CN114168792A - Video recommendation method and device - Google Patents
Video recommendation method and device Download PDFInfo
- Publication number
- CN114168792A CN114168792A CN202111511395.8A CN202111511395A CN114168792A CN 114168792 A CN114168792 A CN 114168792A CN 202111511395 A CN202111511395 A CN 202111511395A CN 114168792 A CN114168792 A CN 114168792A
- Authority
- CN
- China
- Prior art keywords
- video
- sequence
- recommendation
- videos
- recommendation score
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012163 sequencing technique Methods 0.000 claims abstract description 16
- 238000013528 artificial neural network Methods 0.000 claims description 47
- 238000004590 computer program Methods 0.000 claims description 8
- 239000013598 vector Substances 0.000 description 16
- 230000006870 function Effects 0.000 description 9
- 238000013500 data storage Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000000605 extraction Methods 0.000 description 5
- 230000004044 response Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 239000002131 composite material Substances 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The disclosure provides a video recommendation method and device. The video recommendation method comprises the following steps: acquiring a search keyword in a search request, and determining a video related to the search keyword; sequencing the determined videos to obtain a target sequence; sequencing videos meeting a first preset condition in the determined videos to obtain an exploration sequence; determining a recommended sequence based on the exploration sequence and the target sequence; determining recommended videos corresponding to the search requests according to the sequence of the videos in the recommended sequence; wherein the first preset condition is as follows: the release duration is smaller than the duration threshold and the exposure times under the search request are smaller than the time threshold, wherein the release duration is a difference between the current time and the video release time.
Description
Technical Field
The present disclosure relates generally to the field of electronic technology, and more particularly, to a video recommendation method and apparatus.
Background
In video recommendation for video search, videos related to search text (e.g., search keywords) need to be ranked and recommended to a user according to the ranking result. For example, a user may input a search keyword through a client of an application to search for a desired video, the client may transmit a search request including the search keyword input by the user to a server of the application, the server of the application may search for related videos in response to the search request and rank the videos, and then, the top ranked videos may be transmitted to the application client to be recommended to the user. However, the recommendation effect of the existing video recommendation method still needs to be improved.
Disclosure of Invention
An exemplary embodiment of the present disclosure is directed to a video recommendation method and apparatus to solve at least the problems of the related art described above, and may not solve any of the problems described above.
According to a first aspect of the embodiments of the present disclosure, there is provided a video recommendation method, including: acquiring a search keyword in a search request, and determining a video related to the search keyword; sequencing the determined videos to obtain a target sequence; sequencing videos meeting a first preset condition in the determined videos to obtain an exploration sequence; determining a recommended sequence based on the exploration sequence and the target sequence; determining recommended videos corresponding to the search requests according to the sequence of the videos in the recommended sequence; wherein the first preset condition is as follows: the release duration is smaller than the duration threshold and the exposure times under the search request are smaller than the time threshold, wherein the release duration is a difference between the current time and the video release time.
Optionally, the step of sorting the determined videos to obtain the target sequence includes: sequencing the determined videos according to the estimated first recommendation score of each video; the method for sequencing the videos meeting the first preset condition in the determined videos to obtain the exploration sequence comprises the following steps: according to the estimated second recommendation score of each video meeting the first preset condition, ordering the videos meeting the first preset condition in the determined videos; wherein a higher first recommendation score indicates a higher recommendation degree, and a higher second recommendation score indicates a higher recommendation degree.
Optionally, the step of determining the recommendation sequence based on the exploration sequence and the target sequence comprises: determining an adjustment coefficient, so that the second recommendation score of the video ranked at the m-th position in the exploration sequence is multiplied by the adjustment coefficient to be larger than the first recommendation score of the video ranked at the n-th position in the target sequence; multiplying the second recommendation score of each video in the exploration sequence by the adjustment coefficient to obtain a third recommendation score of the video; and sequencing the determined videos according to the third recommendation scores of the videos ranked in the exploration sequence and the first recommendation scores of the videos ranked in the target sequence to obtain a recommendation sequence.
Optionally, in the step of sorting the determined videos to obtain the recommendation sequence, if any one of the videos has both the first recommendation score and the third recommendation score, the videos are sorted according to a higher score of the first recommendation score and the third recommendation score; if any one video only has the first recommendation score, the videos are sorted according to the first recommendation score.
Optionally, the step of determining the adjustment coefficient includes: acquiring a ratio of a first recommendation score of the video ranked at the nth position in the target sequence to a second recommendation score of the video ranked at the mth position in the exploration sequence, and taking the ratio as a first ratio; acquiring a ratio of a first recommendation score of the video ranked at the n-k position in the target sequence to a second recommendation score of the video ranked at the m position in the exploration sequence as a second ratio, wherein k is an integer greater than or equal to 1; the adjustment coefficient is determined to be a value greater than the first ratio and less than the second ratio.
Optionally, the second recommendation score of each video meeting the first preset condition is estimated by the following method: for each video meeting a first preset condition, inputting a bipartite graph result about the video and a consumer into a graph neural network to obtain the characteristics of the video; and obtaining a second recommendation score of the video based on the characteristics of the video and the characteristics of the search keywords.
Optionally, the step of obtaining a second recommendation score of the video based on the features of the video and the features of the search keyword includes: inputting the characteristics of the video and the characteristics of the search keywords into a double-tower neural network for characteristic extraction, splicing the extracted characteristics to obtain spliced characteristics, and obtaining a second recommendation score of the video based on the spliced characteristics.
Optionally, the double tower neural network comprises: the method comprises the following steps of firstly, inputting the characteristics of a video and the characteristics of search keywords into a double-tower neural network to respectively extract the characteristics, splicing the extracted characteristics to obtain spliced characteristics, and obtaining a second recommendation score of the video based on the spliced characteristics, wherein the first multilayer perceptron, the second multilayer perceptron, the third multilayer perceptron, the splicing layer and the classification layer are used for respectively extracting the characteristics of the video and the characteristics of the search keywords, and the steps of obtaining the second recommendation score of the video based on the spliced characteristics comprise: inputting the characteristics of the video into a first multilayer perceptron; inputting the characteristics of the search keywords into a second multilayer perceptron; splicing the vector output by the first multilayer sensor and the vector output by the second multilayer sensor through a splicing layer and then inputting the spliced vectors into a third multilayer sensor; and inputting the vector output by the third multilayer perceptron into the classification layer to obtain a second recommendation score of the video output by the classification layer.
Optionally, the step of sorting the videos that satisfy the first preset condition in the determined videos to obtain an exploration sequence includes: sequencing videos meeting a first preset condition and videos meeting a second preset condition in the determined videos to obtain an exploration sequence; wherein the second preset condition is as follows: the release duration is less than the duration threshold, the exposure times under the search request are greater than or equal to the times threshold, and the user consumption posterior condition meets the preset consumption condition.
Optionally, the step of the user's consumption posterior condition meeting the preset consumption condition includes: the comprehensive index value determined based on the index values of the consumption posterior indexes is higher than a preset threshold value; wherein the individual post-consumer indicators comprise at least one of: viewing duration, click-through rate, long play rate, and like rate.
Optionally, the double tower neural network is trained by: obtaining training samples, wherein the training samples comprise: video samples with labels having publication durations less than a duration threshold, the labels comprising: the video sample is marked as a positive sample or a negative sample, and the marking of the video sample is determined based on the consumption posterior condition of the video sample under the search request; inputting a bipartite graph result about a video sample and a consumer thereof into a graph neural network to obtain characteristics of the video sample; inputting the characteristics of the video sample and the characteristics of the search keywords in the search request into the double-tower neural network to obtain a second recommendation score of the predicted video sample; determining a loss value of a loss function of the double-tower neural network based on the second recommendation score of the predicted video sample and the label of the video sample; and training the double-tower neural network by adjusting the network parameters of the double-tower neural network according to the loss value.
According to a second aspect of the embodiments of the present disclosure, there is provided a video recommendation apparatus including: the video determining unit is configured to obtain a search keyword in a search request and determine a video related to the search keyword; the target sequence determining unit is configured to sequence the determined videos to obtain a target sequence; the searching sequence determining unit is configured to sequence videos meeting a first preset condition in the determined videos to obtain a searching sequence; a recommended sequence determination unit configured to determine a recommended sequence based on the exploration sequence and the target sequence; the recommended video determining unit is configured to determine a recommended video corresponding to the search request according to the sequence of videos in the recommended sequence; wherein the first preset condition is as follows: the release duration is smaller than the duration threshold and the exposure times under the search request are smaller than the time threshold, wherein the release duration is a difference between the current time and the video release time.
Optionally, the target sequence determination unit is configured to sort the determined videos according to the estimated first recommendation score of each video; the exploration sequence determining unit is configured to sort the videos meeting the first preset condition in the determined videos according to the estimated second recommendation score of each video meeting the first preset condition; wherein a higher first recommendation score indicates a higher recommendation degree, and a higher second recommendation score indicates a higher recommendation degree.
Optionally, the recommendation sequence determining unit is configured to: determining an adjustment coefficient, so that the second recommendation score of the video ranked at the m-th position in the exploration sequence is multiplied by the adjustment coefficient to be larger than the first recommendation score of the video ranked at the n-th position in the target sequence; multiplying the second recommendation score of each video in the exploration sequence by the adjustment coefficient to obtain a third recommendation score of the video; and sequencing the determined videos according to the third recommendation scores of the videos ranked in the exploration sequence and the first recommendation scores of the videos ranked in the target sequence to obtain a recommendation sequence.
Optionally, the recommendation sequence determining unit is configured to: if any one video has the first recommendation score and the third recommendation score, sorting the video according to the higher score in the first recommendation score and the third recommendation score; if any one video only has the first recommendation score, the videos are sorted according to the first recommendation score.
Optionally, the recommendation sequence determining unit is configured to: acquiring a ratio of a first recommendation score of the video ranked at the nth position in the target sequence to a second recommendation score of the video ranked at the mth position in the exploration sequence, and taking the ratio as a first ratio; acquiring a ratio of a first recommendation score of the video ranked at the n-k position in the target sequence to a second recommendation score of the video ranked at the m position in the exploration sequence as a second ratio, wherein k is an integer greater than or equal to 1; the adjustment coefficient is determined to be a value greater than the first ratio and less than the second ratio.
Optionally, the exploration sequence determination unit is configured to predict the second recommendation score of each video meeting the first preset condition by: for each video meeting a first preset condition, inputting a bipartite graph result about the video and a consumer into a graph neural network to obtain the characteristics of the video; and obtaining a second recommendation score of the video based on the characteristics of the video and the characteristics of the search keywords.
Optionally, the exploration sequence determining unit is configured to input the features of the video and the features of the search keywords into a double-tower neural network for feature extraction, splice the features obtained by extraction to obtain spliced features, and obtain a second recommendation score of the video based on the spliced features.
Optionally, the double tower neural network comprises: a first multi-layer perceptron, a second multi-layer perceptron, a third multi-layer perceptron, a splice layer, and a classification layer, wherein the exploration sequence determination unit is configured to: inputting the characteristics of the video into a first multilayer perceptron; inputting the characteristics of the search keywords into a second multilayer perceptron; splicing the vector output by the first multilayer sensor and the vector output by the second multilayer sensor through a splicing layer and then inputting the spliced vectors into a third multilayer sensor; and inputting the vector output by the third multilayer perceptron into the classification layer to obtain a second recommendation score of the video output by the classification layer.
Optionally, the exploration sequence determining unit is configured to sort videos meeting a first preset condition and videos meeting a second preset condition in the determined videos to obtain an exploration sequence; wherein the second preset condition is as follows: the release duration is less than the duration threshold, the exposure times under the search request are greater than or equal to the times threshold, and the user consumption posterior condition meets the preset consumption condition.
Optionally, the step of the user's consumption posterior condition meeting the preset consumption condition includes: the comprehensive index value determined based on the index values of the consumption posterior indexes is higher than a preset threshold value; wherein the individual post-consumer indicators comprise at least one of: viewing duration, click-through rate, long play rate, and like rate.
Optionally, the double tower neural network is trained by: obtaining training samples, wherein the training samples comprise: video samples with labels having publication durations less than a duration threshold, the labels comprising: the video sample is marked as a positive sample or a negative sample, and the marking of the video sample is determined based on the consumption posterior condition of the video sample under the search request; inputting a bipartite graph result about a video sample and a consumer thereof into a graph neural network to obtain characteristics of the video sample; inputting the characteristics of the video sample and the characteristics of the search keywords in the search request into the double-tower neural network to obtain a second recommendation score of the predicted video sample; determining a loss value of a loss function of the double-tower neural network based on the second recommendation score of the predicted video sample and the label of the video sample; and training the double-tower neural network by adjusting the network parameters of the double-tower neural network according to the loss value.
According to a third aspect of the embodiments of the present disclosure, there is provided an electronic apparatus including: at least one processor; at least one memory storing computer-executable instructions, wherein the computer-executable instructions, when executed by the at least one processor, cause the at least one processor to perform a video recommendation method as described above.
According to a fourth aspect of embodiments of the present disclosure, there is provided a computer-readable storage medium, wherein instructions, when executed by at least one processor, cause the at least one processor to perform the video recommendation method as described above.
According to a fifth aspect of embodiments of the present disclosure, there is provided a computer program product comprising computer instructions which, when executed by at least one processor, implement the video recommendation method as described above.
According to the video recommendation method and device, the final recommendation sequence is determined by combining the exploration sequence comprising the newly released videos with few exposure times and the target sequence comprising all the videos, so that the new high-quality videos can have a certain exploration flow, good videos can be displayed, and the accuracy and recommendation effect of video recommendation are improved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure and are not to be construed as limiting the disclosure.
Fig. 1 illustrates an application scenario diagram of a video recommendation method according to an exemplary embodiment of the present disclosure;
fig. 2 illustrates a flow chart of a video recommendation method according to an exemplary embodiment of the present disclosure;
FIG. 3 illustrates a flowchart of a method of determining a recommendation sequence based on an exploration sequence and a target sequence, according to an exemplary embodiment of the present disclosure;
FIG. 4 shows a flow diagram of a method of training a double tower neural network, according to an example embodiment of the present disclosure;
fig. 5 illustrates a block diagram of a video recommendation apparatus according to an exemplary embodiment of the present disclosure;
fig. 6 illustrates a block diagram of an electronic device according to an exemplary embodiment of the present disclosure.
Detailed Description
In order to make the technical solutions of the present disclosure better understood by those of ordinary skill in the art, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.
It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein are capable of operation in sequences other than those illustrated or otherwise described herein. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
In this case, the expression "at least one of the items" in the present disclosure means a case where three types of parallel expressions "any one of the items", "a combination of any plural ones of the items", and "the entirety of the items" are included. For example, "include at least one of a and B" includes the following three cases in parallel: (1) comprises A; (2) comprises B; (3) including a and B. For another example, "at least one of the first step and the second step is performed", which means that the following three cases are juxtaposed: (1) executing the step one; (2) executing the step two; (3) and executing the step one and the step two.
Fig. 1 illustrates an application scenario diagram of a video recommendation method according to an exemplary embodiment of the present disclosure.
Referring to fig. 1, when a user searches a video at a client, the client may transmit a video search request including a search keyword to a server; the server side may execute the video recommendation method according to the exemplary embodiment of the present disclosure in response to a video search request of the client side to recommend a video sequence satisfying a user search requirement to the client side.
It should be understood that the video recommendation method according to the exemplary embodiment of the present disclosure may be applied not only to the above-described scenarios but also to other suitable scenarios, and the present disclosure is not limited thereto.
Fig. 2 illustrates a flowchart of a video recommendation method according to an exemplary embodiment of the present disclosure. As an example, the video recommendation method may be performed by a server side.
Referring to fig. 2, in step S101, a search keyword in a search request is acquired.
As an example, the search request may be a request for searching for videos related to at least one search keyword.
In step S102, videos related to the search keyword are determined.
As an example, a database may be searched for videos for which attribute information is associated with the search keyword. As an example, the attribute information of the video may include, but is not limited to, at least one of: identification information of the video (e.g., video name, etc.), tag information of the video, content information of the video (e.g., actor name, character name, etc. appearing in the video), publisher information of the video.
In step S103, the determined videos are sorted to obtain a target sequence.
In the target sequence, the higher the recommendation of the video, the more forward the ranking may be.
As an example, the determined videos may be sorted according to the estimated first recommendation score for each video. Wherein a higher first recommendation score indicates a higher recommendation.
As an example, a first recommendation score for a video may be pre-estimated based on characteristics of the video. For example, a first recommendation score for a video may be estimated based on post-consumer indicator features of the video.
As an example, the post-consumption indicators may include, but are not limited to, at least one of: viewing duration, click-through rate, long-play rate, like-to-click rate, and collection rate.
In step S104, videos that satisfy a first preset condition in the determined videos are sorted to obtain an exploration sequence.
In the exploration sequence, the higher the recommendation of the video, the more forward the ranking may be.
As an example, the first preset condition may include: the release duration is smaller than the duration threshold and the exposure times under the search request are smaller than the time threshold, wherein the release duration is a difference between the current time and the video release time.
As an example, the number of exposures of a video to a search request can be understood as: the number of times the video has been recommended to, and displayed by, the client in response to the search request.
The threshold number may be 100, for example, and it should be understood that this value is merely an example and that suitable values may be set according to actual conditions and requirements.
As an example, the duration threshold may be 1 day, and it should be understood that this value is merely an example, and an appropriate value may be set according to actual conditions and needs.
Further, as an example, step S104 may include: and sequencing the videos meeting the first preset condition and the videos meeting the second preset condition in the determined videos to obtain an exploration sequence.
As an example, the second preset condition may include: the release duration is less than the duration threshold, the exposure times under the search request are greater than or equal to the times threshold, and the user consumption posterior condition meets the preset consumption condition.
As an example, a user-consumed a posteriori situation for a video may be a situation where the video was consumed by the user (e.g., after having been recommended to the user in response to a search request).
As an example, the user consumption posterior case satisfying the preset consumption condition may include: the integrated index value determined based on the index values of the individual consumption posterior indexes is higher than a preset threshold value. For example, the index values of the individual consumption posteriori indexes may be weighted and summed to obtain a composite index value. For example,the longer the watching duration, the higher the click rate, the higher the long-play rate, the higher the approval rate, and the higher the composite index value. For example, the video index value y ═ ax1+bx2+cx3+dx4Wherein x is1To x4And a, b, c and d respectively represent weighting coefficients of the corresponding consumption posterior indexes.
According to the exemplary embodiment of the disclosure, videos with short publishing time and small exposure times can form an exploration sequence, and in addition, videos with short publishing time, improved exposure times and good user consumption posterior conditions can be continuously kept in the exploration sequence, so that videos with short publishing time, improved exposure times and poor user consumption posterior conditions exit the exploration sequence.
As an example, the videos that satisfy the preset condition in the determined videos may be sorted according to the estimated second recommendation score of each video that satisfies the preset condition (i.e., the video that satisfies the first preset condition; or the video that satisfies the first preset condition and the video that satisfies the second preset condition). Wherein a higher second recommendation score indicates a higher recommendation.
By way of example, the characteristics of the video can be obtained based on the consuming user of the video, and the second recommendation score of the video can be estimated based on the characteristics of the video and the characteristics of the search keywords.
As an example, for each video meeting a preset condition, a bipartite graph result about the video and a consuming user thereof may be input into a graph neural network to obtain a feature of the video; and obtaining a second recommendation score of the video based on the characteristics of the video and the characteristics of the search keywords. For example, for each video meeting a preset condition, a bipartite Graph result about the video and a consuming user thereof may be input into a Graph Neural Network (GNN) to obtain an embedding feature (embedded feature) of the video, and it should be understood that other types of features are also possible, and the disclosure is not limited thereto. For example, the search keyword may be characterized by: features of the search keyword extracted through a BERT (bidirectional Encoder retrieval from transforms) network. It should be understood that the features of the search keyword can be extracted by other means, and the disclosure is not limited thereto. For example, the type of the feature of the search keyword may be an embedding feature of the search keyword. It should be understood that other types of features are possible, and the disclosure is not limited thereto.
As an example, bipartite graph results for a video and its consuming users may be constructed based on user consumption data for the video. For example, a consuming user of a video may be a user who clicked to play the video, or has long played the video, or has complied with the video, or has collected the video. A bipartite graph, also called bipartite graph, is a special model in graph theory, and if a vertex V is divisible into two mutually disjoint subsets (a, B), and two vertices i and j associated with each edge (i, j) in the graph belong to two different sets of vertices (i in a, j in B), the graph G is called a bipartite graph. For example, a video and a user may act as vertices in a bipartite graph, and the video and its consuming users may be connected by edges.
As an example, the features of the video and the features of the search keywords may be input into a double-tower neural network, resulting in a second recommendation score for the video. As an example, the features of the video and the features of the search keyword may be input into a double-tower neural network to perform feature extraction, the features extracted respectively may be spliced to obtain spliced features, and the second recommendation score of the video may be obtained based on the spliced features.
As an example, the double tower neural network may include: a first multi-layer perceptron, a second multi-layer perceptron, a third multi-layer perceptron, a splice layer concat, and a classification layer. A Multilayer Perceptron (MLP) is a neural network model. For example, the taxonomy layer may be taxonomy using the softmax function. It should be understood that this double tower neural network structure is merely an example, and the double tower neural network may be constructed in other forms, and the present disclosure is not limited thereto.
As an example, the characteristics of the video may be input to a first multi-layer perceptron; inputting the characteristics of the search keywords into a second multilayer perceptron; splicing the vector output by the first multilayer sensor and the vector output by the second multilayer sensor through a splicing layer and then inputting the spliced vectors into a third multilayer sensor; and inputting the vector output by the third multilayer perceptron into the classification layer to obtain a second recommendation score of the video output by the classification layer.
An exemplary embodiment of the training method of the double tower neural network will be described below in conjunction with fig. 4.
In step S105, a recommendation sequence is determined based on the search sequence and the target sequence.
As an example, a recommendation sequence may be determined based on the exploration sequence and the target sequence, so that videos ranked in the exploration sequence are ranked more forward in the recommendation sequence, thereby enabling videos satisfying a preset condition among the determined videos to be preferentially recommended.
An exemplary embodiment of step S105 will be described below in conjunction with fig. 3.
In step S106, according to the order of the videos in the recommendation sequence, a recommended video corresponding to the search request is determined. That is, the video is recommended to the user as the recommended video corresponding to the search request with higher priority as the video is ranked higher in the recommendation sequence.
Fig. 3 illustrates a flowchart of a method of determining a recommendation sequence based on an exploration sequence and a target sequence according to an exemplary embodiment of the present disclosure.
Referring to fig. 3, in step S201, an adjustment coefficient is determined such that the second recommendation score of the video ranked at the m-th position in the search sequence is multiplied by the adjustment coefficient and is greater than the first recommendation score of the video ranked at the n-th position in the target sequence.
n is an integer greater than 0. As an example, n may be set to 4, and it should be understood that this value is only an example, and an appropriate value may be set according to actual situations and needs. m is an integer greater than 0. As an example, m may be set to 1, and it should be understood that this value is merely an example, and an appropriate value may be set according to actual situations and needs.
As an example, a ratio of a first recommendation score of a video ranked at the nth position in the target sequence to a second recommendation score of a video ranked at the mth position in the exploration sequence may be obtained as a first ratio; acquiring a ratio of a first recommendation score of the video ranked at the n-k position in the target sequence to a second recommendation score of the video ranked at the m position in the exploration sequence as a second ratio, wherein k is an integer greater than or equal to 1; then, the adjustment coefficient is determined to be a value greater than the first ratio and less than the second ratio.
In step S202, the score obtained by multiplying the second recommendation score of each video ranked in the search sequence by the adjustment coefficient is used as the third recommendation score of the video.
In step S203, the determined videos are ranked according to the third recommendation score of each video ranked in the search sequence and the first recommendation score of each video ranked in the target sequence to obtain a recommendation sequence.
As an example, if any one video has both the first recommendation score and the third recommendation score, the video is ranked by the higher score among the first recommendation score and the third recommendation score; if any one video only has the first recommendation score, the videos are sorted according to the first recommendation score. In other words, whether the first recommendation score or the third recommendation score, the higher the recommendation score of the video, the higher its ranking in the recommendation sequence.
By way of example, each video ranked in the exploration sequence may be ranked into the target sequence according to the level of the third recommendation score of the video and the first recommendation score of each video ranked in the target sequence, and only the ranking position of the video in the target sequence, which is further ahead, is retained to obtain the recommendation sequence.
Specifically, each video ranked in the exploration target is inserted into the target sequence according to the height of the third recommendation score of each video ranked in the exploration sequence and the height of the first recommendation score of each video ranked in the target sequence, and the higher the recommendation score is, the higher the ranking is; and the same video only retains its more advanced ordering position in the target sequence. For example, the target sequence is: video a, video b, video c, video d, video e, … … and video z, wherein the first recommendation score of video a is not less than the first recommendation score of video b, not less than the first recommendation score of video c, not less than … …, not less than the first recommendation score of video z, and the exploration sequence is as follows: video c, video d. Arranging the video c and the video d into the target sequence according to the third recommendation scores of the video c and the video d, wherein the third recommendation score of the video c is higher than the first recommendation score of the video b and lower than the first recommendation score of the video a, and the third recommendation score of the video d is higher than the first recommendation score of the video f and lower than the first recommendation score of the video e, so that the sequence is obtained after arranging the video c and the video d into the target sequence: video a, video c, video b, video c, video d, video e, video d, video f … …, and video z, in order to avoid the same video from appearing repeatedly in the sequence, only the earlier ordering position of the same video in the target sequence is reserved, so as to obtain the recommended sequence: video a, video c, video b, video d, video e, video f … …, video z.
Fig. 4 illustrates a flowchart of a method of training a double tower neural network according to an exemplary embodiment of the present disclosure.
Referring to fig. 4, in step S301, a training sample is obtained, wherein the training sample includes: the method comprises the following steps of (1) carrying out annotation on a video sample with the publication duration less than the preset duration, wherein the annotation comprises the following steps: the video sample is an annotation of a positive sample or a negative sample, the annotation of the video sample being determined based on a posterior consumed condition of the video sample under the search request.
For example, after the video sample is recommended to the user based on the search request of the user, whether the video sample is a positive sample or a negative sample can be determined according to the consumption condition of the video sample by the user. For example, a video sample that a user clicks on for viewing or long-playing may be taken as a positive sample, otherwise as a negative sample.
In step S302, the bipartite graph results for the video sample and its consuming users are input into a graph neural network to obtain features of the video sample.
In step S303, the features of the video sample and the features of the search keyword in the search request are input into the double-tower neural network, so as to obtain a second recommendation score of the predicted video sample.
In step S304, a loss value of the loss function of the double-tower neural network is determined based on the second recommendation score of the predicted video sample and the label of the video sample. The loss value of the loss function characterizes a degree of gap between the predicted second recommended score for the video sample and the annotation of the video sample.
As an example, the loss function may be a cross-entropy loss function. It should be understood that other types of loss functions are possible, and the present disclosure is not limited thereto.
In step S305, the double-tower neural network is trained by adjusting network parameters of the double-tower neural network according to the loss value.
As an example, a gradient of each network parameter of the double-tower neural network may be calculated according to the loss value, and for each network parameter, a parameter value of the network parameter may be updated according to the calculated gradient of the network parameter. The network parameters of the double tower neural network may include: a first multi-layer perceptron, a second multi-layer perceptron, a third multi-layer perceptron, a splice layer concat, and parameters of a classification layer.
It should be appreciated that the double tower neural network may be iteratively trained using a plurality of training samples.
The technical scheme provided by the embodiment of the disclosure can bring the following beneficial effects:
the new high-quality video has certain exploration flow through the exploration sequence, so that the good video can be displayed;
through learning and searching the semantic features of the keywords and the videos, the videos lacking text information can be well matched to generate a better sequence, so that high-quality videos with few text information can be prevented from being effectively displayed and ignored;
after the exposure amount of a new video reaches a certain threshold value, whether the video is reserved in an exploration sequence or not is determined through consumption posterior so that a high-quality video can obtain a better exposure opportunity;
the method solves the problems that the posterior feature of the video is heavier, the Martha effect is generated and the newly published video is not friendly in the video recommendation mode in the related technology.
Fig. 5 illustrates a block diagram of a video recommendation apparatus according to an exemplary embodiment of the present disclosure.
As shown in fig. 5, the video recommendation apparatus 10 according to an exemplary embodiment of the present disclosure includes: video determining unit 101, target sequence determining unit 102, exploration sequence determining unit 103, recommended sequence determining unit 104, and recommended video determining unit 105.
Specifically, the video determination unit 101 is configured to obtain a search keyword in a search request, and determine a video related to the search keyword.
The target sequence determination unit 102 is configured to order the determined videos into a target sequence.
The exploration sequence determination unit 103 is configured to sort videos satisfying a first preset condition from the determined videos to obtain an exploration sequence.
The recommended sequence determination unit 104 is configured to determine a recommended sequence based on the exploration sequence and the target sequence.
The recommended video determining unit 105 is configured to determine a recommended video corresponding to the search request according to the ordering of videos in the recommended sequence.
The first preset condition is as follows: the release duration is smaller than the duration threshold and the exposure times under the search request are smaller than the time threshold, wherein the release duration is a difference between the current time and the video release time.
As an example, the target sequence determining unit 102 may be configured to order the determined videos according to the pre-estimated first recommendation score of each video; the exploration sequence determination unit 103 may be configured to sort videos satisfying the first preset condition among the determined videos according to the estimated second recommendation score of each video satisfying the first preset condition; wherein a higher first recommendation score indicates a higher recommendation degree, and a higher second recommendation score indicates a higher recommendation degree.
As an example, the recommendation sequence determination unit 104 may be configured to: determining an adjustment coefficient, so that the second recommendation score of the video ranked at the m-th position in the exploration sequence is multiplied by the adjustment coefficient to be larger than the first recommendation score of the video ranked at the n-th position in the target sequence; multiplying the second recommendation score of each video in the exploration sequence by the adjustment coefficient to obtain a third recommendation score of the video; and sequencing the determined videos according to the third recommendation scores of the videos ranked in the exploration sequence and the first recommendation scores of the videos ranked in the target sequence to obtain a recommendation sequence.
As an example, the recommendation sequence determination unit 104 may be configured to: if any one video has the first recommendation score and the third recommendation score, sorting the video according to the higher score in the first recommendation score and the third recommendation score; if any one video only has the first recommendation score, the videos are sorted according to the first recommendation score.
As an example, the recommendation sequence determination unit 104 may be configured to: acquiring a ratio of a first recommendation score of the video ranked at the nth position in the target sequence to a second recommendation score of the video ranked at the mth position in the exploration sequence, and taking the ratio as a first ratio; acquiring a ratio of a first recommendation score of the video ranked at the n-k position in the target sequence to a second recommendation score of the video ranked at the m position in the exploration sequence as a second ratio, wherein k is an integer greater than or equal to 1; the adjustment coefficient is determined to be a value greater than the first ratio and less than the second ratio.
As an example, the exploration sequence determination unit 103 may be configured to predict the second recommendation score of each video satisfying the first preset condition by: for each video meeting a first preset condition, inputting a bipartite graph result about the video and a consumer into a graph neural network to obtain the characteristics of the video; and obtaining a second recommendation score of the video based on the characteristics of the video and the characteristics of the search keywords.
As an example, the exploration sequence determination unit 103 may be configured to input the features of the video and the features of the search keyword into a double-tower neural network for feature extraction, splice the features extracted from the two towers respectively to obtain spliced features, and obtain the second recommendation score of the video based on the spliced features.
As an example, the double tower neural network may include: a first multi-layer perceptron, a second multi-layer perceptron, a third multi-layer perceptron, a splice layer, and a classification layer, wherein the exploration sequence determination unit 103 may be configured to: inputting the characteristics of the video into a first multilayer perceptron; inputting the characteristics of the search keywords into a second multilayer perceptron; splicing the vector output by the first multilayer sensor and the vector output by the second multilayer sensor through a splicing layer and then inputting the spliced vectors into a third multilayer sensor; and inputting the vector output by the third multilayer perceptron into the classification layer to obtain a second recommendation score of the video output by the classification layer.
As an example, the exploration sequence determination unit 103 may be configured to sort videos satisfying a first preset condition and videos satisfying a second preset condition in the determined videos, resulting in an exploration sequence; wherein the second preset condition is as follows: the release duration is less than the duration threshold, the exposure times under the search request are greater than or equal to the times threshold, and the user consumption posterior condition meets the preset consumption condition.
As an example, the user consumption posterior case satisfying the preset consumption condition may include: the comprehensive index value determined based on the index values of the consumption posterior indexes is higher than a preset threshold value; wherein the individual post-consumer indicators may comprise at least one of: viewing duration, click-through rate, long play rate, and like rate.
As an example, the double tower neural network may be trained by: obtaining training samples, wherein the training samples comprise: video samples with labels having publication durations less than a duration threshold, the labels comprising: the video sample is marked as a positive sample or a negative sample, and the marking of the video sample is determined based on the consumption posterior condition of the video sample under the search request; inputting a bipartite graph result about a video sample and a consumer thereof into a graph neural network to obtain characteristics of the video sample; inputting the characteristics of the video sample and the characteristics of the search keywords in the search request into the double-tower neural network to obtain a second recommendation score of the predicted video sample; determining a loss value of a loss function of the double-tower neural network based on the second recommendation score of the predicted video sample and the label of the video sample; and training the double-tower neural network by adjusting the network parameters of the double-tower neural network according to the loss value.
With respect to the video recommendation apparatus 10 in the above embodiment, the specific manner of performing the operation thereof has been described in detail in the embodiment of the method, and will not be elaborated herein.
Further, it should be understood that each unit in the video recommendation apparatus 10 in the above embodiments may be implemented as a hardware component and/or a software component. The individual units may be implemented, for example, using Field Programmable Gate Arrays (FPGAs) or Application Specific Integrated Circuits (ASICs), depending on the processing performed by the individual units as defined by the skilled person.
Fig. 6 illustrates a block diagram of an electronic device according to an exemplary embodiment of the present disclosure.
Referring to fig. 6, the electronic device 20 includes: at least one memory 201 and at least one processor 202, the at least one memory 201 having stored therein a set of computer-executable instructions that, when executed by the at least one processor 202, perform a video recommendation method as described in the above exemplary embodiments.
By way of example, the electronic device 20 may be a PC computer, tablet device, personal digital assistant, smart phone, or other device capable of executing the set of instructions described above. The electronic device 20 need not be a single electronic device, but can be any collection of devices or circuits that can execute the above instructions (or sets of instructions) individually or in combination. The electronic device 20 may also be part of an integrated control system or system manager, or may be configured as a portable electronic device that interfaces with local or remote (e.g., via wireless transmission).
In the electronic device 20, the processor 202 may include a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), a programmable logic device, a special purpose processor system, a microcontroller, or a microprocessor. By way of example, and not limitation, processor 202 may also include an analog processor, a digital processor, a microprocessor, a multi-core processor, a processor array, a network processor, or the like.
The processor 202 may execute instructions or code stored in the memory 201, wherein the memory 201 may also store data. The instructions and data may also be transmitted or received over a network via a network interface device, which may employ any known transmission protocol.
In addition, the electronic device 20 may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, mouse, touch input device, etc.). All components of the electronic device 20 may be connected to each other via a bus and/or a network.
According to an exemplary embodiment of the present disclosure, there may also be provided a computer-readable storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform the video recommendation method according to the above-described exemplary embodiment. Examples of the computer-readable storage medium herein include: read-only memory (ROM), random-access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random-access memory (RAM), dynamic random-access memory (DRAM), static random-access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD + R, CD-RW, CD + RW, DVD-ROM, DVD-R, DVD + R, DVD-RW, DVD + RW, DVD-RAM, BD-ROM, BD-R, BD-R LTH, BD-RE, Blu-ray or compact disc memory, Hard Disk Drive (HDD), solid-state drive (SSD), card-type memory (such as a multimedia card, a Secure Digital (SD) card or a extreme digital (XD) card), magnetic tape, a floppy disk, a magneto-optical data storage device, an optical data storage device, a hard disk, a magnetic tape, a magneto-optical data storage device, a hard disk, a magnetic tape, a magnetic data storage device, a magnetic tape, a magnetic data storage device, a magnetic tape, a magnetic data storage device, a magnetic tape, a magnetic data storage device, a magnetic tape, a magnetic data storage device, A solid state disk, and any other device configured to store and provide a computer program and any associated data, data files, and data structures to a processor or computer in a non-transitory manner such that the processor or computer can execute the computer program. The computer program in the computer-readable storage medium described above can be run in an environment deployed in a computer apparatus, such as a client, a host, a proxy device, a server, and the like, and further, in one example, the computer program and any associated data, data files, and data structures are distributed across a networked computer system such that the computer program and any associated data, data files, and data structures are stored, accessed, and executed in a distributed fashion by one or more processors or computers.
According to an exemplary embodiment of the present disclosure, a computer program product may also be provided, in which instructions are executable by at least one processor to perform the video recommendation method according to the above exemplary embodiment.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (10)
1. A method for video recommendation, comprising:
acquiring a search keyword in a search request, and determining a video related to the search keyword;
sequencing the determined videos to obtain a target sequence;
sequencing videos meeting a first preset condition in the determined videos to obtain an exploration sequence;
determining a recommended sequence based on the exploration sequence and the target sequence;
determining recommended videos corresponding to the search requests according to the sequence of the videos in the recommended sequence;
wherein the first preset condition is as follows: the release duration is smaller than the duration threshold and the exposure times under the search request are smaller than the time threshold, wherein the release duration is a difference between the current time and the video release time.
2. The video recommendation method of claim 1, wherein the step of ordering the determined videos to obtain the target sequence comprises: sequencing the determined videos according to the estimated first recommendation score of each video;
the method for sequencing the videos meeting the first preset condition in the determined videos to obtain the exploration sequence comprises the following steps: according to the estimated second recommendation score of each video meeting the first preset condition, ordering the videos meeting the first preset condition in the determined videos;
wherein a higher first recommendation score indicates a higher recommendation degree, and a higher second recommendation score indicates a higher recommendation degree.
3. The video recommendation method of claim 2, wherein the step of determining the recommendation sequence based on the exploration sequence and the target sequence comprises:
determining an adjustment coefficient, so that the second recommendation score of the video ranked at the m-th position in the exploration sequence is multiplied by the adjustment coefficient to be larger than the first recommendation score of the video ranked at the n-th position in the target sequence;
multiplying the second recommendation score of each video in the exploration sequence by the adjustment coefficient to obtain a third recommendation score of the video;
and sequencing the determined videos according to the third recommendation scores of the videos ranked in the exploration sequence and the first recommendation scores of the videos ranked in the target sequence to obtain a recommendation sequence.
4. The video recommendation method according to claim 3, wherein in the step of ordering the determined videos to obtain the recommendation sequence,
if any one video has the first recommendation score and the third recommendation score, sorting the video according to the higher score in the first recommendation score and the third recommendation score;
if any one video only has the first recommendation score, the videos are sorted according to the first recommendation score.
5. The video recommendation method of claim 3, wherein the step of determining the adjustment factor comprises:
acquiring a ratio of a first recommendation score of the video ranked at the nth position in the target sequence to a second recommendation score of the video ranked at the mth position in the exploration sequence, and taking the ratio as a first ratio;
acquiring a ratio of a first recommendation score of the video ranked at the n-k position in the target sequence to a second recommendation score of the video ranked at the m position in the exploration sequence as a second ratio, wherein k is an integer greater than or equal to 1;
the adjustment coefficient is determined to be a value greater than the first ratio and less than the second ratio.
6. The video recommendation method according to claim 2, wherein the second recommendation score of each video satisfying the first preset condition is estimated by:
for each video meeting a first preset condition, inputting a bipartite graph result about the video and a consumer into a graph neural network to obtain the characteristics of the video;
and obtaining a second recommendation score of the video based on the characteristics of the video and the characteristics of the search keywords.
7. A video recommendation apparatus, comprising:
the video determining unit is configured to obtain a search keyword in a search request and determine a video related to the search keyword;
the target sequence determining unit is configured to sequence the determined videos to obtain a target sequence;
the searching sequence determining unit is configured to sequence videos meeting a first preset condition in the determined videos to obtain a searching sequence;
a recommended sequence determination unit configured to determine a recommended sequence based on the exploration sequence and the target sequence;
the recommended video determining unit is configured to determine a recommended video corresponding to the search request according to the sequence of videos in the recommended sequence;
wherein the first preset condition is as follows: the release duration is smaller than the duration threshold and the exposure times under the search request are smaller than the time threshold, wherein the release duration is a difference between the current time and the video release time.
8. An electronic device, comprising:
at least one processor;
at least one memory storing computer-executable instructions,
wherein the computer-executable instructions, when executed by the at least one processor, cause the at least one processor to perform the video recommendation method of any of claims 1-6.
9. A computer-readable storage medium, wherein instructions in the computer-readable storage medium, when executed by at least one processor, cause the at least one processor to perform the video recommendation method of any of claims 1-6.
10. A computer program product comprising computer instructions, wherein the computer instructions, when executed by at least one processor, implement the video recommendation method of any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111511395.8A CN114168792A (en) | 2021-12-06 | 2021-12-06 | Video recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111511395.8A CN114168792A (en) | 2021-12-06 | 2021-12-06 | Video recommendation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114168792A true CN114168792A (en) | 2022-03-11 |
Family
ID=80485701
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111511395.8A Pending CN114168792A (en) | 2021-12-06 | 2021-12-06 | Video recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114168792A (en) |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1999043157A1 (en) * | 1998-02-19 | 1999-08-26 | Motorola Inc. | System and method for non-causal encoding of video information for improved streaming thereof |
| CN103440335A (en) * | 2013-09-06 | 2013-12-11 | 北京奇虎科技有限公司 | Video recommendation method and device |
| US20150095183A1 (en) * | 2013-09-30 | 2015-04-02 | International Business Machines Corporation | Methods and Apparatus for Generating Recommendations |
| CN104951468A (en) * | 2014-03-28 | 2015-09-30 | 阿里巴巴集团控股有限公司 | Data searching and processing method and system |
| CN104991962A (en) * | 2015-07-22 | 2015-10-21 | 无锡天脉聚源传媒科技有限公司 | Method and apparatus for generating recommendation information |
| CN107977405A (en) * | 2017-11-16 | 2018-05-01 | 北京三快在线科技有限公司 | Data reordering method, data sorting device, electronic equipment and readable storage medium storing program for executing |
| CN108108446A (en) * | 2017-12-27 | 2018-06-01 | 暴风集团股份有限公司 | Video recommendation method and device, electronic equipment, storage medium, program |
| CN109684527A (en) * | 2018-12-26 | 2019-04-26 | 北京金山安全软件有限公司 | Data sorting method and device, electronic equipment and storage medium |
| CN110658979A (en) * | 2018-06-29 | 2020-01-07 | 杭州海康威视系统技术有限公司 | Data reconstruction method and device, electronic equipment and storage medium |
| CN110837578A (en) * | 2019-11-06 | 2020-02-25 | 合肥工业大学 | Video clip recommendation method based on graph convolution network |
| CN111444380A (en) * | 2020-03-26 | 2020-07-24 | 腾讯音乐娱乐科技（深圳）有限公司 | Music search sorting method, device, equipment and storage medium |
| CN112287238A (en) * | 2020-12-30 | 2021-01-29 | 腾讯科技（深圳）有限公司 | User characteristic determination method and device, storage medium and electronic equipment |
| CN112395458A (en) * | 2019-08-13 | 2021-02-23 | 必艾奇亚洲有限公司 | Video course recommendation system and video course recommendation method for fitness equipment |
| CN112445996A (en) * | 2020-12-10 | 2021-03-05 | 掌阅科技股份有限公司 | Content distribution method, computing device and computer storage medium |
| CN112925924A (en) * | 2019-12-05 | 2021-06-08 | 北京达佳互联信息技术有限公司 | Multimedia file recommendation method and device, electronic equipment and storage medium |
| CN113377972A (en) * | 2020-03-09 | 2021-09-10 | 北京达佳互联信息技术有限公司 | Multimedia content recommendation method and device, computing equipment and storage medium |
-
2021
- 2021-12-06 CN CN202111511395.8A patent/CN114168792A/en active Pending
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1999043157A1 (en) * | 1998-02-19 | 1999-08-26 | Motorola Inc. | System and method for non-causal encoding of video information for improved streaming thereof |
| CN103440335A (en) * | 2013-09-06 | 2013-12-11 | 北京奇虎科技有限公司 | Video recommendation method and device |
| US20150095183A1 (en) * | 2013-09-30 | 2015-04-02 | International Business Machines Corporation | Methods and Apparatus for Generating Recommendations |
| CN104951468A (en) * | 2014-03-28 | 2015-09-30 | 阿里巴巴集团控股有限公司 | Data searching and processing method and system |
| CN104991962A (en) * | 2015-07-22 | 2015-10-21 | 无锡天脉聚源传媒科技有限公司 | Method and apparatus for generating recommendation information |
| CN107977405A (en) * | 2017-11-16 | 2018-05-01 | 北京三快在线科技有限公司 | Data reordering method, data sorting device, electronic equipment and readable storage medium storing program for executing |
| CN108108446A (en) * | 2017-12-27 | 2018-06-01 | 暴风集团股份有限公司 | Video recommendation method and device, electronic equipment, storage medium, program |
| CN110658979A (en) * | 2018-06-29 | 2020-01-07 | 杭州海康威视系统技术有限公司 | Data reconstruction method and device, electronic equipment and storage medium |
| CN109684527A (en) * | 2018-12-26 | 2019-04-26 | 北京金山安全软件有限公司 | Data sorting method and device, electronic equipment and storage medium |
| CN112395458A (en) * | 2019-08-13 | 2021-02-23 | 必艾奇亚洲有限公司 | Video course recommendation system and video course recommendation method for fitness equipment |
| CN110837578A (en) * | 2019-11-06 | 2020-02-25 | 合肥工业大学 | Video clip recommendation method based on graph convolution network |
| CN112925924A (en) * | 2019-12-05 | 2021-06-08 | 北京达佳互联信息技术有限公司 | Multimedia file recommendation method and device, electronic equipment and storage medium |
| CN113377972A (en) * | 2020-03-09 | 2021-09-10 | 北京达佳互联信息技术有限公司 | Multimedia content recommendation method and device, computing equipment and storage medium |
| CN111444380A (en) * | 2020-03-26 | 2020-07-24 | 腾讯音乐娱乐科技（深圳）有限公司 | Music search sorting method, device, equipment and storage medium |
| CN112445996A (en) * | 2020-12-10 | 2021-03-05 | 掌阅科技股份有限公司 | Content distribution method, computing device and computer storage medium |
| CN112287238A (en) * | 2020-12-30 | 2021-01-29 | 腾讯科技（深圳）有限公司 | User characteristic determination method and device, storage medium and electronic equipment |
Non-Patent Citations (3)
| Title |
|---|
| "Quality-of-experience assessment and its application to video services in LTE networks", 《IN IEEE WIRELESS COMMUNICATIONS》, vol. 22, no. 1, 28 February 2015 (2015-02-28), pages 70 - 78 * |
| 刘君良 等: "个性化推荐系统技术进展", 《计算机科学》, vol. 47, no. 7, 28 June 2020 (2020-06-28), pages 47 - 55 * |
| 黄震华 等: "基于排序学习的推荐算法研究综述", 《软件学报》, vol. 27, no. 3, 30 December 2015 (2015-12-30), pages 691 - 713 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20240078258A1 (en) | Training Image and Text Embedding Models | |
| US11062086B2 (en) | Personalized book-to-movie adaptation recommendation | |
| CN105144164B (en) | Scoring concept terms using a deep network | |
| CN104834729B (en) | Topic recommends method and topic recommendation apparatus | |
| CN109690529B (en) | Compiling documents into a timeline by event | |
| US10217058B2 (en) | Predicting interesting things and concepts in content | |
| US9607267B2 (en) | System and method for mining tags using social endorsement networks | |
| CN110134931B (en) | Medium title generation method, medium title generation device, electronic equipment and readable medium | |
| US12038970B2 (en) | Training image and text embedding models | |
| JP6047550B2 (en) | Search method, client and server | |
| US9645987B2 (en) | Topic extraction and video association | |
| US20200265074A1 (en) | Searching multilingual documents based on document structure extraction | |
| US8856109B2 (en) | Topical affinity badges in information retrieval | |
| US11276099B2 (en) | Multi-perceptual similarity detection and resolution | |
| US11308146B2 (en) | Content fragments aligned to content criteria | |
| CN112528147A (en) | Content recommendation method and apparatus, training method, computing device, and storage medium | |
| CN111291564B (en) | Model training method, device and storage medium for word vector acquisition | |
| CN117011737A (en) | Video classification method and device, electronic equipment and storage medium | |
| CN114817692A (en) | Method, device and equipment for determining recommended object and computer storage medium | |
| CN114168792A (en) | Video recommendation method and device | |
| US10296623B2 (en) | Document curation | |
| CN114254151A (en) | Training method of search term recommendation model, search term recommendation method and device | |
| CN113223017A (en) | Training method of target segmentation model, target segmentation method and device | |
| KR102081553B1 (en) | Big Data-Based Monitoring System of Promotional Content for Cultural Media | |
| JP6604608B2 (en) | Information processing apparatus, information processing method, and information processing program |
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 |