CN108388652B

CN108388652B - Method and device for sending song list identification

Info

Publication number: CN108388652B
Application number: CN201810169174.9A
Authority: CN
Inventors: 黄昕
Original assignee: Tencent Music Entertainment Technology Shenzhen Co Ltd
Current assignee: Tencent Music Entertainment Technology Shenzhen Co Ltd
Priority date: 2018-02-28
Filing date: 2018-02-28
Publication date: 2021-01-12
Anticipated expiration: 2038-02-28
Also published as: CN108388652A

Abstract

The embodiment of the invention discloses a method and a device for sending a song list identification, belonging to the technical field of computers. The method comprises the following steps: when detecting that a song menu recommending trigger event corresponding to a first account occurs, determining account information corresponding to the first account; determining singing list information corresponding to the singing list for each singing list in the candidate singing list set; inputting account information corresponding to the first account and song list information corresponding to the song list into a pre-trained scoring model, and predicting the scoring of the first account on the song list; determining a target song list with a corresponding score meeting a preset recommendation condition in each song list in the candidate song list set; and sending the song list identification of the target song list to a login terminal of the first account. By adopting the invention, the efficiency of selecting the song list can be improved.

Description

Method and device for sending song list identification

Technical Field

The invention relates to the technical field of computers, in particular to a method and a device for sending a song list identifier.

Background

With the development of computer technology, terminals such as mobile phones and computers are widely used, and the types and functions of application programs on the corresponding terminals are more and more abundant. A song playing application is a very common application.

The user may play the song through the song play class application. Currently, song playing applications often provide some song lists, wherein the song lists contain a plurality of songs. When a user wants to play songs through the song playing application program, the user can select a song list which the user likes to listen to in a song list provided by the song playing application program according to information such as labels, titles, brief introduction and the like of the song list, and then can click a playing button to trigger the terminal to play the songs in the selected song list.

In the process of implementing the invention, the inventor finds that the prior art has at least the following problems:

based on the above processing method, each time the user wants to select the song list, the user needs to select the song list that the user likes to listen to from the song list provided by the song playing application program, and the number of the song lists in the song list is often large, so that the efficiency of selecting the song list is low.

Disclosure of Invention

In order to solve the problem of low efficiency of selecting a song list in the related art, the embodiment of the invention provides a method and a device for sending a song list identifier. The technical scheme is as follows:

in a first aspect, a method for transmitting a song list identifier is provided, the method comprising:

when detecting that a song menu recommending trigger event corresponding to a first account occurs, determining account information corresponding to the first account;

determining singing list information corresponding to the singing list for each singing list in the candidate singing list set; inputting account information corresponding to the first account and song list information corresponding to the song list into a pre-trained scoring model, and predicting the scoring of the first account on the song list;

determining a target song list with a corresponding score meeting a preset recommendation condition in each song list in the candidate song list set;

and sending the song list identification of the target song list to a login terminal of the first account.

Optionally, when it is detected that the song list recommendation triggering event corresponding to the first account occurs, determining account information corresponding to the first account includes:

when detecting that a song list recommendation triggering event corresponding to a first account occurs, determining account attribute information corresponding to the first account according to basic attribute information corresponding to the first account and song preference data concerned;

determining account characteristic information corresponding to the first account;

and determining account attribute information corresponding to the first account and account feature information corresponding to the first account as account information corresponding to the first account.

Optionally, the determining the account characteristic information corresponding to the first account includes:

acquiring a first concerned song list corresponding to the first account;

determining song characteristic information corresponding to each song in the first concerned song list from the prestored song characteristic information corresponding to each song;

and determining account characteristic information corresponding to the first account according to song characteristic information corresponding to each song in the first concerned song list.

Optionally, the determining, according to song feature information corresponding to each song in the first interest song list, account feature information corresponding to the first account includes:

determining a type of interest to the first account for each song in the first list of songs of interest; determining the attention duration of each song in the first attention song list from the first account to the current moment;

determining an attention weight corresponding to the attention type of each song in the first attention song list by the first account according to the pre-stored corresponding relationship between different attention types and the attention weight; determining the time attenuation degree corresponding to each song according to a time attenuation formula and the attention duration of the attention moment of the first account to each song in the first attention song list from the current moment;

and taking the product of the attention weight corresponding to the attention type of each song and the time attenuation corresponding to each song as the weight of the song characteristic information corresponding to each song, and calculating the weighted sum of the song characteristic information corresponding to each song in the first attention song list to obtain the account characteristic information corresponding to the first account.

Optionally, the determining, for each song list in the candidate song list set, song list information corresponding to the song list includes:

for each song list in the candidate song list set, determining the song list characteristic information corresponding to the song list in the song list characteristic information corresponding to each pre-stored song list;

determining attribute information of the song list corresponding to the song list according to the concerned information corresponding to the song list;

and determining the attribute information of the song list corresponding to the song list and the characteristic information of the song list corresponding to the song list as the song list information corresponding to the song list.

Optionally, the method further includes:

determining identification information of each song in a song library;

inputting the identification information of each song in the song library into a pre-trained song feature extraction model, and determining the song feature information of each song in the song library;

for each pre-stored song list, determining song characteristic information corresponding to each song included in the song list; and determining the singing sheet characteristic information corresponding to the singing sheet according to the song characteristic information corresponding to each song included in the singing sheet.

Optionally, the method further includes:

for each song contained in each prestored song list, acquiring identification information of the song, identification information of a first preset number of songs in front of the song and identification information of a second preset number of songs behind the song, respectively taking the identification information of the song as model input data, taking the identification information of the first preset number of songs and the identification information of the second preset number of songs as model output data, and storing the model output data and the identification information into a corresponding relation between the model input data and the model output data in a training set;

and training the song feature extraction model containing undetermined parameters according to the corresponding relation between the model input data and the model output data in the training set and on the basis of a training principle that the model output data obtained by the song feature extraction model containing undetermined parameters approaches the model output data in the corresponding relation to obtain the song feature extraction model.

Optionally, the method further includes:

and training the scoring model containing undetermined parameters according to the scoring of each account in the training set on each song bill, the account information corresponding to each account and the song bill information corresponding to each song bill, and on the basis of a training principle that the scoring of the account on the song bill predicted by the scoring model containing undetermined parameters approaches to the scoring of the account on the song bill in the training set, so as to obtain the scoring model.

Optionally, the method further includes:

for each account, acquiring an attention song list corresponding to the account; counting the number of songs with different genre types in the concerned song list corresponding to the account according to the concerned song list corresponding to the account, and calculating the proportion of the number of the songs with different genre types in the concerned song list corresponding to the account in the total number of the songs in the concerned song list corresponding to the account to obtain the proportion of the songs with different genre types corresponding to the account;

selecting a target account with the corresponding song occupation ratio of a preset genre type reaching a preset occupation ratio threshold value from the accounts;

determining account information corresponding to a target account;

for each song in the concerned song list corresponding to the target account, determining song information of the song, inputting the account information of the target account and the song information of the song into a pre-trained scoring model, and predicting the score of the song by the target account;

determining a third preset number of songs with the maximum score in the songs in the concerned song list corresponding to the target account;

creating a menu of said third preset number of songs.

In a second aspect, there is provided an apparatus for transmitting a song list identifier, the apparatus comprising:

the first determining module is used for determining account information corresponding to a first account when detecting that a song list recommending trigger event corresponding to the first account occurs;

the second determining module is used for determining the singing list information corresponding to the singing list for each singing list in the candidate singing list set;

the first prediction module is used for inputting account information corresponding to the first account and song list information corresponding to the song list into a pre-trained scoring model and predicting the scoring of the song list by the first account;

the third determining module is used for determining a target song list of which the corresponding score meets the preset recommendation condition in each song list in the candidate song list set;

and the sending module is used for sending the song list identification of the target song list to the login terminal of the first account.

Optionally, the first determining module is configured to:

acquiring a first concerned song list corresponding to the first account;

Optionally, the first determining module is configured to:

Optionally, the second determining module is configured to:

Optionally, the apparatus further comprises:

the fourth determining module is used for determining the identification information of each song in the song library;

a fifth determining module, configured to input the identification information of each song in the song library into a pre-trained song feature extraction model, and determine song feature information of each song in the song library;

the sixth determining module is used for determining song characteristic information corresponding to each song included in each pre-stored song list; and determining the singing sheet characteristic information corresponding to the singing sheet according to the song characteristic information corresponding to each song included in the singing sheet.

Optionally, the apparatus further comprises:

the storage module is used for acquiring identification information of each song, identification information of a first preset number of songs in front of the song and identification information of a second preset number of songs behind the song, respectively taking the identification information of the songs as model input data, taking the identification information of the first preset number of songs and the identification information of the second preset number of songs as model output data, and storing the model input data and the model output data in a corresponding relation in a training set, wherein the identification information of the songs, the identification information of the first preset number of songs and the identification information of the second preset number of songs are contained in each pre-stored song list;

and the first training module is used for training the song feature extraction model containing undetermined parameters according to the corresponding relation between the model input data and the model output data in the training set and on the basis of a training principle that the model output data obtained by the song feature extraction model containing undetermined parameters approaches the model output data in the corresponding relation, so as to obtain the song feature extraction model.

Optionally, the apparatus further comprises:

and the second training module is used for training the scoring model containing undetermined parameters to obtain a scoring model based on the account scoring of the song tickets predicted by the scoring model containing undetermined parameters and approaching to the training principle of the account scoring of the song tickets in the training set according to the scoring of the accounts in the training set, the account information corresponding to the accounts and the song ticket information corresponding to the song tickets.

Optionally, the apparatus further comprises:

the calculation module is used for acquiring an attention song list corresponding to each account; counting the number of songs with different genre types in the concerned song list corresponding to the account according to the concerned song list corresponding to the account, and calculating the proportion of the number of the songs with different genre types in the concerned song list corresponding to the account in the total number of the songs in the concerned song list corresponding to the account to obtain the proportion of the songs with different genre types corresponding to the account;

the selecting module is used for selecting a target account of which the song proportion of the corresponding preset genre type reaches a preset proportion threshold value from the accounts;

the seventh determining module is used for determining account information corresponding to the target account;

the second prediction module is used for determining song information of the song for each song in the concerned song list corresponding to the target account, inputting the account information of the target account and the song information of the song into a pre-trained scoring model, and predicting the score of the target account on the song;

an eighth determining module, configured to determine, among the songs in the concerned song list corresponding to the target account, a third preset number of songs with a largest score;

and the creating module is used for creating the song list consisting of the third preset number of songs.

In a third aspect, there is provided a server comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by the processor to implement the method of transmitting a song list identification according to the first aspect.

In a fourth aspect, there is provided a computer readable storage medium having stored therein at least one instruction, at least one program, set of codes, or set of instructions, which is loaded and executed by a processor to implement the method of sending a song list identification according to the first aspect.

The technical scheme provided by the embodiment of the invention has the following beneficial effects:

in the embodiment of the invention, when a song list recommending trigger event corresponding to a first account is detected, account information corresponding to the first account is determined; determining the singing list information corresponding to each singing list in the candidate singing list set; inputting account information corresponding to the first account and song list information corresponding to the song list into a pre-trained scoring model, and predicting the scoring of the first account on the song list; determining a target song list with a corresponding score meeting a preset recommendation condition in each song list in the candidate song list set; and sending the song list identification of the target song list to a login terminal of the first account. Thus, the user can select the song list which the user likes to listen to from the song list identification of the target song list sent by the server without selecting from all the song lists, thereby improving the efficiency of selecting the song list.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a flowchart of a method for transmitting a song list identifier according to an embodiment of the present invention;

FIG. 2 is a diagram illustrating a method for determining a feature vector of a song menu according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of determining training samples according to an embodiment of the present invention;

FIG. 4 is a system framework diagram provided by an embodiment of the invention;

fig. 5 is a schematic structural diagram of an apparatus for transmitting a song list identifier according to an embodiment of the present invention;

fig. 6 is a schematic structural diagram of an apparatus for sending a song list identifier according to an embodiment of the present invention;

fig. 7 is a schematic structural diagram of an apparatus for transmitting a song list identifier according to an embodiment of the present invention;

fig. 8 is a schematic structural diagram of an apparatus for transmitting a song list identifier according to an embodiment of the present invention;

fig. 9 is a schematic structural diagram of an apparatus for transmitting a song list identifier according to an embodiment of the present invention;

fig. 10 is a schematic structural diagram of a server according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The embodiment of the invention provides a method for sending a song list identifier, and the execution subject of the method is a server. The server may be a server with a function of sending the song list identifier, and may be a background server of a song playing application program. The server may include a processor, memory, transceiver, etc. The processor may be a CPU (Central Processing Unit) or the like, and may be used for the relevant Processing for determining the target song list. The Memory may be a RAM (Random Access Memory), a Flash (Flash Memory), and the like, and may be configured to store received data, data required by a processing procedure, data generated in the processing procedure, and the like, such as a scoring model. The transceiver may be used for data transmission with a terminal or other server, for example, to send a song list identification of a target song list to the terminal, and may include an antenna, matching circuitry, a modem, and the like.

The process flow shown in fig. 1 will be described in detail below with reference to specific embodiments, and the contents may be as follows:

step 101, when detecting that a song list recommending trigger event corresponding to a first account occurs, determining account information corresponding to the first account.

The account information corresponding to the first account may be information reflecting characteristics or attributes of the first account.

In implementation, the server may preset song list recommendation trigger events corresponding to the accounts, where the song list recommendation trigger events corresponding to each account may be the same, for example, the song list recommendation trigger event may be a preset song list recommendation period, and the song list recommendation trigger events corresponding to each account may also be different, for example, the song list recommendation trigger events corresponding to each account may be song list recommendation requests sent by the terminal where the account is logged in, respectively. In the working process, the server can detect the song list recommending trigger events corresponding to the accounts, when the server detects that the song list recommending trigger events corresponding to the first account occur (for example, when a song list recommending request sent by a terminal logged in by the first account is received), the server can judge whether a target song list to be recommended to the first account is stored in the server currently, if the target song list to be recommended to the first account is stored, the server can directly obtain the song list identification of the target song list to be recommended to the first account, further, the song list identification can be sent to the logging terminal of the first account, and if the target song list to be recommended to the first account is not stored, the server can determine the target song list to be recommended to the first account. Specifically, the server may determine account information corresponding to a first account, where the first account may be any account.

Optionally, the account information corresponding to the first account may include account characteristic information and account attribute information, and correspondingly, the processing procedure in step 101 may be as follows: when detecting that a song menu recommendation trigger event corresponding to a first account occurs, determining account attribute information corresponding to the first account according to basic attribute information corresponding to the first account and song preference data concerned; determining account characteristic information corresponding to a first account; and determining account attribute information corresponding to the first account and account characteristic information corresponding to the first account as account information corresponding to the first account.

The basic attribute information may be demographic attribute information, which may include age information and gender information, and the song-attention preference data may be data for reflecting the attention degree of the corresponding account to various songs.

In implementation, when the occurrence of the song list recommendation triggering event corresponding to the first account is detected, the server may determine account characteristic information corresponding to the first account, and in addition, the server may determine account attribute information corresponding to the first account. Specifically, the server may obtain basic attribute information corresponding to the first account, where the basic attribute information includes age information, gender information, and the like, may obtain focused song preference data corresponding to the first account (for example, the number of songs in different genres in a first focused song list corresponding to the first account), and may determine the basic attribute information and the focused song preference data corresponding to the first account as account attribute information corresponding to the first account, which may be denoted by m 3. After determining the account attribute information and the account feature information corresponding to the first account, the server may determine the account attribute information and the account feature information corresponding to the first account as the account information corresponding to the first account. In addition, the account attribute information may be represented by a vector, which may be referred to as an account attribute vector, in which case, when the occurrence of the song list recommendation trigger event corresponding to the first account is detected, the server may determine the account attribute vector corresponding to the first account, where each dimension element of the account attribute vector may represent a different attribute. Specifically, the server may obtain basic attribute information and song preference attention data of the first account, and further, may determine a numerical value of each dimensional element in the account attribute vector according to the basic attribute information and song preference attention data of the first account, that is, obtain the account attribute vector corresponding to the first account.

For example, each dimension element of the account attribute vector represents the number of songs of different age groups (e.g., [5, 12], [13, 17], [18, 24], [25, 30], [31, 35], [36, 40], [41, 50], [51, 60], respectively, less than 5 and greater than 60, unknown) and genres (including male, female, unknown), in which case, after the server obtains the age and gender corresponding to the first account, the server can determine the corresponding dimension of the age and gender in the account attribute vector, further, the value of the element of the dimension can be set to 1, the others to 0, after obtaining the number of songs of different genres in the first interest song list corresponding to the first account, the value of the element representing the number of different genres can be set to the number of songs of different genres in the first interest song list corresponding to the first account, thus, the account attribute vector corresponding to the first account can be obtained.

Optionally, the processing procedure of determining the account characteristic information corresponding to the first account may be as follows: acquiring a first concerned song list corresponding to a first account; determining song characteristic information corresponding to each song in a first concerned song list in the prestored song characteristic information corresponding to each song; and determining account characteristic information corresponding to the first account according to song characteristic information corresponding to each song in the first concerned song list.

The song characteristic information may be information obtained from the correlation between songs.

In an implementation, a focus song list corresponding to each account may be stored in advance in the server, where the focus song list may include songs collected and listened to by the account within a preset historical period (for example, within the past month). When detecting that the song list recommendation triggering event corresponding to the first account occurs, the server may obtain, from the pre-stored attention song lists corresponding to the respective accounts, an attention song list corresponding to the first account (which may be referred to as a first attention song list and may be denoted by U1). The server may also pre-store song feature information corresponding to each song pre-extracted by the song feature extraction model, where the song feature information corresponding to each song may be represented by a vector, which may be referred to as a song feature vector. After the first interested song list is acquired, song feature information corresponding to each song (which may be represented by i) in the first interested song list may be determined from the song feature information corresponding to each song, and V may be used for determining the song feature information corresponding to each song_iAnd representing that i belongs to the U1, and further determining account characteristic information corresponding to the first account based on the determined song characteristic information corresponding to each song, wherein the account characteristic information can be represented by m 1. That is, account characteristic information corresponding to the first account may be expressed by song characteristic information corresponding to each song in the first list of interesting songs.

Optionally, the server may determine the account characteristic information corresponding to the first account according to each song characteristic information corresponding to each song in the first interest song list and the weight corresponding to each song characteristic information, and accordingly, the processing procedure may be as follows: determining a type of interest of the first account for each song in the first list of interesting songs; determining the attention duration of each song in the first attention song list from the first account to the current moment; determining an attention weight corresponding to the attention type of each song in the first attention song list by the first account according to the pre-stored corresponding relationship between different attention types and the attention weight; determining the time attenuation degree corresponding to each song according to a time attenuation formula and the attention duration of the attention moment of the first account to each song in the first attention song list from the current moment; and taking the product of the attention weight corresponding to the attention type of each song and the time attenuation degree corresponding to each song as the weight of the song characteristic information corresponding to each song, and calculating the weighted sum of the song characteristic information corresponding to each song in the first attention song list to obtain the account characteristic information corresponding to the first account.

In implementation, after the first interested song list is acquired, the server may determine an interested type of each song in the first interested song list by the first account, besides determining song feature information corresponding to each song in the first interested song list, where the interested type may be a collection type or a listening type. The server may pre-store the corresponding relationship between different attention types and attention weights, as shown in table 1. After determining the attention type of the first account for each song in the first attention song list, the attention weight corresponding to the attention type of the first account for each song in the first attention song list may be determined in the correspondence, that is, the attention weight corresponding to each song may be obtained (α may be used)_typeRepresentation). After the first concerned song list is obtained, the server can determine the concerned time of each song in the first concerned song list by the first account in addition to the concerned weight corresponding to the concerned type of each song in the first concerned song list by the first account, and further determine the concerned time (which can be represented by differential time) of each song from the concerned time to the current time to obtain the concerned time corresponding to each song. The server may store a time decay formula (where the time decay formula may be e)^∧-t), after determining the attention duration of each song from the current time, substituting the attention duration corresponding to each song into a time attenuation formula (that is, the attention duration corresponding to each song can be used as the value of the parameter t) to obtain the time attenuation degree corresponding to each song (β can be used as the value of the parameter t)_timeAnd represents), wherein the larger the attention duration from the current time at the attention time, the smaller the corresponding time attenuation. For the first song of interestAfter obtaining the attention weight corresponding to each song and the time attenuation corresponding to the song, each song in the song list may be multiplied, and the product may be determined as the weight of the song characteristic information corresponding to the song. After obtaining the weight of the song feature information corresponding to each song in the first concerned song list and the song feature information corresponding to each song, a weighted sum of the song feature information corresponding to each song in the first concerned song list may be calculated to obtain account feature information corresponding to the first account, that is, the account feature information m1 corresponding to the first account may be Σ_U1α_typeβ_timeV_i。

TABLE 1

Type of interest	Attention weight
		Type of collection	1
Listening type	0.5

Step 102, determining the song list information corresponding to each song list in the candidate song list set; inputting account information corresponding to the first account and song list information corresponding to the song list into a pre-trained scoring model, and predicting the scoring of the first account on the song list.

In implementation, when detecting that the song list recommendation triggering event corresponding to the first account occurs, the server may further determine song list information corresponding to each song list in the candidate song list set for each song list in the candidate song list set. After the account information corresponding to the first account and the song list information corresponding to each song list in the candidate song list set are determined, the score of each song list in the candidate song list set by the first account can be predicted. Specifically, the server may store a pre-trained scoring model in advance, where the input of the scoring model may be account information of an account and song information of a song, the output may be a score of the song of the account, and the scoring model may be a neural network scoring model. For each song list in the candidate song list set, the server may input account information corresponding to the first account and song list information corresponding to the song list into the scoring model to predict the score of the song list by the first account.

Optionally, the training process of the above scoring model may be as follows: and training the scoring model containing undetermined parameters according to the scoring of each account in the training set on each song bill, the account information corresponding to each account and the song bill information corresponding to each song bill, and on the basis of a training principle that the scoring of the account on the song bill predicted by the scoring model containing undetermined parameters approaches to the scoring of the account on the song bill in the training set, so as to obtain the scoring model.

In implementation, the server may store a training set and a scoring model containing pending parameters (which may be referred to as a scoring model to be trained) in advance, each training sample in the training set may be a score of each account for each song list, account information corresponding to the account and song list information corresponding to the song list, the training sample may include a positive sample and a negative sample, and the scoring model to be trained may include an input layer (where the input layer is account information corresponding to the account and song list information corresponding to the song list), a sensing layer 1 (where the sensing layer 1 may include 1024 neurons and an activation function relu), a sensing layer 2 (where the sensing layer 2 may include 512 neurons and an activation function relu), a sensing layer 3 (where the sensing layer 3 may include 256 neurons and an activation function relu), and an output layer (the activation function is sigmoid). For each training sample in the training set, the server can input account information corresponding to an account in the training sample and song list information corresponding to the song list into a scoring model containing undetermined parameters to obtain the score of the account on the song list predicted by the scoring model containing the undetermined parameters, and then the scoring model to be trained can be trained on the basis of the training principle that the account on the song list predicted by the scoring model containing the undetermined parameters approaches the score of the account on the song list of the training sample. For example, the difference between the score of the account pair song slip predicted by the score model containing the undetermined parameter and the score of the account pair song slip in the training set is used as a target function, the target function is trained (for example, the minimum value of the target function can be obtained) based on a gradient descent method, the parameter value of the undetermined parameter is obtained, and the final score model is obtained. In addition, the training sample may not be the score of the account for the song list, when a certain account pays attention to a certain song list, the score of the account for the song list may be 1, otherwise, the score is 0.

Optionally, the song list information may include song list characteristic information and song list attribute information corresponding to the song list, and accordingly, the processing procedure in step 102 may be as follows: for each song list in the candidate song list set, determining the song list characteristic information corresponding to the song list in the song list characteristic information corresponding to each pre-stored song list; determining attribute information of the song list corresponding to the song list according to the concerned information corresponding to the song list; and determining the attribute information of the song list corresponding to the song list and the characteristic information of the song list corresponding to the song list as the song list information corresponding to the song list.

The focused information may be information for reflecting the focused situation of the corresponding song list, the focused information may include the collected amount, the playing amount and the like, and the song list attribute information may be information for reflecting the attribute of the corresponding song list.

In implementation, the server may store the singing sheet feature information corresponding to each singing sheet in advance, wherein the singing sheet feature information corresponding to each singing sheet may be represented by a vector, which may be referred to as a singing sheet feature vector. For each of the singing sheets in the candidate singing sheet set, the server may determine the singing sheet feature information (which may be denoted by m 2) corresponding to the singing sheet among the previously stored singing sheet feature information corresponding to each singing sheet.

The server may store the information to be paid attention corresponding to each song list in advance. For each song form in the candidate song form set, the server may determine the focused information corresponding to the song form in the focused information corresponding to each song form, and further may determine the attribute information of the song form corresponding to the song form according to the focused information (for example, the focused information corresponding to the song form may be determined as the attribute information of each song form corresponding to the song form), which may be represented by m4, where the attribute information of the song form may be represented by a vector, which may also be referred to as an attribute vector of the song form. In addition, the attribute information of the menu may also include basic attribute information of the menu, where the basic attribute information may be information for reflecting basic attributes of the corresponding menu, for example, the basic attribute information may include the number of songs included in the menu, the number of songs in different genres in the menu, and the like, that is, the server may determine the information of interest and the basic attribute information corresponding to the menu as the attribute information of the menu corresponding to the menu.

For the case where the menu attribute information is a menu attribute vector, each dimensional element of the menu attribute vector may represent a preset different attribute (e.g., the preset different attribute may include a collection amount, a song listening amount, the number of songs included in the menu, the number of songs in different genres, etc.). In this case, for each song list in the candidate song list set, after the server obtains the concerned information and the basic attribute information corresponding to the song list, the server may determine the attribute vector of the song list corresponding to the song list according to the concerned information and the basic attribute information corresponding to the song list. After determining the attribute information of the singing list corresponding to each singing list and the characteristic information of the singing list corresponding to the singing list, the attribute information of the singing list corresponding to each singing list and the characteristic information of the singing list corresponding to the singing list can be determined as the singing list information corresponding to the singing list.

Optionally, the pre-stored song feature information may be obtained by the server in advance through a song feature extraction model, the menu feature information may be obtained by the server through song feature information corresponding to a song included in the menu, and accordingly, the processing procedure may be as follows: determining identification information of each song in a song library; inputting the identification information of each song in the song library into a pre-trained song feature extraction model, and determining the song feature information of each song in the song library; for each pre-stored song list, determining song characteristic information corresponding to each song included in the song list; and determining the song list characteristic information corresponding to the song list according to the song characteristic information corresponding to each song included in the song list.

In implementation, each song may be stored in the song library in advance, and for each song in the song library, the server may encode each song to obtain identification information of each song, where the identification information may be an identification vector, the dimension of the identification vector of each song is the total number of songs in the song library, and the value of one-dimensional element in the identification vector corresponding to each song is 1, the values of other dimensional elements are 0, and the dimension of each identification vector value of 1 is different. For example, if the total number of songs in the song library is n1, the identification vector for each song may be an n 1-dimensional vector, the identification vector for the first song may be (1, 0, 0, …, 0, 0), the identification vector for the second song may be (0, 1, 0, …, 0, 0), and so on, resulting in an identification vector for each song in the song library.

After the identification information of each song in the song library is determined, the identification information of the song can be input into a pre-trained song feature extraction model, wherein the song feature extraction model can be a neural network model and can be a calculation framework of item2vec, and then the server can determine the output result of the last hidden layer in the song feature extraction model as the song feature information corresponding to the song. The server can obtain the song characteristic information corresponding to each song in the song library according to the processing mode and store the song characteristic information. After determining the song characteristic information corresponding to each song, the server may determine the pre-stored song list characteristic information of each song list according to the song characteristic information corresponding to each song. Specifically, for each pre-stored song list, the server may determine songs included in the song list, and further may determine song characteristic information corresponding to the songs included in the song list from the song characteristic information corresponding to each song. After determining the song feature information corresponding to the song included in the song list, an average value of the song feature information corresponding to the song included in the song list can be calculated, and then the obtained average value can be determined as the song list feature information corresponding to the song list, wherein the song feature information can be a song feature vector, and the song feature information can be a song list feature vector. For example, the menu a includes song 1, song 2, and song 3, the song feature vector corresponding to song 1 is (99, 170, 50), the song feature vector corresponding to song 2 is (70, 199, 50), and the song feature vector corresponding to song 3 is (80, 180, 50), then for each dimension element of the menu feature vector, an average value of the values of the dimension element in the song feature vector corresponding to the song included in the menu may be calculated, and the value of the dimension element of the menu feature vector is obtained, that is, the menu feature vector is (83, 183, 50), as shown in fig. 2.

Optionally, the song feature extraction model may be pre-trained by the server, and accordingly, the processing procedure may be as follows: for each song contained in each prestored song list, acquiring identification information of the song, identification information of a first preset number of songs in front of the song and identification information of a second preset number of songs behind the song, respectively taking the identification information of the song as model input data, taking the identification information of the first preset number of songs and the identification information of the second preset number of songs as model output data, and storing the model output data and the identification information into a corresponding relation between the model input data and the model output data in a training set; and training the song feature extraction model containing undetermined parameters according to the corresponding relation between the model input data and the model output data in the training set and on the basis of a training principle that the model output data obtained by the song feature extraction model containing undetermined parameters approaches the model output data in the corresponding relation to obtain the song feature extraction model.

In implementation, the server may first determine each training sample in the training set before training the song feature extraction model. Specifically, each pre-stored song list is obtained, for each song in each song list, the server may obtain identification information of the song, identification information of a first preset number of songs in the song list before the song, and identification information of a second preset number of songs in the song list after the song, and further, the identification information of the song may be used as model input data, and the identification information of the first preset number of songs and the identification information of the second preset number of songs may be used as model output data, which are stored in a corresponding relationship between the model input data and the model output data in the training set, respectively, where the first preset number and the second preset number may be the same or different. For example, the first preset number and the second preset number are 2, and the song list a includes the following songs: for song 4, the server may use the identification information of song 4 as model input data, and the identification information of song 2, song 3, song 5, and song 6 as model output data, as a training sample, and store the training sample in the corresponding relationship between the model input data and the model output data in the training set, as shown in fig. 3.

Songs in the same song list have certain correlation, so that the song feature extraction model can be trained through the correlation of adjacent songs in the same song list. Specifically, after obtaining the training samples, for each training sample in the training set, the server may input the model input data in the training sample into a song feature extraction model (which may be referred to as a to-be-trained song feature extraction model) including undetermined parameters, to obtain model output data corresponding to the model input data determined by the to-be-trained song feature extraction model (where the model output data includes the to-be-determined parameters), and further, may obtain a training rule based on which the model output data obtained by the song feature extraction model including the undetermined parameters approaches the model output data corresponding to the model input data in the corresponding relationship (for example, a difference between the model output data obtained by the song feature extraction model including the undetermined parameters and the model output data corresponding to the model input data in the corresponding relationship is obtained, as an objective function, further, the minimum value of the objective function may be found), and the song feature extraction model to be trained is trained. And when training of all the training samples is finished, obtaining a training value of the undetermined parameter, namely obtaining a final song feature extraction model, wherein the song feature extraction model can be a neural network model.

Optionally, the server may further create the song list through the scoring model, where the created song list may be used to provide a training sample of the song feature extraction model, and may also be used as the song list in the candidate set, and accordingly, the processing procedure may be as follows: for each account, acquiring an attention song list corresponding to the account; according to the concerned song list corresponding to the account, counting the number of songs of different genre types in the concerned song list corresponding to the account, and calculating the ratio of the number of the songs of different genre types in the concerned song list corresponding to the account to the total number of the songs in the concerned song list corresponding to the account to obtain the song ratios of different genre types corresponding to the account; selecting a target account with the corresponding song occupation ratio of a preset genre type reaching a preset occupation ratio threshold value from the accounts; determining account information corresponding to a target account; for each song in the concerned song list corresponding to the target account, determining song information of the song, inputting the account information of the target account and the song information of the song into a pre-trained scoring model, and predicting the score of the song by the target account; determining a third preset number of songs with the maximum score in the songs in the concerned song list corresponding to the target account; a menu is created that consists of a third preset number of songs.

In an implementation, the song sheets pre-stored in the server may be created by the server through a scoring model, that is, the server may create song sheets of a certain preset genre type through the scoring model. In order to create a song list with preset genre types, the server may count the song proportions of different genre types corresponding to each account. Specifically, for each account, the server may obtain a focus song list corresponding to the account, further, may count the number of songs of different genre types included in the focus song list, and calculate a ratio of the number of songs of different genre types to a total number of songs included in the focus song list, to obtain a song ratio of different genre types corresponding to the account. For example, account 1 includes, in focus song list 1: song 1 (where the corresponding genre type is x), song 2 (where the corresponding genre type is x), song 3 (where the corresponding genre type is x), song 4 (where the corresponding genre type is y), and song 5 (where the corresponding genre type is x), then the server counts the number of songs of genre type x in the interest song list corresponding to account 1 as 4, the number of songs of genre type y in the interest song list corresponding to account 1 as 1, and accordingly, the song duty of genre type x corresponding to account 1 is 0.8(4/5), and the song duty of genre type y is 0.2 (1/5). After obtaining the song occupation ratios of different genre types corresponding to each account, the server may select, from the accounts, an account (which may be referred to as a target account, where the target account may be one or more) whose song occupation ratio of the corresponding preset genre type (where the preset genre type may be one or more) reaches a preset occupation ratio threshold. For example, if the song proportion of the genre type x corresponding to the account 1 is 0.8, the song proportion of the genre type x corresponding to the account 2 is 0.6, the song proportion of the genre type x corresponding to the account 3 is 0.4, the preset genre type is the genre x, and the preset proportion threshold is 0.5, the server may determine the account 1 and the account 2 as the target account.

After the target account is determined, the server can determine the score of each song in the concerned song list corresponding to the target account by the target account through a scoring model. Specifically, the server may determine the account information corresponding to the target account according to a processing manner of determining the account information corresponding to the first account. In addition, for each song in the concerned song list corresponding to the target account, the server may determine song information corresponding to the song according to a processing manner of determining song list information corresponding to the song list, and further, may input the account information of the target account and the song information of the song into a scoring model to predict a score of the song by the target account. After the scores of the target account for the songs in the concerned song list corresponding to the target account are obtained, the server may select a third preset number (for example, the third preset number is 20) of songs with the largest corresponding scores from the songs in the concerned song list corresponding to the target account, and further may create a song list composed of the selected third preset number of songs.

And 103, determining a target song list with a corresponding score meeting a preset recommendation condition in each song list in the candidate song list set.

The preset recommendation condition may be a condition for judging whether the song list is recommended according to the score of the first account on the song list.

In an implementation, after the server determines the scores of the first account for the singing sheets in the candidate singing sheet set, a singing sheet (which may be called a target singing sheet) with a corresponding score meeting a preset recommendation condition may be determined in each singing sheet in the candidate singing sheet set. Specifically, the server may determine a preset number of target sings with the highest corresponding scores from among the sings in the candidate singing list set, or the server may determine a target singing list with a corresponding score greater than a preset score threshold from among the sings in the candidate singing list set.

And 104, sending the song list identification of the target song list to the login terminal of the first account.

In implementation, after the server determines the target song list, the server may send the song list identifier of the target song list to the login terminal of the first account, and correspondingly, the terminal may receive the song list identifier of the target song list sent by the server and may display the song list identifier, so that the user may select the song list that the user likes to listen to, as shown in fig. 4.

Based on the same technical concept, an embodiment of the present invention further provides an apparatus for sending a song list identifier, as shown in fig. 5, the apparatus includes:

the first determining module 510 is configured to determine account information corresponding to a first account when detecting that a song menu recommendation triggering event corresponding to the first account occurs;

a second determining module 520, configured to determine, for each song menu in the candidate song menu set, song menu information corresponding to the song menu;

a first prediction module 530, configured to input account information corresponding to the first account and song list information corresponding to the song list into a pre-trained scoring model, and predict a score of the song list by the first account;

a third determining module 540, configured to determine, from the song sheets in the candidate song sheet set, a target song sheet whose corresponding score meets a preset recommendation condition;

a sending module 550, configured to send the song list identifier of the target song list to the login terminal of the first account.

Optionally, the first determining module 510 is configured to:

acquiring a first concerned song list corresponding to the first account;

Optionally, the first determining module 510 is configured to:

Optionally, the second determining module 520 is configured to:

Optionally, as shown in fig. 6, the apparatus further includes:

a fourth determining module 560 for determining identification information of each song in the song library; inputting the identification information of each song in the song library into a pre-trained song feature extraction model, and determining the song feature information of each song in the song library; for each pre-stored song list, determining song characteristic information corresponding to each song included in the song list; and determining the singing sheet characteristic information corresponding to the singing sheet according to the song characteristic information corresponding to each song included in the singing sheet.

Optionally, as shown in fig. 7, the apparatus further includes:

the storage module 570 is configured to, for each song included in each pre-stored song list, obtain identification information of the song, identification information of a first preset number of songs before the song, and identification information of a second preset number of songs after the song, respectively use the identification information of the song as model input data, use the identification information of the first preset number of songs and the identification information of the second preset number of songs as model output data, and store the model output data and the identification information into a corresponding relationship between each model input data and each model output data in a training set;

the first training module 580 is configured to train, according to a corresponding relationship between each model input data and each model output data in the training set, the song feature extraction model including the undetermined parameter based on a training principle that model output data obtained by the song feature extraction model including the undetermined parameter approaches model output data in the corresponding relationship, so as to obtain the song feature extraction model.

Optionally, as shown in fig. 8, the apparatus further includes:

the second training module 590 is configured to train the scoring model including the undetermined parameter based on a training principle that the score of the account pair song slip predicted by the scoring model including the undetermined parameter approaches to the score of the account pair song slip in the training set, according to the score of each account in the training set to each song slip, the account information corresponding to each account, and the song slip information corresponding to each song slip, so as to obtain a scoring model.

Optionally, as shown in fig. 9, the apparatus further includes:

the calculation module 5100 is configured to, for each account, obtain a list of interested songs corresponding to the account; counting the number of songs with different genre types in the concerned song list corresponding to the account according to the concerned song list corresponding to the account, and calculating the proportion of the number of the songs with different genre types in the concerned song list corresponding to the account in the total number of the songs in the concerned song list corresponding to the account to obtain the proportion of the songs with different genre types corresponding to the account;

a selecting module 5110, configured to select, from the accounts, a target account whose song proportion of the corresponding preset genre type reaches a preset proportion threshold;

a fifth determining module 5120, configured to determine account information corresponding to the target account;

a second prediction module 5130, configured to determine song information of each song in the concerned song list corresponding to the target account, input the account information of the target account and the song information of the song into a pre-trained scoring model, and predict a score of the song by the target account;

a sixth determining module 5140, configured to determine, among the songs in the attention song list corresponding to the target account, a third preset number of songs with a largest score;

a creating module 5150, configured to create a song list composed of the third preset number of songs.

It should be noted that: the apparatus for sending a song list identifier provided in the above embodiment is only illustrated by the division of the above functional modules when sending the song list identifier, and in practical applications, the above function distribution may be completed by different functional modules according to needs, that is, the internal structure of the server is divided into different functional modules to complete all or part of the above described functions. In addition, the apparatus for sending a song list identifier and the method for sending a song list identifier provided by the above embodiments belong to the same concept, and specific implementation processes thereof are detailed in the method embodiments and are not described herein again.

Fig. 10 is a schematic structural diagram of a server according to an embodiment of the present invention. The server 1900 may vary widely by configuration or performance and may include one or more Central Processing Units (CPUs) 1922 (e.g., one or more processors) and memory 1932, one or more storage media 1930 (e.g., one or more mass storage devices) storing applications 1942 or data 1944. Memory 1932 and storage medium 1930 can be, among other things, transient or persistent storage. The program stored in the storage medium 1930 may include one or more modules (not shown), each of which may include a series of instructions operating on a server. Still further, a central processor 1922 may be provided in communication with the storage medium 1930 to execute a series of instruction operations in the storage medium 1930 on the server 1900.

The server 1900 may also include one or more power supplies 1926, one or more wired or wireless network interfaces 1950, one or more input-output interfaces 1958, one or more keyboards 1956, and/or one or more operating systems 1941, such as Windows Server, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

Server 1900 may include memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors include instructions for:

acquiring a first concerned song list corresponding to the first account;

Optionally, the method further includes:

determining identification information of each song in a song library;

Optionally, the method further includes:

determining account information corresponding to a target account;

creating a menu of said third preset number of songs.

It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A method of transmitting a song list identification, the method comprising:

when detecting that a song menu recommendation triggering event corresponding to a first account occurs, determining account information corresponding to the first account, wherein the account information corresponding to the first account comprises account characteristic information corresponding to the first account; the account characteristic information is obtained according to song characteristic information and weight corresponding to each song in a first concerned song list corresponding to the first account; the weight corresponding to each song is obtained according to the attention weight and the time attenuation corresponding to each song; determining the time attenuation degree corresponding to each song based on the attention duration from the attention moment of each song to the current moment; determining an attention weight value corresponding to each song based on the attention type corresponding to each song;

determining singing list information corresponding to the singing list for each singing list in the candidate singing list set;

inputting account information corresponding to the first account and song list information corresponding to the song list into a pre-trained scoring model, and predicting scoring of the song list by the first account, wherein the scoring model is trained on the basis of scoring of each song list by each account in a training set, account information corresponding to each account and song list information corresponding to each song list, and the pre-trained scoring model is obtained;

2. The method of claim 1, wherein determining account information corresponding to the first account when detecting that the song menu recommendation triggering event corresponding to the first account occurs comprises:

3. The method of claim 2, wherein the determining account characteristic information corresponding to the first account comprises:

acquiring a first concerned song list corresponding to the first account;

4. The method of claim 3, wherein determining account characteristic information corresponding to the first account according to song characteristic information corresponding to each song in the first list of interesting songs comprises:

5. The method of claim 1, wherein said determining, for each of the singing slips in the candidate singling slip set, the singing slip information corresponding to the singing slip comprises:

6. The method according to claim 3 or 5, characterized in that the method further comprises:

determining identification information of each song in a song library;

7. The method of claim 6, further comprising:

8. The method of claim 1, further comprising:

and training the scoring model containing undetermined parameters according to the scoring of each account in the training set on each song bill, account information corresponding to each account and song bill information corresponding to each song bill, and based on a training principle that the scoring of the account on the song bill predicted by the scoring model containing undetermined parameters approaches to the scoring of the account on the song bill in the training set, so as to obtain a pre-trained scoring model.

9. The method of claim 1 or 7, further comprising:

determining account information corresponding to a target account;

creating a menu of said third preset number of songs.

10. An apparatus for transmitting a song list identifier, the apparatus comprising:

the second determining module is used for determining the singing list information corresponding to the singing list for each singing list in the candidate singing list set; the account information corresponding to the first account comprises account characteristic information corresponding to the first account; the account characteristic information is obtained according to song characteristic information and weight corresponding to each song in a first concerned song list corresponding to the first account; the weight corresponding to each song is obtained according to the attention weight and the time attenuation corresponding to each song; determining the time attenuation degree corresponding to each song based on the attention duration from the attention moment of each song to the current moment; determining an attention weight value corresponding to each song based on the attention type corresponding to each song;

a first prediction module, configured to input account information corresponding to the first account and song list information corresponding to the song list into a pre-trained scoring model, and predict scoring of the song list by the first account, where the scoring model is trained based on scoring of each song list by each account in a training set, account information corresponding to each account, and song list information corresponding to each song list, so as to obtain the pre-trained scoring model;

11. The apparatus of claim 10, wherein the first determining module is configured to:

12. The apparatus of claim 11, wherein the first determining module is configured to:

acquiring a first concerned song list corresponding to the first account;

13. The apparatus of claim 12, wherein the first determining module is configured to:

14. The apparatus of claim 10, wherein the second determining module is configured to:

15. The apparatus of claim 12 or 14, further comprising:

the fourth determining module is used for determining the identification information of each song in the song library; inputting the identification information of each song in the song library into a pre-trained song feature extraction model, and determining the song feature information of each song in the song library; for each pre-stored song list, determining song characteristic information corresponding to each song included in the song list; and determining the singing sheet characteristic information corresponding to the singing sheet according to the song characteristic information corresponding to each song included in the singing sheet.

16. The apparatus of claim 15, further comprising:

17. The apparatus of claim 10, further comprising:

and the second training module is used for training the scoring model containing undetermined parameters according to the scoring of each account to each song bill in the training set, the account information corresponding to each account and the song bill information corresponding to each song bill and on the basis of a training principle that the scoring of the account to the song bill predicted by the scoring model containing undetermined parameters approaches to the scoring of the account to the song bill in the training set, so as to obtain a pre-trained scoring model.

18. The apparatus of claim 10 or 16, further comprising:

the fifth determining module is used for determining account information corresponding to the target account;

a sixth determining module, configured to determine, among the songs in the concerned song list corresponding to the target account, a third preset number of songs with a largest score;

19. A server, characterized in that it comprises a processor and a memory, in which at least one instruction, at least one program, set of codes or set of instructions is stored, which is loaded and executed by the processor to implement the method of sending a singing sheet identification according to any one of claims 1 to 9.

20. A computer readable storage medium having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by a processor to implement the method of sending a song list identification according to any one of claims 1 to 9.