CN113407763B - Hot music mining method, electronic device and computer readable storage medium - Google Patents
Hot music mining method, electronic device and computer readable storage medium Download PDFInfo
- Publication number
- CN113407763B CN113407763B CN202110707822.3A CN202110707822A CN113407763B CN 113407763 B CN113407763 B CN 113407763B CN 202110707822 A CN202110707822 A CN 202110707822A CN 113407763 B CN113407763 B CN 113407763B
- Authority
- CN
- China
- Prior art keywords
- music
- seed
- user
- feature vector
- mined
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000005065 mining Methods 0.000 title claims abstract description 51
- 239000013598 vector Substances 0.000 claims abstract description 105
- 238000003062 neural network model Methods 0.000 claims abstract description 37
- 230000006399 behavior Effects 0.000 claims description 87
- 238000012549 training Methods 0.000 claims description 48
- 238000013145 classification model Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 description 19
- 238000012545 processing Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000009471 action Effects 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000005474 detonation Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/63—Querying
- G06F16/635—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/68—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/683—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/68—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/686—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, title or artist information, time, location or usage information, user ratings
-
- 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
-
- 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)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Library & Information Science (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
本申请公开了一种热点音乐挖掘方法、装置及一种电子设备和计算机可读存储介质，该方法包括：在音乐平台的所有用户中确定种子用户，并基于种子用户的操作行为数据构建第一特征向量；基于种子用户的操作行为数据确定待挖掘音乐，并基于待挖掘音乐的基本信息构建第二特征向量；对第一特征向量和第二特征向量进行拼接得到待挖掘音乐的目标特征向量；将目标特征向量输入训练完成的神经网络模型中，得到潜力热点音乐的预测结果。本申请提供的热点音乐挖掘方法，提高了热点音乐挖掘的准确性。
The present application discloses a hot music mining method, device, electronic device and computer-readable storage medium, the method comprising: determining a seed user among all users of a music platform, and constructing a first feature vector based on the operation behavior data of the seed user; determining the music to be mined based on the operation behavior data of the seed user, and constructing a second feature vector based on the basic information of the music to be mined; splicing the first feature vector and the second feature vector to obtain a target feature vector of the music to be mined; inputting the target feature vector into a trained neural network model to obtain a prediction result of potential hot music. The hot music mining method provided by the present application improves the accuracy of hot music mining.
Description
Technical Field
The present application relates to the field of hotspot mining technology, and more particularly, to a hotspot music mining method, an electronic device, and a computer-readable storage medium.
Background
In the related art, the hot spot mining technology is mainly applied to the fields of videos, texts and the like, statistics and monitoring are carried out based on the content, and the change of the heat of related words can only be monitored passively, so that the mining of the content of musical compositions cannot be carried out, and the content which is cold at present but can become hot in the future cannot be mined.
Therefore, how to improve the accuracy of mining hot music is a technical problem that needs to be solved by those skilled in the art.
Disclosure of Invention
The application aims to provide a hot music mining method, electronic equipment and a computer readable storage medium, which improve the accuracy of hot music mining.
In order to achieve the above object, a first aspect of the present application provides a hot music mining method, including:
determining a seed user in all users of a music platform, and constructing a first feature vector based on operation behavior data of the seed user;
determining music to be mined based on the operation behavior data of the seed user, and constructing a second feature vector based on the basic information of the music to be mined;
splicing the first feature vector and the second feature vector to obtain a target feature vector of the music to be mined;
And inputting the target feature vector into a trained neural network model to obtain a prediction result of the potential hot spot music.
To achieve the above object, a second aspect of the present application provides an electronic device, including:
a memory for storing a computer program;
and the processor is used for realizing the steps of the hot music mining method when executing the computer program.
To achieve the above object, a third aspect of the present application provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of a hot music mining method as described above.
According to the scheme, the hot spot music mining method comprises the steps of determining a seed user in all users of a music platform, constructing a first feature vector based on operation behavior data of the seed user, determining music to be mined based on the operation behavior data of the seed user, constructing a second feature vector based on basic information of the music to be mined, splicing the first feature vector and the second feature vector to obtain a target feature vector of the music to be mined, and inputting the target feature vector into a trained neural network model to obtain a prediction result of potential hot spot music. According to the hot spot music mining method provided by the application, the seed users are firstly screened from the music platform, and the probability that the music to be mined is determined to be potential hot spot music based on the seed operation behavior data is larger because the appreciation capability or influence of the seed users on the music is stronger. And constructing a target feature vector based on the operation behavior data of the seed user to be mined and the basic information of the music to be mined, and inputting the target feature vector into a neural network model to obtain potential hot music. The construction of the target feature vector based on the operation behavior data of the seed user ensures that music has a certain propagation degree and a detonation point, the construction of the target feature vector based on the basic information of the music to be mined ensures that the music production quality is high enough, and the combination of the two can improve the accuracy of mining hot music. The application also discloses an electronic device and a computer readable storage medium, which can realize the technical effects.
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 as claimed.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. The accompanying drawings are included to provide a further understanding of the disclosure, and are incorporated in and constitute a part of this specification, illustrate the disclosure and together with the description serve to explain, but do not limit the disclosure. In the drawings:
fig. 1 is a schematic diagram of a hotspot music mining system according to an embodiment of the present application;
Fig. 2 is a flowchart of a hot music mining method according to an embodiment of the present application;
fig. 3 is a flowchart of another hot music mining method according to an embodiment of the present application;
fig. 4 is a flowchart of another hotspot music mining method according to an embodiment of the present application;
Fig. 5 is a block diagram of a hot music mining apparatus according to an embodiment of the present application;
fig. 6 is a block diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In order to facilitate understanding of the hot music mining method provided by the present application, a system used by the hot music mining method will be described below. Referring to fig. 1, an architecture diagram of a hotspot music mining system according to an embodiment of the present application is shown, and as shown in fig. 1, the architecture diagram includes a server 10.
In the application, the server 10 is used for executing the hot music mining method, which comprises the steps of determining a seed user from all users of a music platform, determining music to be mined based on operation behavior data of the seed user, constructing a target feature vector of the music to be mined based on the operation behavior data and basic information of the music to be mined, and inputting the target feature vector into a trained neural network model to obtain a prediction result of potential hot music.
It may be appreciated that the server 10 may be further provided with a neural network model for predicting potential hot music, where an input of the neural network model is a target feature vector constructed based on the operation behavior data of the music to be mined by the seed user and the basic information of the music to be mined, and an output of the neural network model may be a potential value corresponding to each music to be mined, where the potential value is greater than a preset potential value, and the music to be mined is the potential hot music.
Further, the hot music mining system may further include a plurality of clients 20 that establish communication connection with the server 10, where the clients 20 may include a fixed terminal such as a PC (chinese full name: personal computer, english full name: personal Computer) and a mobile terminal such as a mobile phone, for displaying the potential hot music.
The embodiment of the application discloses a hot music mining method, which improves the accuracy of hot music mining.
Referring to fig. 2, a flowchart of a hot music mining method provided by an embodiment of the present application, as shown in fig. 2, includes:
S101, determining a seed user in all users of a music platform, and constructing a first feature vector based on operation behavior data of the seed user;
The implementation main body of the embodiment is the server, and aims to screen potential hot music in the music platform. In this step, the seed user is first determined among all users in the music platform, and since the seed user has a strong appreciation capability or influence on music, the music related to the seed user has a greater chance of becoming potential hot music.
As a possible implementation manner, the method for determining the seed users in all users of the music platform comprises the steps of determining current hot music in the music platform and a cold time period corresponding to the current hot music, and determining the user with operation behaviors on the current hot music in the cold time period as the seed user. In the present embodiment, a user who successfully predicts the current popular music, which is a user having a prospective operation behavior, is regarded as a high-quality user having a priori knowledge, and the high-quality user is regarded as a seed user. The standard of the current hot music may be music with a current click rate greater than a preset click rate, and the prospective operation behavior includes playing, commonly playing, commenting, collecting or sharing the current hot music in a cold time period of the current hot music, where the click rate of the current hot music is less than the preset click rate. In specific implementation, the current hot music is analyzed in the music platform, the corresponding cold time period is determined by tracing backwards in time, and the user with the operation behavior of the current hot music in the cold time period is the high-grade user. As a preferred embodiment, the number and duration of the aforementioned prospective operation actions may be analyzed, and the user whose number of prospective operation actions is greater than a preset number of times, or whose duration is greater than a preset time, or a combination of both may be determined as the seed user.
As another possible implementation, the determining the seed user among all users of the music platform includes determining the quality music content in the music platform and determining the user having an operation behavior on the quality music content as the seed user. In this embodiment, content (User Generated Content, UGC) produced by an ordinary user in the music platform is analyzed to screen out high-quality music content. The high-quality music content can include the number of high-quality comments produced by the user, the number of high-quality singing videos or podcasts released, and the like, and the user producing the high-quality music content or the user commenting, collecting or analyzing the music content is regarded as a high-creativity user, and the high-creativity user is regarded as a seed user.
As a further possible implementation manner, the step of determining the seed user in all users of the music platform comprises the step of determining the user with the influence greater than the preset influence in the music platform as the seed user, wherein the influence comprises the influence in the music platform and/or the influence in other platforms. In this embodiment, the influence of all users in the music platform is analyzed, and the user with the larger influence is used as the seed user. In a specific implementation, the influence of the user may be represented by the attention amount, the number of works or the interaction data of the user, where the influence may include the influence of the user in the music platform, or may include the influence of the user in other platforms, that is, in this embodiment, the influence of the user on the music platform may be analyzed, or the influence of the user on other platforms may be analyzed.
Furthermore, the embodiment may further analyze the popularity, freshness, coldness, etc. of the song listening preference of the user, so as to screen out the seed users having the opportunity to predict the popular music, which is not limited in detail.
On the basis of the step, the method is used as a preferred implementation mode, after the seed users are determined in all users of the music platform, the method further comprises the steps of constructing user characteristics of the seed users based on user information and music preference portraits of the seed users, selecting non-seed training users from non-seed users, constructing user characteristics of the non-seed training users based on the user information and the music preference portraits of the non-seed training users, training a classification model by utilizing the user characteristics of the seed users and the user characteristics of the non-seed training users, inputting the user information and the music preference portraits of the non-seed users except the non-seed training users into the classification model after training to predict extended seed users, and adding the extended seed users into the seed users.
In implementations, the user characteristics are built based on user information of the user, which may include age, gender, region, school, etc., and music preference portraits for describing music preferred by the user, which may include a preferred music style, a preferred singer type, etc., of the user, without limitation. The training of the classification model is performed with the user features of the seed users as positive samples and the user features of some non-seed users as negative samples, where the classification model may include a logistic regression model, etc., and is not specifically limited herein. The trained classification model is used for predicting the rest non-seed users, and the expanded seed users with low taste, creativity and influence but comprehensive capability are added into the seed users. Furthermore, the attention relationship network and the sharing propagation network of the seed user can be analyzed, and the seed user can be expanded in the attention relationship network and the sharing propagation network of the seed user.
Next, a first feature vector is constructed based on operational behavior data of the seed user, each piece of operational behavior data including an operational behavior, the seed user performing the operational behavior, and music being operated. The method for constructing the first feature vector based on the operation behavior data of the seed user comprises the steps of constructing a network topology based on the operation behavior data of the seed user, wherein nodes in the network topology represent seed users or operated music executing the operation behaviors, edges between the nodes represent the operation behaviors, traversing the network topology according to preset rules to obtain a music sequence to be mined, and carrying out feature representation on the music sequence to be mined to obtain the first feature vector. In specific implementation, a network topology is constructed based on operation behavior data of a seed user on music, nodes of the network topology are the seed user and the music operated by the seed user, namely the music to be mined, and edges between the nodes represent operation behaviors of the seed user on the music to be mined, so that a music sequence to be mined can be determined by traversing the network topology through deepwalk or node2vec and other algorithms, wherein the music sequence comprises a plurality of identifications of the music to be mined. Further, the word2vec algorithm is used for carrying out feature representation on the music sequence to be mined to obtain a second feature vector.
S102, determining music to be mined based on the operation behavior data of the seed user, and constructing a second feature vector based on the basic information of the music to be mined;
In this step, music to be mined is determined based on the operation behavior data of the seed user, that is, music related to the seed user is determined as music to be mined, and the music to be mined may be determined based on the music sequence to be mined obtained after traversing the network topology in the previous step. The basic information of the music to be mined, which may include content knowledge maps, audio, lyrics text, etc., which may include a composer, singer, language, genre tag, etc., of the musical piece, may be characterized using a convolutional neural network (CNN, convolutionalNeuralNetworks), a transformer-based bi-directional encoder representation technique (BERT, bidirectional Encoder Representations from Transformers), a graph neural network (GNN, convolutional Neural Networks), etc., respectively, to construct a second feature vector, without being specifically limited thereto.
S103, splicing the first feature vector and the second feature vector to obtain a target feature vector of the music to be mined;
It should be noted that, constructing the first feature vector based on the operation behavior data of the seed user ensures that the music has a certain propagation degree and a detonation point, constructing the second feature vector based on the basic information of the music to be mined ensures that the music production quality is high enough, and the target feature vector obtained by splicing the first feature vector and the second feature vector can be used for evaluating whether the music is potential hot music.
And S104, inputting the target feature vector into a trained neural network model to obtain a prediction result of the potential hot spot music.
In implementations, a neural network model is used to predict potential hot music. In the step, a target feature vector constructed based on the operation behavior data of the seed user and the basic information of the music to be mined is input into a neural network model after training is completed, and a prediction result of the potential hot spot music is obtained.
As a possible implementation manner, the method can include the steps of obtaining a potential value corresponding to each piece of music to be mined from a prediction result neural network model input by the target feature vector, and determining the piece of music to be mined with the potential value being larger than a preset potential value as potential hot spot music. In a specific implementation, each row in the target feature vector corresponds to one piece of music to be mined, the neural network model is used for predicting potential values of the music to be mined, namely, the target feature vector is input into the neural network model to obtain the potential value corresponding to each piece of music to be mined, and the music to be mined with the potential value larger than a preset potential value is determined to be potential hot music.
According to the hot spot music mining method provided by the embodiment of the application, the seed users are firstly selected from the music platform, and the probability that the music to be mined is determined to be potential hot spot music based on the seed operation behavior data is larger because the appreciation capability or influence of the seed users on the music is stronger. And constructing a target feature vector based on the operation behavior data of the seed user to be mined and the basic information of the music to be mined, and inputting the target feature vector into a neural network model to obtain potential hot music. The construction of the target feature vector based on the operation behavior data of the seed user ensures that music has a certain propagation degree and a detonation point, the construction of the target feature vector based on the basic information of the music to be mined ensures that the music production quality is high enough, and the combination of the two can improve the accuracy of mining hot music.
The embodiment of the application discloses a hot music mining method, which further describes and optimizes a technical scheme relative to the previous embodiments. Specific:
Referring to fig. 3, a flowchart of another hot music mining method provided by an embodiment of the present application, as shown in fig. 3, includes:
s201, determining seed users in all users of a music platform;
S202, distributing corresponding weights for each type of operation behavior data, and constructing a network topology based on the operation behavior data of the seed user and the weights corresponding to each type of operation behavior data;
in this embodiment, a corresponding weight may be allocated to each type of operation behavior data, for example, the weight of the complete playing is greater than that of the normal playing, the shared weight is greater than that of the comment, and specific weight allocation criteria may be flexibly set by those skilled in the art according to the actual situation. And constructing a network topology based on the operation behavior data of the seed user and the weight corresponding to each type of operation behavior data, namely each side in the network topology has the corresponding weight.
S203, traversing the network topology according to the weight corresponding to each type of operation behavior data to obtain a music sequence to be mined, and carrying out feature representation on the music sequence to be mined to obtain a first feature vector;
In the step, traversing is carried out based on the weight of each edge in the network topology to obtain a music sequence to be mined, and feature representation is carried out on the music sequence to be mined so as to construct a first feature vector.
S204, determining music to be mined based on the music sequence to be mined, and constructing a second feature vector based on the basic information of the music to be mined;
S205, splicing the first feature vector and the second feature vector to obtain a target feature vector of the music to be mined;
S206, inputting the target feature vector into the trained neural network model to obtain a prediction result of the potential hot spot music.
Therefore, according to the embodiment, different weights are distributed to different types of operation behavior data, the weighted network topology can more accurately represent the operation behavior of the seed user, more accurate target feature vectors can be obtained by traversing the weighted network topology, and accordingly, the accuracy of the predicted potential hot spot music is higher.
The present embodiment describes a training process of a neural network model, specifically:
Referring to fig. 4, a flowchart of a training method of a neural network model according to an embodiment of the present application, as shown in fig. 4, includes:
S301, determining seed users in all users of a music platform, and determining current hot music, current cold music and cold time periods corresponding to the current hot music in the music platform;
In this embodiment, the current trending music in the music platform is taken as a positive sample, a certain proportion of the current trending music in the music platform is randomly extracted as a negative sample to train the neural network model, and the trained neural network model is used for predicting potential trending music.
S302, constructing training feature vectors corresponding to the current popular music based on the operation behavior data of the seed user on the current popular music and the basic information of the current popular music in the cold time period;
S303, constructing a training feature vector corresponding to the current cold music based on the operation behavior data of the seed user on the current cold music and the basic information of the current cold music;
and S304, training a neural network model by taking the training feature vector corresponding to the current hot music as a positive sample and the training feature vector corresponding to the current cold music as a negative sample, wherein the trained neural network model is used for predicting potential hot music based on the target feature vector of the music to be mined.
In specific implementation, a training feature vector corresponding to the current popular music, namely a training feature vector of a positive sample, is constructed based on the operation behavior data of the current popular music and the basic information of the current popular music of the seed user in the cold time period. And constructing a training feature vector corresponding to the current cold music, namely a training feature vector of a negative sample, based on the operation behavior data of the seed user on the current cold music and the basic information of the current cold music. And training the neural network model by utilizing training feature vectors respectively corresponding to the positive sample and the negative sample, so that the trained neural network model can predict whether the music to be mined corresponding to the input target feature vector is potential hot spot music.
The following describes a hot music mining apparatus according to an embodiment of the present application, and the hot music mining apparatus and the hot music mining method described above may be referred to each other.
Referring to fig. 5, a structure diagram of a hot music mining apparatus according to an embodiment of the present application, as shown in fig. 5, includes:
A first construction module 100, configured to determine a seed user among all users of the music platform, and construct a first feature vector based on operational behavior data of the seed user;
A second construction module 200, configured to determine music to be mined based on the operation behavior data of the seed user, and construct a second feature vector based on the basic information of the music to be mined;
The stitching module 300 is configured to stitch the first feature vector and the second feature vector to obtain a target feature vector of the music to be mined;
And the input module 400 is used for inputting the target feature vector into the trained neural network model to obtain a prediction result of the potential hot music.
According to the hot spot music mining device provided by the embodiment of the application, the seed users are firstly screened in the music platform, and the probability that the music to be mined is determined to be potential hot spot music based on the seed operation behavior data is larger because the appreciation capability or influence of the seed users on the music is stronger. And constructing a target feature vector based on the operation behavior data of the seed user to be mined and the basic information of the music to be mined, and inputting the target feature vector into a neural network model to obtain potential hot music. The construction of the target feature vector based on the operation behavior data of the seed user ensures that music has a certain propagation degree and a detonation point, the construction of the target feature vector based on the basic information of the music to be mined ensures that the music production quality is high enough, and the combination of the two can improve the accuracy of mining hot music.
On the basis of the foregoing embodiment, as a preferred implementation manner, the first construction module 100 is specifically a module for determining current hot music in a music platform and a cold time period corresponding to the current hot music, determining a user having an operation behavior on the current hot music in the cold time period as a seed user, and constructing a first feature vector based on operation behavior data of the seed user;
on the basis of the above embodiment, as a preferred implementation manner, the first construction module 100 is specifically a module for determining high-quality music content in a music platform, determining a user having an operation behavior on the high-quality music content as a seed user, and constructing a first feature vector based on operation behavior data of the seed user;
On the basis of the foregoing embodiment, as a preferred implementation manner, the first construction module 100 specifically determines a user with an influence greater than a preset influence in a music platform as a seed user, and constructs a first feature vector based on operation behavior data of the seed user, where the influence includes the influence in the music platform and/or the influence in other platforms.
On the basis of the above embodiment, as a preferred implementation manner, the method further includes:
A third construction module for constructing user characteristics of the seed user based on the user information of the seed user and the music preference portrait;
a fourth construction module, configured to select a non-seed training user from among non-seed users, and construct user features of the non-seed training user based on user information and music preference portraits of the non-seed training user;
the first training module is used for training a classification model by utilizing the user characteristics of the seed users and the user characteristics of the non-seed training users;
a prediction module for inputting user information of non-seed users except the non-seed training users and music preference portraits into the trained classification model to predict extended seed users;
and the adding module is used for adding the expanded seed users into the seed users.
On the basis of the above embodiment, as a preferred implementation manner, the first building module 100 includes:
A determining unit for determining a seed user among all users of the music platform;
The system comprises a construction unit, a storage unit and a processing unit, wherein the construction unit is used for constructing a network topology based on operation behavior data of the seed users, each piece of operation behavior data comprises operation behaviors, the seed users executing the operation behaviors and operated music, nodes in the network topology represent the seed users executing the operation behaviors or the operated music, and edges between the nodes represent the operation behaviors;
the traversing unit is used for traversing the network topology according to a preset rule to obtain a music sequence to be mined, and carrying out feature representation on the music sequence to be mined to obtain a first feature vector.
On the basis of the foregoing embodiment, as a preferred implementation manner, the second construction module 200 is specifically a module that determines music to be mined based on the sequence of music to be mined, and constructs a second feature vector based on the basic information of the music to be mined.
On the basis of the above embodiment, as a preferred implementation manner, the construction unit specifically allocates a corresponding weight for each type of operation behavior data, constructs a network topology based on the operation behavior data of the seed user and the weight corresponding to each type of operation behavior data, and determines a unit to be mined according to the network topology;
correspondingly, the traversing unit is specifically a unit for traversing the network topology according to the weight corresponding to each type of operation behavior data to obtain a music sequence to be mined, and carrying out feature representation on the music sequence to be mined to obtain a first feature vector.
On the basis of the foregoing embodiment, as a preferred implementation manner, the input module 400 is specifically a module that inputs the target feature vector into a trained neural network model to obtain a potential value corresponding to each piece of music to be mined, and determines the piece of music to be mined with the potential value greater than a preset potential value as potential hot music.
On the basis of the above embodiment, as a preferred implementation manner, the method further includes:
the determining module is used for determining seed users among all users of the music platform and determining current hot music, current cold music and cold time periods corresponding to the current hot music in the music platform;
A fifth construction module, configured to construct a training feature vector corresponding to the current popular music based on the operation behavior data of the seed user on the current popular music and the basic information of the current popular music in the cold time period;
A sixth construction module, configured to construct a training feature vector corresponding to the current cold music based on the operation behavior data of the seed user on the current cold music and the basic information of the current cold music;
the second training module is used for taking the training feature vector corresponding to the current hot music as a positive sample and the training feature vector corresponding to the current cold music as a negative sample to train the neural network model, and the trained neural network model is used for predicting potential hot music based on the target feature vector of the music to be mined.
The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.
The present application also provides an electronic device, referring to fig. 6, and a block diagram of an electronic device 60 provided in an embodiment of the present application, as shown in fig. 6, may include a processor 61 and a memory 62.
Processor 61 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 61 may be implemented in at least one hardware form of DSP (DIGITAL SIGNAL Processing), FPGA (Field-Programmable gate array), PLA (Programmable Logic Array ). The processor 61 may also include a main processor, which is a processor for processing data in a wake-up state, also called a CPU (Central Processing Unit ), and a coprocessor, which is a low-power processor for processing data in a standby state. In some embodiments, the processor 61 may integrate a GPU (Graphics Processing Unit, image processor) for rendering and drawing of content required to be displayed by the display screen. In some embodiments, the processor 61 may also include an AI (ARTIFICIAL INTELLIGENCE ) processor for processing computing operations related to machine learning.
Memory 62 may include one or more computer-readable storage media, which may be non-transitory. Memory 62 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 this embodiment, the memory 62 is at least used for storing a computer program 621, where the computer program, after being loaded and executed by the processor 61, can implement relevant steps in the hot spot music mining method performed by the server side as disclosed in any of the foregoing embodiments. In addition, the resources stored by the memory 62 may also include an operating system 622, data 623, and the like, and the storage manner may be transient storage or permanent storage. Wherein the operating system 622 may include Windows, unix, linux, etc.
In some embodiments, the electronic device 60 may further include a display 63, an input-output interface 64, a communication interface 65, a sensor 66, a power supply 67, and a communication bus 68.
Of course, the structure of the electronic device shown in fig. 6 is not limited to the electronic device in the embodiment of the present application, and the electronic device may include more or fewer components than those shown in fig. 6 or may combine some components in practical applications.
In another exemplary embodiment, a computer readable storage medium is also provided, comprising program instructions which, when executed by a processor, implement the steps of the hotspot music mining method performed by the electronic device of any of the embodiments described above.
In the description, each embodiment is described in a progressive manner, and each embodiment is mainly described by the differences from other embodiments, so that the same similar parts among the embodiments are mutually referred. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section. It should be noted that it will be apparent to those skilled in the art that various modifications and adaptations of the application can be made without departing from the principles of the application and these modifications and adaptations are intended to be within the scope of the application as defined in the following claims.
It should also be noted that in this specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises an element.
Claims (9)
1. A method of hotspot music mining, comprising:
determining a seed user in all users of a music platform, and constructing a first feature vector based on operation behavior data of the seed user;
determining music to be mined based on the operation behavior data of the seed user, and constructing a second feature vector based on the basic information of the music to be mined;
splicing the first feature vector and the second feature vector to obtain a target feature vector of the music to be mined;
inputting the target feature vector into a trained neural network model to obtain a prediction result of potential hot music;
the constructing a first feature vector based on the operation behavior data of the seed user includes:
Constructing a network topology based on the operation behavior data of the seed users, wherein each piece of operation behavior data comprises operation behaviors, the seed users executing the operation behaviors and operated music, nodes in the network topology represent the seed users executing the operation behaviors or the operated music, the operated music is music to be mined, and edges between the nodes represent the operation behaviors;
Traversing the network topology according to a preset rule to obtain a music sequence to be mined, and carrying out feature representation on the music sequence to be mined to obtain a first feature vector.
2. The hotspot music mining method of claim 1, wherein after determining the seed user among all users of the music platform, further comprising:
Constructing user features of the seed user based on user information of the seed user and the music preference portrait;
Selecting a non-seed training user from non-seed users, and constructing user characteristics of the non-seed training users based on user information and music preference portraits of the non-seed training users;
Training a classification model using the user characteristics of the seed users and the user characteristics of the non-seed training users;
inputting user information of non-seed users except the non-seed training users and music preference portraits into a trained classification model to predict extended seed users;
and adding the expanded seed users into the seed users.
3. The hotspot music mining method of claim 1, wherein the determining a seed user among all users of the music platform comprises:
determining current hot music and a cold time period corresponding to the current hot music in a music platform, and determining a user with operation behaviors on the current hot music in the cold time period as a seed user;
And/or determining high-quality music content in a music platform, and determining a user with operation behaviors on the high-quality music content as a seed user;
And/or determining the user with the influence greater than the preset influence in the music platform as a seed user, wherein the influence comprises the influence in the music platform and/or the influence in other platforms.
4. The hotspot music mining method according to claim 1, wherein the determining music to be mined based on the operational behavior data of the seed user comprises:
And determining the music to be mined based on the music sequence to be mined.
5. The hotspot music mining method of claim 1, wherein the constructing a network topology based on operational behavior data of the seed user comprises:
distributing corresponding weights for each type of operation behavior data, and constructing a network topology based on the operation behavior data of the seed user and the weights corresponding to each type of operation behavior data;
Correspondingly, traversing the network topology according to a preset rule to obtain a music sequence to be mined, including:
Traversing the network topology according to the weight corresponding to each type of operation behavior data to obtain a music sequence to be mined.
6. The method for mining hot music according to claim 1, wherein inputting the target feature vector into a trained neural network model to obtain a predicted result of potential hot music comprises:
Inputting the target feature vector into a trained neural network model to obtain a potential value corresponding to each piece of music to be mined;
and determining the music to be mined with the potential value larger than a preset potential value as potential hot music.
7. A method for training a neural network model, comprising:
determining seed users in all users of a music platform, and determining current hot music, current cold music and cold time periods corresponding to the current hot music in the music platform;
constructing training feature vectors corresponding to the current popular music based on the operation behavior data of the seed user on the current popular music and the basic information of the current popular music in the cold time period;
Constructing training feature vectors corresponding to the current cold music based on the operation behavior data of the seed user on the current cold music and the basic information of the current cold music;
and training a neural network model by taking the training feature vector corresponding to the current hot music as a positive sample and the training feature vector corresponding to the current cold music as a negative sample, wherein the trained neural network model is used for predicting potential hot music based on the target feature vector of the music to be mined.
8. An electronic device, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the hotspot music mining method according to any one of claims 1 to 6 or the training method of the neural network model according to claim 7 when executing the computer program.
9. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the hot spot music mining method according to any one of claims 1 to 6 or the neural network model training method according to claim 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110707822.3A CN113407763B (en) | 2021-06-24 | 2021-06-24 | Hot music mining method, electronic device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110707822.3A CN113407763B (en) | 2021-06-24 | 2021-06-24 | Hot music mining method, electronic device and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113407763A CN113407763A (en) | 2021-09-17 |
| CN113407763B true CN113407763B (en) | 2025-04-04 |
Family
ID=77683153
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110707822.3A Active CN113407763B (en) | 2021-06-24 | 2021-06-24 | Hot music mining method, electronic device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113407763B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114491241B (en) * | 2022-01-04 | 2025-03-21 | 杭州网易云音乐科技有限公司 | Preference calculation method, device, electronic device and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107766449A (en) * | 2017-09-26 | 2018-03-06 | 杭州云赢网络科技有限公司 | Focus method for digging, device, electronic equipment and storage medium |
| CN112667714A (en) * | 2021-03-17 | 2021-04-16 | 腾讯科技（深圳）有限公司 | User portrait optimization method and device based on deep learning and storage medium |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015016784A1 (en) * | 2013-08-01 | 2015-02-05 | National University Of Singapore | A method and apparatus for tracking microblog messages for relevancy to an entity identifiable by an associated text and an image |
| US10636320B2 (en) * | 2017-11-20 | 2020-04-28 | International Business Machines Corporation | Musical instrument tutor system |
| EP3776477A4 (en) * | 2018-04-09 | 2022-01-26 | Nokia Technologies Oy | An apparatus, a method and a computer program for video coding and decoding |
| CN109903086B (en) * | 2019-02-14 | 2020-12-18 | 北京奇艺世纪科技有限公司 | Similar crowd expansion method and device and electronic equipment |
| CN110457387B (en) * | 2019-08-19 | 2023-11-10 | 腾讯科技（深圳）有限公司 | Method and related device applied to user tag determination in network |
| CN112925973B (en) * | 2019-12-06 | 2024-06-18 | 北京沃东天骏信息技术有限公司 | Data processing method and device |
| CN111144936B (en) * | 2019-12-19 | 2024-01-05 | 有米科技股份有限公司 | Similar crowd expansion method and device based on user labels |
| CN111444379B (en) * | 2020-03-30 | 2023-08-08 | 腾讯科技（深圳）有限公司 | Audio feature vector generation method and audio fragment representation model training method |
| CN112231511B (en) * | 2020-10-20 | 2025-06-13 | 腾讯音乐娱乐科技（深圳）有限公司 | A neural network model training method and song mining method and device |
| CN112950276B (en) * | 2021-03-12 | 2022-03-29 | 华扬联众数字技术(深圳)有限公司 | Seed population expansion method based on multi-order feature combination |
-
2021
- 2021-06-24 CN CN202110707822.3A patent/CN113407763B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107766449A (en) * | 2017-09-26 | 2018-03-06 | 杭州云赢网络科技有限公司 | Focus method for digging, device, electronic equipment and storage medium |
| CN112667714A (en) * | 2021-03-17 | 2021-04-16 | 腾讯科技（深圳）有限公司 | User portrait optimization method and device based on deep learning and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113407763A (en) | 2021-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113261003B (en) | Generate integrated circuit floorplans using neural networks | |
| CN111046286B (en) | Object recommendation method and device and computer storage medium | |
| US20200410392A1 (en) | Task-aware command recommendation and proactive help | |
| CN111242310B (en) | Feature validity evaluation method and device, electronic equipment and storage medium | |
| JP7488871B2 (en) | Dialogue recommendation method, device, electronic device, storage medium, and computer program | |
| CN112489676B (en) | Model training method, device, equipment and storage medium | |
| US9552399B1 (en) | Displaying information about distributed group discussions | |
| JP7498248B2 (en) | Content recommendation and sorting model training method, device, equipment, storage medium, and computer program | |
| CN105637507A (en) | Smart selection of text spans | |
| CN108875090A (en) | A kind of song recommendations method, apparatus and storage medium | |
| US10740397B2 (en) | User modelling and metadata of transmedia content data | |
| CN110175678A (en) | System and method for simulating complicated intensified learning environment | |
| CN119721259B (en) | Model reasoning method and device based on visual language model | |
| CN113407763B (en) | Hot music mining method, electronic device and computer readable storage medium | |
| Naugle et al. | Feedback density and causal complexity of simulation model structure | |
| CN113077815B (en) | Audio evaluation method and assembly | |
| CN115905616A (en) | Short video multi-behavior click prediction method based on multi-task learning | |
| Hong | How much is a “feedback” worth? User engagement and interaction for computer-supported adaptive quizzing | |
| CN113761193A (en) | Log classification method, apparatus, computer equipment and storage medium | |
| CN118585668A (en) | Song recommendation model training method, computer device and storage medium | |
| Berjawi et al. | Understanding radicalization pathways: a framework for assessing diversity in YouTube recommendation systems | |
| CN112329943A (en) | Combined index selection method and device, computer equipment and medium | |
| CN118505320B (en) | Data processing method, device, product and medium | |
| WO2022199261A1 (en) | Model recommendation method and apparatus, and computer device | |
| CN111931058B (en) | Sequence recommendation method and system based on self-adaptive network depth |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |