duration prediction model when the consistency ratio is greater than a preset threshold value.
According to the method and the device, the combined data of the historical multimedia sample data and the tag-associated multimedia sample data are input through the training duration prediction model, and the predicted playing duration of the multimedia data is output, so that the method and the device are not limited to the traditional mode of recommending according to the historical viewing data of the user, and the problems that recommended content is occupied by single type of multimedia data and the recommended content is single are effectively avoided. In addition, the multimedia data recommendation content is various, the problems that users feel tired and the playing time length is reduced due to single recommendation content are avoided, and further, the retention rate of the users is improved.
The second embodiment of the present application provides a recommendation method, which can be applied to a client device, for example, a mobile phone, a computer, a tablet, a television, etc., and also can be applied to an application program installed on the client device, for example, a video application program, a news application program, etc., and also can be applied to a server.
The following description will take the application of the method to a server as an example, however, the method is merely illustrative, and is not intended to limit the scope of the present application, and the method is not specifically recited herein. Moreover, some other examples in the present application are not intended to limit the scope of the present application, and are not described in a one-to-one manner.
Specifically, an application program installed on the client device sends an instruction for acquiring a video list to the server through a recommendation interface, wherein the instruction for acquiring the video list comprises: request identification, unique identification of the target account, etc., e.g., the unique identification includes: the user logs into an account of the application, physical parameters of the client device on which the application is installed, etc.
The server acquires the multimedia data combination with the longest playing time according to the unique identifier of the target account, and returns the target multimedia data content combination corresponding to the multimedia data combination with the longest playing time to the application program, and the application program receives and displays the target multimedia data content combination table returned by the server.
The specific implementation of the method is as shown in fig. 3:
step 301, a multimedia database of a target account is obtained.
Wherein the multimedia database comprises: the method includes the steps of associating historical multimedia data associated with a target account, and tag-associated multimedia data based on the historical multimedia data, the tag-associated multimedia data being multimedia data of interest to a user obtained from the historical multimedia data.
Specifically, an application program installed on the client device sends an instruction for acquiring a video list to a server through a recommendation interface, and the server firstly acquires a multimedia database of a user from a user movie watching behavior library according to the unique identifier.
Wherein, user's viewing action library includes: the video types include a fun video, a face value video, a neighborhood video, a disfigure video, an animation video, a food video, a lover video and the like. The historical multimedia data includes: video contents watched by a user in a preset time period, wherein the video contents comprise: the name of the video, the point in time of viewing, the length of time of viewing.
For example, the user has watched television episode a for a week, beginning at 8 pm for each day, 2 hours each time.
Step 302, extracting a preset amount of multimedia data from the multimedia database to generate a plurality of multimedia data combinations.
In one embodiment, as shown in fig. 4, the specific implementation manner of filtering the multimedia database to obtain the final multimedia database is as follows:
step 401, filtering out historical multimedia data with a playing duty ratio smaller than a first preset ratio from a multimedia database to obtain first filtered multimedia data, wherein the playing duty ratio is as follows: a ratio of a played duration to an unplayed duration of the historical multimedia data.
Specifically, when the historical multimedia data is a video collection, the first preset ratio is 20%, for example, a certain television series, the total of 50 sets is watched by the user, the play ratio of the television series is 16%, and 16% is less than 20%, so that the television series is filtered. When the historical multimedia data is a single video, such as a movie, the first preset ratio is 10%, the movie is 120 minutes, the user watches for 30 minutes, the playing ratio of the movie is 25%, and 10% is less than 25%, so the movie is not filtered.
In addition, regarding the first filtered multimedia data, the user can select from the near to the far according to the sequence of time.
Of course, the user may set the first preset ratio according to the actual requirement of the user, which is not used for limiting the protection scope.
The video filtering method and the video filtering device have the advantages that videos smaller than the first preset ratio are filtered, and interest appeal of users can be better met.
Step 402, filtering the historical multimedia data having a continuous relationship and a playing duty ratio larger than the second preset ratio in the second preset time period, to obtain second filtered multimedia data, and using the second filtered multimedia data as a multimedia database.
Specifically, the second preset time period is taken as an example of one week.
The multimedia data having a continuous relationship is a continuously loaded video, such as a television series, a continuously loaded animation, a continuously loaded variety program, etc., which includes a video that has been updated and a video that has not been updated.
When the user watches the video as the updated TV series, the second preset ratio is 90%, the total of 50 sets of TV series is watched by the user, the play ratio of the TV series is 96%,96% is more than 90%, and therefore the TV series is filtered out. When the user watches the video as the television series which is not updated, the second preset ratio is 90%, the television series is updated by 8 sets, the user watches 5 sets, and the playing ratio of the television series is 62.5% and 62.5% is less than 90%, so that the television series is not filtered. For example, 80 pieces of first multimedia data are selected.
Of course, the user may set the second preset ratio, the second preset time period, and the number of the first multimedia data according to the actual needs of the user, which is not limited by the protection scope. However, it is required to prompt that the second preset time period is the current time period.
In a specific embodiment, the specific implementation of extracting a preset amount of multimedia data from the multimedia database and generating a plurality of multimedia data combinations is shown in fig. 5:
in step 501, a first data pool and a second data pool are constructed.
The first data pool is used for storing historical multimedia data, and the second data pool is used for storing tag-associated multimedia data.
In a specific embodiment, after the first data pool and the second data pool are constructed, historical multimedia data is stored in the first data pool, and tag-associated multimedia data is stored in the second data pool, and the specific implementation is as shown in fig. 6:
and step 601, inputting the multimedia database into a scoring model, and respectively outputting scores corresponding to the historical multimedia data and the tag-associated multimedia data through the scoring model.
Specifically, the scoring model is obtained through historical multimedia sample data, label-associated multimedia sample data and score training respectively corresponding to the historical multimedia sample data and the label-associated multimedia sample data.
Specifically, taking 80 historical multimedia data and 300 tag-related multimedia data as examples, scoring is performed, and then the process needs to be performed 380 times to obtain 380 scores, wherein the scoring model can be a pointwise model.
Step 602, determining the playing probability of each historical multimedia data and each tag associated multimedia data according to the scores.
Specifically, the score is proportional to the play probability.
Step 603, extracting a first amount of historical multimedia data from the multimedia database according to the playing probability, storing the historical multimedia data in the first data pool, extracting a second amount of tag-associated multimedia data from the multimedia database, and storing the second amount of tag-associated multimedia data in the second data pool.
Wherein the upper limit of the first quantity is greater than or equal to a preset quantity, and the upper limit of the second quantity of the tag-associated multimedia data is greater than or equal to the preset quantity.
Specifically, a first amount of historical multimedia data is extracted from a multimedia database, and a second amount of tag-associated multimedia data is obtained by utilizing a collaborative filtering algorithm according to the historical multimedia data. The collaborative filtering algorithm based on the user discovers which multimedia data the user likes through the historical multimedia data of the user, measures and scores the favorite multimedia data of the user, then determines the preference degree of the user on each multimedia data according to the scoring result, and recommends the multimedia data to the user according to the preference degree.
The first amount of historical multimedia data extracted is a video that the user is performing hot tracking in a first preset time period, and may also be a collection of videos, that is, what collection a certain television series is instead of a certain television series, what period a certain variety of shows is instead of a certain variety of shows, and so on. The extracted second amount of tag-associated multimedia data is video that the user has not watched but is of interest, such as a television series, a show, a movie, a game video album, a cartoon, etc.
Specifically, according to the historical multimedia data, the second number of tag-associated multimedia data is obtained by utilizing a collaborative filtering algorithm, specifically: the obtaining is performed according to the attribute of the video, for example, the attribute of the video includes: what subject the video belongs to, e.g., movies, television shows, documentaries, shows, cartoons, games, etc., actor lists of the video, director lists of the video, content styles of the video, e.g., police officers, lovers, antiques, etc., content age of the video, e.g., ancient, modern, future, etc. Videos with attributes similar to videos, such as a television show of the same starburst, police films or films crossing the subject, for example, 300 tag-related multimedia data are selected from a video total pool according to historical multimedia data of a user.
However, the number of tag-associated multimedia data is merely illustrative and is not intended to limit the scope of protection.
The method and the device filter the videos with continuous relations, which are being hot-traced by the users in the current time period, and solve the problem that the multimedia data in the final multimedia data combination take the hot-traced by the users as the leading part, so that the video recommendation ignores the interest appeal of the users.
Step 502, extracting a first sub-number of historical multimedia data from the first data pool, and extracting a second sub-number of tag-associated multimedia data from the second data pool, to generate a plurality of multimedia data combinations.
Wherein the sum of the first sub-number and the second sub-number is equal to a preset amount.
In one embodiment, in the combination of multimedia data, the duty cycle of the first sub-amount of the extracted historical multimedia data is sequentially incremented.
In particular, the application program is exemplified by 15 recommended bits, so the preset amount is equal to 15 at this time, which is, of course, only illustrative and not intended to limit the scope of the present application.
In a specific embodiment, according to the playing probability of the first sub-number of historical multimedia data, the first multimedia data is ordered from high to low according to the playing probability, M historical multimedia data are selected, and the M historical multimedia data are combined to obtain at least one first combination result. And the same, according to the playing probability of the second sub-number of the tag-associated multimedia data, sorting the tag-associated multimedia data from high to low according to the playing probability, selecting N tag-associated multimedia data, and combining the N tag-associated multimedia data to obtain at least one second combination result. And generating a plurality of multimedia volume data combinations according to the first combination result and the second combination result.
Wherein, the value of M is 0-15, and the corresponding value of N is 15-0.
Specifically, 15 multimedia data combinations are generated according to the first combination result and the second combination result, wherein the chasing content belongs to historical multimedia data, and the interest content belongs to tag-associated multimedia data.
Specifically, scoring is performed on the first number of historical multimedia data through a scoring model, playing probability corresponding to the score of each historical multimedia data output by the scoring model is determined, and the playing probabilities are ordered from large to small to obtain the first 15 historical multimedia data. Similarly, the first 15 tag-associated multimedia data are obtained.
Associating the historical multimedia data ranked 1 with the tag of the top ranking 14 to form a first multimedia data combination; the second multimedia data combination is composed of the top-ranked 2 historical multimedia data and the top-ranked 13 tag-associated multimedia data, and so forth, the fourteenth multimedia data combination is composed of the top-ranked 14 historical multimedia data and the top-ranked 1 tag-associated multimedia data, and the fifteenth multimedia data combination is composed of the top-ranked 15 historical multimedia data.
The method comprises the following steps:
1 chasing content +14 interest content;
2 chasing content +13 interest content;
3 chasing content +12 interest content;
...
m chasing content+ (15-M) interest content;
...
14 chasing content+1 interest content;
15 chasing content.
Wherein, the 15 multimedia data combinations are represented by the situation that the chasing content is greater than or equal to 15. When the chasing content is smaller than 15, there may be less than 15 multimedia volume data combinations, because the case that the chasing content is larger than 15 does not exist, and the following is taken as an example of the chasing content being 5:
1 chasing content +14 interest content;
2 chasing content +13 interest content;
3 chasing content +12 interest content;
4 chasing content +11 interest content;
5 chasing content +10 interest content.
Step 303, inputting the multiple multimedia data combinations into the duration prediction model respectively, and obtaining the predicted playing duration of each multimedia data combination.
And step 304, determining the target multimedia data content combination according to the predicted playing time length.
Specifically, 15 multimedia data combinations are illustrated as an example.
In a specific embodiment, 15 multimedia data combinations are input into a duration prediction model, analysis is carried out through the duration prediction model, and the predicted playing duration of each multimedia data combination is output; and acquiring the multimedia data combination with the longest predicted playing time length, and determining the target multimedia volume data content combination according to the multimedia data combination with the longest predicted playing time length.
The method and the device recommend the multimedia data combination with the longest predicted playing time length by acquiring the multimedia data combination, so as to improve the playing time length of the video and the retention rate of the user. In addition, the problem that the recommended content is single and the user feels tired is solved, and in addition, the problem that the interest content is ignored in the traditional recommendation is solved, so that the interest appeal of the user can be better met.
And 305, generating recommendation information of the target multimedia data content combination, and sending the recommendation information to the client device corresponding to the target account so as to enable the client device to display the recommendation information.
Next, a specific description is made of the recommended method of the present application through fig. 7, as follows:
in step 701, an application installed on a client device issues an instruction to acquire a video list.
Step 702, executing a recommendation process according to an instruction for acquiring a video list.
Step 703 recalls the user's chasing content.
Step 704 recalls the user's interest content.
Step 705, scoring the recalled chasing content and interest content through the pointwise model, and outputting the score.
Step 706, combining the chasing content and the interest content according to the scores to generate a plurality of multimedia data combinations.
And step 707, predicting the playing time length of each multimedia data combination through the listwise model, and selecting the multimedia data combination with the longest predicted playing time length for recommendation.
According to the recommendation method provided by the embodiment of the application, a multimedia database of a target account is obtained, wherein the multimedia database comprises: the historical multimedia data associated with the target account and the tag-associated multimedia data based on the historical multimedia data can meet the follow-up appeal of the user and the interest discovery of the user.
Further, extracting a preset amount of multimedia data from a multimedia database, generating a plurality of multimedia data combinations, respectively inputting the plurality of multimedia data combinations into a duration prediction model, obtaining a predicted playing duration of each multimedia data combination, determining a target multimedia data content combination according to the predicted playing duration, generating recommendation information of the target multimedia data content combination, and sending the recommendation information to client equipment corresponding to a target account so that the client equipment can display the recommendation information.
According to the method and the device, the combined data of the historical multimedia data and the tag-associated multimedia data are input by using the duration prediction model, and the predicted playing duration of the multimedia data combination is output, so that the method and the device are not limited to the traditional mode of recommending according to the historical viewing data of the user, and the problem that the recommended content is occupied by single type of multimedia data and the recommended content is single is effectively avoided. In addition, the multimedia data recommendation content is various, the problems that users feel tired and the playing time length is reduced due to single recommendation content are avoided, and further, the retention rate of the users is improved.
The third embodiment of the present application provides a training device for a duration prediction model, as shown in fig. 8, and the specific implementation of the device may refer to the descriptions of the first embodiment part and the second embodiment part, and the repetition is omitted, which includes:
a construction module 801, configured to construct a duration prediction model.
An obtaining module 802, configured to obtain a multimedia data sample library, where the multimedia data sample library includes: historical multimedia sample data, and tags based on the historical multimedia sample data.
A generating module 803, configured to extract a preset amount of multimedia data samples from the multimedia data sample library, and generate a plurality of multimedia data sample combinations.
Training module 804 is configured to perform the following training procedure for each multimedia data sample combination: and inputting the multimedia data sample combination into a duration prediction model, and outputting the predicted playing duration of the multimedia data sample combination through the duration prediction model.
The calculating module 805 is configured to calculate a coincidence rate of the predicted playing duration and the verification value.
And an adjusting module 806, configured to complete training the duration prediction model if the consistency ratio is greater than a preset threshold.
The fourth embodiment of the present application further provides a recommendation device, as shown in fig. 9, and the specific implementation of the device may refer to the descriptions of the first embodiment part and the second embodiment part, and the repetition is omitted, which includes:
the obtaining module 901 is configured to obtain a multimedia database of a target account, where the multimedia database includes: historical multimedia data associated with the target account, and tag-associated multimedia data based on the historical multimedia data.
A generating module 902, configured to extract a preset amount of multimedia data from the multimedia database, and generate a plurality of multimedia data combinations.
The prediction module 903 is configured to input a plurality of multimedia data combinations to a duration prediction model, respectively, to obtain a predicted play duration of each multimedia data combination.
A determining module 904, configured to determine a target multimedia data content combination according to the predicted playing duration.
And the recommendation module 905 is configured to generate recommendation information of the target multimedia data content combination, and send the recommendation information to the client device corresponding to the target account, so that the client device displays the recommendation information.
Based on the same concept, there is also provided an electronic device in a fifth embodiment of the present application, as shown in fig. 10, the electronic device mainly including: processor 1001, communication module 1002, memory 1003 and communication bus 1004, wherein processor 1001, communication module 1002 and memory 1003 accomplish each other's communication through communication bus 1004. The memory 1003 stores therein a program executable by the processor 1001, and the processor 1001 executes the program stored in the memory 1003 to realize the following steps: constructing a duration prediction model; acquiring a multimedia data sample library, wherein the multimedia data sample library comprises: historical multimedia sample data, and tag-associated multimedia sample data based on the historical multimedia sample data; extracting a preset amount of multimedia data samples from a multimedia data sample library to generate a plurality of multimedia data sample combinations; the following training procedure is performed separately for each multimedia data sample combination: inputting the multimedia data sample combination into a duration prediction model, and outputting the predicted playing duration of the multimedia data sample combination through the duration prediction model; calculating the consistency ratio of the predicted playing time length and the verification value; if the consistency ratio is greater than a preset threshold, training the duration prediction model is completed, and/or a multimedia database of the target account is acquired; extracting a preset amount of multimedia data from a multimedia database to generate a plurality of multimedia data combinations; respectively inputting a plurality of multimedia data combinations into a duration prediction model to obtain predicted playing duration of each multimedia data combination; determining a target multimedia data content combination according to the predicted playing time length; and generating recommendation information of the target multimedia data content combination, and sending the recommendation information to the client equipment corresponding to the target account so as to enable the client equipment to display the recommendation information.
The communication bus 1004 mentioned in the above electronic device may be a peripheral component interconnect standard (Peripheral Component Interconnect, abbreviated to PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, abbreviated to EISA) bus, or the like. The communication bus 1004 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in fig. 10, but not only one bus or one type of bus.
The communication component 1002 is utilized for communication between the electronic device described above and other devices.
The memory 1003 may include a random access memory (Random Access Memory, simply RAM) or may include a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. Optionally, the memory may also be at least one memory device located remotely from the processor 1001.
The processor 1001 may be a general-purpose processor including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a digital signal processor (Digital Signal Processing, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a Field programmable gate array (Field-Programmable Gate Array, FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware components.
In a sixth embodiment of the present application, there is also provided a computer-readable storage medium having stored therein a computer program which, when run on a computer, causes the computer to perform the training method of the duration prediction model described in the above embodiments, and/or to recommend the method.
In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the processes or functions described in accordance with the embodiments of the present application are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, by a wired (e.g., coaxial cable, optical fiber, digital Subscriber Line (DSL)), or wireless (e.g., infrared, microwave, etc.) means from one website, computer, server, or data center to another. The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape, etc.), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk), etc.
It should be noted that in this document, 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.
The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (11)
1. A method for training a duration prediction model, comprising:
constructing a duration prediction model;
obtaining a multimedia data sample library, the multimedia data sample library comprising: historical multimedia sample data, and tag-associated multimedia sample data based on the historical multimedia sample data;
extracting a preset amount of multimedia data samples from the multimedia data sample library to generate a plurality of multimedia data sample combinations;
the following training procedure is performed for each of the multimedia data sample combinations separately: inputting the multimedia data sample combination into the duration prediction model, and outputting the predicted playing duration of the multimedia data sample combination through the duration prediction model;
calculating the consistency ratio of the predicted playing time length and the verification value;
if the consistency ratio is greater than a preset threshold, training the duration prediction model is completed;
the extracting a preset amount of multimedia data samples from the multimedia data sample library, and generating a plurality of multimedia data sample combinations comprises: inputting the multimedia data sample library into a scoring model, and respectively outputting scores corresponding to the historical multimedia sample data and the tag-associated multimedia sample data through the scoring model; determining the playing probability of each historical multimedia sample data and each tag-associated multimedia sample data according to the scores; extracting a first amount of historical multimedia sample data from the multimedia data sample library according to the play probability, storing the historical multimedia sample data in a first data pool, extracting a second amount of tag-related multimedia sample data from the multimedia data sample library, and storing the second amount of tag-related multimedia sample data in a second data pool; wherein an upper limit of the first number is greater than or equal to the preset amount, and an upper limit of the second number of the tag-associated multimedia sample data is greater than or equal to the preset amount; generating a plurality of multimedia data sample combinations according to the first data pool and the second data pool.
2. The method of claim 1, wherein the number of historical multimedia sample data in the combination of multimedia data samples is sequentially increased in proportion to the number of historical multimedia sample data.
3. A recommendation method, applied to a server, comprising:
acquiring a multimedia database of a target account, wherein the multimedia database comprises: historical multimedia data associated with the target account, and tag-associated multimedia data based on the historical multimedia data;
extracting a preset amount of multimedia data from the multimedia database to generate a plurality of multimedia data combinations;
respectively inputting a plurality of multimedia data combinations into a duration prediction model to obtain predicted playing duration of each multimedia data combination;
determining a target multimedia data content combination according to the predicted playing time length;
generating recommendation information of the target multimedia data content combination, and sending the recommendation information to client equipment corresponding to the target account so that the client equipment displays the recommendation information;
the extracting a preset amount of multimedia data from the multimedia database, and generating a plurality of multimedia data combinations includes: inputting the multimedia database into a scoring model, and respectively outputting scores corresponding to the historical multimedia data and the tag-associated multimedia data through the scoring model; determining the playing probability of each historical multimedia data and each tag-associated multimedia data according to the scores; extracting a first amount of the historical multimedia data from the multimedia database according to the play probability, storing the historical multimedia data in a first data pool, extracting a second amount of the tag-related multimedia data from the multimedia database, and storing the second amount of the tag-related multimedia data in a second data pool; wherein an upper limit of the first number is greater than or equal to the preset amount, and an upper limit of the second number of the tag-associated multimedia data is greater than or equal to the preset amount; generating a plurality of multimedia data combinations according to the first data pool and the second data pool.
4. A recommendation method according to claim 3, wherein extracting a preset amount of multimedia data from the multimedia database, generating a plurality of multimedia data combinations, comprises:
constructing a first data pool and a second data pool, wherein the first data pool is used for storing the historical multimedia data, and the second data pool is used for storing the tag-associated multimedia data;
extracting a first sub-number of the historical multimedia data from the first data pool, and extracting a second sub-number of the tag-associated multimedia data from the second data pool to generate a plurality of multimedia data combinations;
wherein the sum of the first sub-quantity and the second sub-quantity is equal to the preset quantity.
5. The recommendation method of claim 4 wherein the ratio of the first sub-amounts of the historical multimedia data extracted in the multimedia data combination is sequentially incremented.
6. A recommendation method according to claim 3, further comprising, prior to entering the multimedia database into a scoring model:
filtering the historical multimedia data with the playing duty ratio smaller than a first preset ratio from the multimedia database to obtain first filtered multimedia data; the playing duty ratio is as follows: a ratio of a played duration to an unplayed duration of the historical multimedia data;
And filtering the historical multimedia data which has a continuous relation and has a playing duty ratio larger than a second preset ratio in the first filtered multimedia data within a second preset time period to obtain second filtered multimedia data, and taking the second filtered multimedia data as the multimedia database.
7. The recommendation method according to any one of claims 3-6, wherein determining a target multimedia data content combination based on said predicted play duration comprises:
acquiring the multimedia data combination with the longest predicted playing duration;
and determining the target multimedia data content combination according to the multimedia data combination with the longest predicted playing duration.
8. A training device for a duration prediction model, comprising:
the construction module is used for constructing a duration prediction model;
an acquisition module, configured to acquire a multimedia data sample library, where the multimedia data sample library includes: historical multimedia sample data, and tag-associated multimedia sample data based on the historical multimedia sample data;
the generation module is used for extracting a preset amount of multimedia data samples from the multimedia data sample library and generating a plurality of multimedia data sample combinations;
The training module is used for respectively executing the following training process for each multimedia data sample combination: inputting the multimedia data sample combination into the duration prediction model, and outputting the predicted playing duration of the multimedia data sample combination through the duration prediction model;
the calculation module is used for calculating the consistency ratio of the predicted playing time length and the verification value;
the adjustment module is used for finishing training the duration prediction model if the consistency ratio is greater than a preset threshold value;
the generating module is specifically configured to: inputting the multimedia data sample library into a scoring model, and respectively outputting scores corresponding to the historical multimedia sample data and the tag-associated multimedia sample data through the scoring model; determining the playing probability of each historical multimedia sample data and each tag-associated multimedia sample data according to the scores; extracting a first amount of historical multimedia sample data from the multimedia data sample library according to the play probability, storing the historical multimedia sample data in a first data pool, extracting a second amount of tag-related multimedia sample data from the multimedia data sample library, and storing the second amount of tag-related multimedia sample data in a second data pool; wherein an upper limit of the first number is greater than or equal to the preset amount, and an upper limit of the second number of the tag-associated multimedia sample data is greater than or equal to the preset amount; generating a plurality of multimedia data sample combinations according to the first data pool and the second data pool.
9. A recommendation device, comprising:
the acquisition module is used for acquiring a multimedia database of the target account, and the multimedia database comprises: historical multimedia data associated with the target account, and tag-associated multimedia data based on the historical multimedia data;
the generation module is used for extracting a preset amount of multimedia data from the multimedia database and generating a plurality of multimedia data combinations;
the prediction module is used for respectively inputting a plurality of multimedia data combinations into a duration prediction model to obtain a predicted playing duration of each multimedia data combination;
the determining module is used for determining a target multimedia data content combination according to the predicted playing time length;
the recommendation module is used for generating recommendation information of the target multimedia data content combination, and sending the recommendation information to the client equipment corresponding to the target account so that the client equipment can display the recommendation information;
the generating module is specifically configured to: inputting the multimedia database into a scoring model, and respectively outputting scores corresponding to the historical multimedia data and the tag-associated multimedia data through the scoring model; determining the playing probability of each historical multimedia data and each tag-associated multimedia data according to the scores; extracting a first amount of the historical multimedia data from the multimedia database according to the play probability, storing the historical multimedia data in a first data pool, extracting a second amount of the tag-related multimedia data from the multimedia database, and storing the second amount of the tag-related multimedia data in a second data pool; wherein an upper limit of the first number is greater than or equal to the preset amount, and an upper limit of the second number of the tag-associated multimedia data is greater than or equal to the preset amount; generating a plurality of multimedia data combinations according to the first data pool and the second data pool.
10. An electronic device, comprising: the device comprises a processor, a communication assembly, a memory and a communication bus, wherein the processor, the communication assembly and the memory are communicated with each other through the communication bus;
the memory is used for storing a computer program;
the processor is configured to execute the program stored in the memory, implement the training method of the duration prediction model according to any one of claims 1 to 2, and/or implement the recommendation method according to any one of claims 3 to 7.
11. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements a training method of a duration prediction model according to any one of claims 1-2 and/or implements a recommendation method according to any one of claims 3-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011405977.3A CN112507163B (en) | 2020-12-02 | 2020-12-02 | Duration prediction model training method, recommendation method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011405977.3A CN112507163B (en) | 2020-12-02 | 2020-12-02 | Duration prediction model training method, recommendation method, device, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112507163A CN112507163A (en) | 2021-03-16 |
| CN112507163B true CN112507163B (en) | 2023-07-21 |
Family
ID=74968548
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011405977.3A Active CN112507163B (en) | 2020-12-02 | 2020-12-02 | Duration prediction model training method, recommendation method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112507163B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113111217B (en) * | 2021-04-22 | 2024-02-27 | 北京达佳互联信息技术有限公司 | Training method of play duration prediction model, video recommendation method and device |
| CN113132803B (en) * | 2021-04-23 | 2022-09-16 | Oppo广东移动通信有限公司 | Video watching time length prediction method, device, storage medium and terminal |
| CN113656681B (en) * | 2021-07-08 | 2023-08-11 | 北京奇艺世纪科技有限公司 | Object evaluation method, device, equipment and storage medium |
| CN116109354A (en) * | 2021-11-09 | 2023-05-12 | 腾讯科技（深圳）有限公司 | Content recommendation method, apparatus, device, storage medium, and computer program product |
| CN116828265B (en) * | 2023-08-28 | 2023-11-28 | 湖南快乐阳光互动娱乐传媒有限公司 | Video control method, system, electronic equipment and readable storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020135193A1 (en) * | 2018-12-27 | 2020-07-02 | 深圳Tcl新技术有限公司 | Deep neural network-based video recommendation method and system, and storage medium |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101459852B (en) * | 2008-12-22 | 2011-04-13 | 华为技术有限公司 | Method and device for predicting video service occurrence delay |
| CN107333153B (en) * | 2016-04-28 | 2020-06-02 | 华为技术有限公司 | Video transmission method, base station and system |
| CN106528813B (en) * | 2016-11-18 | 2018-12-11 | 腾讯科技（深圳）有限公司 | A kind of multimedia recommendation method and device |
| CN106649655A (en) * | 2016-12-13 | 2017-05-10 | 宁夏宁信信息科技有限公司 | Personalized recommending method and system in video app |
| US10225603B2 (en) * | 2017-03-13 | 2019-03-05 | Wipro Limited | Methods and systems for rendering multimedia content on a user device |
| US10390060B2 (en) * | 2017-05-04 | 2019-08-20 | Facebook, Inc. | Guaranteed delivery of video content items based on received constraints |
| CN108304441B (en) * | 2017-11-14 | 2022-04-26 | 腾讯科技（深圳）有限公司 | Network resource recommendation method and device, electronic equipment, server and storage medium |
| CN108763314B (en) * | 2018-04-26 | 2021-01-19 | 深圳市腾讯计算机系统有限公司 | Interest recommendation method, device, server and storage medium |
| CN110149540B (en) * | 2018-04-27 | 2021-08-24 | 腾讯科技（深圳）有限公司 | Recommendation processing method and device for multimedia resources, terminal and readable medium |
| CN110209843A (en) * | 2019-05-31 | 2019-09-06 | 腾讯科技（深圳）有限公司 | Multimedia resource playback method, device, equipment and storage medium |
| CN110933492B (en) * | 2019-12-10 | 2022-03-04 | 北京爱奇艺科技有限公司 | Method and device for predicting playing time |
| CN111353068A (en) * | 2020-02-28 | 2020-06-30 | 腾讯音乐娱乐科技（深圳）有限公司 | Video recommendation method and device |
-
2020
- 2020-12-02 CN CN202011405977.3A patent/CN112507163B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020135193A1 (en) * | 2018-12-27 | 2020-07-02 | 深圳Tcl新技术有限公司 | Deep neural network-based video recommendation method and system, and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112507163A (en) | 2021-03-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112507163B (en) | Duration prediction model training method, recommendation method, device, equipment and medium | |
| US11899637B2 (en) | Event-related media management system | |
| US11017024B2 (en) | Media content rankings for discovery of novel content | |
| US11997176B2 (en) | Systems and methods for resolving ambiguous terms in social chatter based on a user profile | |
| US11636335B2 (en) | System and method for content discovery | |
| US9892109B2 (en) | Automatically coding fact check results in a web page | |
| US20180152767A1 (en) | Providing related objects during playback of video data | |
| KR101816113B1 (en) | Estimating and displaying social interest in time-based media | |
| US8887058B2 (en) | Media management for multi-user group | |
| US20150293928A1 (en) | Systems and Methods for Generating Personalized Video Playlists | |
| JP4370850B2 (en) | Information processing apparatus and method, program, and recording medium | |
| CN110430476A (en) | Direct broadcasting room searching method, system, computer equipment and storage medium | |
| JP2005056361A (en) | Information processor and method, program, and storage medium | |
| CN103686344A (en) | Enhanced video system and method | |
| CN103984740A (en) | Combination label based search page display method and system | |
| KR20130083829A (en) | Automatic image discovery and recommendation for displayed television content | |
| CN104685899A (en) | Dynamic media segment pricing | |
| CN107592572B (en) | Video recommendation method, device and equipment | |
| US9542395B2 (en) | Systems and methods for determining alternative names | |
| CN106231428A (en) | A kind of video recommendation method and device | |
| CN104053063A (en) | Intuitive image-based program guide for controlling display device | |
| US10592831B2 (en) | Methods and systems for recommending actors | |
| WO2016206035A1 (en) | Information recommendation method and user terminal | |
| Sumiyoshi et al. | CurioView: TV recommendations related to content being viewed | |
| CN110942070B (en) | Content display method, device, electronic equipment and computer readable storage medium |
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 |