Music recommendation method, server and storage medium

Abstract

The embodiment of the present invention discloses a music recommendation method, a server and a readable storage medium, the method comprising: obtaining user feature information and song listening feature information of a user; inputting the user feature information and the song listening feature information into a pre-trained target calculation model to obtain a song listening intention probability value of the user; screening out a first song set from the user's song collection list according to the song listening intention probability value; extracting feature information of songs to be recommended from each song to be recommended in the recommended song set, and inputting the user feature information, the song listening feature information, the first song set and the song listening feature information into a pre-trained song recommendation model to obtain a second song set; fusing the first song set and the second song set to obtain a target song set, and pushing the target song set to a user terminal, thereby satisfying the user's song listening needs in different environments and more effectively recommending songs to the user.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a music recommendation method, a server and a storage medium. Background Art

[0002] With the continuous development of Internet technology, users can obtain information that users may be interested in from a large amount of information recommended by recommendation systems. Recommendation systems play an increasingly important role in many applications. Among them, in song recommendation applications such as music software, songs are usually recommended to users based on the historical songs that users have listened to. However, in actual use, users have different needs for listening to songs in different environments. The current song recommendation method mainly recommends songs to users based on the historical songs that users have listened to, which is relatively simple and has poor recommendation effects. Therefore, it is very important to recommend songs to users more effectively. Summary of the invention

[0003] The embodiments of the present invention provide a music recommendation method, a server and a storage medium, which recommend songs to users by combining a favorite song list and a recommended song collection, thereby meeting the user's listening needs in different environments and recommending songs to users more effectively.

[0004] In a first aspect, an embodiment of the present invention provides a music recommendation method, comprising:

[0005] Obtaining user characteristic information and music listening characteristic information of the user;

[0006] Inputting the user characteristic information and the song listening characteristic information into a pre-trained target calculation model to obtain a song listening intention probability value of the user, and filtering out a first song set from the user's song collection list according to the song listening intention probability value;

[0007] Extracting feature information of the song to be recommended from each song to be recommended in the recommended song set, and inputting the user feature information, the song listening feature information, the first song set and the feature information of the song to be recommended into a pre-trained song recommendation model to obtain a second song set;

[0008] The first song set and the second song set are merged to obtain a target song set, and the target song set is pushed to a user terminal corresponding to the user.

[0009] In a second aspect, an embodiment of the present invention provides a music recommendation device, including:

[0010] An acquisition unit, used to acquire user characteristic information and song listening characteristic information of the user;

[0011] A screening unit, configured to input the user characteristic information and the song listening characteristic information into a pre-trained target calculation model, obtain a song listening intention probability value of the user, and screen a first song set from the user's song collection list according to the song listening intention probability value;

[0012] A determination unit, configured to extract feature information of songs to be recommended from each song to be recommended in the recommended song set, and input the user feature information, the song listening feature information, the first song set, and the feature information of songs to be recommended into a pre-trained song recommendation model to obtain a second song set;

[0013] A push unit is used to merge the first song set and the second song set to obtain a target song set, and push the target song set to a user terminal corresponding to the user.

[0014] In a third aspect, an embodiment of the present invention provides a server, the server comprising: a processor and a memory, the processor being configured to execute:

[0015] Obtaining user characteristic information and music listening characteristic information of the user;

[0016] Inputting the user characteristic information and the song listening characteristic information into a pre-trained target calculation model to obtain a song listening intention probability value of the user, and filtering out a first song set from the user's song collection list according to the song listening intention probability value;

[0017] Extracting feature information of the song to be recommended from each song to be recommended in the recommended song set, and inputting the user feature information, the song listening feature information, the first song set and the feature information of the song to be recommended into a pre-trained song recommendation model to obtain a second song set;

[0018] The first song set and the second song set are merged to obtain a target song set, and the target song set is pushed to a user terminal corresponding to the user.

[0019] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, in which program instructions are stored. When the program instructions are executed, they are used to implement the method described in the first aspect above.

[0020] The embodiment of the present invention can obtain the user's user characteristic information and song listening characteristic information; input the user characteristic information and song listening characteristic information into a pre-trained target calculation model to obtain the user's song listening intention probability value, and filter out a first song set from the user's song collection list according to the song listening intention probability value; extract the song characteristic information to be recommended from each song to be recommended in the recommended song set, and input the user characteristic information, song listening characteristic information, the first song set and the song characteristic information to be recommended into the pre-trained song recommendation model to obtain a second song set; perform fusion processing on the first song set and the second song set to obtain a target song set, and push the target song set to the user terminal corresponding to the user. This method of recommending songs to users by combining the collection song list and the recommended song set meets the user's listening needs in different environments, and can more effectively recommend songs to users. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 is a structural diagram of a music recommendation system provided by an embodiment of the present invention;

[0023] Figure 2 is a flowchart of a music recommendation method provided by an embodiment of the present invention;

[0024] Figure 3 It is a schematic diagram of the structure of a DNN;

[0025] Figure 4 is a training schematic diagram of a target computing model provided by an embodiment of the present invention;

[0026] Figure 5 is a training diagram of a song recommendation model provided by an embodiment of the present invention;

[0027] Figure 6 is a structural schematic diagram of a music recommendation device provided by an embodiment of the present invention;

[0028] Figure 7 It is a structural diagram of a server provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present invention will be described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] At present, in the scenario where a user listens to songs through music software, the user can select songs to listen to from the music software's music listening function. Music software generally has two types of functions: a song collection list and a song recommendation list (i.e., a recommended song collection). The songs in the song collection list are added by the user, and the songs in the song recommendation list are recommended by the music software based on the user's listening preferences. In some embodiments, the user's listening preferences can be determined by the music software based on the user's historical listening songs. The user can listen to the songs in the music software's song collection list, and the user can also listen to the songs recommended by the music software's song recommendation list.

[0031] However, in actual use, users sometimes listen to different types of songs as time changes. For example, users tend to listen to songs in the song recommendation list from 7 to 9 in the morning, and tend to listen to songs in the song collection list from 7 to 9 in the evening. This shows that users sometimes tend to listen to favorite songs in the song collection list, and sometimes tend to listen to recommended songs in the song recommendation list. This shows that users have needs for both song collection lists and song recommendation lists.

[0032] In view of the above-mentioned situation where users have demands for both a song collection list and a song recommendation list, the embodiment of the present invention provides a music recommendation method.

[0033] In the process of the music recommendation method, the user can obtain the number of user assets of the user in the music software by clicking the music recommendation mode in the music software in the user terminal, and generate a music recommendation request according to the number of user assets, and send the music recommendation request to the server. In some embodiments, the number of user assets can be used to refer to the number of songs in the user's song collection list. After receiving the music recommendation request, the server can determine whether the number of user assets in the music recommendation request is greater than the preset number threshold. If the judgment result is that the number of user assets is greater than the preset number threshold, the user's listening intention probability value can be calculated by the target calculation model. In some embodiments, the listening intention probability value can be a ratio value of the first number of songs collected in the song collection list and the second number of songs recommended in the recommended song list. After calculating the user's listening intention probability value, the first song set can be screened from the song collection list according to the listening intention probability value through a shuffle algorithm. In some embodiments, the shuffle algorithm can be a fisher-yates shuffle algorithm; at the same time, the user feature information, the listening feature information, the first song set, and the feature information of the songs to be recommended can be input into the pre-trained song recommendation model to obtain the second song set. Finally, the first song set and the second song set are merged to obtain the target song set. The server can send the target song set to the user terminal and recommend the target song set to the user corresponding to the user terminal. In the embodiment of the present invention, songs can be recommended to users more effectively by combining the collection song list and the recommended song set to recommend songs to users.

[0034] The music recommendation method provided in the embodiment of the present invention can be applied to a music recommendation system, wherein the system includes a music recommendation device and a user terminal. The music recommendation device can be set in a server. In some embodiments, the server can include but is not limited to smart phones, tablet computers, laptop computers, desktop computers, vehicle-mounted smart terminals, smart watches and other smart terminal devices. In some embodiments, the server includes one or more databases, which can be used to store audio files such as songs and other contents. In some embodiments, the server can be a cloud server. In some embodiments, the server can be a single server or a cluster of one or more servers composed of a series of servers. In some embodiments, the server can also be other devices with computing capabilities. In some embodiments, the user terminal can include but is not limited to smart phones, tablet computers, laptop computers, desktop computers, vehicle-mounted smart terminals, smart watches and other smart terminal devices; in some embodiments, a music client such as music software can be installed on the user terminal.

[0035] The following is combined with Figure 1 The music recommendation system provided by the embodiment of the present invention is schematically described.

[0036] See also Figure 1 , Figure 1 It is a structural diagram of a music recommendation system provided in an embodiment of the present invention. The system includes a user terminal 11 and a server 12. In some embodiments, the user terminal 11 and the server 12 can establish a communication connection through wireless communication; wherein, in some scenarios, the user terminal 11 and the server 12 can also establish a communication connection through wired communication.

[0037] In the embodiment of the present invention, the server 12 can obtain the user characteristic information and the song listening characteristic information of the user sent by the user terminal 11, the user characteristic information includes the user identification information and / or the user portrait characteristic information, the song listening characteristic information includes one or more of the song listening type information, the song listening time information, the song switching information, and the song listening quantity information, the song listening type information includes the collection song information and the recommended song information, and the song listening time information includes one or more of the song listening time, the current time, and the time interval between the start time of the song listening and the current time; the server 12 can input the user characteristic information and the song listening characteristic information into the pre-trained target calculation model to obtain the user The server 12 can filter out the first song set from the user's song collection list according to the probability value of the intention to listen to songs, and extract the feature information of the songs to be recommended from each song to be recommended in the recommended song set, and input the user feature information, the listening feature information, the first song set and the feature information of the songs to be recommended into the pre-trained song recommendation model to obtain the second song set; the server 12 can fuse the first song set and the second song set to obtain the target song set, and push the target song set to the user terminal 11 corresponding to the user, and the user terminal 11 can output the recommendation result, i.e., the target song set, to the user. In some embodiments, the output method can be an interface displayed on the user terminal 11, in some embodiments, the output method can also be a voice prompt, in other embodiments, the output method can also be other methods, and the embodiment of the present invention does not specifically limit the output method.

[0038] It can be seen that the embodiment of the present invention can more accurately and effectively recommend songs preferred by users by combining user favorite songs with recommended songs, thereby improving the accuracy and effectiveness of song recommendations.

[0039] The following is combined with Figure 2 The music recommendation method provided by the embodiment of the present invention is schematically described.

[0040] For details, please see Figure 2 , Figure 21 is a flowchart of a music recommendation method provided by an embodiment of the present invention. The music recommendation method of the embodiment of the present invention can be executed by a music recommendation device, wherein the music recommendation device is set in a server, wherein the specific explanation of the server is as above. Specifically, the method of the embodiment of the present invention includes the following steps.

[0041] S201: Obtain user characteristic information and music listening characteristic information of the user.

[0042] In an embodiment of the present invention, a music recommendation device can obtain user characteristic information and song listening characteristic information of a user. The user characteristic information includes user identification information and / or user portrait characteristic information. In some embodiments, the song listening characteristic information includes one or more of song listening type information, song listening time information, song switching information, and song listening quantity information. The song listening type information includes collection song information and recommended song information. The song listening time information includes one or more of song listening time, current time, and the time interval between the start time of the song listening and the current time. In some embodiments, the user characteristic information and the song listening characteristic information are both vector information.

[0043] In some embodiments, user identification information includes but is not limited to user name, and user portrait feature information includes but is not limited to one or more of user age, gender, user listening preferences, etc. In some embodiments, the listening type information is 1 or 0, and the listening type information is determined based on the user's historical listening records within a preset time range from the current time, and is used to indicate whether it is a song in the song collection list. If the historical listening record is a song in the song collection list, the listening type information is the collection song information, and takes 1. If the historical listening record is not a song in the song collection list, the listening type information is the recommended song information, and takes 0. In some embodiments, the listening time information can be obtained through a timestamp. In some embodiments, the listening feature information can also include basic song information, such as the basic song information can include but is not limited to one or more of the song identification, language identification, singer identification, etc.

[0044] In one embodiment, before obtaining user characteristic information and listening characteristic information, the music recommendation device can obtain the number of songs in the user's song collection list when receiving a music recommendation request sent by a user terminal corresponding to the user; and determine whether the number of songs in the user's song collection list is greater than a preset number threshold; if the judgment result is yes, the step of obtaining user characteristic information and listening characteristic information can be executed; if the judgment result is no, a prompt message can be output, and the prompt message is used to instruct the user to add songs to the song collection list, and the music recommendation request is not sent until the number of songs in the song collection list is greater than the preset number threshold.

[0045] For example, assuming that the preset number threshold is 5, when the server receives a music recommendation request sent by a user terminal corresponding to the user, it can determine whether the number of songs in the user's song collection list is greater than 5. If the number of songs in the song collection list is greater than 5, it can be determined to execute the steps of obtaining user characteristic information and listening characteristic information. If the number of songs in the song collection list is less than or equal to 5, a prompt message can be output to instruct the user to add songs to the song collection list until the number of songs in the song collection list is greater than 5.

[0046] In one embodiment, a music recommendation request carries user information and user song listening information; when the music recommendation device obtains user feature information and song listening feature information, it can extract user vector information and song listening vector information from the user information and user song listening information carried in the music recommendation request, that is, determine the user vector information as user feature information, and determine the song listening vector information as song listening feature information.

[0047] In some embodiments, when the music recommendation device extracts the user vector information and the song vector information from the user information and the song information carried in the music recommendation request, it can use a preset feature extraction model to obtain the information. In some embodiments, the preset feature extraction model can be obtained by training a neural network model, and the present invention does not specifically limit the feature extraction model. In some embodiments, the neural network model used in the present invention can be a deep neural network (DNN), and the structure of the DNN can be Figure 3 Take this as an example to illustrate: Figure 3 It is a structural diagram of a DNN, such as Figure 7 The figure shows a schematic diagram of a three-layer DNN structure. The neural network layers inside the DNN can be divided into three categories: input layer, hidden layer and output layer. Generally speaking, the first layer is the input layer, the last layer is the output layer, and the layers in between are all hidden layers. The layers are fully connected, that is, any neuron in the i-th layer is connected to any neuron in the i+1-th layer.

[0048] S202: Inputting user feature information and song listening feature information into a pre-trained target calculation model to obtain a probability value of the user's song listening intention, and filtering out a first song set from the user's song collection list according to the probability value of the song listening intention.

[0049] In an embodiment of the present invention, the music recommendation device may input user feature information and song listening feature information into a pre-trained target calculation model to obtain a probability value of the user's song listening intention, and filter out a first song set from the user's song collection list according to the probability value of the song listening intention. In some embodiments, the pre-trained target calculation model may be obtained by training a neural network model. In some embodiments, the neural network model used in the embodiment of the present invention may include but is not limited to DNN.

[0050] In one embodiment, when the music recommendation device inputs the user characteristic information and the song listening characteristic information into the pre-trained target calculation model to obtain the user's listening intention probability value, the user characteristic information and the song listening characteristic information can be input into the pre-trained target calculation model to predict the first number of songs in the user's song collection list and the second number of songs in the user's recommended song list; and the ratio of the first number of songs to the second number of songs is determined as the user's listening intention probability value. In some embodiments, the listening intention probability value is used to indicate the probability value of the user's tendency to listen to the songs in the song collection list.

[0051] For example, the music recommendation device inputs the user feature information and song listening feature information obtained from the music recommendation request sent by user 1 into the pre-trained target calculation model, and predicts that the number of the first songs in user 1's song collection list is 10, and the number of the second songs in user 1's song recommendation list is 20. It can be determined that the ratio of the first song number 10 to the second song number 20 is 1 / 2, and therefore the probability value of user 1's intention to listen to songs can be determined to be 0.5.

[0052] In one embodiment, when the probability value of the user's intention to listen to songs is greater than a first preset threshold, the probability value of the intention to listen to songs is confirmed to be the first threshold, and when the probability value of the user's intention to listen to songs is less than a second preset threshold, the probability value of the intention to listen to songs is confirmed to be the second threshold, wherein the first preset probability is greater than the second preset probability. For example, assuming that the probability value of the intention to listen to songs is P, the first preset threshold is 0.8, and the second preset threshold is 0.2, when P is greater than 0.8, it is determined that the P value is 0.8, and when P is less than 0.2, it is confirmed that the P value is 0.2.

[0053] In one embodiment, before inputting user feature information and song listening feature information into a pre-trained target computing model to obtain the user's probability value of listening intention, the music recommendation device can obtain a first training data set, the first training data set including multiple first training data, the multiple first training data including first training user feature information and first training song listening feature information; input the multiple first training data in the first training data set into a preset first deep neural network model for training to obtain a first loss function value; adjust the first model parameters of the first deep neural network model according to the first loss function value; input the multiple first training data in the first training data set into the first deep neural network model after adjusting the first model parameters for retraining; when the first loss function value obtained by retraining is less than the first preset threshold, it is determined that the target computing model is obtained.

[0054] Specific can Figure 4 Take this as an example to illustrate: Figure 4 It is a training schematic diagram of a target computing model provided by an embodiment of the present invention, wherein a first training data set 41 is input into a preset first feature extraction model 42 to obtain first training user feature information and first training song listening feature information, wherein the first training data set 41 includes user identification information, user portrait information, song listening type information, song listening time information, song switching information, song listening quantity information and other information, the first user feature information and the first training song listening feature information are input into a first deep neural network model 43 for training to obtain a first loss function value 44; the first model parameters of the first deep neural network model 43 are adjusted according to the first loss function value 44; multiple first training data are input into the first deep neural network model 43 after the first model parameters are adjusted for retraining; when the first loss function value 44 obtained by retraining is less than the first preset threshold value, it is determined that the target computing model is obtained.

[0055] In one embodiment, when the music recommendation device filters out the first song set from the user's song collection list based on the probability value of listening to songs, it can use a shuffling algorithm to randomly select a song priority list from the song collection list; determine the number of songs to be filtered out based on the probability value of listening to songs and the number of collected songs in the song priority list; and filter out the first song set from the song priority list based on the number of songs to be filtered out.

[0056] In one embodiment, when the music recommendation device randomly selects a song priority list from a song collection list using a shuffling algorithm, the last song may be exchanged with one of the previous n-1 songs (or not), and then the second-to-last song may be exchanged with one of the previous n-2 songs, and so on until the last element (i.e., song), and the shuffling is completed. The algorithm ensures that the probability of each element in each position is equal.

[0057] For example, assuming that the songs in the song collection list are initially sorted as song 0, song 1, song 2, song 3, song 4, which is represented by 01234, the steps for randomly selecting a song priority list from the song collection list are:

[0058] Randomly select a number (for example, 3) from the 5 positions [0,4] (including 0 and 4) and exchange it with number 4, and we get 01243. The probability of putting 3 in position 4 is 1 / 5; randomly select a number (for example, 0) from the 4 positions [0,3] (including 0 and 3) and exchange it with number 3, and we get 41203. The probability of putting 0 in position 3 is (4 / 5)*(1 / 4)=1 / 5; randomly select a number (for example, 0) from the 3 positions [0,2] (including 0 and 2) and exchange it with number 2, and we get 21403. The probability of putting 4 in position 2 is (4 / 5)*(3 / 4)*(1 / 3)=1 / 5; randomly select a number (for example, 0) from the two positions [0,1] (including 0 and 1) and exchange it with 1, and we get 12403. The probability of putting 2 in the first position is (4 / 5)*(3 / 4)*(2 / 3)*(1 / 2)=1 / 5; randomly select a number (for example, 0) from the one position [0,0] (including 0 and 0) and exchange it with 0, and we get 12403. The probability of putting 1 in the 0th position is (4 / 5)*(3 / 4)*(2 / 3)*(1 / 2)=1 / 5. Therefore, the song priority list can be obtained as song 1, song 2, song 4, song 0, song 3.

[0059] In one embodiment, when the music recommendation device determines the number to be screened based on the probability value of listening to songs and the number of collected songs in the song priority list, it can determine the number to be screened based on the product of the probability value of listening to songs and the number of collected songs in the song priority list.

[0060] S203: Extract feature information of the song to be recommended from each song to be recommended in the recommended song set, and input the user feature information, the song listening feature information, the first song set and the feature information of the song to be recommended into a pre-trained song recommendation model to obtain a second song set.

[0061] In an embodiment of the present invention, the music recommendation device may obtain a recommended song set, extract feature information of the song to be recommended from each song to be recommended in the recommended song set, and input the user feature information, the listening feature information, the first song set, and the feature information of the song to be recommended into a pre-trained song recommendation model to obtain a second song set. In some embodiments, the pre-trained song recommendation model may include but is not limited to a DNN.

[0062] In one embodiment, when the music recommendation device inputs user characteristic information, listening characteristic information, a first song set, and characteristic information of songs to be recommended into a pre-trained song recommendation model to obtain a second song set, the user characteristic information, listening characteristic information, a first song set, and characteristic information of songs to be recommended can be input into the pre-trained song recommendation model to obtain a recommendation score value corresponding to each song to be recommended in the recommended song set; the songs to be recommended are sorted in descending order according to the recommendation score value corresponding to each song to be recommended, and the songs to be recommended ranked in the top M positions are determined to form the second song set, where M is a positive integer.

[0063] For example, assuming that the value of M is 5, the music recommendation device can input the user feature information of user 1, the listening feature information, the first song set of user 1, and the feature information of the songs to be recommended into the pre-trained song recommendation model, and obtain the recommendation score value corresponding to song 1 in the 10 songs in the song set to be recommended is 80, the recommendation score value corresponding to song 2 is 82, the recommendation score value corresponding to song 3 is 83, the recommendation score value corresponding to song 4 is 81, the recommendation score value corresponding to song 5 is 79, the recommendation score value corresponding to song 6 is 84, the recommendation score value corresponding to song 7 is 78, the recommendation score value corresponding to song 8 is 85, the recommendation score value corresponding to song 9 is 86, and the recommendation score value corresponding to song 10 is 88. According to the recommendation score value corresponding to each song to be recommended, the songs to be recommended are sorted in descending order as follows: song 10, song 9, song 8, song 6, song 3, song 2, song 4, song 1, song 5, song 7, and it can be determined that the top 5 songs to be recommended, song 10, song 9, song 8, song 6, song 3, song 7, are the second song set.

[0064] In one embodiment, before inputting user characteristic information, listening characteristic information, a first song set, and characteristic information of songs to be recommended into a pre-trained song recommendation model to obtain a second song set, the music recommendation device may obtain a second training data set, the second training data set including multiple second training data, the multiple second training data including one or more of second training user characteristic information, second training listening characteristic information, and training listening intention song information; input multiple second training data in the second training data set into a second deep neural network model for training to obtain a second loss function value; adjust the second model parameters of the second deep neural network model according to the second loss function value; input multiple second training data in the second training data set into the second deep neural network model after adjusting the second model parameters for retraining; when the second loss function value obtained by retraining is less than a second preset threshold, it is determined that a song recommendation model is obtained. In some embodiments, the training listening intention song information may include but is not limited to the listening intention probability value of the training user.

[0065] Specific can Figure 5 Take this as an example to illustrate: Figure 5 It is a training diagram of a song recommendation model provided by an embodiment of the present invention, wherein a second training data set 51 is input into a preset second feature extraction model 52 to obtain second training user feature information, second training song listening feature information, and a training song listening intention probability value, wherein the second training data set 51 includes user identification information, user portrait information, song listening type information, song listening time information, song switching information, song listening quantity information, and song listening intention song information, and the second user feature information, the second training song listening feature information, and the training song listening intention probability value are input into a second deep neural network model 53 for training to obtain a second loss function value 54; the second model parameters of the second deep neural network model 53 are adjusted according to the second loss function value 54; multiple second training data are input into the second deep neural network model 53 after the second model parameters are adjusted for retraining; when the second loss function value 54 obtained by retraining is less than the second preset threshold value, it is determined that a song recommendation model is obtained.

[0066] S204: The first song set and the second song set are merged to obtain a target song set, and the target song set is pushed to a user terminal corresponding to the user.

[0067] In the embodiment of the present invention, the music recommendation device can perform fusion processing on the first song set and the second song set to obtain a target song set, and push the target song set to a user terminal corresponding to the user.

[0068] In one embodiment, when the music recommendation device fuses the first song set and the second song set to obtain the target song set, the first song set and the second song set may be fused according to the first preset weight of the first song set and the second preset weight of the second song set to obtain the target song set. In other embodiments, the music recommendation device may also fuse the first song set and the second song set in other ways, and the embodiment of the present invention does not specifically limit the way in which the first song set and the second song set are fused.

[0069] The embodiment of the invention can input user characteristic information and song listening characteristic information into a pre-trained target calculation model to obtain the user's listening intention probability value; filter out a first song set from the user's song collection list according to the listening intention probability value; input user characteristic information, song listening characteristic information, the first song set and the characteristic information of the songs to be recommended into the pre-trained song recommendation model to obtain a second song set; fuse the first song set and the second song set to obtain a target song set, and push the target song set to the user terminal corresponding to the user. By combining the user's listening intention at different times, the user's collection of songs and the recommended songs, songs that meet the user's preferences in different time periods can be recommended to the user more accurately and effectively, meeting the user's listening needs in different time periods and improving the accuracy and effectiveness of song recommendations.

[0070] See also Figure 6 , Figure 6 6 is a structural diagram of a music recommendation device provided by an embodiment of the present invention. Specifically, the music recommendation device is set in a server, and the device includes: an acquisition unit 601, a screening unit 602, a determination unit 603, and a push unit 604;

[0071] An acquisition unit 601 is used to acquire user characteristic information and song listening characteristic information of a user;

[0072] A screening unit 602 is used to input the user characteristic information and the song listening characteristic information into a pre-trained target calculation model to obtain a probability value of the user's song listening intention, and screen out a first song set from the user's song collection list according to the probability value of the song listening intention;

[0073] The determination unit 603 is used to extract the feature information of the song to be recommended from each song to be recommended in the recommended song set, and input the user feature information, the song listening feature information, the first song set and the feature information of the song to be recommended into a pre-trained song recommendation model to obtain a second song set;

[0074] The push unit 604 is used to merge the first song set and the second song set to obtain a target song set, and push the target song set to the user terminal corresponding to the user.

[0075] Furthermore, when the screening unit 602 inputs the user characteristic information and the song listening characteristic information into the pre-trained target calculation model to obtain the probability value of the user's song listening intention, it is specifically used to:

[0076] Inputting the user characteristic information and the song listening characteristic information into a pre-trained target computing model, predicting a first number of songs in the user's song collection list and a second number of songs in the user's song recommendation list;

[0077] Determine the ratio of the first number of songs to the second number of songs as the probability value of the user's intention to listen to songs.

[0078] Further, when the screening unit 602 screens out the first song set from the user's song collection list according to the song listening intention probability value, it is specifically used to:

[0079] A song priority list is randomly selected from the song collection list using a shuffling algorithm;

[0080] Determine the number of songs to be screened according to the probability value of the intention to listen to songs and the number of songs collected in the song priority list;

[0081] The first song set is filtered out from the song priority list according to the number to be filtered.

[0082] Further, the determining unit 603 inputs the user characteristic information, the song listening characteristic information, the first song set and the characteristic information of the songs to be recommended into the pre-trained song recommendation model to obtain the second song set, which is specifically used for:

[0083] Input the user characteristic information, the song listening characteristic information, the first song set and the characteristic information of the songs to be recommended into a pre-trained song recommendation model to obtain a recommendation score value corresponding to each song to be recommended in the song set to be recommended;

[0084] The songs to be recommended are sorted in descending order according to the recommendation score value corresponding to each song to be recommended, and the songs to be recommended ranked in the first M positions are determined to form the second song set, where M is a positive integer.

[0085] Furthermore, when the push unit 604 performs a fusion process on the first song set and the second song set to obtain a target song set, it is specifically used to:

[0086] According to the first preset weight of the first song set and the second preset weight of the second song set, the first song set and the second song set are fused to obtain the target song set.

[0087] Furthermore, before the screening unit 602 inputs the user characteristic information and the song listening characteristic information into the pre-trained target calculation model to obtain the user's song listening intention probability value, it is also used to:

[0088] Acquire a first training data set, where the first training data set includes a plurality of first training data, where the plurality of first training data includes first training user feature information and first training song listening feature training information;

[0089] Inputting the plurality of first training data in the first training data set into a preset first deep neural network model for training to obtain a first loss function value;

[0090] Adjusting a first model parameter of the first deep neural network model according to the first loss function value;

[0091] Inputting the plurality of first training data in the first training data set into the first deep neural network model after adjusting the first model parameters for retraining;

[0092] When the first loss function value obtained by retraining is less than the first preset threshold, it is determined that the target computing model is obtained.

[0093] Furthermore, the determining unit 603 inputs the user characteristic information, the song listening characteristic information, the first song set and the characteristic information of the songs to be recommended into a pre-trained song recommendation model to obtain the second song set, and is further used for:

[0094] Acquire a second training data set, where the second training data set includes a plurality of second training data, where the plurality of second training data includes one or more of second training user feature information, second training song listening feature information, and training song listening intention song information;

[0095] Inputting the plurality of second training data in the second training data set into a second deep neural network model for training to obtain a second loss function value;

[0096] Adjusting a second model parameter of the second deep neural network model according to the second loss function value;

[0097] Inputting the plurality of second training data in the second training data set into the second deep neural network model after adjusting the second model parameters for retraining;

[0098] When the second loss function value obtained by retraining is less than the second preset threshold, it is determined that the song recommendation model is obtained.

[0099] Furthermore, before the acquisition unit 601 acquires the user characteristic information and the song listening characteristic information of the user, it is also used to:

[0100] When receiving a music recommendation request sent by a user terminal corresponding to the user, obtaining the number of songs in the song collection list of the user, and determining whether the number of songs in the song collection list of the user is greater than a preset number threshold;

[0101] If the judgment result is yes, then the step of obtaining the user characteristic information and the song listening characteristic information is executed;

[0102] If the judgment result is no, a prompt message is output, wherein the prompt message is used to instruct the user to add songs to the song collection list, and the music recommendation request is not sent until the number of songs in the song collection list exceeds the preset number threshold.

[0103] The embodiment of the invention can input user feature information and song listening feature information into a pre-trained target calculation model to obtain the user's listening intention probability value; filter out a first song set from the user's song collection list according to the listening intention probability value; input user feature information, song listening feature information, the first song set, and the feature information of songs to be recommended into a pre-trained song recommendation model to obtain a second song set; perform fusion processing on the first song set and the second song set to obtain a target song set, and push the target song set to the user terminal corresponding to the user. In this way, the user's listening needs in different environments are met, and songs can be recommended to users more effectively.

[0104] See also Figure 7 , Figure 7 701 and 702. The server is a schematic diagram of a server provided by an embodiment of the present invention. Specifically, the server includes: a memory 701 and a processor 702.

[0105] In one embodiment, the server further includes a data interface 703, and the data interface 703 is used to transfer data information between the server and other devices.

[0106] The memory 701 may include a volatile memory; the memory 701 may also include a non-volatile memory; the memory 701 may also include a combination of the above-mentioned types of memories. The processor 702 may be a central processing unit (CPU). The processor 702 may further include a hardware chip. The above-mentioned hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD) or a combination thereof. The above-mentioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA) or any combination thereof.

[0107] The memory 701 is used to store programs, and the processor 702 can call the program stored in the memory 701 to perform the following steps:

[0108] Obtaining user characteristic information and music listening characteristic information of the user;

[0109] Inputting the user characteristic information and the song listening characteristic information into a pre-trained target calculation model to obtain a probability value of the user's song listening intention, and filtering out a first song set from the user's song collection list according to the song listening intention probability value;

[0110] Extracting feature information of the song to be recommended from each song to be recommended in the recommended song set, and inputting the user feature information, the song listening feature information, the first song set and the feature information of the song to be recommended into a pre-trained song recommendation model to obtain a second song set;

[0111] The first song set and the second song set are merged to obtain a target song set, and the target song set is pushed to a user terminal corresponding to the user.

[0112] Further, when the processor 702 inputs the user characteristic information and the song listening characteristic information into the pre-trained target calculation model to obtain the probability value of the user's song listening intention, it is specifically used to:

[0113] Inputting the user characteristic information and the song listening characteristic information into a pre-trained target computing model, predicting a first number of songs in the user's song collection list and a second number of songs in the user's song recommendation list;

[0114] Determine the ratio of the first number of songs to the second number of songs as the probability value of the user's intention to listen to songs.

[0115] Further, when the processor 702 filters out the first song set from the user's song collection list according to the song listening intention probability value, it is specifically configured to:

[0116] A song priority list is randomly selected from the song collection list using a shuffling algorithm;

[0117] Determine the number of songs to be screened according to the probability value of the intention to listen to songs and the number of songs collected in the song priority list;

[0118] The first song set is filtered out from the song priority list according to the number to be filtered.

[0119] Further, the processor 702 inputs the user characteristic information, the song listening characteristic information, the first song set and the characteristic information of the songs to be recommended into the pre-trained song recommendation model to obtain the second song set, specifically for:

[0120] Input the user characteristic information, the song listening characteristic information, the first song set and the characteristic information of the songs to be recommended into a pre-trained song recommendation model to obtain a recommendation score value corresponding to each song to be recommended in the song set to be recommended;

[0121] The songs to be recommended are sorted in descending order according to the recommendation score value corresponding to each song to be recommended, and the songs to be recommended ranked in the first M positions are determined to form the second song set, where M is a positive integer.

[0122] Further, when the processor 702 performs fusion processing on the first song set and the second song set to obtain a target song set, it is specifically used to:

[0123] According to the first preset weight of the first song set and the second preset weight of the second song set, the first song set and the second song set are fused to obtain the target song set.

[0124] Furthermore, before the processor 702 inputs the user characteristic information and the song listening characteristic information into the pre-trained target calculation model to obtain the user's song listening intention probability value, it is also used to:

[0125] Acquire a first training data set, where the first training data set includes a plurality of first training data, where the plurality of first training data includes first training user feature information and first training song listening feature training information;

[0126] Inputting the plurality of first training data in the first training data set into a preset first deep neural network model for training to obtain a first loss function value;

[0127] Adjusting a first model parameter of the first deep neural network model according to the first loss function value;

[0128] Inputting the plurality of first training data in the first training data set into the first deep neural network model after adjusting the first model parameters for retraining;

[0129] When the first loss function value obtained by retraining is less than the first preset threshold, it is determined that the target computing model is obtained.

[0130] Furthermore, the processor 702 inputs the user characteristic information, the song listening characteristic information, the first song set, and the characteristic information of the songs to be recommended into a pre-trained song recommendation model to obtain the second song set, and is further used for:

[0131] Acquire a second training data set, where the second training data set includes a plurality of second training data, where the plurality of second training data includes one or more of second training user feature information, second training song listening feature information, and training song listening intention song information;

[0132] Inputting the plurality of second training data in the second training data set into a second deep neural network model for training to obtain a second loss function value;

[0133] Adjusting a second model parameter of the second deep neural network model according to the second loss function value;

[0134] Inputting the plurality of second training data in the second training data set into the second deep neural network model after adjusting the second model parameters for retraining;

[0135] When the second loss function value obtained by retraining is less than the second preset threshold, it is determined that the song recommendation model is obtained.

[0136] Furthermore, before the processor 702 obtains the user characteristic information and the song listening characteristic information of the user, it is also used to:

[0137] When receiving a music recommendation request sent by a user terminal corresponding to the user, obtaining the number of songs in the song collection list of the user, and determining whether the number of songs in the song collection list of the user is greater than a preset number threshold;

[0138] If the judgment result is yes, then the step of obtaining the user characteristic information and the song listening characteristic information is executed;

[0139] If the judgment result is no, a prompt message is output, wherein the prompt message is used to instruct the user to add songs to the song collection list, and the music recommendation request is not sent until the number of songs in the song collection list exceeds the preset number threshold.

[0140] The embodiment of the present invention can input user feature information and song listening feature information into a pre-trained target calculation model to obtain a user's listening intention probability value; filter out a first song set from the user's song collection list according to the listening intention probability value; input user feature information, song listening feature information, the first song set, and the feature information of songs to be recommended into a pre-trained song recommendation model to obtain a second song set; perform fusion processing on the first song set and the second song set to obtain a target song set, and push the target song set to the user terminal corresponding to the user. In this way, the user's listening needs in different environments are met, and songs can be recommended to users more effectively.

[0141] The embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the present invention is implemented. Figure 2 The method described in the corresponding embodiment can also be implemented Figure 6 The equipment corresponding to the embodiment of the present invention will not be described in detail here.

[0142] The computer-readable storage medium may be an internal storage unit of the device described in any of the foregoing embodiments, such as a hard disk or memory of the device. The computer-readable storage medium may also be an external storage device of the device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of the device and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the terminal. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0143] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0144] The above disclosure is only part of the embodiments of the present invention, which certainly cannot be used to limit the scope of the rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention are still within the scope of the invention.

Claims

1. A music recommendation method, characterized in that: include: Obtaining user characteristic information and music listening characteristic information of the user; Inputting the user characteristic information and the song listening characteristic information into a pre-trained target calculation model to obtain a song listening intention probability value of the user, and filtering out a first song set from the user's song collection list according to the song listening intention probability value; Extracting feature information of the song to be recommended from each song to be recommended in the recommended song set, and inputting the user feature information, the song listening feature information, the first song set and the feature information of the song to be recommended into a pre-trained song recommendation model to obtain a second song set; The first song set and the second song set are merged to obtain a target song set, and the target song set is pushed to a user terminal corresponding to the user.

2. The method according to claim 1, characterized in that The step of inputting the user characteristic information and the song listening characteristic information into a pre-trained target calculation model to obtain a probability value of the user's song listening intention includes: Inputting the user characteristic information and the song listening characteristic information into a pre-trained target computing model, predicting a first number of songs in the user's song collection list and a second number of songs in the user's song recommendation list; Determine the ratio of the first number of songs to the second number of songs as the probability value of the user's intention to listen to songs.

3. The method according to claim 2, characterized in that The step of selecting a first song set from the user's song collection list according to the song listening intention probability value comprises: A song priority list is randomly selected from the song collection list using a shuffling algorithm; Determine the number of songs to be screened according to the probability value of the intention to listen to songs and the number of songs collected in the song priority list; The first song set is filtered out from the song priority list according to the number to be filtered.

4. The method according to claim 1, characterized in that The step of inputting the user characteristic information, the song listening characteristic information, the first song set, and the to-be-recommended song characteristic information into a pre-trained song recommendation model to obtain a second song set includes: Input the user characteristic information, the song listening characteristic information, the first song set and the characteristic information of the songs to be recommended into a pre-trained song recommendation model to obtain a recommendation score value corresponding to each song to be recommended in the song set to be recommended; The songs to be recommended are sorted in descending order according to the recommendation score value corresponding to each song to be recommended, and the songs to be recommended ranked in the first M positions are determined to form the second song set, where M is a positive integer.

5. The method according to claim 1, characterized in that The step of fusing the first song set and the second song set to obtain a target song set includes: According to the first preset weight of the first song set and the second preset weight of the second song set, the first song set and the second song set are fused to obtain the target song set.

6. The method according to claim 1, characterized in that Before inputting the user characteristic information and the song listening characteristic information into the pre-trained target calculation model to obtain the user's song listening intention probability value, the method further includes: Acquire a first training data set, where the first training data set includes a plurality of first training data, and the plurality of first training data includes first training user feature information and first training song listening feature information; Input the multiple first training data in the first training data set into a preset first deep neural network model for training to obtain a first loss function value; adjust the first model parameters of the first deep neural network model according to the first loss function value; input the multiple first training data in the first training data set into the first deep neural network model after adjusting the first model parameters for retraining; When the first loss function value obtained by retraining is less than the first preset threshold, it is determined that the target computing model is obtained.

7. The method according to claim 1, characterized in that Before inputting the user characteristic information, the song listening characteristic information, the first song set, and the characteristic information of the songs to be recommended into a pre-trained song recommendation model to obtain the second song set, the method further includes: Acquire a second training data set, where the second training data set includes a plurality of second training data, where the plurality of second training data includes one or more of second training user feature information, second training song listening feature information, and training song listening intention song information; Input the multiple second training data in the second training data set into a second deep neural network model for training to obtain a second loss function value; adjust the second model parameters of the second deep neural network model according to the second loss function value; input the multiple second training data in the second training data set into the second deep neural network model after adjusting the second model parameters for retraining; When the second loss function value obtained by retraining is less than the second preset threshold, it is determined that the song recommendation model is obtained.

8. The method according to claim 1, characterized in that Before obtaining the user characteristic information and the song listening characteristic information of the user, the method further includes: When receiving a music recommendation request sent by a user terminal corresponding to the user, obtaining the number of songs in the song collection list of the user, and determining whether the number of songs in the song collection list of the user is greater than a preset number threshold; If the judgment result is yes, then the step of obtaining the user characteristic information and the song listening characteristic information is executed; If the judgment result is no, a prompt message is output, wherein the prompt message is used to instruct the user to add songs to the song collection list, and the music recommendation request is not sent until the number of songs in the song collection list exceeds the preset number threshold.

9. A server, characterized in that: The method comprises a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises a program, and the processor is configured to call the program to execute the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, which, when executed, are used to implement the method according to any one of claims 1 to 8.

Images