Training method and device for playing parameter recommendation model

Abstract

The embodiment of the invention provides a playing parameter recommendation model training method and device, and the method comprises the steps: carrying out the coding processing of a training sample, obtaining a sample feature, carrying out the correlation feature detection of the sample feature, obtaining a sample correlation feature, and carrying out the playing parameter recommendation according to the sample correlation feature, and finally, screening the recommended playing parameters, training the to-be-trained model according to the target playing parameters obtained through screening and the corresponding playing parameter labels, obtaining the playing parameter recommendation model after training is completed, and by adopting the playing parameter recommendation model, according to the network environment and the equipment capability of the user, playing the to-be-trained model according to the network environment and the equipment capability of the user. The appropriate target playing parameters are recommended for the user, so that the problem of unsmooth video playing is avoided, the watching experience of the user is improved, and the user retention rate and the user satisfaction degree are increased.

Description

Technical Field

[0001] This document relates to the field of artificial intelligence, and in particular to a training method and device for a playback parameter recommendation model. Background Art

[0002] With the development of information technology, people have higher and higher requirements for video playback quality. Any factors that affect video playback quality may reduce user experience and even lead to business loss. Therefore, it is very important to adjust the video playback parameters according to the network environment in a timely manner while users are watching videos. The existing technical solution recommends appropriate clarity through network bandwidth and buffering rate. This solution fails to consider the impact of objective factors such as user's historical usage habits and the quality of the resource itself on user behavior. At the same time, the setting and changes of this solution are all set up and jumped manually, and it is unable to respond to changes quickly. For example, some users are accustomed to pausing buffering and playing high-definition videos, while some users are unwilling to wait and accept low-resolution video playback and similar situations. Therefore, how to improve the video playback quality has become a focus of future attention. Summary of the invention

[0003] An object of an embodiment of the present specification is to provide a method and device for training a playback parameter recommendation model to solve the problems of low efficiency and low accuracy of playback parameter recommendation.

[0004] To solve the above technical problems, an embodiment of this specification is implemented as follows: In a first aspect, an embodiment of the present specification provides a method for training a playback parameter recommendation model, the method comprising: Encoding the training samples to obtain sample features; the training samples include expanded samples obtained by expanding the initial samples using a sample expansion algorithm; Performing correlation feature detection on the sample features to obtain sample correlation features; Recommending playback parameters according to the sample-related features to obtain recommended playback parameters; The recommended playback parameters are screened, and the to-be-trained model is trained according to the target playback parameters obtained through the screening and the corresponding playback parameter labels, and a playback parameter recommendation model is obtained after the training is completed.

[0005] In a second aspect, another embodiment of the present specification provides a playback parameter recommendation method, the method comprising: Obtain video playback information, video playback parameters, and viewing user information of the video played on the playback end; Inputting the video playback information, the video playback parameters, and the viewing user information into a playback parameter recommendation model to perform playback parameter recommendation, thereby obtaining first recommended playback parameters; Determine a second recommended playback parameter based on the preset playback parameter and the video playback information, and calculate and obtain a target playback parameter according to the first recommended playback parameter and the second recommended playback parameter; The target playback parameters are returned to the playback end.

[0006] In a third aspect, an embodiment of the present specification provides a training device for a playback parameter recommendation model, the device comprising: A sample encoding module is configured to encode training samples to obtain sample features; the training samples include expanded samples obtained by expanding initial samples using a sample expansion algorithm; A feature detection module is configured to perform associated feature detection on the sample features to obtain sample associated features; A parameter recommendation module is configured to recommend playback parameters according to the sample association features to obtain recommended playback parameters; The model training module is configured to screen the recommended playback parameters, train the to-be-trained model according to the target playback parameters obtained by the screening and the corresponding playback parameter labels, and obtain a playback parameter recommendation model after the training is completed.

[0007] In a fourth aspect, another embodiment of the present specification provides a playback parameter recommendation device, the device comprising: An information acquisition module is configured to acquire video playback information, video playback parameters, and viewing user information of the video played on the playback end; A parameter recommendation module is configured to input the video playback information, the video playback parameters and the viewing user information into a playback parameter recommendation model to perform playback parameter recommendation and obtain a first recommended playback parameter; a parameter determination module, configured to determine a second recommended playback parameter based on a preset playback parameter and the video playback information, and to calculate a target playback parameter according to the first recommended playback parameter and the second recommended playback parameter; The parameter returning module is configured to return the target playback parameter to the playback end.

[0008] In a fifth aspect, an embodiment of the present specification provides a training device for a playback parameter recommendation model, comprising: a memory, a processor, and computer executable instructions stored on the memory and executable on the processor, wherein the computer executable instructions, when executed by the processor, implement the steps of the training method for the playback parameter recommendation model as described in the first aspect above.

[0009] In the sixth aspect, another embodiment of the present specification provides a playback parameter recommendation device, comprising: a memory, a processor, and computer executable instructions stored on the memory and executable on the processor, wherein the computer executable instructions, when executed by the processor, implement the steps of the playback parameter recommendation method as described in the first aspect above.

[0010] In a seventh aspect, an embodiment of the present specification provides a computer-readable storage medium, which is used to store computer-executable instructions. When the computer-executable instructions are executed by a processor, the steps of the training method of the playback parameter recommendation model as described in the first aspect above are implemented.

[0011] In an eighth aspect, another embodiment of the present specification provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store computer-executable instructions, and when the computer-executable instructions are executed by a processor, the steps of the playback parameter recommendation method as described in the first aspect above are implemented.

[0012] In a ninth aspect, an embodiment of the present specification provides a computer program product, wherein the computer program product includes a training program for a playback parameter recommendation model, and the training program for the playback parameter recommendation model is executed by a processor to implement the steps of the training method for the playback parameter recommendation model as described in the first aspect above.

[0013] In the tenth aspect, another embodiment of the present specification provides a computer program product, wherein the computer program product includes a playback parameter recommendation program, and the playback parameter recommendation program is executed by a processor to implement the steps of the playback parameter recommendation method as described in the first aspect above.

[0014] The training method of the playback parameter recommendation model provided in this embodiment, during the playback parameter recommendation model training process, firstly, the training samples are encoded to obtain sample features, then the sample features are detected for associated features to obtain sample associated features, and playback parameters are recommended again based on the sample associated features to obtain recommended playback parameters, and finally the recommended playback parameters are screened, and the training model is trained based on the target playback parameters obtained by screening and the corresponding playback parameter labels. After the training is completed, the playback parameter recommendation model is obtained. By using the playback parameter recommendation model, suitable target playback parameters can be recommended to users based on their network environment and device capabilities, so as to avoid the problem of unsmooth video playback, improve the user viewing experience, and increase user retention rate and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in one or more embodiments of the present specification, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0016] Figure 1 A processing flow chart of a method for training a playback parameter recommendation model provided in one embodiment of this specification; Figure 2 A schematic diagram of a model encoding process provided for one embodiment of this specification; Figure 3 A flowchart of an associated feature detection provided for one embodiment of this specification; Figure 4 A schematic diagram of the architecture of a model to be trained provided in one embodiment of this specification; Figure 5 A processing flow chart of a training method for a playback parameter recommendation model applied to a training scenario of a playback parameter recommendation model provided by one embodiment of the present specification; Figure 6 A processing flow chart of a playback parameter recommendation method provided in one embodiment of this specification; Figure 7 A schematic diagram of a training device for a playback parameter recommendation model provided by an embodiment of this specification; Figure 8 A schematic diagram of a playback parameter recommendation device provided in one embodiment of this specification; Fig. 9 A schematic diagram of the structure of a training device for a playback parameter recommendation model provided by one embodiment of this specification; Fig.10 A schematic diagram of the structure of a playback parameter recommendation device provided in one embodiment of this specification. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.

[0018] One or more embodiments of a playback parameter recommendation method provided in this specification are as follows: Reference Figure 1 , which shows a processing flow chart of a playback parameter recommendation method provided by this embodiment, specifically including the following steps S102 to S108.

[0019] Step S102: Encode the training samples to obtain sample features.

[0020] In this embodiment, the training samples can be obtained by embedding points at the playback end, where the playback end can be a playback application, a browser, a PC client and / or a mini-program; specifically, by embedding points at the playback end, when the embedding point is triggered, the training samples are obtained.

[0021] Optionally, the training sample includes an expanded sample obtained by expanding the initial sample using a sample expansion algorithm; the initial sample includes at least one of the following: video information, user information, and playback parameters; wherein the video information may be video playback environment information (such as IP, network operator, and region), or may also be video playback information (such as buffer data information, CDN, video ID, current playback time, resolution, decoder type, bit rate, video format, and audio format); the playback parameter may be the playback clarity of the video, or a selection record of the playback clarity of the video (such as a switching record of the video clarity); the user information may be information of the user viewing the video (such as age, gender, personality classification, user portrait, current emotion, activity, preference, and occupation).

[0022] In specific implementation, there may be a situation where the richness of the initial sample is not enough, so the initial sample needs to be expanded. In an optional implementation provided by this embodiment, the initial sample is expanded in the following manner: The video information, user information and playback parameters are expanded based on a sample expansion algorithm and sample expansion parameters to obtain the expanded sample.

[0023] Specifically, the sample expansion algorithm may be a sample expansion curve, and the sample expansion parameter may be an expansion adjustment factor, an expansion rate, or an expansion number. The expanded sample may include expanded video information, expanded user information, and / or expanded playback parameters. After the video information, user information, and playback parameters are expanded, the obtained expanded sample may be used together with the video information, user information, and playback parameters as a training sample.

[0024] The following example is used to illustrate the process of sample expansion. In the process of sample expansion, the total number of adjustments c during playback is first obtained. When c>2, expansion is performed, otherwise the expansion is terminated. Then, the playback records of other users with the same user tag as the current user and other videos with the same video tag as the current video are obtained. R={ }, determine the expansion adjustment factor λ, which is configured through system parameters, and λ∈(0,1000), and calculate the expansion curve as follows :

[0025] Adjust the expansion rate :

[0026] in , Represents a set of positive integers from 1 to c-2, and gets the hash point , in the curve Go up, take a point The corresponding value is

[0027] Get T = { },in: , then traverse T, , ,in (R,k) means to find the kth record in R, k∈ ,get , last updated The user information is the current user information, the played video is the current video, and the selected value is the currently selected definition option, completing the sample expansion.

[0028] In the specific implementation process, after obtaining the training samples, in order to facilitate model recognition and processing, the training samples are encoded to obtain sample features. Here, since the training samples contain multiple types of samples, the training samples can be encoded according to different sample types; Specifically, for the user information contained in the training sample, in the process of encoding the user information sample, the sub-user information contained in the user information can be transformed to form a data quantization expression U=[ ], i.e., the first sample feature, ensures that the characteristics of the user information remain unchanged after conversion; for the video information contained in the training sample, each sub-video information contained in the video information can be converted to form a data quantitative expression r=[ ], i.e., the second sample feature, ensures that the characteristics of the video playback information remain unchanged after conversion.

[0029] In addition, for the playback parameters included in the training samples, in an optional implementation provided by this embodiment, the playback parameters are encoded in the following manner: Pre-encoding the playback parameters contained in the training sample to obtain playback parameter features; Constructing a feature matrix based on the playback parameter feature, and calculating a position vector and an attention weight vector of the playback parameter feature; The feature matrix, the position vector, the attention weight vector, and the playback parameter feature are concatenated to obtain the sample feature.

[0030] Specifically, for the playback parameter samples included in the training samples, the playback parameter samples can be first pre-encoded to obtain playback parameter features, and then a feature matrix is ​​constructed based on the playback parameter features, and the position vector and attention weight vector of the playback parameter features are calculated. The feature matrix, position vector, attention weight vector and playback parameter features are spliced ​​to obtain a third sample feature.

[0031] For example, refer to Figure 2 As shown in the schematic diagram of the model encoding process, the playback parameters are first pre-encoded to obtain playback parameter features, and then a feature matrix is ​​constructed based on the playback parameter features, and the position vector and attention weight vector of the playback parameter features are calculated. The feature matrix, position vector, attention weight vector and playback parameter features are spliced, and the spliced ​​results are mean pooled to obtain sample features.

[0032] Among them, the position vector can be calculated using the following formula : ,(in: , d represents the dimension of encoding) The attention weight vector can be calculated using the following formula: : , (where: d also represents the dimension of the encoding) Furthermore, in the process of constructing the feature matrix, associated parameter features that are associated with the playback parameter features may be obtained, and the feature matrix may be constructed based on the associated parameter features. In an optional implementation provided in this embodiment, the feature matrix may be constructed based on the playback parameter features in the following manner: Acquire associated parameter features that are associated with the playback parameter features, and group the associated parameter features to obtain feature groups; Calculating the similarity between the playback parameter feature and the feature grouping, sorting the feature grouping according to the similarity, and obtaining a feature grouping sequence; The feature groupings in a preset position in the feature grouping sequence are concatenated with the video playback features to obtain the feature matrix.

[0033] The associated parameter feature may be a playback parameter feature of other users who watch the video corresponding to the playback parameter feature.

[0034] For example, the playback parameter feature corresponds to video A, and the playback parameter features of video A by other viewing users are obtained:

[0035] Where m=min(c,1000), where c is the number of playback history records of all users who played video A.

[0036] Will Group by user, where i is the number of users, and get the grouped list =[ ] Among them, 0 <i<m; 0< <m; =m The similarity is calculated according to the following formula:

[0037]

[0038]

[0039] Calculate the similarity between each user's historical record data and other user group data to get:

[0040]

[0041]

[0042] By group: Take the maximum value of each to get the most similar historical records for each user , and Splicing, we get:

[0043] Will It serves as the user operation record encoding input for the model.

[0044] Among them, there are n operation records. After encoding, an operation record encoding matrix with n rows and p columns is formed. ,according to , generate the operation sort position encoding matrix, where for any record i, (1) Generate position vector The rules are: ,(in: , d represents the dimension of encoding), (2) Generate weight vector The rules are: , (where: d also represents the dimension of the encoding) Will Add to the end of the operation record vector to form a new operation record o with dimension n, 2p+2.

[0045] Step S104: performing correlation feature detection on the sample features to obtain sample correlation features.

[0046] After the training samples are encoded to obtain sample features, in this step, the sample features are subjected to correlation feature detection to obtain sample correlation features. Specifically, the sample features can be input into the correlation feature detection module of the model to be trained to perform correlation feature detection to obtain sample correlation features.

[0047] In the specific execution process, when the correlation feature detection module performs correlation feature detection, the first sample correlation feature can be queried through the correlation detection matrix q on the one hand, and the second sample correlation feature can be queried through the correlation detection matrix k on the other hand, and then the sample correlation feature is obtained by splicing the first sample correlation feature, the second sample correlation feature and the sample feature. In an optional implementation manner provided by this embodiment, the following method is used to perform correlation feature detection based on the sample feature to obtain the sample correlation feature, including: Multiplying the sample feature with a first correlation detection matrix to obtain a first sample correlation feature, and multiplying the sample feature with a second correlation detection matrix to obtain a second sample correlation feature; The sample feature, the first sample association feature, and the second sample association feature are concatenated to obtain the sample association feature.

[0048] For example, see Figure 3 The process of correlation feature detection shown in FIG. 1 is to perform correlation feature detection, wherein the sample feature is multiplied by the first correlation detection matrix to obtain the first sample correlation feature, which can be expressed as: , multiply the sample feature by the second association detection matrix to obtain the second sample association feature, which can be expressed as: , the sample association features are: =[ ] Step S106: Recommend playback parameters based on the sample correlation features to obtain the recommended playback parameters.

[0049] After the above-mentioned correlation feature detection of the sample features to obtain the sample correlation features, in this step, recommend playback parameters based on the sample correlation features to obtain the recommended playback parameters. Specifically, the sample correlation features can be input into the parameter recommendation module of the model to be trained for playback parameter recommendation. Among them, the parameter recommendation module can include a recurrent neural network (RNN) composed of a residual connection layer, a regularization layer, and an activation layer, and a linear layer. Optionally, the model to be trained includes a recurrent neural network, where the recurrent neural network includes a residual connection layer, a regularization layer, and an activation layer.

[0050] Step S108: Screen the recommended playback parameters, and train the model to be trained according to the target playback parameters obtained by screening and the corresponding playback parameter tags. After the training is completed, a playback parameter recommendation model is obtained.

[0051] In practical applications, the recommended playback parameters output by the pre-trained model may not be within the applicable range of playback parameters. After the above-mentioned recommendation of playback parameters based on the sample correlation features to obtain the recommended playback parameters, it is necessary to screen the recommended playback parameters to obtain the target playback parameters. In an optional implementation provided in this embodiment, the following method is used to screen the recommended playback parameters: Eliminate the abnormal recommended playback parameters that are inconsistent with the preset playback parameters in the recommended playback parameters to obtain the target playback parameters.

[0052] Specifically, the preset playback parameters can be the playback parameters supported by the playback end. In the process of eliminating the abnormal recommended playback parameters that are inconsistent with the preset playback parameters in the recommended playback parameters, the vector representation corresponding to the recommended playback parameters can be multiplied by the vector representation corresponding to the preset playback parameters to obtain the target playback parameters.

[0053] For example, let the number of playback parameters supported by the system be n. When there are m playback parameters, use an m-dimensional vector to represent the set of playback parameters. , for the kth playback parameter, its corresponding playback parameter vector is , where the kth dimension value is 1 and the rest are 0. The adjustment factor output by the model is , where: , 0 < m ≤ n, where the adjustment factor is generated according to the resource information and does not participate in the backpropagation during the model training process.

[0054] In the specific implementation process, after the recommended playback parameters are screened and the target playback parameters are obtained, model training can be performed. In an optional implementation provided in this embodiment, the model training is performed in the following manner: Calculating the loss according to the target playback parameter and the playback parameter label; The parameters of the model to be trained are adjusted based on the calculated loss to obtain the playback parameter recommendation model.

[0055] The following Figure 4 Taking as an example, the architecture of the training model is explained. First, the training samples are obtained. Then, the video information samples, user information samples, and playback parameter samples in the training samples are encoded and processed respectively to obtain sample features. Then, the sample features are input into the associated feature detection module for associated feature detection. The obtained sample associated features are input into the parameter recommendation module composed of the residual link layer, the regularization layer, the activation layer, and the linear layer to obtain the recommended playback parameters. Then, the recommended playback parameters are screened to obtain the target playback parameters.

[0056] To summarize, the training method of the playback parameter recommendation model provided in this embodiment first encodes the training samples to obtain sample features, then performs associated feature detection on the sample features to obtain sample associated features, and then recommends playback parameters based on the sample associated features to obtain recommended playback parameters, and finally screens the recommended playback parameters, and trains the training model based on the target playback parameters obtained by screening and the corresponding playback parameter labels. After the training is completed, a playback parameter recommendation model is obtained. By using the playback parameter recommendation model, appropriate target playback parameters can be recommended to users based on their network environment and device capabilities to avoid the problem of unsmooth video playback, improve user viewing experience, and increase user retention rate and user satisfaction.

[0057] The following combination Figure 5 , taking the application of the training method of the playback parameter recommendation model provided by this embodiment in the training scenario of the playback parameter recommendation model as an example, the training method of the playback parameter recommendation model provided by this embodiment is further described. Figure 5 The training method of the playback parameter recommendation model applied to the training scenario of the playback parameter recommendation model specifically includes steps S502 to S514.

[0058] Step S502: obtaining training samples.

[0059] Optionally, the training sample includes at least one of the following: video information, user information, playback parameters, and extended samples.

[0060] Step S504: Encode the training samples to obtain sample features.

[0061] Step S506: multiply the sample feature by the first association detection matrix to obtain the first sample association feature, and multiply the sample feature by the second association detection matrix to obtain the second sample association feature.

[0062] Step S508: concatenate the sample feature, the first sample association feature, and the second sample association feature to obtain the sample association feature.

[0063] Step S510: recommending playback parameters based on sample-related features to obtain recommended playback parameters.

[0064] Step S512: Screen the recommended playback parameters to obtain target playback parameters.

[0065] Step S514, training the to-be-trained model according to the target playback parameters and the playback parameter labels corresponding to the target playback parameters, and obtaining a playback parameter recommendation model after the training is completed.

[0066] This specification provides an example of a playback parameter recommendation method: Reference Figure 6 , which shows a processing flow chart of a training method for a playback parameter recommendation model provided in this embodiment. The training method for a playback parameter recommendation model provided in this embodiment specifically includes the following steps S602 to S608.

[0067] Step S602, obtaining video playback information, video playback parameters, and viewing user information of the video played on the playback end.

[0068] The video playback information may be playback environment information (e.g., IP, network operator, region), or may be video playback information (e.g., buffer data information, CDN, video ID, current playback time, resolution, decoder type, bit rate, video format, audio format); the viewing user information may be information of the user viewing the video (e.g., age, gender, personality classification, user portrait, current emotion, activity, preference, occupation); the video playback parameters include video clarity.

[0069] In an optional implementation manner provided by this embodiment, the video playback information, video playback parameters and viewing user information are obtained in the following manner: Determine a response event according to the operation instruction sent by the playback end; Acquire the target tracking data corresponding to the response event; the tracking data includes the video playback information, the video playback parameters and the viewing user information.

[0070] The response event includes: a play event for the play video, a switch event for the play state of the play video, and an adjustment event for the play parameter of the play video.

[0071] Step S604: input the video playback information, video playback parameters, and viewing user information into a playback parameter recommendation model to perform playback parameter recommendation and obtain first recommended playback parameters.

[0072] Step S606: determining second recommended playback parameters based on preset playback parameters and the video playback information, and calculating target playback parameters according to the first recommended playback parameters and the second recommended playback parameters.

[0073] In an optional implementation manner provided by this embodiment, the second playback recommendation parameter is determined in the following manner: Calculating a preset bandwidth corresponding to the preset playback parameter, and filtering the preset bandwidth based on the network bandwidth included in the video playback information to obtain a target bandwidth; The target bandwidths are sorted to obtain a bandwidth sequence, and the preset playback parameters corresponding to the target bandwidths in a preset position in the bandwidth sequence are used as the second recommended playback parameters.

[0074] For example, to get the clarity options collection, , and get the optional selection adjustment factor , obtain the user's current network bandwidth NS, obtain the minimum bandwidth of each definition in the definition option set C corresponding to the resource playback, and obtain , according to the following formula:

[0075] Where max represents getting the maximum value:

[0076] Traversal , get the optional items in the tuple middle Corresponding , which is the recommended value output by the model.

[0077] In an optional implementation manner provided by this embodiment, the target playback parameter is calculated according to the first playback parameter and the second playback parameter in the following manner: Determining a parameter weight of the first playback recommendation parameter, and calculating a parameter recommendation success rate of the first playback recommendation parameter; The target playback parameter is calculated based on the parameter weight, the parameter recommendation success rate, the first playback recommendation parameter, and the second playback recommendation parameter.

[0078] For example, the target playback parameters can be calculated using the following formula:

[0079] in, ,

[0080]

[0081] The number of records where the resolution selected by the user on that day is consistent with the resolution recommended by the AI ​​clarity selection recommendation model. The recommended number of records for the total resolution for the day.

[0082] Step S608: Return the target playback parameters to the playback end.

[0083] In an optional implementation manner provided by this embodiment, the target playback parameters are returned to the playback end in the following manner: A parameter recommendation message including the target playback parameter is generated, and the parameter recommendation message is returned to the playback end.

[0084] To summarize, the training method of the playback parameter recommendation model provided in this embodiment first encodes the training samples to obtain sample features, then performs associated feature detection on the sample features to obtain sample associated features, and then recommends playback parameters based on the sample associated features to obtain recommended playback parameters, and finally screens the recommended playback parameters, and trains the training model based on the target playback parameters obtained by screening and the corresponding playback parameter labels. After the training is completed, a playback parameter recommendation model is obtained. By using the playback parameter recommendation model, appropriate target playback parameters can be recommended to users based on their network environment and device capabilities to avoid the problem of unsmooth video playback, improve user viewing experience, and increase user retention rate and user satisfaction.

[0085] Figure 7 A schematic diagram of a playback parameter recommendation device provided by an embodiment of the present invention is shown in FIG. Figure 7 As shown, the device comprises: The sample encoding module 702 is configured to encode the training samples to obtain sample features; the training samples include expanded samples obtained by expanding the initial samples using a sample expansion algorithm; The feature detection module 704 is configured to perform correlation feature detection on the sample feature to obtain sample correlation features; The parameter recommendation module 706 is configured to recommend playback parameters according to the sample association features to obtain recommended playback parameters; The model training module 708 is configured to screen the recommended playback parameters, train the to-be-trained model according to the target playback parameters obtained through the screening and the corresponding playback parameter labels, and obtain a playback parameter recommendation model after the training is completed.

[0086] The playback parameter recommendation device provided in this embodiment first runs the sample encoding module 702 to encode the training sample to obtain the sample feature, then runs the feature detection module 704 to perform associated feature detection on the sample feature to obtain the sample associated feature, then runs the parameter recommendation module 706 to recommend the playback parameter according to the sample associated feature to obtain the recommended playback parameter, and finally runs the model training module 708 to screen the recommended playback parameters, and trains the to-be-trained model according to the screened target playback parameters and the corresponding playback parameter labels. After the training is completed, the playback parameter recommendation model is obtained. The playback parameter recommendation model can be used to recommend appropriate target playback parameters to the user according to the user's network environment and device capabilities, so as to avoid the problem of unsmooth video playback, improve the user's viewing experience, and increase the user retention rate and user satisfaction.

[0087] The training device for the playback parameter recommendation model provided in an embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.

[0088] Figure 8 A schematic diagram of another training device for a playback parameter recommendation model provided by an embodiment of the present invention, such as Figure 8 As shown, the device comprises: The information acquisition module 802 is configured to acquire video playback information, video playback parameters, and viewing user information of the video played on the playback end; The parameter recommendation module 804 is configured to input the video playback information, the video playback parameters and the viewing user information into a playback parameter recommendation model to perform playback parameter recommendation and obtain a first recommended playback parameter; A parameter determination module 806 is configured to determine a second recommended playback parameter based on the preset playback parameter and the video playback information, and calculate a target playback parameter according to the first recommended playback parameter and the second recommended playback parameter; The parameter returning module 808 is configured to return the target playback parameter to the playback end.

[0089] The training device of the playback parameter recommendation model provided in this embodiment, the training device of the playback parameter recommendation model provided in this embodiment, first runs the information acquisition module 802 to obtain the video playback information, video playback parameters and viewing user information of the video played in the playback end, and then runs the parameter recommendation module 804 to input the video playback information, video playback parameters and the viewing user information into the playback parameter recommendation model to recommend playback parameters and obtain the first recommended playback parameters, and then runs the parameter determination module 806 to determine the second recommended playback parameters based on the preset playback parameters and the video playback information, and calculates the target playback parameters according to the first recommended playback parameters and the second recommended playback parameters, and finally runs the parameter return module 808 to return the target playback parameters to the playback end. The playback parameter recommendation model can be used to recommend appropriate target playback parameters to users according to the user's network environment and device capabilities, so as to avoid the problem of unsmooth video playback, improve the user's viewing experience, and increase user retention rate and user satisfaction.

[0090] The playback parameter recommendation device provided in an embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.

[0091] Furthermore, an embodiment of the present specification also provides a training device for a playback parameter recommendation model. Fig. 9 A structural diagram of a training device for a playback parameter recommendation model provided in an embodiment of this specification, such as Fig. 9 As shown, the device includes: a memory 901, a processor 902, a bus 903 and a communication interface 904. The memory 901, the processor 902 and the communication interface 904 communicate through the bus 903, and the communication interface 904 may include an input and output interface, which includes but is not limited to a keyboard, a mouse, a display, a microphone, a loudspeaker, etc.

[0092] Fig. 9 In the embodiment, the memory 901 stores computer executable instructions that can be run on the processor 902. When the computer executable instructions are executed by the processor 902, the following process is implemented: Encoding the training samples to obtain sample features; the training samples include expanded samples obtained by expanding the initial samples using a sample expansion algorithm; Performing correlation feature detection on the sample features to obtain sample correlation features; Recommending playback parameters according to the sample-related features to obtain recommended playback parameters; The recommended playback parameters are screened, and the to-be-trained model is trained according to the target playback parameters obtained through the screening and the corresponding playback parameter labels, and a playback parameter recommendation model is obtained after the training is completed.

[0093] The training device of the playback parameter recommendation model provided in this embodiment, through the cooperation of the memory 901, the processor 902, the bus 903 and the communication interface 904, the training device of the playback parameter recommendation model provided in this embodiment, first encodes the training sample to obtain the sample characteristics, then detects the associated characteristics of the sample characteristics to obtain the sample associated characteristics, and then recommends the playback parameters according to the sample associated characteristics to obtain the recommended playback parameters, and finally screens the recommended playback parameters, and trains the training model according to the target playback parameters obtained by screening and the corresponding playback parameter labels. After the training is completed, the playback parameter recommendation model is obtained. The playback parameter recommendation model can be used to recommend appropriate target playback parameters to users according to the user's network environment and device capabilities, so as to avoid the problem of unsmooth video playback, improve the user's viewing experience, and increase the user retention rate and user satisfaction.

[0094] The playback parameter recommendation device provided in an embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.

[0095] Furthermore, an embodiment of the present specification also provides a playback parameter recommendation device, Fig.10 A schematic diagram of a playback parameter recommendation device provided in an embodiment of this specification, such as Fig.10 As shown, the device includes: a memory 1001, a processor 1002, a bus 1003 and a communication interface 1004. The memory 1001, the processor 1002 and the communication interface 1004 communicate via the bus 1003, and the communication interface 1004 may include an input and output interface, which includes but is not limited to a keyboard, a mouse, a display, a microphone, a loudspeaker, etc.

[0096] Fig.10 In the embodiment, the memory 1001 stores computer executable instructions that can be run on the processor 1002, and when the computer executable instructions are executed by the processor 1002, the following process is implemented: Obtain video playback information, video playback parameters, and viewing user information of the video played on the playback end; Inputting the video playback information, the video playback parameters, and the viewing user information into a playback parameter recommendation model to perform playback parameter recommendation, thereby obtaining first recommended playback parameters; Determine a second recommended playback parameter based on the preset playback parameter and the video playback information, and calculate and obtain a target playback parameter according to the first recommended playback parameter and the second recommended playback parameter; The target playback parameters are returned to the playback end.

[0097] The playback parameter recommendation device provided in this embodiment, through the cooperation of the memory 1001, the processor 1002, the bus 1003 and the communication interface 1004, first encodes the training sample to obtain the sample characteristics, then detects the associated characteristics of the sample characteristics to obtain the sample associated characteristics, and then recommends the playback parameters according to the sample associated characteristics to obtain the recommended playback parameters, and finally screens the recommended playback parameters, and trains the to-be-trained model according to the target playback parameters obtained by screening and the corresponding playback parameter labels, and obtains the playback parameter recommendation model after the training is completed. The playback parameter recommendation model can be used to recommend appropriate target playback parameters to the user according to the user's network environment and device capabilities, so as to avoid the problem of unsmooth video playback, improve the user's viewing experience, and increase the user retention rate and user satisfaction.

[0098] The playback parameter recommendation device provided in an embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.

[0099] Furthermore, an embodiment of the present specification also provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store computer-executable instructions, and when the computer-executable instructions are executed by a processor, the following process is implemented: Encoding the training samples to obtain sample features; the training samples include expanded samples obtained by expanding the initial samples using a sample expansion algorithm; Performing correlation feature detection on the sample features to obtain sample correlation features; Recommending playback parameters according to the sample-related features to obtain recommended playback parameters; The recommended playback parameters are screened, and the to-be-trained model is trained according to the target playback parameters obtained through the screening and the corresponding playback parameter labels, and a playback parameter recommendation model is obtained after the training is completed.

[0100] The computer-readable storage medium provided in this embodiment first encodes the training samples to obtain sample features, then performs associated feature detection on the sample features to obtain sample associated features, then recommends playback parameters based on the sample associated features to obtain recommended playback parameters, and finally screens the recommended playback parameters, and trains the to-be-trained model based on the target playback parameters obtained by screening and the corresponding playback parameter labels. After the training is completed, a playback parameter recommendation model is obtained. By using the playback parameter recommendation model, appropriate target playback parameters can be recommended to users based on their network environment and device capabilities, so as to avoid the problem of unsmooth video playback, improve user viewing experience, and increase user retention rate and user satisfaction.

[0101] The computer-readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0102] The computer-readable storage medium provided in an embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.

[0103] Furthermore, another embodiment of the present specification provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store computer-executable instructions, and when the computer-executable instructions are executed by a processor, the following process is implemented: Obtain video playback information, video playback parameters, and viewing user information of the video played on the playback end; Inputting the video playback information, the video playback parameters, and the viewing user information into a playback parameter recommendation model to perform playback parameter recommendation, thereby obtaining first recommended playback parameters; Determine a second recommended playback parameter based on the preset playback parameter and the video playback information, and calculate and obtain a target playback parameter according to the first recommended playback parameter and the second recommended playback parameter; The target playback parameters are returned to the playback end.

[0104] The computer-readable storage medium provided in this embodiment first encodes the training samples to obtain sample features, then performs associated feature detection on the sample features to obtain sample associated features, then recommends playback parameters based on the sample associated features to obtain recommended playback parameters, and finally screens the recommended playback parameters, and trains the to-be-trained model based on the target playback parameters obtained by screening and the corresponding playback parameter labels. After the training is completed, a playback parameter recommendation model is obtained. By using the playback parameter recommendation model, appropriate target playback parameters can be recommended to users based on their network environment and device capabilities, so as to avoid the problem of unsmooth video playback, improve user viewing experience, and increase user retention rate and user satisfaction.

[0105] The computer-readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0106] The computer-readable storage medium provided in an embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.

[0107] Furthermore, an embodiment of the present specification also provides a computer program product, which implements the following process when executed by a processor: Encoding the training samples to obtain sample features; the training samples include expanded samples obtained by expanding the initial samples using a sample expansion algorithm; Performing correlation feature detection on the sample features to obtain sample correlation features; Recommending playback parameters according to the sample-related features to obtain recommended playback parameters; The recommended playback parameters are screened, and the to-be-trained model is trained according to the target playback parameters obtained through the screening and the corresponding playback parameter labels, and a playback parameter recommendation model is obtained after the training is completed.

[0108] The computer program product provided in this embodiment first encodes the training samples to obtain sample features, then detects associated features of the sample features to obtain sample associated features, then recommends playback parameters based on the sample associated features to obtain recommended playback parameters, and finally screens the recommended playback parameters, and trains the training model based on the target playback parameters obtained by screening and the corresponding playback parameter labels. After the training is completed, a playback parameter recommendation model is obtained. The playback parameter recommendation model can be used to recommend appropriate target playback parameters to users based on the user's network environment and device capabilities to avoid the problem of unsmooth video playback, improve the user's viewing experience, and increase user retention and user satisfaction. The computer program product provided in one embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.

[0109] Furthermore, another embodiment of the present specification also provides a computer program product, which implements the following process when executed by a processor: Obtain video playback information, video playback parameters, and viewing user information of the video played on the playback end; Inputting the video playback information, the video playback parameters, and the viewing user information into a playback parameter recommendation model to perform playback parameter recommendation, thereby obtaining first recommended playback parameters; Determine a second recommended playback parameter based on the preset playback parameter and the video playback information, and calculate and obtain a target playback parameter according to the first recommended playback parameter and the second recommended playback parameter; The target playback parameters are returned to the playback end.

[0110] The computer program product provided in this embodiment first encodes the training samples to obtain sample features, then detects associated features of the sample features to obtain sample associated features, then recommends playback parameters based on the sample associated features to obtain recommended playback parameters, and finally screens the recommended playback parameters, and trains the training model based on the target playback parameters obtained by screening and the corresponding playback parameter labels. After the training is completed, a playback parameter recommendation model is obtained. The playback parameter recommendation model can be used to recommend appropriate target playback parameters to users based on the user's network environment and device capabilities to avoid the problem of unsmooth video playback, improve the user's viewing experience, and increase user retention and user satisfaction. The computer program product provided in one embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.

[0111] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0115] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0116] Memory may include non-permanent storage in a computer-readable storage medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable storage medium.

[0117] Computer-readable storage media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable storage media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0118] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0119] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0120] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for training a playback parameter recommendation model, characterized in that: The method comprises: Encoding the training samples to obtain sample features; the training samples include expanded samples obtained by expanding the initial samples using a sample expansion algorithm; Performing correlation feature detection on the sample features to obtain sample correlation features; Recommending playback parameters according to the sample-related features to obtain recommended playback parameters; The recommended playback parameters are screened, and the to-be-trained model is trained according to the target playback parameters obtained through the screening and the corresponding playback parameter labels, and a playback parameter recommendation model is obtained after the training is completed.

2. The method for training a playback parameter recommendation model according to claim 1, characterized in that: The encoding process of the training samples to obtain sample features includes: Pre-encoding the playback parameters contained in the training sample to obtain playback parameter features; Constructing a feature matrix based on the playback parameter feature, and calculating a position vector and an attention weight vector of the playback parameter feature; The feature matrix, the position vector, the attention weight vector, and the playback parameter feature are concatenated to obtain the sample feature.

3. The method for training a playback parameter recommendation model according to claim 2, characterized in that: The constructing a feature matrix based on the playback parameter features includes: Acquire associated parameter features that are associated with the playback parameter features, and group the associated parameter features to obtain feature groups; Calculating the similarity between the playback parameter feature and the feature grouping, sorting the feature grouping according to the similarity, and obtaining a feature grouping sequence; The feature groupings in a preset position in the feature grouping sequence are concatenated with the playback parameter features to obtain the feature matrix.

4. The method for training a playback parameter recommendation model according to claim 1, characterized in that: The performing correlation feature detection on the sample features to obtain sample correlation features includes: Multiplying the sample feature with a first correlation detection matrix to obtain a first sample correlation feature, and multiplying the sample feature with a second correlation detection matrix to obtain a second sample correlation feature; The sample feature, the first sample association feature, and the second sample association feature are concatenated to obtain the sample association feature.

5. The method for training a playback parameter recommendation model according to claim 1, characterized in that: The screening of the recommended playback parameters includes: Abnormal recommended playback parameters that are inconsistent with preset playback parameters are eliminated from the recommended playback parameters to obtain the target playback parameters.

6. The method for training a playback parameter recommendation model according to claim 1, characterized in that: The training of the to-be-trained model according to the target playback parameters obtained through screening and the corresponding playback parameter labels includes: Calculating the loss according to the target playback parameter and the playback parameter label; The parameters of the model to be trained are adjusted based on the calculated loss to obtain the playback parameter recommendation model.

7. A method for recommending playback parameters, characterized in that: The method comprises: Obtain video playback information, video playback parameters, and viewing user information of the video played on the playback end; Inputting the video playback information, the video playback parameters, and the viewing user information into a playback parameter recommendation model to perform playback parameter recommendation, thereby obtaining first recommended playback parameters; Determine a second recommended playback parameter based on the preset playback parameter and the video playback information, and calculate and obtain a target playback parameter according to the first recommended playback parameter and the second recommended playback parameter; The target playback parameters are returned to the playback end.

8. The method for training a playback parameter recommendation model according to claim 7, characterized in that: The step of obtaining video playback information, video playback parameters, and viewing user information of the video played on the playback end includes: Determine a response event according to the operation instruction sent by the playback end; the response event includes: a playback event of the playback video, a switching event of the playback state of the playback video, and an adjustment event of the playback parameters of the playback video; Acquire the target tracking data corresponding to the response event; the tracking data includes the video playback information, the video playback parameters and the viewing user information.

9. The method for recommending playback parameters according to claim 7, characterized in that: The determining the second recommended playback parameter based on the preset playback parameter and the video playback information includes: Calculating a preset bandwidth corresponding to the preset playback parameter, and filtering the preset bandwidth based on the network bandwidth included in the video playback information to obtain a target bandwidth; The target bandwidths are sorted to obtain a bandwidth sequence, and the preset playback parameters corresponding to the target bandwidths in a preset position in the bandwidth sequence are used as the second recommended playback parameters.

10. The method for recommending playback parameters according to claim 7, characterized in that: The calculating and obtaining the target playback parameter according to the first recommended playback parameter and the second recommended playback parameter includes: Determining a parameter weight of the first playback recommendation parameter, and calculating a parameter recommendation success rate of the first playback recommendation parameter; The target playback parameter is calculated based on the parameter weight, the parameter recommendation success rate, the first playback recommendation parameter, and the second playback recommendation parameter.

11. The method for recommending playback parameters according to claim 7, characterized in that: The step of returning the target playback parameter to the playback end includes: A parameter recommendation message including the target playback parameter is generated, and the parameter recommendation message is returned to the playback end.

Images