Disclosure of Invention
The disclosure provides a data recommendation method, a data recommendation device, a server and a storage medium, and accuracy of data recommendation is improved.
According to an aspect of the embodiments of the present disclosure, there is provided a data recommendation method, the method including:
Acquiring target data characteristics corresponding to target multimedia data to be recommended;
Invoking a feature conversion model, and performing feature conversion processing on the target data features to obtain converted predicted operation features, wherein the predicted operation features are used for representing predicted operations performed on the target multimedia data;
Determining the matching degree of each user and the target multimedia data according to the user characteristics corresponding to the users and the prediction operation characteristics;
And selecting at least one target user from the plurality of users according to the matching degree corresponding to the plurality of users, and recommending the target multimedia data to the at least one target user.
In some embodiments, the method further comprises:
invoking the feature conversion model, and performing feature conversion processing on first sample data features corresponding to the first sample multimedia data to obtain converted first sample operation features;
Invoking a feature discrimination model to discriminate the first sample operation feature to obtain a discrimination mark of the first sample operation feature, wherein the discrimination mark is used for indicating whether the first sample operation feature belongs to a data feature or an operation feature;
And training the feature conversion model according to the first sample operation feature and the distinguishing mark.
In some embodiments, the training the feature transformation model according to the first sample operation feature and the discrimination indicator includes:
Determining a first loss value of the feature conversion model according to the first sample operation feature and the distinguishing mark;
And training the feature conversion model according to the first loss value.
In some embodiments, after the training of the feature transformation model according to the first sample operation feature and the discrimination indicator, the method further includes:
Invoking the feature conversion model, and performing feature conversion processing on second sample data features corresponding to the second sample multimedia data to obtain converted second sample operation features;
Invoking the feature discrimination model to respectively discriminate the second sample data feature and the second sample operation feature, and determining discrimination identifications of the second sample data feature and the second sample operation feature, wherein the discrimination identifications are used for indicating whether the corresponding features belong to data features or operation features;
and training the characteristic discrimination model according to the second sample data characteristic, the second sample operation characteristic, the second sample data characteristic and the discrimination identification of the second sample operation characteristic.
In some embodiments, the training the feature discrimination model according to the second sample data feature, the second sample operational feature, the second sample data feature, and the discrimination identification of the second sample operational feature comprises:
determining a second loss value of the feature discrimination model according to the second sample data feature, the second sample operation feature, the second sample data feature and the discrimination identification of the second sample operation feature;
and training the characteristic distinguishing model according to the second loss value.
In some embodiments, after the training of the feature discrimination model according to the second sample data feature, the second sample operational feature, the second sample data feature, and the discrimination identification of the second sample operational feature, the method further comprises:
invoking the feature conversion model, and performing feature conversion processing on the third sample data features corresponding to the third sample multimedia data to obtain converted third sample operation features;
invoking the characteristic distinguishing model, distinguishing the third sample operation characteristic, and determining distinguishing identification of the third sample operation characteristic, wherein the distinguishing identification is used for indicating whether the third sample operation characteristic belongs to a data characteristic or an operation characteristic;
And training the feature conversion model according to the third sample operation feature and the distinguishing mark.
In some embodiments, the obtaining the target data feature corresponding to the target multimedia data to be recommended includes:
and calling a feature extraction model to perform feature extraction on the target multimedia data to obtain target data features corresponding to the target multimedia data.
In some embodiments, the feature extraction model includes a text feature extraction sub-model and an image feature extraction sub-model, and the invoking the feature extraction model to perform feature extraction on the target multimedia data to obtain a target data feature corresponding to the target multimedia data includes:
invoking the text feature extraction sub-model to perform feature extraction on the target multimedia data to obtain text features of the target multimedia data;
invoking the image feature extraction sub-model to perform feature extraction on the target multimedia data to obtain image features of the target multimedia data;
And performing splicing processing on the text features and the image features to obtain target data features of the target multimedia data.
In some embodiments, the determining the matching degree between each user and the target multimedia data according to the user characteristics corresponding to the users and the predicted operation characteristics includes:
and determining the matching degree of each user and the target multimedia data according to the user characteristics, the target data characteristics and the prediction operation characteristics corresponding to the plurality of users.
In some embodiments, the determining the matching degree of each user and the target multimedia data according to the user features, the target data features and the predicted operation features corresponding to the plurality of users includes:
And for the user characteristics of any user, carrying out matching processing on the user characteristics, the target data characteristics and the predicted operation characteristics to obtain the matching degree of the user and the target multimedia data.
In some embodiments, the selecting at least one target user from the plurality of users according to the matching degrees corresponding to the plurality of users, and recommending the target multimedia data to the at least one target user includes:
Determining recommendation parameters corresponding to the plurality of users according to the matching degree corresponding to the plurality of users and the resource quantity corresponding to the target multimedia data;
And selecting at least one target user from the plurality of users according to the recommendation parameters corresponding to the plurality of users, and recommending the target multimedia data to the at least one target user.
According to still another aspect of the embodiments of the present disclosure, there is provided a data recommendation apparatus, the apparatus including:
The characteristic acquisition unit is configured to acquire target data characteristics corresponding to target multimedia data to be recommended;
a transformation processing unit configured to perform feature transformation processing on the target data feature to obtain a transformed predicted operation feature, where the predicted operation feature is used to represent a predicted operation performed on the target multimedia data;
a determining unit configured to determine a degree of matching between each user and the target multimedia data according to user characteristics corresponding to the plurality of users and the predicted operation characteristics;
and the recommending unit is configured to select at least one target user from the plurality of users according to the matching degree corresponding to the plurality of users and recommend the target multimedia data to the at least one target user.
In some embodiments, the apparatus further comprises:
The transformation processing unit is configured to call the feature transformation model, perform feature transformation processing on first sample data features corresponding to the first sample multimedia data, and obtain transformed first sample operation features;
the distinguishing processing unit is configured to call a feature distinguishing model, distinguish the first sample operation feature to obtain distinguishing identification of the first sample operation feature, wherein the distinguishing identification is used for indicating whether the first sample operation feature belongs to a data feature or an operation feature;
The first training unit is configured to train the feature conversion model according to the first sample operation feature and the distinguishing mark.
In some embodiments, the first training unit comprises:
a first determining subunit configured to determine a first loss value of the feature transformation model according to the first sample operation feature and the discrimination indicator;
A first training subunit configured to train the feature transformation model according to the first loss value.
In some embodiments, the apparatus further comprises:
The transformation processing unit is configured to call the feature transformation model, perform feature transformation processing on the second sample data features corresponding to the second sample multimedia data, and obtain transformed second sample operation features;
The distinguishing processing unit is further configured to call the feature distinguishing model, distinguish the second sample data feature and the second sample operation feature, and determine distinguishing marks of the second sample data feature and the second sample operation feature, wherein the distinguishing marks are used for indicating whether the corresponding feature belongs to the data feature or the operation feature;
And the second training unit is configured to train the feature discrimination model according to the second sample data feature, the second sample operation feature, the second sample data feature and the discrimination identification of the second sample operation feature.
In some embodiments, the second training unit comprises:
A second determining subunit configured to determine a second loss value of the feature discrimination model according to the second sample data feature, the second sample operational feature, the second sample data feature, and a discrimination identity of the second sample operational feature;
And a second training subunit configured to train the feature discrimination model according to the second loss value.
In some embodiments, the apparatus further comprises:
The transformation processing unit is configured to call the feature transformation model, perform feature transformation processing on the third sample data features corresponding to the third sample multimedia data, and obtain transformed third sample operation features;
The distinguishing processing unit is further configured to call the feature distinguishing model, distinguish the third sample operation feature, and determine distinguishing identification of the third sample operation feature, where the distinguishing identification is used to indicate whether the third sample operation feature belongs to a data feature or an operation feature;
And the third training unit is configured to train the feature conversion model according to the third sample operation feature and the distinguishing mark.
In some embodiments, the feature acquisition unit includes:
And the feature extraction subunit is configured to call a feature extraction model to perform feature extraction on the target multimedia data so as to obtain target data features corresponding to the target multimedia data.
In some embodiments, the feature extraction model includes a text feature extraction sub-model and an image feature extraction sub-model, and the feature extraction sub-unit is configured to call the text feature extraction sub-model, perform feature extraction on the target multimedia data to obtain text features of the target multimedia data, call the image feature extraction sub-model, perform feature extraction on the target multimedia data to obtain image features of the target multimedia data, and perform stitching processing on the text features and the image features to obtain target data features of the target multimedia data.
In some embodiments, the determining unit includes:
And the third determining subunit is configured to determine the matching degree of each user and the target multimedia data according to the user characteristics, the target data characteristics and the prediction operation characteristics corresponding to the plurality of users.
In some embodiments, the third determining subunit is configured to perform matching processing on the user feature, the target data feature and the predicted operation feature of any user, so as to obtain a matching degree between the user and the target multimedia data.
In some embodiments, the recommending unit is configured to determine recommending parameters corresponding to the plurality of users according to the matching degree corresponding to the plurality of users and the number of resources corresponding to the target multimedia data, select at least one target user from the plurality of users according to the recommending parameters corresponding to the plurality of users, and recommend the target multimedia data to the at least one target user.
According to still another aspect of the embodiments of the present disclosure, there is provided a server including:
One or more processors;
Volatile or non-volatile memory for storing the one or more processor-executable instructions;
wherein the one or more processors are configured to perform the data recommendation method of the first aspect.
According to still another aspect of the embodiments of the present disclosure, there is provided a non-transitory computer-readable storage medium, which when executed by a processor of a server, enables the server to perform the data recommendation method of the above aspect.
According to yet another aspect of embodiments of the present disclosure, there is provided a computer program product, which when executed by a processor of a server, enables the server to perform the data recommendation method of the above aspect.
According to the data recommendation method, device, server and storage medium provided by the embodiment of the disclosure, the operation possibly performed on the multimedia data by a user is represented by the prediction operation characteristics of the multimedia data, so that the user possibly performing the operation on the multimedia data is determined to be the user matched with the multimedia data according to the prediction operation characteristics, the multimedia data is subsequently recommended to the matched user, the accurate recommendation of the multimedia data is realized, the accuracy of data recommendation is ensured no matter whether the operation performed on the multimedia data can be acquired or not, and the multimedia data can be recommended even if the operation performed on the multimedia data is not acquired, namely the cold start of the multimedia data is realized.
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.
Detailed Description
In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions of 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 of the present disclosure and the claims and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.
The terms "at least one," "a plurality," "each," "any" as used herein, at least one includes one, two or more, a plurality includes two or more, and each refers to each of a corresponding plurality, any of which refers to any of the plurality. For example, the plurality of users includes 3 users, and each refers to each of the 3 users, and any one of the 3 users may be a first user, a second user, or a third user.
The user information (including but not limited to user equipment information, user personal information, etc.) related to the present disclosure is information authorized by the user or sufficiently authorized by each party.
Fig. 1 is a schematic diagram of an implementation environment provided according to an exemplary embodiment, where the implementation environment includes a terminal 101 and a server 102, where the terminal 101 is connected to the server 102 through a network, and may interact with the server 102 through the network connection.
In some embodiments, the terminal is a mobile phone, a tablet computer, or a computer. In some embodiments, the server is a server, or a server cluster comprising a plurality of servers, or a cloud computing service center.
The server 102 has a data processing function and a data recommending function, and is capable of processing multimedia data and recommending the multimedia data to a user. The terminal 101 has a multimedia data display function, and can display multimedia data recommended by the server 102 for viewing by a user.
In some embodiments, terminal 101 installs an application served by server 102, e.g., a shopping application, a video application, etc. The terminal 101 is capable of running an application and presenting multimedia data in the application to a user. Server 102 is capable of recommending multimedia data to a user for presentation in an application for viewing by the user.
The method provided by the embodiment of the disclosure can be applied to various scenes.
For example, in an advertisement recommendation scenario.
When a new advertisement is put, because the new advertisement is not put or the putting time is short, the operation of the user on the advertisement cannot be obtained, for example, the checking times of the user on the multimedia data cannot be obtained, or the click rate of the user on the multimedia data cannot be obtained, etc., by adopting the method provided by the embodiment of the disclosure, the advertisement is recommended to the matched user through the obtained prediction operation characteristics, and the accuracy of advertisement recommendation is improved, so that the cold start of the advertisement is realized.
Fig. 2 is a flowchart illustrating a data recommendation method, see fig. 2, applied to a server, according to an exemplary embodiment, including the steps of:
in step 201, the server acquires a target data feature corresponding to target multimedia data to be recommended.
In step 202, the server invokes the feature transformation model to perform feature transformation processing on the target data feature to obtain a transformed predicted operation feature, where the predicted operation feature is used to represent a predicted operation performed on the target multimedia data.
In step 203, the server determines the matching degree between each user and the target multimedia data according to the user characteristics and the predicted operation characteristics corresponding to the plurality of users.
In step 204, the server selects at least one target user from the plurality of users according to the matching degrees corresponding to the plurality of users, and recommends the target multimedia data to the at least one target user.
According to the method provided by the embodiment of the disclosure, the operation which a user may execute on the multimedia data is represented by the prediction operation characteristics of the multimedia data, so that the user which may execute the operation on the multimedia data is determined according to the prediction operation characteristics, namely, the user matched with the multimedia data, and the multimedia data is subsequently recommended to the matched user, so that the accurate recommendation of the multimedia data is realized, the accuracy of data recommendation is ensured no matter whether the operation executed on the multimedia data can be acquired or not, and the multimedia data can be recommended even if the operation executed on the multimedia data is not acquired, namely, the cold start of the multimedia data is realized.
In some embodiments, the method further comprises:
Calling a feature conversion model, and performing feature conversion processing on first sample data features corresponding to the first sample multimedia data to obtain converted first sample operation features;
invoking a feature discrimination model to discriminate the first sample operation feature to obtain a discrimination mark of the first sample operation feature, wherein the discrimination mark is used for indicating whether the first sample operation feature belongs to the data feature or the operation feature;
and training the feature conversion model according to the first sample operation feature and the discrimination mark.
In some embodiments, training the feature transformation model based on the first sample operational feature and the discrimination indicator includes:
determining a first loss value of the feature conversion model according to the first sample operation feature and the discrimination mark;
And training the feature conversion model according to the first loss value.
In some embodiments, after training the feature transformation model according to the first sample operation feature and the discrimination indicator, the method further includes:
Invoking a feature conversion model, and performing feature conversion processing on second sample data features corresponding to the second sample multimedia data to obtain converted second sample operation features;
Invoking a feature discrimination model to respectively discriminate the second sample data feature and the second sample operation feature, and determining discrimination marks of the second sample data feature and the second sample operation feature, wherein the discrimination marks are used for indicating whether the corresponding feature belongs to the data feature or the operation feature;
And training the feature discrimination model according to the second sample data feature, the second sample operation feature, the second sample data feature and the discrimination identification of the second sample operation feature.
In some embodiments, training the feature discrimination model based on the second sample data feature, the second sample operational feature, the second sample data feature, and the discrimination identification of the second sample operational feature comprises:
determining a second loss value of the feature discrimination model according to the second sample data feature, the second sample operation feature, the second sample data feature and the discrimination identification of the second sample operation feature;
and training the feature discrimination model according to the second loss value.
In some embodiments, after training the feature discrimination model based on the second sample data feature, the second sample operational feature, the second sample data feature, and the discrimination identification of the second sample operational feature, the method further comprises:
Invoking a feature conversion model, and performing feature conversion processing on third sample data features corresponding to third sample multimedia data to obtain converted third sample operation features;
invoking a feature discrimination model to discriminate the third sample operation feature, determining a discrimination mark of the third sample operation feature, wherein the discrimination mark is used for indicating whether the third sample operation feature belongs to the data feature or the operation feature;
And training the feature conversion model according to the third sample operation feature and the discrimination mark.
In some embodiments, obtaining the target data feature corresponding to the target multimedia data to be recommended includes:
And calling a feature extraction model to perform feature extraction on the target multimedia data to obtain target data features corresponding to the target multimedia data.
In some embodiments, the feature extraction model includes a text feature extraction sub-model and an image feature extraction sub-model, and the feature extraction model is called to perform feature extraction on the target multimedia data to obtain a target data feature corresponding to the target multimedia data, including:
invoking a text feature extraction sub-model to perform feature extraction on the target multimedia data to obtain text features of the target multimedia data;
invoking an image feature extraction sub-model to perform feature extraction on the target multimedia data to obtain image features of the target multimedia data;
And performing splicing processing on the text features and the image features to obtain target data features of the target multimedia data.
In some embodiments, determining the matching degree of each user and the target multimedia data according to the user characteristics corresponding to the plurality of users and the predicted operation characteristics includes:
And determining the matching degree of each user and the target multimedia data according to the user characteristics, the target data characteristics and the predicted operation characteristics corresponding to the plurality of users.
In some embodiments, determining the matching degree of each user and the target multimedia data according to the user characteristics, the target data characteristics and the predicted operation characteristics corresponding to the plurality of users includes:
and processing the user characteristics, the target data characteristics and the predicted operation characteristics of any user to obtain the matching degree of the user and the target multimedia data.
In some embodiments, selecting at least one target user from the plurality of users according to the matching degree corresponding to the plurality of users, recommending the target multimedia data to the at least one target user, including:
Determining recommendation parameters corresponding to a plurality of users according to the matching degree corresponding to the plurality of users and the resource quantity corresponding to the target multimedia data;
And selecting at least one target user from the plurality of users according to the recommendation parameters corresponding to the plurality of users, and recommending the target multimedia data to the at least one target user.
Fig. 3 is a flowchart illustrating a data recommendation method, see fig. 3, applied to a server, according to an exemplary embodiment, including the steps of:
in step 301, the server invokes the feature extraction model to perform feature extraction on the target multimedia data, so as to obtain a target data feature corresponding to the target multimedia data.
In the embodiment of the present disclosure, the target multimedia data is multimedia data to be recommended, where the delivery time is short or multimedia data that has not been delivered yet. In some embodiments, the targeted multimedia data is advertisements, videos, novels, and the like. The target multimedia data can not be accurately recommended according to the operation performed by the user on the target multimedia data because the target multimedia data is put in a short time, the operation performed by the user on the target multimedia data is obtained, or the operation performed by the user on the target multimedia data, such as the number of times the user views the multimedia data or the click rate of the user on the multimedia data, can not be obtained because the target multimedia data is not put in yet. Therefore, the operation of the user on the target multimedia data is predicted to obtain the operation characteristics corresponding to the target multimedia data, so that the target multimedia data can be accurately recommended according to the operation characteristics, and the cold start of the target multimedia data is realized.
Wherein the feature extraction model is used for extracting data features of the multimedia data, and representing the corresponding multimedia data by the data features, and in some embodiments, the target data features comprise data feature vectors or data feature matrices. And extracting target data characteristics corresponding to the target multimedia data through the characteristic extraction model so as to ensure the accuracy of the target data characteristics, and acquiring the prediction operation characteristics of the target multimedia data through the target data characteristics.
Because the server stores a plurality of multimedia data, the process of selecting the target multimedia from the stored plurality of multimedia data comprises the following two modes:
in the first mode, a plurality of multimedia data and the throwing time length of each multimedia data are stored in a database, and then the multimedia data with the throwing time length smaller than the reference time length are used as target multimedia data according to the throwing time length of the plurality of multimedia data.
The delivery duration represents the duration of delivering the multimedia data. The reference time period is any time period, such as 10 hours, 2 days, etc. After any multimedia data is released, the user can view the multimedia data.
When the multimedia data is just released, the operation performed on the multimedia data by the user in the process of releasing the multimedia data is limited, and the operation performed on the multimedia data by the user cannot be accurately acquired, so that the accurate recommendation of the multimedia data cannot be realized according to the acquired operation, and therefore the multimedia data with the release duration smaller than the reference duration is used as target multimedia data.
In the second mode, a plurality of multimedia data and the operation times corresponding to each multimedia data are stored in a database, and then the multimedia data with the operation times smaller than the reference times are used as target multimedia data according to the operation times corresponding to the plurality of multimedia data.
After any multimedia data is put out, any user performs an operation on the multimedia data, and the operation times corresponding to the multimedia data are increased by 1. In some embodiments, the number of operations is a number of clicks.
Because the accurate recommendation of the multimedia data cannot be realized according to the operation performed by the user on the multimedia data when the operation times performed by the user on the multimedia data are low, the multimedia data with the throwing duration less than the reference duration are used as target multimedia data.
In some embodiments, the feature extraction model includes a text feature extraction sub-model and an image feature extraction sub-model, and the step 301 includes the steps 3011-3013 of:
3011. And calling a text feature extraction sub-model, and carrying out feature extraction on the target multimedia data to obtain the text features of the target multimedia data.
In the embodiment of the disclosure, the target multimedia data includes text and images, for example, the multimedia data is video, and the video includes text and images. And respectively extracting the characteristics of the target multimedia data from the text dimension and the image dimension, and performing splicing processing on the obtained text characteristics and the obtained image characteristics to enrich the data contained in the target data characteristics corresponding to the target multimedia data, thereby improving the accuracy of the target data characteristics. The process of obtaining the target data features through the text feature extraction sub-model and the image feature extraction sub-model is shown in fig. 4.
The text feature extraction sub-model is used for extracting text features in the multimedia data. In some embodiments, the text feature extraction sub-model is BERT (Bidirectional Encoder Representations from Transformers, bi-directional encoder representation), or other model, or the like. The text feature is used to represent text contained in the target multimedia data, and in some embodiments, the text feature comprises a text feature vector or text feature matrix.
Since the text feature extraction sub-model is a model for extracting text features in the multimedia data, the text features in the target multimedia data are extracted by the text feature extraction sub-model, and the accuracy of the text features is ensured.
3012. And calling the image feature extraction sub-model to perform feature extraction on the target multimedia data to obtain the image features of the target multimedia data.
The image feature extraction sub-model is used for extracting image features in the target multimedia data. In some embodiments, the image feature extraction sub-model is VGG-16 (Visual Geometry Group Network ), or other model, or the like. The image features are used to represent images contained in the target multimedia data, and in some embodiments, the image features comprise image feature vectors or image feature matrices.
Because the image feature extraction sub-model is a model for extracting image features in the multimedia data, the image features in the target multimedia data are extracted by using the image feature extraction sub-model, and the accuracy of the image features is ensured.
3013. And performing splicing processing on the text features and the image features to obtain target data features of the target multimedia data.
Since the text feature and the image feature describe the target multimedia data from different dimensions, the text feature and the image feature are spliced together to serve as the target data feature of the target multimedia data, so that the data contained in the target data feature is enriched, and the accuracy of the target data feature is improved.
In some embodiments, the text feature is a text feature vector and the image feature is an image feature vector, and step 3013 includes stitching the text feature vector with the image feature vector to obtain a target data feature vector of the target multimedia data.
In some embodiments, the sum of the number of feature dimensions included in the text feature vector and the number of feature dimensions included in the image feature vector is equal to the number of feature dimensions included in the target data feature vector. For example, a text feature vector includes 4 feature dimensions, the text feature vector is (1, 1), an image feature vector includes 3 feature dimensions, the image feature vector is (2, 2), then the resulting target data feature vector includes 7 feature dimensions, the target data feature vector is (1,1,1,1,2,2,2), or the target data feature vector is (2,2,2,1,1,1,1).
It should be noted that, in the embodiment of the present disclosure, the target data features corresponding to the target multimedia data are obtained through the feature extraction model, and in another embodiment, the server can obtain the target data features corresponding to the target multimedia data to be recommended in other manners.
In step 302, the server invokes the feature transformation model to perform feature transformation processing on the target data features to obtain transformed predicted operational features.
The feature conversion model is used for acquiring predicted operation features of the multimedia data, wherein the predicted operation features are used for representing predicted operations performed on the target multimedia data. For example, the predicted operational characteristic may represent a number of views of the multimedia data by the user, or the operational characteristic may represent a click rate of the multimedia data by the user. In some embodiments, the form of the predicted operational characteristics includes, but is not limited to, an operational characteristic vector or an operational characteristic matrix.
And transforming the target data characteristics used for representing the target multimedia data through the characteristic transformation model to obtain transformed prediction characteristics so as to represent the predicted operation of the user on the target multimedia data, so that the subsequent user matched with the target multimedia data can be determined according to the predicted operation of the user on the target multimedia data, the accurate recommendation of the multimedia data is ensured, and the cold start of the multimedia data is realized.
In some embodiments, the target data feature is a data feature vector, the predicted operational feature is an operational feature vector, and the number of feature dimensions of the data feature vector is equal to the number of feature dimensions of the operational feature vector.
In some embodiments, the target data feature is a data feature vector, the feature transformation model includes at least two hidden layers, and the step 302 includes performing feature transformation processing on the data feature based on a first hidden layer of the at least two hidden layers, inputting the feature output by the first hidden layer to a next hidden layer, performing feature transformation processing on the feature output by the previous hidden layer based on the next hidden layer, and taking the feature output by a most significant hidden layer of the at least two hidden layers as the prediction operation feature.
And carrying out feature transformation on the target data features through a plurality of hidden layers in the feature transformation model so as to predict the operation executed by a user on the target multimedia data, and obtaining the predicted operation features corresponding to the target data features, namely obtaining the predicted operation features corresponding to the target multimedia data. As shown in fig. 5, the predicted operation features corresponding to the target data features are obtained through two hidden layers in the feature conversion model.
In some embodiments, each hidden layer corresponds to a feature transformation matrix, and the feature transformation matrices corresponding to different hidden layers are different. And carrying out feature transformation on the target data features through feature transformation matrixes corresponding to the plurality of hidden layers so as to obtain the prediction operation features of the target multimedia data.
It should be noted that, in the embodiment of the present disclosure, the predicted operation feature corresponding to the target multimedia data is obtained through the feature conversion model is described as an example, and in another embodiment, the step 302 is not required to be executed, and other manners can be adopted to perform feature conversion processing on the target data feature to obtain the converted predicted operation feature.
In step 303, the server determines a matching degree between each user and the target multimedia data according to the user characteristics, the target data characteristics, and the predicted operation characteristics corresponding to the plurality of users.
The matching degree is used for indicating the preference degree of the user to the multimedia data, and the higher the matching degree is, the higher the preference degree of the user to the multimedia data is, the lower the matching degree is, and the preference degree of the user to the multimedia data is low. User features are used to describe a user, and in some embodiments, include user feature vectors or user feature matrices. In some embodiments, the user characteristics of the user are obtained by extracting characteristics of the user information. Wherein the user information includes gender, age, residence, occupation, preference, etc.
Through the user characteristics, the target data characteristics and the prediction operation characteristics, the preference degree of each user on the target multimedia data can be determined, namely, the matching degree of each user and the target multimedia data is determined, so that the target multimedia data is recommended to the matched user according to the matching degree of each user and the target multimedia data.
In some embodiments, the step 303 includes, for the user characteristics of any user, performing matching processing on the user characteristics, the target data characteristics and the predicted operation characteristics, so as to obtain a matching degree between the user and the target multimedia data.
Because the target data features and the prediction operation features describe the target multimedia data from different dimensions, and the user features describe the user, the matching processing is performed through the user features, the target data features and the prediction operation features so as to determine the matching degree among the user features, the target data features and the prediction operation features, and the matching degree between the user and the target multimedia data is obtained.
In some embodiments, a feature matching model is invoked to match the user features, the target data features, and the predicted operational features to obtain a degree of match between the user and the target multimedia data. The feature matching model is used for acquiring the matching degree of the user and the multimedia data. After the user characteristics, the target data characteristics corresponding to the multimedia data and the prediction operation characteristics are obtained, the matching degree of the user and the multimedia data can be obtained through the characteristic matching model. As shown in fig. 6, the user characteristics, the target data characteristics and the predicted operation characteristics are input into the characteristic matching model, the characteristics are matched through a hidden layer in the characteristic matching model, and the matching degree of the user and the target multimedia data is output through an output layer.
In some embodiments, the user feature is a user feature vector, the target data feature is a data feature vector, the predicted operation feature is an operation feature vector, the user feature vector, the data feature vector and the operation feature vector all include feature values with the same dimension, and then the feature values with the same dimension of the user feature vector, the data feature vector and the operation feature vector are counted to obtain a fused feature vector, and the fused feature values with the multiple feature dimensions of the fused feature vector are counted to obtain the matching degree of the user and the target multimedia data.
In some embodiments, when the feature values of the same dimension of the user feature vector, the data feature vector and the operation feature vector are summed, the sum of the feature values of the same dimension is used as the fusion feature value of the corresponding dimension, or the feature values of the same dimension are subjected to weighted average processing, so that the fusion feature value of the corresponding dimension is obtained.
In some embodiments, when the fused feature values of the feature dimensions of the fused feature vector are processed, the fused feature values of the feature dimensions are weighted and averaged to obtain the matching degree, or the sum of the fused feature values of the feature dimensions is used as the matching degree.
In some embodiments, the target multimedia data corresponds to a specified user feature representing that the target multimedia data is to be recommended to a user conforming to the specified user feature, and before step 303, the method includes selecting, as the user to be recommended of the target multimedia data, a reference user whose user feature conforms to the specified user feature according to the user features corresponding to the plurality of reference users and the specified user feature.
For example, the specified user characteristic is that a female user is selected from a plurality of reference users as a user to be recommended for the target multimedia data, or the specified user characteristic is that a user residing in XX city is selected from a plurality of reference users as a user to be recommended for the target multimedia data.
In step 304, the server selects at least one target user from the plurality of users according to the matching degrees corresponding to the plurality of users, and recommends the target multimedia data to the at least one target user.
And recommending the target multimedia data to the target user by selecting the user matched with the target multimedia data from a plurality of users as the target user according to the matching degree of each user and the multimedia data.
In some embodiments, in the matching degrees corresponding to the plurality of users, the matching degree corresponding to the target user is greater than the matching degrees corresponding to other users in the plurality of users.
In some embodiments, the step 304 includes determining recommendation parameters corresponding to the plurality of users according to the number of resources corresponding to the target multimedia data and matching the matching degrees corresponding to the plurality of users, selecting at least one target user from the plurality of users according to the recommendation parameters corresponding to the plurality of users, and recommending the target multimedia data to the at least one target user.
Wherein the number of resources is the number of resources available for recommending the target multimedia data. The recommendation parameter is used to indicate a possibility of recommending to the user, and the higher the recommendation parameter is, the greater the possibility of recommending the target multimedia resource to the corresponding user is, and the lower the recommendation parameter is, the smaller the possibility of recommending the target multimedia resource to the corresponding user is. The target user to be recommended is selected from a plurality of users according to the recommendation parameters, so that a large number of resources can be acquired under the condition of ensuring accurate recommendation of the target multimedia data.
In some embodiments, for any user, the product of the user matching degree and the number of resources corresponding to the target multimedia data is used as a recommendation parameter corresponding to the user.
In some embodiments, recommendation parameters corresponding to a plurality of users are arranged in order from large to small, and at least one target user is selected. The recommendation parameters of the target user are larger than those of other users in the plurality of users.
It should be noted that, in the embodiment of the present disclosure, the target multimedia data is recommended to the target user according to the data feature, the operation feature and the user feature, and in another embodiment, the target multimedia data is recommended directly according to the operation feature without executing step 304.
When recommending the multimedia data, if the multimedia data has been put in for a long time, the operation of the user on the multimedia data, for example, the number of times the user views the multimedia data, or the click rate of the user on the multimedia data, etc., can be obtained, and when recommending the multimedia data to other users, the matched user can be determined according to the obtained operation of the user on the multimedia data, so as to recommend the multimedia data to the determined user. If the multimedia data is multimedia data which is not yet put in, or multimedia data which is put in but is put in for a period of time, at the moment, operation of a user on the multimedia data cannot be obtained, or operation of the user on the multimedia data is obtained to be less, so that the multimedia data cannot be accurately recommended according to the operation of the user on the multimedia data, and further, the operation of the user on the multimedia data cannot be further lacked, so that vicious circle is formed, and the multimedia data cannot be accurately recommended all the time, namely cold start of the multimedia data cannot be realized. Therefore, in order to realize the cold start of the multimedia data, through the method provided by the disclosure, under the condition that the operation performed on the multimedia data cannot be obtained, the operation of the user on the multimedia data is predicted, the matched user is determined according to the predicted operation of the user on the multimedia data, the multimedia data is recommended to the matched user, the accurate recommendation of the multimedia data is realized, the accuracy of the data recommendation is ensured, and the cold start of the multimedia data is realized.
According to the method provided by the embodiment of the disclosure, the operation which a user may execute on the multimedia data is represented by the prediction operation characteristics of the multimedia data, so that the user which may execute the operation on the multimedia data is determined according to the prediction operation characteristics, namely, the user matched with the multimedia data, and the multimedia data is subsequently recommended to the matched user, so that the accurate recommendation of the multimedia data is realized, the accuracy of data recommendation is ensured no matter whether the operation executed on the multimedia data can be acquired or not, and the multimedia data can be recommended even if the operation executed on the multimedia data is not acquired, namely, the cold start of the multimedia data is realized.
And recommending the multimedia data according to the matching degree of the user and the multimedia data, so as to improve the accuracy of data recommendation.
Based on the above embodiments, the embodiments of the present disclosure provide a method for recommending data to a user, as shown in fig. 7, applied to a server, where the method includes:
In step 701, the server selects, from the plurality of multimedia data, multimedia data whose specified user characteristics match the target user characteristics according to the target user characteristics corresponding to the target user and the specified user characteristics corresponding to the plurality of multimedia data, as first reference multimedia data to be recommended.
The step is similar to the scheme of selecting the user to be recommended for the target multimedia data from the plurality of reference users by the server in step 303, and will not be described herein.
In step 702, the server selects a plurality of second reference multimedia data according to the number of resources corresponding to the selected plurality of first reference multimedia data.
The number of resources corresponding to the second reference multimedia data is greater than the number of resources corresponding to other reference multimedia data in the plurality of first reference multimedia data.
In the process of selecting the second reference multimedia resource by the server, a lightweight model can be called, and second reference multimedia resource data with large resource quantity is selected from the resource quantity corresponding to the plurality of first reference multimedia data.
In step 703, the server invokes the feature extraction model to perform feature extraction on each second reference multimedia data, so as to obtain a target data feature corresponding to each second reference multimedia data.
This step is similar to step 301 described above and will not be described again here.
In step 704, the server invokes the feature transformation model to perform feature transformation processing on the target data feature corresponding to each second reference multimedia data, so as to obtain a predicted operation feature corresponding to each second reference multimedia data.
This step is similar to step 302 described above and will not be described again.
In step 705, for any second reference multimedia data, the server determines the matching degree between the target user and the second reference multimedia data according to the target user characteristic, the target data characteristic corresponding to the second reference multimedia data, and the predicted operation characteristic.
This step is similar to step 303 described above and will not be described again here.
In step 706, the server determines recommended parameters of each second reference multimedia resource according to the matching degree and the number of resources corresponding to each second reference multimedia data.
The step is similar to the above-mentioned step 304, and the scheme for obtaining the recommended parameters corresponding to the plurality of users is not described herein.
In step 707, the server selects at least one target multimedia resource from the plurality of second reference multimedia resources to recommend the target user according to the recommendation parameters of the plurality of second reference multimedia resources.
The recommendation parameters corresponding to the target multimedia resources are larger than recommendation parameters of other multimedia resources in the second reference multimedia resources. And selecting the target multimedia resources to be recommended from the plurality of second reference multimedia resources according to the recommendation parameters, so that the matched target multimedia resources are recommended to the user under the condition of ensuring that a plurality of resource quantities are acquired.
In the embodiment of the disclosure, after the multimedia data is accurately recommended to the user, the corresponding resource quantity can be obtained, for example, after the advertisement is accurately recommended to the user, a certain benefit can be obtained. Therefore, when recommending the multimedia data to the target user, under the condition that the operation executed by the multimedia data to be recommended cannot be acquired, in order to realize the cold start of the multimedia data and acquire a large number of resources, the multimedia data with the large number of corresponding resources is selected, and the operation executed by the user on each multimedia resource data is predicted, so that the multimedia resources matched with the target user can be determined, the matched multimedia resources are recommended to the target user, the accuracy of recommending the multimedia data is ensured, the large number of resources can be acquired, and the cold start of the multimedia resources is realized.
According to the method provided by the embodiment of the disclosure, the operation which a user may execute on the multimedia data is represented by the prediction operation characteristics of the multimedia data, so that the user which may execute the operation on the multimedia data is determined according to the prediction operation characteristics, namely, the user matched with the multimedia data, and the multimedia data is subsequently recommended to the matched user, so that the accurate recommendation of the multimedia data is realized, the accuracy of data recommendation is ensured no matter whether the operation executed on the multimedia data can be acquired or not, and the multimedia data can be recommended even if the operation executed on the multimedia data is not acquired, namely, the cold start of the multimedia data is realized.
Based on the steps in the foregoing embodiments, the embodiment of the present disclosure provides a data recommendation process, as shown in fig. 8, including:
1. and orienting a plurality of first reference multimedia data conforming to the target user through the user characteristics of the target user.
2. And determining a plurality of second reference multimedia data with large resource quantity by recalling the plurality of first reference multimedia data.
3. And recalling each second reference multimedia data through the neural network model, and determining recommended parameters of each second reference multimedia data.
4. And bidding the second reference multimedia data according to the recommendation parameters of the second reference multimedia data, selecting target multimedia data, and recommending the target multimedia data to the target user.
Based on the embodiments shown in fig. 3 and 7, training of the feature transformation model is required before invoking the feature transformation model, and the training process is described in detail in the following embodiments.
Fig. 9 is a training method of a feature transformation model provided in an embodiment of the present disclosure, and referring to fig. 9, the method is applied to a server, and includes:
in step 901, a server invokes a feature conversion model to perform feature conversion processing on a first sample data feature corresponding to the first sample multimedia data, so as to obtain a converted first sample operation feature.
In the disclosed embodiment, the feature conversion model and the feature discrimination model are alternately and iteratively trained by adopting the feature conversion model and the feature discrimination model to perform countermeasure training, so as to obtain an accurate feature conversion model. For example, the feature conversion model is trained first, then the feature discrimination model is trained based on the trained feature conversion model, then the feature conversion model is trained through the trained feature discrimination model, and the above process is repeated to train the feature conversion model and the feature discrimination model respectively, so as to improve the accuracy of the feature conversion model.
Wherein the first sample multimedia data is arbitrary multimedia data. Acquiring first sample data characteristics corresponding to the first sample multimedia data, similar to step 301 described above; this step 901 is similar to the above step 302 and will not be described again.
In step 902, the server invokes a feature discrimination model to discriminate the first sample operation feature, and obtains a discrimination flag of the first sample operation feature.
The feature discrimination model is used for discriminating whether the input features are data features or operation features. The discrimination indicator is used to indicate whether the first sample operational feature belongs to the data feature or the operational feature. In some embodiments, the discrimination is identified as a probability, representing a probability that the input feature is a data feature, or representing a probability that the input feature is an operational feature. For example, the range of the probability is (0, 1), if the probability is 0, the feature indicating the discrimination is the operation feature, if the probability is 1, the feature indicating the discrimination is the data feature, or if the probability is 0, the feature indicating the discrimination is the data feature, if the probability is 1, the feature indicating the discrimination is the operation feature. A process of acquiring the discrimination identification of the inputted feature by the feature discrimination model is shown in fig. 10.
And judging the first sample operation feature through the feature judgment model to determine whether the first sample operation feature generated by the feature conversion model belongs to the data feature or the operation feature, so as to determine whether the first sample operation feature generated by the feature conversion model is accurate or not according to the judging identification.
In step 903, the server trains the feature transformation model according to the first sample operating feature and the discrimination indicator.
And determining whether the first sample operation feature generated by the feature conversion model is accurate or not through the sample operation feature and the discrimination mark so as to adjust the feature conversion model according to the discrimination mark later so as to improve the accuracy of the feature conversion model.
In some embodiments, this step 903 includes determining a first loss value for the feature transformation model based on the first sample operational feature and the discriminant identification, and training the feature transformation model based on the first loss value.
Wherein the first loss value is used to represent the inaccuracy of the current feature transformation model. And training the feature conversion model through the first loss value to reduce the loss value of the feature conversion model, thereby improving the accuracy of the feature conversion model.
In some embodiments, the first sample data characteristic, the first sample operational characteristic and the first loss value satisfy the following relationship:
Wherein,The method comprises the steps of representing a first loss value, representing a first sample operation characteristic by G (z), representing a distinguishing identification of the first sample operation characteristic by D (G (z)), representing a distribution situation of the first sample data characteristic z of a plurality of first sample multimedia data by z-Pz (z), and representing a mean value of the loss values of the plurality of sample multimedia data by E ().
In the embodiment of the disclosure, only one training is taken as an example to train the feature transformation model, and in another embodiment, the feature transformation model is trained multiple times through multiple first sample multimedia data.
In some embodiments, the feature transformation model is iteratively trained with a plurality of first sample multimedia data, and training of the feature transformation model is stopped in response to a loss value obtained from a current training round being less than a reference loss value, or in response to the training round of the feature transformation model reaching the reference round.
The feature conversion model is trained once through a first sample multimedia data, and a training round is represented. The reference loss value is any value, such as 0.3 or 0.4. The reference round is any round, for example, 10 times or 15 times, etc.
After the feature conversion model is trained for a plurality of times through a plurality of first sample multimedia data, when the training stopping condition of the feature conversion model is reached, the condition indicates that whether the sample operation feature generated by the feature conversion model belongs to the operation feature or the data feature cannot be recognized based on the current feature judgment model, so that the feature judgment model needs to be trained subsequently, and the feature conversion model can be trained continuously through the trained feature judgment model subsequently.
In step 904, the server invokes a feature conversion model to perform feature conversion processing on the second sample data feature corresponding to the second sample multimedia data, so as to obtain a converted second sample operation feature.
Wherein the second sample multimedia data is different from the first sample multimedia data. This step 904 is similar to step 901 described above and will not be described again.
In step 905, the server invokes a feature discrimination model to respectively discriminate the second sample data feature and the second sample operation feature, and determine discrimination identifications of the second sample data feature and the second sample operation feature.
The distinguishing mark is used for indicating whether the corresponding feature belongs to the data feature or the operation feature.
Because the feature discriminating model is used for discriminating whether the input features belong to the data features or the operation features, in order to ensure the accuracy of the feature discriminating model, the discriminating model is required to discriminate the sample data features and the sample operation features when the discriminating model is trained, so that the discriminating model after subsequent training can distinguish whether the features belong to the data features or the operation features.
The process of calling the feature discrimination model to discriminate the second sample data feature and the second sample operation feature is similar to the above step 902, and will not be described herein.
In step 906, the server trains the feature discrimination model according to the second sample data feature, the second sample operational feature, the second sample data feature, and the discrimination identification of the second sample operational feature.
And determining whether the feature discrimination model discrimination features are accurate or not through the second sample data features, the second sample operation features, the second sample data features and discrimination marks of the second sample operation features, so that the feature discrimination model can be adjusted to enhance the discrimination performance of the feature discrimination model and improve the accuracy of the feature discrimination model.
In some embodiments, step 906 includes taking the second sample data feature as a positive sample, taking the second sample operational feature, and training the feature discrimination model based on the second sample data feature, the second sample operational feature, the second sample data feature, and the discrimination identification of the second sample operational feature.
In some embodiments, step 906 includes determining a second loss value for the feature discrimination model based on the second sample data feature, the second sample operational feature, the second sample data feature, and the discrimination identification for the second sample operational feature, and training the feature discrimination model based on the second loss value.
Wherein the second loss value is used to represent inaccuracy of the current feature discrimination model. Training the feature discrimination model through the second loss value to reduce the loss value of the feature discrimination model and enhance the discrimination performance of the feature discrimination model, thereby improving the accuracy of the feature discrimination model.
In some embodiments, the second sample data characteristic, the second sample operational characteristic, and the second loss value satisfy the following relationship:
Wherein,The method comprises the steps of representing a second loss value, x-Pdata (z) representing distribution conditions of a plurality of second sample multimedia data x and second sample data characteristics of the second sample multimedia data, z-Pz (z) representing distribution conditions of the second sample data characteristics z of the plurality of second sample multimedia data, D (x) distinguishing marks of the second sample data characteristics, G (z) representing second sample operation characteristics, D (G (z)) representing distinguishing marks of the second sample operation characteristics, and E () representing an average value of the loss values of the plurality of sample multimedia data.
In the embodiment of the disclosure, only one training is taken as an example to train the feature discrimination model, and in another embodiment, the feature discrimination model is trained multiple times through multiple second sample multimedia data.
In some embodiments, the feature discrimination model is iteratively trained with a plurality of second sample multimedia data, and training of the feature discrimination model is stopped in response to a loss value obtained from a current training round being less than a reference loss value, or in response to the training round of the feature discrimination model reaching the reference round.
And training the feature discrimination model once through a second sample multimedia data, wherein the training is represented by a training round. After the feature discrimination model is trained for a plurality of times through a plurality of second sample multimedia data, when the training stopping condition of the feature discrimination model is reached, the feature discrimination model is based on the current feature conversion model, and the feature discrimination model can accurately distinguish sample data features from sample operation features generated by the current feature conversion model, so that the feature conversion model can be trained continuously through the currently obtained feature discrimination model.
In step 907, the server invokes the feature transformation model to perform feature transformation processing on the third sample data feature corresponding to the third sample multimedia data, so as to obtain a transformed third sample operation feature.
The third sample multimedia data is different from the first sample multimedia data and the second sample multimedia data.
This step 907 is similar to step 901 described above and will not be described again.
In step 908, the server invokes the feature discrimination model to discriminate the third sample operational feature, and determines a discrimination indicator for the third sample operational feature.
Wherein the discrimination indicator is used to indicate whether the third sample operational feature belongs to the data feature or the operational feature. This step 908 is similar to step 902 described above and will not be described again.
In step 909, the server trains the feature transformation model according to the third sample operating feature and the discrimination indicator.
This step 909 is similar to step 903 described above and will not be described again here.
It should be noted that, in the embodiment of the present disclosure, the feature conversion model is trained first, then the feature discrimination model is trained, and finally the feature conversion model is trained, and in another embodiment, after step 909, the feature discrimination model is trained again based on the trained feature conversion model, then the feature conversion model is trained based on the trained feature discrimination model, the above process is repeated, the feature discrimination model and the feature conversion model are alternately and iteratively trained, and the alternately and iteratively trained feature conversion model and the feature discrimination model is stopped in response to the feature conversion model and the feature discrimination model reaching the equilibrium state.
In some embodiments, after the feature conversion model and the feature classification model are subjected to a plurality of alternating iterative training rounds, when the feature conversion model is trained, the loss value of the feature conversion model converges, and when the feature classification model is trained, the loss value of the feature classification model converges, which indicates that the current feature conversion model and the feature classification model have reached an equilibrium state, so that the alternating iterative training of the feature conversion model and the feature classification model is stopped.
In the process of training the feature conversion model, the feature conversion model is used as a Generator, the feature discrimination model is used as a Discriminator (discriminator), the feature conversion model and the feature discrimination model form a GAN (GENERATIVE ADVERSARIAL Network, generating an countermeasure Network), and the generated countermeasure Network is trained to obtain an accurate feature conversion model, so that the obtained feature conversion model can accurately predict the operation of multimedia data, and the accurate recommendation of the multimedia data can be ensured according to the predicted operation of the multimedia data, so as to realize the cold start of the multimedia data.
According to the method provided by the embodiment of the disclosure, the feature conversion model and the feature discrimination model are alternately and iteratively trained in a manner of performing countermeasure training by adopting the feature conversion model and the feature discrimination model, so that the feature conversion model and the feature discrimination model reach an equilibrium state, and an accurate feature conversion model is obtained.
FIG. 11 is a block diagram illustrating a data recommendation device, according to an example embodiment. Referring to fig. 11, the apparatus includes:
A feature acquisition unit 1101 configured to acquire a target data feature corresponding to target multimedia data to be recommended;
The transformation processing unit 1102 is configured to invoke a feature transformation model, perform feature transformation processing on the target data features, and obtain transformed predicted operation features, where the predicted operation features are used to represent predicted operations performed on the target multimedia data;
a determining unit 1103 configured to determine a matching degree between each user and the target multimedia data according to user characteristics corresponding to the plurality of users and the prediction operation characteristics;
And a recommending unit 1104 configured to select at least one target user from the plurality of users according to the matching degree corresponding to the plurality of users, and recommend the target multimedia data to the at least one target user.
In some embodiments, referring to fig. 12, the apparatus further comprises:
The transformation processing unit 1102 is configured to call a feature conversion model, perform feature transformation processing on first sample data features corresponding to the first sample multimedia data, and obtain transformed first sample operation features;
A distinguishing processing unit 1105 configured to invoke a feature distinguishing model, and distinguish the first sample operation feature to obtain a distinguishing identifier of the first sample operation feature, where the distinguishing identifier is used to indicate whether the first sample operation feature belongs to a data feature or an operation feature;
The first training unit 1106 is configured to train the feature transformation model according to the first sample operation feature and the discrimination indicator.
In some embodiments, referring to fig. 12, first training unit 1106 comprises:
A first determining subunit 1161 configured to determine a first loss value of the feature transformation model according to the first sample operating feature and the discrimination indicator;
The first training subunit 1162 is configured to train the feature transformation model according to the first loss value.
In some embodiments, referring to fig. 12, the apparatus further comprises:
the transformation processing unit 1102 is configured to call a feature conversion model, perform feature transformation processing on the second sample data features corresponding to the second sample multimedia data, and obtain transformed second sample operation features;
The distinguishing processing unit 1105 is further configured to invoke a feature distinguishing model to distinguish the second sample data feature and the second sample operation feature, and determine distinguishing identifiers of the second sample data feature and the second sample operation feature, where the distinguishing identifiers are used to indicate whether the corresponding features belong to the data feature or the operation feature;
the second training unit 1107 is configured to train the feature discrimination model according to the second sample data feature, the second sample operation feature, the second sample data feature, and the discrimination identification of the second sample operation feature.
In some embodiments, referring to fig. 12, the second training unit 1107 includes:
A second determining subunit 1171 configured to determine a second loss value of the feature discrimination model based on the second sample data feature, the second sample operational feature, the second sample data feature, and the discrimination identity of the second sample operational feature;
a second training subunit 1172 configured to train the feature discrimination model according to the second loss value.
In some embodiments, referring to fig. 12, the apparatus further comprises:
The transformation processing unit 1102 is configured to call a feature conversion model, perform feature transformation processing on the third sample data feature corresponding to the third sample multimedia data, and obtain a transformed third sample operation feature;
The distinguishing processing unit 1105 is further configured to invoke a feature distinguishing model, distinguish the third sample operation feature, determine a distinguishing identifier of the third sample operation feature, and the distinguishing identifier is used for indicating whether the third sample operation feature belongs to the data feature or the operation feature;
The third training unit 1108 is configured to train the feature transformation model according to the third sample operation feature and the discrimination indicator.
In some embodiments, referring to fig. 12, the feature acquisition unit 1101 includes:
The feature extraction subunit 1111 is configured to invoke a feature extraction model to perform feature extraction on the target multimedia data, so as to obtain a target data feature corresponding to the target multimedia data.
In some embodiments, the feature extraction model includes a text feature extraction sub-model and an image feature extraction sub-model, and the feature extraction sub-unit 1111 is configured to call the text feature extraction sub-model, perform feature extraction on the target multimedia data to obtain text features of the target multimedia data, call the image feature extraction sub-model, perform feature extraction on the target multimedia data to obtain image features of the target multimedia data, and perform stitching processing on the text features and the image features to obtain target data features of the target multimedia data.
In some embodiments, referring to fig. 12, the determining unit 1103 includes:
the third determining subunit 1131 is configured to determine, according to the user characteristics, the target data characteristics, and the predicted operation characteristics corresponding to the plurality of users, a matching degree between each user and the target multimedia data.
In some embodiments, the third determining subunit 1131 is configured to perform matching processing on the user characteristics, the target data characteristics and the predicted operation characteristics of any user, so as to obtain a matching degree between the user and the target multimedia data.
In some embodiments, the recommendation unit 1104 is configured to determine recommendation parameters corresponding to the plurality of users according to the matching degree corresponding to the plurality of users and the number of resources corresponding to the target multimedia data, select at least one target user from the plurality of users according to the recommendation parameters corresponding to the plurality of users, and recommend the target multimedia data to the at least one target user.
The specific manner in which the individual units perform the operations in relation to the apparatus of the above embodiments has been described in detail in relation to the embodiments of the method and will not be described in detail here.
Fig. 13 is a block diagram illustrating a structure of a terminal 1300 according to an exemplary embodiment. The terminal 1300 may be a portable mobile terminal such as a smart phone, tablet, MP3 player (Moving Picture Experts Group Audio Layer III, motion picture expert compression standard audio plane 3), MP4 (Moving Picture Experts Group Audio Layer IV, motion picture expert compression standard audio plane 4) player, notebook or desktop. Terminal 1300 may also be referred to by other names of user devices, portable terminals, laptop terminals, desktop terminals, etc.
Terminal 1300 includes a processor 1301 and a memory 1302.
Processor 1301 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. Processor 1301 may be implemented in at least one hardware form of DSP (DIGITAL SIGNAL Processing), FPGA (Field-Programmable gate array) GATE ARRAY, PLA (Programmable Logic Array ). Processor 1301 may also include a main processor, which is a processor for processing data in a wake-up state, also referred to as a CPU (Central Processing Unit ), and a coprocessor, which is a low-power processor for processing data in a standby state. In some embodiments, processor 1301 may integrate a GPU (Graphics Processing Unit, image processor) for rendering and drawing of content that is required to be displayed by the display screen. In some embodiments, processor 1301 may also include an AI (ARTIFICIAL INTELLIGENCE ) processor for processing computing operations related to machine learning.
Memory 1302 may include one or more computer-readable storage media, which may be non-transitory. Memory 1302 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, a non-transitory computer readable storage medium in memory 1302 is used to store at least one program code for execution by processor 1301 to implement the data recommendation method provided by the method embodiments in the present disclosure.
In some embodiments, terminal 1300 may also optionally include a peripheral interface 1303 and at least one peripheral. The processor 1301, the memory 1302, and the peripheral interface 1303 may be connected by a bus or signal lines. The respective peripheral devices may be connected to the peripheral device interface 1303 through a bus, a signal line, or a circuit board. Specifically, the peripheral devices include at least one of radio frequency circuitry 1304, a display screen 1305, a camera assembly 1306, audio circuitry 1307, a positioning assembly 1308, and a power supply 1309.
A peripheral interface 1303 may be used to connect I/O (Input/Output) related at least one peripheral to the processor 1301 and the memory 1302. In some embodiments, processor 1301, memory 1302, and peripheral interface 1303 are integrated on the same chip or circuit board, and in some other embodiments, either or both of processor 1301, memory 1302, and peripheral interface 1303 may be implemented on separate chips or circuit boards, which is not limited in this embodiment.
The Radio Frequency circuit 1304 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 1304 communicates with a communication network and other communication devices via electromagnetic signals. The radio frequency circuit 1304 converts an electrical signal to an electromagnetic signal for transmission, or converts a received electromagnetic signal to an electrical signal. In some embodiments, radio frequency circuit 1304 includes an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a subscriber identity module card, and so forth. The radio frequency circuit 1304 may communicate with other terminals via at least one wireless communication protocol. The wireless communication protocols include, but are not limited to, the world wide web, metropolitan area networks, intranets, various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks, and/or WiFi (WIRELESS FIDELITY ) networks. In some embodiments, the radio frequency circuit 1304 may also include NFC (NEAR FIELD Communication) related circuits, which is not limited by the present disclosure.
The display screen 1305 is used to display a UI (User Interface). The UI may include graphics, text, icons, video, and any combination thereof. When the display 1305 is a touch display, the display 1305 also has the ability to capture touch signals at or above the surface of the display 1305. The touch signal may be input to the processor 1301 as a control signal for processing. At this point, the display 1305 may also be used to provide virtual buttons and/or a virtual keyboard, also referred to as soft buttons and/or a soft keyboard. In some embodiments, the display 1305 may be one and disposed on the front panel of the terminal 1300, in other embodiments, the display 1305 may be at least two and disposed on different surfaces or in a folded configuration of the terminal 1300, respectively, and in other embodiments, the display 1305 may be a flexible display disposed on a curved surface or a folded surface of the terminal 1300. Even more, the display screen 1305 may be arranged in a non-rectangular irregular pattern, i.e., a shaped screen. The display screen 1305 may be made of LCD (Liquid CRYSTAL DISPLAY), OLED (Organic Light-Emitting Diode) or other materials.
The camera assembly 1306 is used to capture images or video. In some embodiments, camera assembly 1306 includes a front camera and a rear camera. The front camera is arranged on the front panel of the terminal, and the rear camera is arranged on the back of the terminal. In some embodiments, the at least two rear cameras are any one of a main camera, a depth camera, a wide-angle camera and a tele camera, so as to realize that the main camera and the depth camera are fused to realize a background blurring function, and the main camera and the wide-angle camera are fused to realize a panoramic shooting and Virtual Reality (VR) shooting function or other fusion shooting functions. In some embodiments, camera assembly 1306 may also include a flash. The flash lamp can be a single-color temperature flash lamp or a double-color temperature flash lamp. The dual-color temperature flash lamp refers to a combination of a warm light flash lamp and a cold light flash lamp, and can be used for light compensation under different color temperatures.
The audio circuit 1307 may include a microphone and a speaker. The microphone is used for collecting sound waves of users and environments, converting the sound waves into electric signals, and inputting the electric signals to the processor 1301 for processing, or inputting the electric signals to the radio frequency circuit 1304 for voice communication. For purposes of stereo acquisition or noise reduction, a plurality of microphones may be provided at different portions of the terminal 1300, respectively. The microphone may also be an array microphone or an omni-directional pickup microphone. The speaker is then used to convert electrical signals from the processor 1301 or the radio frequency circuit 1304 into sound waves. The speaker may be a conventional thin film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, not only the electric signal can be converted into a sound wave audible to humans, but also the electric signal can be converted into a sound wave inaudible to humans for ranging and other purposes. In some embodiments, the audio circuit 1307 may also comprise a headphone jack.
The location component 1308 is utilized to locate the current geographic location of the terminal 1300 to enable navigation or LBS (Location Based Service, location-based services). The positioning component 1308 may be a positioning component based on the United states GPS (Global Positioning System ), the Beidou system of China, or the Galileo system of Russia.
A power supply 1309 is used to power the various components in the terminal 1300. The power supply 1309 may be an alternating current, a direct current, a disposable battery, or a rechargeable battery. When the power supply 1309 comprises a rechargeable battery, the rechargeable battery may be a wired rechargeable battery or a wireless rechargeable battery. The wired rechargeable battery is a battery charged through a wired line, and the wireless rechargeable battery is a battery charged through a wireless coil. The rechargeable battery may also be used to support fast charge technology.
In some embodiments, terminal 1300 also includes one or more sensors 1310. The one or more sensors 1310 include, but are not limited to, an acceleration sensor 1311, a gyroscope sensor 1312, a pressure sensor 1313, a fingerprint sensor 1314, an optical sensor 1315, and a proximity sensor 1316.
The acceleration sensor 1311 can detect the magnitudes of accelerations on three coordinate axes of the coordinate system established with the terminal 1300. For example, the acceleration sensor 1311 may be used to detect components of gravitational acceleration in three coordinate axes. Processor 1301 may control display screen 1305 to display a user interface in either a landscape view or a portrait view based on gravitational acceleration signals acquired by acceleration sensor 1311. The acceleration sensor 1311 may also be used for the acquisition of motion data of a game or user.
The gyro sensor 1312 may detect a body direction and a rotation angle of the terminal 1300, and the gyro sensor 1312 may collect a 3D motion of the user on the terminal 1300 in cooperation with the acceleration sensor 1311. Based on the data collected by gyro sensor 1312, processor 1301 can realize functions such as motion sensing (e.g., changing UI according to a tilting operation by a user), image stabilization at photographing, game control, and inertial navigation.
Pressure sensor 1313 may be disposed on a side frame of terminal 1300 and/or below display screen 1305. When the pressure sensor 1313 is disposed at a side frame of the terminal 1300, a grip signal of the terminal 1300 by a user may be detected, and the processor 1301 performs left-right hand recognition or shortcut operation according to the grip signal collected by the pressure sensor 1313. When the pressure sensor 1313 is disposed at the lower layer of the display screen 1305, the processor 1301 realizes control of the operability control on the UI interface according to the pressure operation of the user on the display screen 1305. The operability controls include at least one of a button control, a scroll bar control, an icon control, and a menu control.
The fingerprint sensor 1314 is used to collect a fingerprint of the user, and the processor 1301 identifies the identity of the user based on the fingerprint collected by the fingerprint sensor 1314, or the fingerprint sensor 1314 identifies the identity of the user based on the collected fingerprint. Upon recognizing that the user's identity is a trusted identity, the user is authorized by processor 1301 to perform relevant sensitive operations including unlocking the screen, viewing encrypted information, downloading software, paying for and changing settings, etc. The fingerprint sensor 1314 may be disposed on the front, back, or side of the terminal 1300. When a physical key or vendor Logo is provided on the terminal 1300, the fingerprint sensor 1314 may be integrated with the physical key or vendor Logo.
The optical sensor 1315 is used to collect ambient light intensity. In one embodiment, processor 1301 may control the display brightness of display screen 1305 based on the intensity of ambient light collected by optical sensor 1315. Specifically, the display brightness of the display screen 1305 is turned up when the ambient light intensity is high, and the display brightness of the display screen 1305 is turned down when the ambient light intensity is low. In another embodiment, processor 1301 may also dynamically adjust the shooting parameters of camera assembly 1306 based on the intensity of ambient light collected by optical sensor 1315.
A proximity sensor 1316, also referred to as a distance sensor, is provided on the front panel of the terminal 1300. The proximity sensor 1316 is used to collect the distance between the user and the front of the terminal 1300. In one embodiment, the processor 1301 controls the display screen 1305 to switch from the on-screen state to the off-screen state when the proximity sensor 1316 detects that the distance between the user and the front surface of the terminal 1300 becomes gradually smaller, and the processor 1301 controls the display screen 1305 to switch from the off-screen state to the on-screen state when the proximity sensor 1316 detects that the distance between the user and the front surface of the terminal 1300 becomes gradually larger.
Those skilled in the art will appreciate that the structure shown in fig. 13 is not limiting of terminal 1300 and may include more or fewer components than shown, or may combine certain components, or may employ a different arrangement of components.
Fig. 14 is a schematic diagram of a server according to an exemplary embodiment, where the server 1400 may have a relatively large difference between configurations or performances, and may include one or more processors (Central Processing Units, CPU) 1401 and one or more memories 1402, where at least one program code is stored in the memories 1402, and the at least one program code is loaded and executed by the processors 1401 to implement the methods provided in the respective method embodiments described above. Of course, the server may also have a wired or wireless network interface, a keyboard, an input/output interface, and other components for implementing the functions of the device, which are not described herein.
The server 1400 may be used to perform the steps performed by the server in the data recommendation method described above.
In an exemplary embodiment, a non-transitory computer readable storage medium is also provided, which when executed by a processor of a server, enables the server to perform the steps performed by the server in the data recommendation method described above. In some embodiments, the storage medium may be a non-transitory computer readable storage medium, for example, a ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.
In an exemplary embodiment, there is also provided a server, characterized in that the server includes:
One or more processors;
volatile or non-volatile memory for storing one or more processor-executable instructions;
Wherein the one or more processors are configured to perform the steps performed by the server in the data recommendation method described above.
In an exemplary embodiment, a non-transitory computer readable storage medium is also provided, which when executed by a processor of a server, enables the server to perform the steps performed by the server in the data recommendation method described above.
In an exemplary embodiment, a computer program product is also provided, which, when executed by a processor of a server, enables the server to perform the steps performed by the server in the data recommendation method described above.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure herein. This disclosure is intended to cover any adaptations, uses, or adaptations of the disclosure following the general 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 is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.