CN110633376A - Media object sorting method, device, equipment and storage medium - Google Patents
Media object sorting method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN110633376A CN110633376A CN201910779717.3A CN201910779717A CN110633376A CN 110633376 A CN110633376 A CN 110633376A CN 201910779717 A CN201910779717 A CN 201910779717A CN 110633376 A CN110633376 A CN 110633376A
- Authority
- CN
- China
- Prior art keywords
- media object
- ranking
- data
- value
- media
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/44—Browsing; Visualisation therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Databases & Information Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention provides a method, a device, equipment and a storage medium for ordering media objects, wherein the method for ordering the media objects comprises the following steps: acquiring characteristic data of a media object; inputting the characteristic data into a media object recognition model, and determining a ranking value; and sorting the media objects according to the sorting value. The media object recognition model can extract information with stronger relevance with the media object to generate the ranking value based on different feature data of different media objects, so that the ranking value can be closer to the characteristics of the media object, compared with a ranking list generated by the ranking value determined according to a fixed rule, the ranking list generated based on the ranking value in the embodiment of the invention can reduce the relevance between the ranking list and the fixed rule to a certain extent, the rule does not need to be adjusted manually when the ranking list is generated, and the ranking in the ranking list is more objective due to comprehensive consideration of the feature information of the media objects.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a method, an apparatus, a device, and a storage medium for ordering media objects.
Background
With the popularization and development of networks, many self-media emerge, for example, the 'public number' of WeChat, the 'microblog number' of microblog, the 'first number' of today and the 'big fish number' of UC are self-media, a platform for publishing self-opinions and sharing personal events is provided for users, the users can obtain the latest information and have more personalized opinions, and the lives of people are greatly enriched.
At present, the list of the self-media on each platform is presented according to a certain ranking rule, such as ranking according to user index score, index grade, credit score, credit grade, and the like, and specifically, according to a fixed rule, ranking is performed on values obtained by weighted accumulation of the user index score, index grade, credit score, and credit grade, so as to obtain the list. The relevance between the list obtained by the method and the ranking rule is strong, so that the adaptability of the ranking rule is poor, and when the list is generated by using the method, the rule needs to be manually adjusted according to the condition of the self-media, which wastes time and labor.
Disclosure of Invention
The technical problem to be solved by the embodiments of the present invention is to provide a method, an apparatus, a device and a storage medium for ranking media objects, so as to solve the technical problems that manual adjustment is required according to the situation of a self-media and time and labor are wasted when a list is generated according to a fixed rule.
Correspondingly, the embodiment of the invention also provides a media object sequencing device, equipment and a storage medium, which are used for ensuring the realization and the application of the method.
In order to solve the problems, the invention is realized by the following technical scheme:
a first aspect provides a method of ordering media objects, the method comprising:
acquiring characteristic data of a media object;
inputting the characteristic data into a media object recognition model, and determining a ranking value;
and sorting the media objects according to the sorting value.
A second aspect provides an apparatus for ranking media objects, the apparatus comprising:
the characteristic data acquisition module is used for acquiring the characteristic data of the media object;
the sequence value determining module is used for inputting the characteristic data into a media object recognition model and determining a sequence value;
and the sequencing module is used for sequencing the media objects according to the sequencing values.
A third aspect provides an apparatus comprising: a memory, a processor and a computer program stored on the memory and executable on the processor, the computer program, when executed by the processor, implementing the steps of the method of ordering media objects as described above.
A fourth aspect provides a storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps in the method of ranking media objects as described above.
Compared with the prior art, the embodiment of the invention has the following advantages:
the media object sorting method provided by the embodiment of the invention can acquire the characteristic data of the media objects, input the characteristic data into the media object recognition model, determine the sorting value, and sort the media objects according to the sorting value. The media object recognition model can extract information with stronger relevance with the media object to generate the ranking value based on different feature data of different media objects, so that the ranking value can be closer to the characteristics of the media object, compared with a list generated by the ranking value determined according to a fixed rule, the list generated based on the ranking value in the embodiment of the invention can reduce the relevance between the list and the fixed rule to a certain extent, the rule does not need to be adjusted manually when the list is generated, and the ranking in the list is more objective due to comprehensive consideration of the feature information of the media objects.
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 application.
Drawings
FIG. 1 is a flow chart of a method for ordering media objects according to an embodiment of the present invention;
FIG. 2 is a flow chart of establishing a media object model according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of the multilayer perceptron model provided by the embodiments of the present invention;
FIG. 4 is a schematic diagram of a neural network model training process in an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a media object sorting apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Referring to fig. 1, a flowchart of a media object sorting method according to an embodiment of the present invention may specifically include the following steps 100 to 300:
step 100: characteristic data of the media object is obtained.
In this step, the media object may be a media to be ranked, for example, the media object may be a self-media to be ranked, and the feature data may be data that may affect the ranking of the media object, for example, the feature data may include: at least one of user registration time, user level, user index score, user credit level, user credit score, number of uploaded media objects, number of online media objects, industry in which the user is located, number of channels for uploading media objects, content level, and user nickname.
Specifically, when the feature data of the media object is acquired, the data output by the media object can be acquired through the back-end server as the media data, because the data which can affect the ranking of the media object generally has multiple types, therefore, the feature data under different feature dimensions can be extracted from the media data based on multiple different feature dimensions, and thus, by extracting the feature data under multiple feature dimensions, when the ranking is performed based on the feature data in the subsequent steps, the media object can be comprehensively ranked from the angle embodied by multiple different feature dimensions, and further the final ranking result can be more reasonable to a certain extent.
Further, it should be noted that other obtaining manners may also be adopted to obtain the information according to actual requirements, which is not limited in the embodiment of the present invention.
Further, after the feature data is obtained, the value of the feature data in the feature dimension of the preset type may be encoded, where the preset type may be a discrete type, and since the value of the feature data in the feature dimension of the discrete type is often a discrete value, and for the model, the difficulty in extracting the numerical feature from the discrete value is often large, in this step, the value in the feature data of the discrete type may be encoded to vectorize the value, so that the model may conveniently extract the feature in the value, and further the processing effect is improved. Specifically, the encoding process may be implemented based on One-Hot encoding (One-Hot).
For example, as the vector dimension of the encoding process, the possible value situations of the values of the feature data in the preset type of feature dimension may be assumed to be the vector dimension of the encoding process, for example, for a user level, it is assumed that there are 7 possible value situations of the values of the user level, that is, the values of the user level may be 1, may be 2, …, and may be 7, then the vector dimension may be determined to be 7, each vector dimension represents one possible value situation, accordingly, when the encoding process is performed, the value of the vector dimension of the value situation corresponding to the value of the user level may be set to 1, and the values of the other vector dimensions may be set to 0, thereby implementing the encoding process.
Of course, other encoding methods may be used for encoding, which is not limited in the embodiment of the present invention.
And 200, inputting the characteristic data into a media object recognition model, and determining a ranking value.
In this step, the feature data may be input into the media object recognition model, and the media object recognition model generates the ranking value of the media object based on the feature data, where the media object recognition model may be obtained by pre-training based on training sample data.
And step 300, sequencing the media objects according to the sequencing values.
In this step, the media objects may be sorted according to a predetermined sorting rule according to the sorting value, for example, the media objects may be sorted in a descending order of the sorting value, and the media objects may be sorted to corresponding positions of the list, so as to obtain a corresponding list of the media objects. Further, in the process of ranking, if the ranking values of a plurality of media objects are the same and the ranking values are in the media object ranking list, the plurality of media objects can be displayed in the ranking list according to a predetermined rule. The condition that the ranking value is in the media object ranking list means that the ranking value is not greater than the ranking number in the media object ranking list, for example, if the ranking number in the media object ranking list is 10 and the ranking value is 7, then the ranking value can be considered to be in the media object ranking list. Further, the display according to the predetermined rule may be a parallel display or a display according to the first letter of the pinyin of the name of the media object. For example, if the ranking value of the media object a and the ranking value of the media object B are 5 and 5, respectively, according to the calculation, the media object a and the media object B may be displayed in parallel on the fifth place of the media list.
In summary, the media object sorting method provided in the embodiments of the present invention may obtain the feature data of the media object, input the feature data into the media object recognition model, determine the sorting value, and sort the media object according to the sorting value. The media object recognition model can extract information with stronger relevance with the media object to generate the ranking value based on different feature data of different media objects, so that the ranking value can be closer to the characteristics of the media object, compared with a list generated by the ranking value determined according to a fixed rule, the list generated based on the ranking value in the embodiment of the invention can reduce the relevance between the list and the fixed rule to a certain extent, the rule does not need to be adjusted manually when the list is generated, and the ranking in the list is more objective due to comprehensive consideration of the feature information of the media objects.
Optionally, prior to step 100, a media object recognition model may also be pre-established. Specifically, referring to fig. 2, the process of establishing the media object model may include:
The training sample data may include feature data of the sample media object and tag data, where the feature data may be feature data of a plurality of different feature dimensions of the sample media object. The characteristic data may be data that may affect the ranking of the sample media object, which may include, for example: at least one of user registration time, user rating, user index score, user credit rating, user credit score, number of uploaded media objects, number of online media objects, industry in which the user is located, number of channels for uploading media objects, content rating, and user nickname, the tag data indicates the true ranking value of the sample media object, which may include historical ranking of media objects, the media object historical ranking condition may be a ranking in a media object ranking list that has a number of rankings less than a third predetermined threshold and, accordingly, because some sample media objects may never appear on the media object leader board, the user may, therefore, the historical ranking of the sample media object may be set to be next to the last name in the list, that is, the historical ranking of the sample media object may be set to be the third preset threshold + 1. In this way, by setting the label data for the sample media object that does not appear in the media object ranking list, the integrity of the training sample data can be ensured, and the effect of training based on the training sample data can be further improved.
And 102, extracting characteristic data from the training sample data.
Specifically, the manner of extracting the feature data in this step may refer to the step 100, which is not described herein again in this embodiment of the present invention.
And 103, inputting the characteristic data into the neural network model for calculation to obtain a calculation result.
In this step, the calculation result may represent a ranking value predicted by the neural network model based on the feature data of the sample media object. Specifically, for the convenience of computer identification and calculation, feature vectors of corresponding dimensions may be constructed according to the feature data acquired in step 102, and then the feature vectors corresponding to the training samples are input into the neural network model for calculation to obtain a calculation result.
When the feature vector is constructed according to the acquired feature data, the feature data may be cleaned first, and specifically, the cleaning may be realized by performing error correction processing on the feature dimension in which an abnormal value occurs in the feature data.
The abnormal value refers to a value that is Not within a reasonable range, an ambiguous numerical result (Not a Number, NaN), and a null value. For example, the user rating may not be provided for some media objects, i.e. the value in the feature dimension is null, and at this time, the value may be determined to be an abnormal value, and further, if the normal age range is 0 to 120, if the age is 130, the value in the feature dimension of the age may be considered to be an abnormal value. Accordingly, the value can be subjected to error correction processing. Specifically, the error correction process may replace the abnormal value with a median of the values in the feature dimension. Compared with a mode of correcting the abnormal value according to a default value, in the step, the abnormal value is corrected in a median filling mode, so that the corrected value is more representative, and meanwhile, the data concentration condition in the characteristic dimension can be reflected due to the median of each value in the characteristic dimension, so that the characteristic data can reflect the concentration condition of the training data in the mode of correcting the abnormal value in the median filling mode.
Meanwhile, by cleaning the characteristic data, the data can be ensured to be input into the neural network model, so that the computer can conveniently identify and calculate, and a more reasonable model can be obtained.
Alternatively, referring to fig. 3, the neural network model may be a multi-layered perceptron model, and specifically, the multi-layered perceptron model may include an input layer 1, a hidden layer 2, and an output layer 3, where X is1～XMRepresenting M input level nodes, H1～HNRepresenting N hidden layer nodes, the number of the nodes of the input layer 1 is the same as the number of the feature data of each training sample, namely the number of the dimensions of the feature vector. Referring to fig. 4, when inputting the feature data into the neural network model for calculation, the following steps 201 to 204 may be implemented:
In this step, the feature data may be input to the input layer for weighted calculation to obtain the first data. For example, the data is composed of X1～XMInput, H1Are in turn W1～WNThen the input value at H1 may be X1*W1+X2*W2+X3*W3+X4*W4+…+XN*WN。
In the above step, an activation function is provided in the hidden layer, and the input first data may be calculated by the activation function to obtain N second data. Wherein the activation function is preset, and for example, the activation function may be a ReLu function.
And 203, setting the second data of the hidden layer node with the preset proportion in the N hidden layer nodes as 0.
Specifically, the preset ratio may be preset, and the preset ratio may be controlled according to the keep _ prob parameter, for example, if the preset ratio is to be controlled to be 10%, then keep _ prob may be controlled to be 0.9. Specifically, after the hidden layer is calculated to obtain N second data, the second data of the preset proportion in the N second data may be randomly set to 0. For example, assuming that the preset ratio is 10% and N is 100, the 10 second data may be randomly set to 0 after the hidden layer performs calculation to obtain 100 second data.
In this step, through setting the second data of the hidden layer nodes with the preset proportion to 0, overfitting of training data can be prevented, and then problems caused by overfitting of data in the model training process are avoided, and further the model can be successfully trained, meanwhile, through setting the second data of the hidden layer nodes with the preset proportion to 0, the randomness of the data can be increased, and further the finally obtained model is more flexible.
And step 204, outputting the data of the nodes of the hidden layers to an output layer for calculation to obtain a calculation result.
In this step, when outputting the N second data to the output layer, the weighting may be implemented, for example, by assuming that H is used as the second data1～HNThe weighting coefficient of the input Y is V1～VNThen, one may: h1*V1+H2*V2+H3*V3+H4*V4+…+HN*VNAnd outputting to an output layer.
Further, the output layer may process the input content to obtain a calculation result. Wherein the calculation result can represent a ranking value predicted by the neural network model.
Specifically, the output layer may generate a probability distribution of ranking values of one media object in the list based on the input content, and then output the ranking value with the highest probability as a calculation result.
And 104, inputting the calculation result and the label data into a loss function, and determining a loss value.
In this step, the tag data representing the historical ranking extracted in step 102 and the calculation result calculated in step 103 may be input into the loss function, and then the output value of the loss function may be used as the loss value. Specifically, the loss function may be selected according to actual requirements, and for example, the loss function may be a loss function based on cross information entropy.
And 105, under the condition that the loss value does not reach a first preset threshold value, continuously inputting sample characteristic data for calculation according to parameters in the neural network model adjusted by the optimizer.
In this step, the optimizer may be predefined according to actual requirements, and for example, the optimizer may be an adaptive optimizer Adagrad. If the loss value calculated in step 104 does not reach the first preset threshold, it indicates that the difference between the calculation result and the true value represented by the tag data is far, i.e. it indicates that the current parameters in the neural network model are not reasonable, and the current neural network model cannot accurately determine the sequencing value, so that the parameters can be adjusted according to the optimizer, and the adjusted neural network model continues to be trained.
Specifically, in this step, under the condition that the loss value does not reach the first preset threshold, the optimizer controls two hyper-parameters, namely the learning _ rate and the keep _ prob, in the neural network model according to a preset step size to traverse all possible combinations in a grid search manner, and calculates input sample characteristic data of the neural network model under each learning _ rate and keep _ prob combination to find the combination of the learning _ rate and keep _ prob, which enables the loss value to reach the first preset threshold.
And 106, determining the neural network model as a media object recognition model under the condition that the loss value reaches a first preset threshold value.
In the above steps, if the loss value calculated in step 104 reaches the first preset threshold, it indicates that the calculation result is close to the true value, that is, it indicates that the current parameter in the neural network model is reasonable, and the model training can accurately determine the ranking value based on the feature data, so that the neural network model can be determined as the media object recognition model.
Optionally, after determining the neural network model as a media object recognition model, the media object recognition model may also be tested. Specifically, the test process can be realized through the following steps 301 to 303:
step 301, obtaining test sample data.
In this step, the test sample data may be the feature data of the test sample and the true ranking value of the test sample, and the feature dimension of the feature data of the test sample may be the same as the feature dimension of the feature data of the training sample data, so that the comparability of the test sample data and the true ranking value of the test sample can be increased by using the same dimension, and the representativeness of the test result can be further improved. Specifically, the manner of obtaining the test sample data is similar to the manner of obtaining the training sample data in the foregoing step, and reference may be made to the description in the foregoing step, which is not repeated herein in this embodiment of the present invention.
And 302, testing the accuracy of the media object recognition model according to the test training sample data.
In this step, the feature data of each test training sample data may be input into the media object recognition model to obtain a predicted ranking value of each test training sample data, and then, the predicted ranking value of each test training sample data is compared with the real ranking value of each test training sample data to determine the number of correct predicted ranking values, and finally, the ratio of the number of correct predicted ranking values to the total number of test training sample data may be determined as an accuracy.
It should be noted that, in the process of determining the predicted ranking value according to the media object identification model, the calculation may be performed with the output data of all hidden layers, that is, keep _ prob is controlled to be 1, so as to ensure that the determined predicted ranking value can embody all the features of the test sample data, and further improve the accuracy of the predicted ranking value.
Step 303, retraining the neural network model to obtain the media object recognition model with the accuracy reaching a second preset threshold value under the condition that the accuracy is lower than the second preset threshold value.
In this step, if the accuracy is lower than the second preset threshold, it is determined that the media object recognition model is not qualified, and the neural network model needs to be retrained to obtain the media object recognition model with the accuracy reaching the second preset threshold, so as to ensure that the trained media object recognition model can determine the ranking value more accurately.
In summary, the method for ranking media objects according to the embodiments of the present invention may obtain training sample data, train a neural network model according to the training sample data, obtain a media object recognition model, determine the accuracy of the media object recognition model, rank the media object recognition model only if the accuracy meets a preset requirement, and further ensure that a ranking result based on the media object recognition model is more accurate. The media object recognition model can extract information with stronger relevance with the media object to generate the ranking value based on different feature data of different media objects, so that the ranking value can be closer to the characteristics of the media object, compared with a ranking list generated by the ranking value determined according to a fixed rule, the ranking list generated based on the ranking value in the embodiment of the invention can reduce the relevance between the ranking list and the fixed rule to a certain extent, the rule does not need to be adjusted manually when the ranking list is generated, and the ranking in the ranking list is more objective due to comprehensive consideration of the feature information of the media objects.
It should be noted that, for simplicity of description, the method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the illustrated acts or combination of acts, as some steps may occur in other orders or concurrently according to the present invention. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no particular act is required to implement the invention.
Referring to fig. 5, a media object sorting apparatus according to an embodiment of the present invention is characterized in that the apparatus includes:
a characteristic data obtaining module 10, configured to obtain characteristic data of a media object;
an order value determining module 20, configured to input the feature data into a media object recognition model, and determine an order value;
a sorting module 30, configured to sort the media objects according to the sorting value.
Optionally, in another embodiment, on the basis of the above embodiment, the apparatus further includes:
and the model training module is used for acquiring training sample data of the media object, training the neural network model according to the training sample data and acquiring the media object recognition model.
Optionally, in another embodiment, on the basis of the above embodiment, the training sample data includes feature data and label data; the model training module comprises:
a training sample data acquisition unit for acquiring training sample data of the media object;
the characteristic extraction unit is used for extracting characteristic data from the training sample data;
the first calculation unit is used for inputting the characteristic data into the neural network model for calculation to obtain a calculation result;
a loss value determining unit, configured to input the calculation result and the tag data into a loss function, and determine a loss value;
the second calculation unit is used for adjusting parameters in the neural network model according to the optimizer and continuously inputting sample characteristic data to calculate under the condition that the loss value does not reach a first preset threshold value;
and the model generation unit is used for determining the neural network model as a media object recognition model under the condition that the loss value reaches a first preset threshold value.
Optionally, in another embodiment, on the basis of the above embodiment, the model training module further includes a testing unit, configured to test the media object recognition model determined by the model generating unit;
the test unit includes:
the data acquisition subunit is used for acquiring test sample data;
the accuracy rate calculating subunit is used for testing the accuracy rate of the media object identification model according to the test training sample data;
and the training subunit is used for retraining the neural network model to obtain the media object recognition model with the accuracy reaching a second preset threshold under the condition that the accuracy is lower than the second preset threshold.
Optionally, in another embodiment, on the basis of the above embodiment, the neural network model is a multilayer perceptron model; the multilayer perceptron model comprises an input layer, a hidden layer and an output layer; the hidden layer comprises N hidden layer nodes;
the first calculation unit includes:
the first calculation subunit is used for inputting the characteristic data into the input layer for calculation to obtain first data;
the second calculation subunit is configured to input the first data into the N hidden layer nodes to perform calculation, so as to obtain N second data;
the zero setting module is used for setting the second data of the hidden layer node with the preset proportion in the N hidden layer nodes as 0;
and the third computation submodule is used for outputting the data of each hidden layer node to the output layer for computation to obtain a computation result.
Optionally, in another embodiment, on the basis of the above embodiment, the model training module further includes:
and the error correction unit is used for carrying out error correction processing on the characteristic dimension of the abnormal value in the characteristic data.
Alternatively, in another embodiment, on the basis of the above embodiment, the error correction unit is configured to replace the abnormal value with a median of each value in the feature dimension.
Optionally, in another embodiment, on the basis of the above embodiment, the feature extraction unit includes:
and the coding subunit is used for coding the value of the feature data under the feature dimension of the preset type.
Optionally, in another embodiment, on the basis of the above embodiment, the characteristic dimensions include: the tag data comprises at least one of user registration time, user grades, user index scores, user credit grades, user credit scores, uploaded media object quantity, online media object quantity, industry of the user, uploaded media object channel quantity, content grades and user nicknames, and the tag data comprises historical ranking conditions of the media objects.
Optionally, in another embodiment, on the basis of the above embodiment, the historical ranking condition of the media object is a ranking in a ranking list of the media object, and the ranking number of the ranking list of the media object is smaller than a third preset value; and when the media object does not enter the ranking list of the media object, setting the historical ranking condition of the media object as the next name of the last name in the ranking list.
Optionally, in another embodiment, on the basis of the above embodiment, the ranking value determining module displays the plurality of media objects in parallel in the ranking list when the ranking values of the plurality of media objects are the same and the ranking values are in the ranking list of the media objects.
Optionally, an embodiment of the present invention further provides an apparatus, including: the computer program is executed by the processor to implement the steps of the media object sorting method described above, and can achieve the same technical effects, and the details are not repeated here in order to avoid repetition.
Optionally, an embodiment of the present invention further provides a storage medium, where a computer program is stored on the storage medium, and when the computer program is executed by a processor, the steps in the media object sorting method described above are implemented, and the same technical effect can be achieved, and in order to avoid repetition, details are not repeated here. The storage medium may be a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the true scope of the embodiments of the invention.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.
The method, apparatus, device and storage medium for ordering media objects provided by the present invention are introduced in detail, and a specific example is applied in the present document to explain the principle and the implementation of the present invention, and the description of the above embodiment is only used to help understand the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (16)
1. A method for ranking media objects, comprising:
acquiring characteristic data of a media object;
inputting the characteristic data into a media object recognition model, and determining a ranking value;
and sorting the media objects according to the sorting value.
2. The method of claim 1, prior to obtaining the characterizing data of the media object, further comprising:
acquiring training sample data of a media object; the training sample data comprises feature data and label data;
extracting feature data from the training sample data;
inputting the characteristic data into the neural network model for calculation to obtain a calculation result;
inputting the calculation result and the label data into a loss function to determine a loss value;
under the condition that the loss value does not reach a first preset threshold value, parameters in the neural network model are adjusted according to an optimizer, and sample characteristic data are continuously input for calculation;
determining the neural network model as a media object recognition model if the loss value reaches a first preset threshold.
3. The method of claim 2, wherein after determining the neural network model as a media object recognition model if the loss value reaches a preset threshold, further comprising:
obtaining test sample data;
testing the accuracy of the media object recognition model according to the test training sample data;
and under the condition that the accuracy is lower than a second preset threshold, retraining the neural network model to obtain a media object recognition model with the accuracy reaching the second preset threshold.
4. The method according to claim 2, wherein after extracting feature data from the training sample data, comprising:
and encoding the value of the feature data under the feature dimension of the preset type.
5. The method according to any of claims 2-4, wherein the characterization data comprises: the system comprises at least one item of user registration time, user grades, user index scores, user credit grades, user credit scores, uploaded media object quantity, online media object quantity, industry of the user, uploaded media object channel quantity, content grades and user nicknames, wherein the label data is the ranking of the media objects in a media object ranking list.
6. The method of claim 5, wherein the number of rankings in the media object leaderboard is less than a third preset threshold; and when the media object does not enter the ranking list of the media object, setting the ranking of the media object in the ranking list of the media object to be next to the last ranking in the ranking list.
7. The method of claim 5, wherein said sorting the media objects according to the sorting value comprises:
and when the ranking values of a plurality of media objects are the same and the ranking values are in the ranking list of the media objects, the plurality of media objects are displayed in parallel in the ranking list.
8. An apparatus for ranking media objects, comprising:
the characteristic data acquisition module is used for acquiring the characteristic data of the media object;
the sequence value determining module is used for inputting the characteristic data into a media object recognition model and determining a sequence value;
and the sequencing module is used for sequencing the media objects according to the sequencing values.
9. The apparatus of claim 8, further comprising a model training module, the model training module comprising:
the training sample data acquisition unit is used for acquiring training sample data of the media object, wherein the training sample data comprises characteristic data and label data;
the characteristic extraction unit is used for extracting characteristic data from the training sample data;
the first calculation unit is used for inputting the characteristic data into the neural network model for calculation to obtain a calculation result;
a loss value determining unit, configured to input the calculation result and the tag data into a loss function, and determine a loss value;
the second calculation unit is used for adjusting parameters in the neural network model according to the optimizer and continuously inputting sample characteristic data to calculate under the condition that the loss value does not reach a first preset threshold value;
and the model generation unit is used for determining the neural network model as a media object recognition model under the condition that the loss value reaches a first preset threshold value.
10. The apparatus of claim 9, wherein the model training module further comprises a testing unit for testing the media object recognition model determined by the model generation unit;
the test unit includes:
the data acquisition subunit is used for acquiring test sample data;
the accuracy rate calculating subunit is used for testing the accuracy rate of the media object identification model according to the test training sample data;
and the training subunit is used for retraining the neural network model to obtain the media object recognition model with the accuracy reaching a second preset threshold under the condition that the accuracy is lower than the second preset threshold.
11. The apparatus of claim 9, wherein the feature extraction unit comprises:
and the coding subunit is used for coding the value of the characteristic dimension under the characteristic dimension of the preset type.
12. The apparatus according to any one of claims 9 to 11, wherein the feature data comprises: the method comprises the steps of obtaining user registration time, user grades, user index scores, user credit grades, user credit scores, uploaded media object quantity, online media object quantity, industry where the user is located, uploaded media object channel quantity, content grades and user nicknames, wherein label data is at least one of ranking of the media objects in a ranking list of the media objects.
13. The apparatus of claim 12, wherein the number of rankings in the media object leaderboard is less than a third preset value; and when the media object does not enter the ranking list of the media object, setting the ranking of the media object in the ranking list of the media object to be next to the last ranking in the ranking list.
14. The apparatus of claim 12, wherein the ranking value determination module is configured to display the plurality of media objects in a ranking list in parallel if the ranking values of the plurality of media objects are the same and the ranking values are in the ranking list of media objects.
15. An apparatus, comprising: memory, processor and computer program stored on the memory and executable on the processor, which computer program, when executed by the processor, carries out the steps of the media object ranking method according to any of claims 1 to 7.
16. A storage medium, characterized in that the storage medium has stored thereon a computer program which, when being executed by a processor, carries out the steps in the method of ranking media objects according to any of the claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910779717.3A CN110633376A (en) | 2019-08-22 | 2019-08-22 | Media object sorting method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910779717.3A CN110633376A (en) | 2019-08-22 | 2019-08-22 | Media object sorting method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110633376A true CN110633376A (en) | 2019-12-31 |
Family
ID=68970469
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910779717.3A Pending CN110633376A (en) | 2019-08-22 | 2019-08-22 | Media object sorting method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110633376A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111597179A (en) * | 2020-05-18 | 2020-08-28 | 北京思特奇信息技术股份有限公司 | Method and device for automatically cleaning data, electronic equipment and storage medium |
| CN111984866A (en) * | 2020-08-20 | 2020-11-24 | 北京奇艺世纪科技有限公司 | Ranking list generation method and device for data |
| CN112650952A (en) * | 2020-12-25 | 2021-04-13 | 北京达佳互联信息技术有限公司 | Search sorting method and device |
| CN113554048A (en) * | 2020-04-26 | 2021-10-26 | 中移(成都)信息通信科技有限公司 | Data identification method, device, equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104834731A (en) * | 2015-05-15 | 2015-08-12 | 百度在线网络技术（北京）有限公司 | Recommendation method and device for self-media information |
| CN105824946A (en) * | 2016-03-21 | 2016-08-03 | 乐视网信息技术（北京）股份有限公司 | Method and system for multimedia recommendation on basis of data grading |
| US20170085929A1 (en) * | 2015-09-18 | 2017-03-23 | Spotify Ab | Systems, methods, and computer products for recommending media suitable for a designated style of use |
| CN108494851A (en) * | 2018-03-21 | 2018-09-04 | 广东欧珀移动通信有限公司 | Application program recommends method, apparatus and mobile terminal |
| CN109062963A (en) * | 2018-06-27 | 2018-12-21 | 广州优视网络科技有限公司 | From Media Recommendation Method, device and electronic equipment |
-
2019
- 2019-08-22 CN CN201910779717.3A patent/CN110633376A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104834731A (en) * | 2015-05-15 | 2015-08-12 | 百度在线网络技术（北京）有限公司 | Recommendation method and device for self-media information |
| US20170085929A1 (en) * | 2015-09-18 | 2017-03-23 | Spotify Ab | Systems, methods, and computer products for recommending media suitable for a designated style of use |
| CN105824946A (en) * | 2016-03-21 | 2016-08-03 | 乐视网信息技术（北京）股份有限公司 | Method and system for multimedia recommendation on basis of data grading |
| CN108494851A (en) * | 2018-03-21 | 2018-09-04 | 广东欧珀移动通信有限公司 | Application program recommends method, apparatus and mobile terminal |
| CN109062963A (en) * | 2018-06-27 | 2018-12-21 | 广州优视网络科技有限公司 | From Media Recommendation Method, device and electronic equipment |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113554048A (en) * | 2020-04-26 | 2021-10-26 | 中移(成都)信息通信科技有限公司 | Data identification method, device, equipment and storage medium |
| CN113554048B (en) * | 2020-04-26 | 2024-02-02 | 中移(成都)信息通信科技有限公司 | Data identification method, device, equipment and storage medium |
| CN111597179A (en) * | 2020-05-18 | 2020-08-28 | 北京思特奇信息技术股份有限公司 | Method and device for automatically cleaning data, electronic equipment and storage medium |
| CN111597179B (en) * | 2020-05-18 | 2023-12-05 | 北京思特奇信息技术股份有限公司 | Method and device for automatically cleaning data, electronic equipment and storage medium |
| CN111984866A (en) * | 2020-08-20 | 2020-11-24 | 北京奇艺世纪科技有限公司 | Ranking list generation method and device for data |
| CN111984866B (en) * | 2020-08-20 | 2023-09-05 | 北京奇艺世纪科技有限公司 | Ranking list generation method and device for data |
| CN112650952A (en) * | 2020-12-25 | 2021-04-13 | 北京达佳互联信息技术有限公司 | Search sorting method and device |
| CN112650952B (en) * | 2020-12-25 | 2023-07-25 | 北京达佳互联信息技术有限公司 | Search ordering method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110633376A (en) | Media object sorting method, device, equipment and storage medium | |
| CN109408731B (en) | Multi-target recommendation method, multi-target recommendation model generation method and device | |
| CN108829822B (en) | Media content recommendation method and device, storage medium and electronic device | |
| CN107274020B (en) | Learner subject total measured result prediction system and method based on collaborative filtering thought | |
| CN112699283B (en) | Test paper generation method and device | |
| CN111160606B (en) | Test question difficulty prediction method and related device | |
| CN110334275B (en) | News popularity prediction method, equipment and storage medium | |
| CN109816438A (en) | Information-pushing method and device | |
| CN113282832A (en) | Search information recommendation method and device, electronic equipment and storage medium | |
| Ikawati et al. | Student behavior analysis to predict learning styles based felder silverman model using ensemble tree method | |
| CN114169869A (en) | Attention mechanism-based post recommendation method and device | |
| KR20220098698A (en) | Learning content recommendation system that predicts the user's correct answer probability using collaborative filtering based on latent factors and operation method thereof | |
| CN114742319A (en) | Method, system and storage medium for predicting scores of law examination objective questions | |
| CN113987161A (en) | Text sorting method and device | |
| CN117540336A (en) | Time sequence prediction method and device and electronic equipment | |
| CN110415081A (en) | A kind of matching recommended method of the user individual product based on content | |
| CN115600017A (en) | Feature coding model training method and device and media object recommendation method and device | |
| KR20200142871A (en) | Method and apparatus for recommending items using explicit and implicit feedback | |
| CN113312445B (en) | Data processing method, model construction method, classification method and computing equipment | |
| CN116956171A (en) | Classification method, device, equipment and storage medium based on AI model | |
| CN110069783B (en) | Answer content evaluation method and device | |
| CN112365302B (en) | Product recommendation network training method, device, equipment and medium | |
| CN114186646A (en) | Block chain abnormal transaction identification method and device, storage medium and electronic equipment | |
| CN112069806B (en) | Resume screening method and device, electronic equipment and storage medium | |
| CN114138932A (en) | Method, device and equipment for determining explanatory information and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191231 |
|
| RJ01 | Rejection of invention patent application after publication |