Music pushing method and device, equipment and medium

Abstract

The invention provides a music pushing method and device, equipment and a medium, and the method comprises the steps: obtaining a music recommendation demand of a user side, and extracting the demand key information in the music recommendation demand; obtaining a key information type to which the demand key information belongs so as to determine a music retrieval strategy of the music recommendation demand, and obtaining a candidate music set matched with the demand key information from a music resource library according to the music retrieval strategy; according to the candidate music set and a pre-acquired large model, screening out a target music list matched with the music recommendation demand from the candidate music set, and generating music description information of each piece of target music in the target music list; and pushing the target music list and the music description information of each piece of target music to the user side. According to the method and the device, the flexibility of acquiring the candidate music set is improved, the accuracy of music retrieval is improved, the accuracy of pushing the target music is improved, and a user side does not need to search the music by himself / herself.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and in particular to a music push method, apparatus, device, and medium. Background Art

[0002] With the development of technology, people can now achieve tasks such as continuous conversation and obtaining objective answers by inputting prompt words or question text into intelligent robots. In related technologies, pre-trained language models can be used to capture the demand information contained in user input information, thereby obtaining and pushing corresponding response information based on this demand information. However, pre-trained language models may not fully extract the demand information contained in the user input information, resulting in inaccurate extracted demand information and, consequently, a deviation between the pushed response information and the user's demand.

[0003] Therefore, it is very important to push accurate content to the user based on the demand information input by the user. Summary of the Invention

[0004] The present disclosure aims to solve one of the technical problems in the related art at least to a certain extent.

[0005] To this end, the first aspect of the present disclosure proposes a music push method.

[0006] A second aspect of the present disclosure provides a music pushing device.

[0007] A third aspect of the present disclosure provides an electronic device.

[0008] A fourth aspect of the present disclosure provides a computer-readable storage medium.

[0009] In a first aspect, the present disclosure proposes a music push method, which includes: obtaining music recommendation needs of a user terminal and extracting key information of the needs in the music recommendation needs; obtaining the key information type to which the key information of the needs belongs to determine a music retrieval strategy for the music recommendation needs, and obtaining a candidate music set that matches the key information of the needs from a music resource library according to the music retrieval strategy; screening a target music list that is adapted to the music recommendation needs from the candidate music set according to the candidate music set and a pre-acquired large model, and generating music description information for each target music in the target music list; and pushing the target music list and the music description information for each target music to the user terminal.

[0010] In addition, the music push method proposed in the first aspect of the present disclosure may also have the following additional technical features:

[0011] According to an embodiment of the present disclosure, the step of obtaining the key information type to which the key information of the requirement belongs so as to determine the music retrieval strategy of the music recommendation requirement, and obtaining a candidate music set matching the key information of the requirement from a music resource library according to the music retrieval strategy, includes: in response to the key information type to which the key information of the requirement belongs being a scene information type, determining that the scene resource retrieval strategy is the retrieval strategy of the music recommendation requirement; in response to the retrieval strategy being the scene resource retrieval strategy, obtaining an extended keyword of the key information of the requirement; obtaining a scene resource tag of the key information of the requirement from a tag resource library according to the key information of the requirement and the extended keyword; obtaining resource metadata of each music resource in the music resource library, and obtaining a plurality of first candidate music of the music recommendation requirement through the resource metadata of each music resource according to the key information of the requirement, the extended keyword and the scene resource tag, so as to obtain the candidate music set.

[0012] According to an embodiment of the present disclosure, the method of obtaining the key information type to which the key information of the requirement belongs to determine the music retrieval strategy of the music recommendation requirement, and obtaining a candidate music set matching the key information of the requirement from a music resource library according to the music retrieval strategy, includes: in response to the key information type to which the key information of the requirement belongs being a text keyword information type, determining a text keyword resource retrieval strategy as the retrieval strategy of the music recommendation requirement, and determining the key information of the requirement as a resource retrieval keyword; in response to the retrieval strategy being the text keyword resource retrieval strategy, obtaining resource metadata of each music resource in the music resource library, and according to the resource retrieval keyword, obtaining a plurality of second candidate music of the music recommendation requirement through the resource metadata of each music resource to obtain the candidate music set.

[0013] According to an embodiment of the present disclosure, the key information type to which the key information of the demand belongs is obtained to determine the music retrieval strategy of the music recommendation demand, and according to the music retrieval strategy, a candidate music set matching the key information of the demand is obtained from a music resource library, including: in response to the key information type to which the key information of the demand belongs being a fuzzy information type, determining that the intention resource retrieval strategy is the retrieval strategy of the music recommendation demand; in response to the retrieval strategy being the intention resource retrieval strategy, identifying the demand intent information of the music recommendation demand, and obtaining an estimated scenario vector of the music recommendation demand based on the demand intent information; obtaining a backup resource tag from a tag resource library based on the estimated scenario vector and the demand intent information; obtaining resource metadata of each music resource in the music resource library, and according to the demand intent information, the estimated scenario vector and the backup resource tag, obtaining multiple third candidate music of the music recommendation demand through the resource metadata of each music resource to obtain the candidate music set.

[0014] According to one embodiment of the present disclosure, the target music list adapted to the music recommendation requirement is screened out from the candidate music set based on the candidate music set and the pre-acquired big model, including: acquiring at least one group of multiple first candidate music, multiple second candidate music and multiple third candidate music to obtain the candidate music set that matches the key information of the requirement; generating model prompt words of the big model based on the candidate music set; and performing adaptation evaluation on each candidate music in the candidate music set in the model prompt words based on the model prompt words and the big model, so as to obtain the target music list that passes the adaptation evaluation from the candidate matching music set.

[0015] According to one embodiment of the present disclosure, generating the model prompt words of the large model based on the candidate music set includes: obtaining a prompt word construction template of the large model; and filling the candidate matching music set into the prompt word construction template based on the prompt word construction template to generate the model prompt words of the large model.

[0016] According to one embodiment of the present disclosure, the adaptation evaluation is performed on each candidate music in the candidate music set in the model prompt word based on the model prompt word and the big model to obtain the target music list that passes the adaptation evaluation from the candidate matching music set, including: inputting the model prompt word into the big model, and obtaining multiple target music that pass the adaptation evaluation from the candidate music set carried in the model prompt word through the big model; obtaining the music heat value of each of the multiple target music, and sorting the multiple target music by heat based on the order from high to low to form the target music list for the music recommendation requirement.

[0017] According to an embodiment of the present disclosure, the generation of music description information for each target music in the target music list includes: obtaining resource description information of each music resource in the music resource library to obtain target resource description information of each target music in the target music list; for any target music, integrating text information of the target resource description information of the target music and the music recommendation requirements to generate the music description information of the target music.

[0018] According to an embodiment of the present disclosure, the step of obtaining the key information type to which the required key information belongs, and obtaining a candidate music set that matches the required key information from a music resource library according to the key information type, includes: obtaining a singer list, singer introduction details of each singer in the singer list, and album information of each singer from an open source music information library, wherein, for any singer, the singer's singer introduction details and the album information are obtained based on the open source introduction information on each music APP; obtaining a song list and song information of each song in the song list from the open source music information library, wherein the song information includes at least lyrics information, comment information, and song details information of the song; constructing the music resource library based on the singer list, singer introduction details of each singer, album information of each singer, the song list, and song information of each song.

[0019] According to one embodiment of the present disclosure, obtaining the music recommendation demand of the user terminal and extracting the key information of the demand in the music recommendation demand include: obtaining input information of the user terminal and identifying the input intention of the input information; in response to identifying that the input intention is a music recommendation intention, obtaining the music recommendation demand of the user terminal; performing semantic recognition on the music recommendation demand and extracting the key information of the demand in the music recommendation demand.

[0020] The second aspect of the present disclosure proposes a music push device, which includes: a first acquisition module, used to obtain the music recommendation needs of a user terminal and extract the key information of the needs in the music recommendation needs; a second acquisition module, used to obtain the key information type to which the key information of the needs belongs, so as to determine the music retrieval strategy of the music recommendation needs, and obtain a candidate music set that matches the key information of the needs from a music resource library according to the music retrieval strategy; a third acquisition module, used to screen out a target music list adapted to the music recommendation needs from the candidate music set based on the candidate music set and a pre-acquired large model, and generate music description information of each target music in the target music list; a push module, used to push the target music list and the music description information of each target music to the user terminal.

[0021] In addition, the music push device proposed in the second aspect of the present disclosure may also have the following additional technical features:

[0022] According to one embodiment of the present disclosure, the second acquisition module is further used to: in response to the key information type to which the demand key information belongs being the scene information type, determine that the scene resource retrieval strategy is the retrieval strategy for the music recommendation demand; in response to the retrieval strategy being the scene resource retrieval strategy, obtain the extended keyword of the demand key information; obtain the scene resource tag of the demand key information from the tag resource library based on the demand key information and the extended keyword; obtain the resource metadata of each music resource in the music resource library, and according to the demand key information, the extended keyword and the scene resource tag, obtain multiple first candidate music for the music recommendation demand through the resource metadata of each music resource to obtain the candidate music set.

[0023] According to one embodiment of the present disclosure, the second acquisition module is further used to: in response to the key information type to which the key information of the requirement belongs being a text keyword information type, determine the text keyword resource retrieval strategy as the retrieval strategy for the music recommendation requirement, and determine the key information of the requirement as the resource retrieval keyword; in response to the retrieval strategy being the text keyword resource retrieval strategy, obtain resource metadata of each music resource in the music resource library, and according to the resource retrieval keyword, obtain multiple second candidate music for the music recommendation requirement through the resource metadata of each music resource to obtain the candidate music set.

[0024] According to one embodiment of the present disclosure, the second acquisition module is also used to: in response to the key information type to which the key information of the demand is a fuzzy information type, determine that the intention resource retrieval strategy is the retrieval strategy of the music recommendation demand; in response to the retrieval strategy being the intention resource retrieval strategy, identify the demand intention information of the music recommendation demand, and obtain the estimated scenario vector of the music recommendation demand based on the demand intention information; obtain a backup resource tag from a tag resource library based on the estimated scenario vector and the demand intention information; obtain resource metadata of each music resource in the music resource library, and obtain multiple third candidate music of the music recommendation demand through the resource metadata of each music resource based on the demand intention information, the estimated scenario vector and the backup resource tag to obtain the candidate music set.

[0025] According to one embodiment of the present disclosure, the third acquisition module is further used to: acquire at least one group of multiple first candidate music, multiple second candidate music and multiple third candidate music to obtain the candidate music set that matches the required key information; generate model prompt words of the large model based on the candidate music set; and perform adaptation evaluation on each candidate music in the candidate music set in the model prompt words based on the model prompt words and the large model to obtain the target music list that passes the adaptation evaluation from the candidate matching music set.

[0026] According to one embodiment of the present disclosure, the third acquisition module is further used to: obtain a prompt word construction template of the large model; and fill the candidate matching music set into the prompt word construction template according to the prompt word construction template to generate the model prompt words of the large model.

[0027] In the embodiment of the present disclosure, the third acquisition module is further used to: input the model prompt word into the large model, and obtain multiple target music that pass the adaptation evaluation from the candidate music set carried in the model prompt word through the large model; obtain the music popularity value of each of the multiple target music, and sort the multiple target music by popularity in order from high to low to form the target music list for the music recommendation requirement.

[0028] According to one embodiment of the present disclosure, the third acquisition module is also used to: obtain resource description information of each music resource in the music resource library to obtain target resource description information of each target music in the target music list; for any target music, integrate the target resource description information of the target music and the music recommendation requirements into text information to generate the music description information of the target music.

[0029] According to one embodiment of the present disclosure, the device also includes a construction module, which is used to: obtain a singer list, singer introduction details of each singer in the singer list and album information of each singer from an open source music information library, wherein, for any singer, the singer detailed introduction and the album information of the singer are obtained based on the open source introduction information on each music application software; obtain a song list and song information of each song in the song list from the open source music information library, wherein the song information includes at least lyrics information, comment information and song details information of the song; construct the music resource library based on the singer list, singer introduction details of each singer, album information of each singer, the song list and song information of each song.

[0030] According to one embodiment of the present disclosure, the first acquisition module is further used to: obtain input information from the user terminal and identify the input intention of the input information; in response to identifying that the input intention is a music recommendation intention, obtain the music recommendation demand of the user terminal; perform semantic recognition on the music recommendation demand, and extract key demand information in the music recommendation demand.

[0031] A third aspect of the present disclosure proposes an electronic device, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the instructions to implement the music push method proposed in the first aspect above.

[0032] In a fourth aspect, the present disclosure proposes a computer-readable storage medium. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the music push method proposed in the first aspect above.

[0033] The music push method and device proposed in the present disclosure obtain the music recommendation needs of the user end, extract the key information of the needs in the music recommendation needs, obtain the key information type to which the key information of the needs belongs, so as to obtain a candidate music set from a music resource library, obtain a target music list adapted to the music recommendation needs from the candidate music set based on the candidate music set and the big model, and generate music description information of each target music, and further push the target music list and the music description information of each target music to the user end. In the present disclosure, a music retrieval strategy is determined according to the key information type to which the required key information belongs, so as to obtain a candidate music set from a music resource library, and a target music list is obtained according to the candidate music set and a large model. The music retrieval strategy is determined according to the key information type to which the required key information belongs to obtain the candidate music set, thereby improving the flexibility of obtaining the candidate music set and improving the accuracy of music retrieval. In the scenario where the music resource library is updated, there is no need to train and optimize the model, which reduces the training cost of the model, improves the degree of adaptation between the target music list and the music recommendation requirements input by the user end, and thus improves the accuracy of the target music push. The user end only needs to input the music recommendation requirements to obtain the target music list from the music resource library. Compared with people's daily search for target music from their usual music application software, the music search range of the user end is expanded, the user end does not need to search for music by itself, and the time it takes for the user end to obtain the target music list is shortened, the music search experience of the user end is optimized, and user stickiness is increased.

[0034] It should be understood that the contents described in the present disclosure are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The above and / or additional aspects and advantages of the present disclosure will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0036] Figure 1 Schematic diagram of a music push method according to an embodiment of the present disclosure;

[0037] Figure 2 This is a flowchart of a music push method according to another embodiment of the present disclosure;

[0038] Figure 3 This is a flowchart of a music push method according to another embodiment of the present disclosure;

[0039] Figure 4 This is a schematic diagram of the vertical categories to which the large model belongs according to an embodiment of the present disclosure;

[0040] Figure 5 This is a flowchart of a music push method according to another embodiment of the present disclosure;

[0041] Figure 6 This is a flowchart of a music push method according to another embodiment of the present disclosure;

[0042] Figure 7 A schematic diagram of a process for constructing a music resource library according to an embodiment of the present disclosure;

[0043] Figure 8 This is a flowchart of a music push method according to another embodiment of the present disclosure;

[0044] Figure 9 This is a structural diagram of a music push device according to an embodiment of the present disclosure;

[0045] Figure 10 A block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0046] The following describes in detail embodiments of the present disclosure, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0047] The following describes a music push method, apparatus, device, and medium proposed in an embodiment of the present disclosure with reference to the accompanying drawings.

[0048] Figure 1 FIG. 1 is a flow chart of a music push method according to an embodiment of the present disclosure. Figure 1 As shown, the method includes:

[0049] S101, obtaining music recommendation requirements of the user end, and extracting key information of the requirements in the music recommendation requirements.

[0050] In the embodiment of the present disclosure, the user terminal can obtain the recommended music returned by the server through the input demand information, wherein the demand information input by the user terminal requiring the server to make music recommendations can be marked as the user terminal's music recommendation demand.

[0051] It should be noted that the music recommendation requirements input by the user can be input in the form of text or voice, and there is no specific limitation here.

[0052] As an example, Figure 2 As shown, the user can Figure 2 The text input box shown can be used to input music recommendation requirements in text form, or you can use Figure 2 The microphone icon shown allows for voice-based driving input of music recommendation requests.

[0053] Optionally, the music recommendation demand carries the user's key demand information for recommended music. The music recommendation demand can be semantically identified and parsed using a preset semantic parsing algorithm, and the key information therein can be extracted and marked as the key demand information in the music recommendation demand.

[0054] It can be understood that based on the key information of the demand, adaptive music recommendation can be achieved based on the music recommendation demand of the user end.

[0055] S102, obtaining the key information type to which the required key information belongs, to determine a music retrieval strategy for the music recommendation requirement, and obtaining a candidate music set matching the required key information from a music resource library according to the music retrieval strategy.

[0056] In the embodiment of the present disclosure, there are multiple ways to input the music recommendation requirements input by the user terminal. Among them, the user terminal can directly input clear indication information such as the style, name or singer of the required music, and the server can obtain suitable music for the user terminal based on the clear indication information input by the user terminal.

[0057] Correspondingly, the user terminal can also input the scene-type indication information of the current environment or mood. The server analyzes the music style currently required by the user terminal based on the scene information input by the user terminal, and pushes adapted music to the user terminal.

[0058] In this scenario, based on the music recommendation requirements input in different ways, the key information types of the extracted key information of the requirements are different.

[0059] As an example, assume that the music recommendation demand input by the user carries clear indication information of "Song Name 1", then the key information extracted from the music recommendation demand can be "Song Name 1", and the key information type to which the key information of the demand belongs can be marked as the information type corresponding to the clear indication information.

[0060] Also, if the music recommendation demand input by the user carries scene information expressing the emotion of "bad mood", the key information extracted from the music recommendation demand can be "sad, sad and quiet", and the key information type to which the key information of the demand belongs can be marked as the information type corresponding to the emotional scene information.

[0061] From the above examples, it can be seen that the key information types of the extracted key information of music recommendation requirements under different input methods are different.

[0062] In the embodiment of the present disclosure, there are differences between the music retrieval strategies for different key information types. In this scenario, the corresponding music search strategy can be determined based on the key information type to which the required key information belongs, and at least one matching music can be obtained from the music resource library and marked as at least one candidate music, thereby obtaining a candidate music set consisting of at least one candidate music.

[0063] As an example, based on the above example, for the information type corresponding to the clear indication information, the candidate music set can be obtained from the music resource library based on the relevant text retrieval strategy according to the clear key information of the requirements; and for the information type corresponding to the emotional scene information, the candidate music set can be obtained from the music resource library based on the emotional information retrieval strategy.

[0064] S103: Based on the candidate music set and the pre-acquired large model, a target music list that meets the music recommendation requirement is screened out from the candidate music set, and music description information of each target music in the target music list is generated.

[0065] In an embodiment of the present disclosure, the candidate music collection may include multiple candidate music to be recommended. In this scenario, at least one music whose adaptability to the music recommendation requirements input by the user terminal meets preset conditions can be screened out from the candidate music collection as the target music pushed to the user terminal, and a list consisting of the at least one screened target music can be marked as a target music list.

[0066] Optionally, input information corresponding to a large model can be generated based on the candidate music set, and the input information can be input into the large model. Based on the large model, the degree of adaptation between each candidate music in the candidate music set and the music recommendation requirements can be screened, and then a target music list consisting of at least one target music that is adapted to the music recommendation requirements can be screened out from the candidate music set.

[0067] As an example, setting Figure 2 The music song 1 of singer 1 is input by the server based on the user terminal. Figure 2 The target music pushed by “I want to listen to songs suitable for i people” is shown. Figure 2 As shown, when the server pushes the target music, it will also push the text description content of the music song 1 as the target music, wherein the text description content of the music song 1 can be Figure 2 The song states, "Singer 1's restrained voice, combined with the subtle melancholy and nostalgic atmosphere in the lyrics, is perfect for those with introverted personalities to reminisce about the past in quiet time."

[0068] In this example, you can Figure 2 The above text description content shown is determined to be Figure 2 The music description information of the music song 1 as the target music is shown.

[0069] Optionally, for any target music in the target music list, metadata information of the target music, such as Meta information, etc., can be obtained, and input information of the large model can be generated based on the obtained metadata information and input into the large model to extract and integrate the text content of the metadata information of the target music, thereby obtaining the music description information of the target music.

[0070] The music description information of the target music may include the reason information for pushing the target music, the description information of the emotional value that the target music can provide to the user terminal, and the reminder information of the server to the user terminal, which is not specifically limited here.

[0071] S104: Push the target music list and the music description information of each target music to the user terminal.

[0072] In the disclosed embodiment, there may be an interactive client between the server and the user end. In this scenario, the server can push each music item in the target music list it has obtained, as well as the music description information of each target music, to the user end through the display interface of the interactive client.

[0073] As an example, setting Figure 2 The interface shown is the client interface for interaction between the server and the user. Figure 2As shown, when the server receives the music recommendation request "What song to listen to when having poop" from the user, it will Figure 2 The music song 2 of the singer 2 is shown as the target music. Figure 2 On the client display interface shown, music song 2 and Figure 2 The music description information of the music song 2 shown, "The classic voice of singer 2 and the gentle melody of this song can create a calm atmosphere, help you concentrate and successfully complete the "big thing"." is pushed to the user terminal.

[0074] The music push method proposed in the present invention obtains the music recommendation demand of the user end, extracts the key information of the demand in the music recommendation demand, obtains the key information type to which the key information of the demand belongs, obtains a candidate music set from a music resource library, obtains a target music list adapted to the music recommendation demand from the candidate music set based on the candidate music set and a large model, and generates music description information of each target music, and further pushes the target music list and the music description information of each target music to the user end. In the present disclosure, a music retrieval strategy is determined according to the key information type to which the required key information belongs, so as to obtain a candidate music set from a music resource library, and a target music list is obtained according to the candidate music set and a large model. The music retrieval strategy is determined according to the key information type to which the required key information belongs to obtain the candidate music set, thereby improving the flexibility of obtaining the candidate music set and improving the accuracy of music retrieval. In the scenario where the music resource library is updated, there is no need to train and optimize the model, which reduces the training cost of the model, improves the degree of adaptation between the target music list and the music recommendation requirements input by the user end, and thus improves the accuracy of the target music push. The user end only needs to input the music recommendation requirements to obtain the target music list from the music resource library. Compared with people's daily search for target music from their usual music application software, the music search range of the user end is expanded, the user end does not need to search for music by itself, and the time it takes for the user end to obtain the target music list is shortened, the music search experience of the user end is optimized, and user stickiness is increased.

[0075] In the above embodiment, the acquisition of the target music list can be combined with Figure 3 Further understanding, Figure 3 FIG. 1 is a flow chart of a music push method according to another embodiment of the present disclosure. Figure 3 As shown, the method includes:

[0076] S301, obtaining the music recommendation demand of the user end, and extracting the key demand information in the music recommendation demand.

[0077] In an embodiment of the present disclosure, the recommendation vertical category to which the demand information input by the user terminal belongs may not belong to the music recommendation vertical category. In this scenario, the input information of the user terminal can be analyzed and identified, wherein the input information of the user terminal is obtained, and the input intention of the input information is identified. wherein, in response to identifying that the input intention is a music recommendation intention, the music recommendation demand of the user terminal is obtained, and the music recommendation demand is semantically identified to extract key demand information in the music recommendation demand.

[0078] As an example, Figure 4 As shown, the base model of the large model may include Figure 4 The content type big model, control type big model and life service type big model are shown, among which different types of big models belong to different vertical categories.

[0079] In this example, if the recommended vertical category of the demand information input by the user is the vertical category of the content information, Figure 4 The content class model shown provides corresponding push content to the user terminal.

[0080] And, if the recommended vertical category of the demand information input by the user is the vertical category of the control information, then Figure 4 The control class model shown provides corresponding push content to the user terminal.

[0081] And, if the recommended vertical category of the demand information input by the user is the vertical category of life service information, then Figure 4 The large model of life services shown provides corresponding push content to the user end.

[0082] Optionally, after obtaining the input information from the user side, the input information can be algorithmically processed based on the intent recognition algorithm in the relevant technology, and then the information intent carried in the input information from the user side can be obtained according to the result of the algorithm processing, and it can be marked as the input intention of the user side.

[0083] In this scenario, the vertical category to which the user's recommendation needs belong can be identified through input intention. The identification strategy of each vertical category can be obtained and compared with the user's input intention. When it is identified that the user's input intention matches the identification strategy of the music recommendation vertical category, it can be determined that the vertical category to which the user's input intention belongs is the music recommendation vertical category, and further, it can be determined that the user's input intention is a music recommendation intention.

[0084] As an example, Figure 4 As shown, the large model used in the music recommendation vertical category is set to Figure 4The content category model shown can identify and analyze the demand information input by the user, obtain the input intention of the user's input information, and identify the recommendation vertical category to which the input intention belongs as the music recommendation vertical category.

[0085] In this example, you can Figure 4 The content type big model used in the music recommendation vertical category is obtained from the corresponding vertical categories of the content type big model, the corresponding vertical categories of the control type big model, and the corresponding vertical categories of the life service type big model shown, and the corresponding push content is provided for the demand information input by the user end.

[0086] Optionally, when the user's input and the intention to recommend music are identified, the user's input information can be used as a music recommendation requirement, and key information of the requirement can be extracted.

[0087] Among them, the music recommendation demand can be processed by a semantic recognition algorithm based on the semantic recognition algorithm in the relevant technology. Through the processing results of the semantic recognition algorithm, the key information is extracted from the semantic information of the music recommendation demand as the key information of the demand in the music recommendation demand.

[0088] Correspondingly, information extraction processing can also be performed on the music recommendation demand based on the key information extraction algorithm in the relevant technology, so as to extract the key information in the music recommendation demand as the demand key information.

[0089] S302, obtaining the key information type to which the required key information belongs, to determine a music retrieval strategy for the music recommendation requirement, and obtaining a candidate music set matching the required key information from a music resource library according to the music retrieval strategy.

[0090] As a possible implementation method, the key information type of the demand key information in the music recommendation demand input by the user side can be a scene information type, wherein, in response to the key information type of the demand key information being a scene information type, the scene resource retrieval strategy is determined as the retrieval strategy for the music recommendation demand.

[0091] In the embodiment of the present disclosure, the music recommendation requirement input by the user end can be the scene atmosphere information of the song that the user wants to listen to, such as Figure 2 As shown, when the music recommendation requirement input by the user is Figure 2 When the message "I want to listen to songs suitable for person i" is displayed, it can be determined that the scene atmosphere information of the song the user wants to listen to is the scene atmosphere information "suitable for person i", and then the key information of the demand "suitable for person i" can be extracted from the music recommendation demand, and the key information type to which the key information of the demand belongs can be determined to be the scene information type.

[0092] Optionally, in response to the retrieval strategy being a scene resource retrieval strategy, an extended keyword of the requirement key information is obtained, and based on the requirement key information and the extended keyword, a scene resource tag of the requirement key information is obtained from a tag resource library.

[0093] In the disclosed embodiment, when the key information of the music recommendation requirement is a scene information type, the search strategy of the music resource corresponding to the scene information type can be used as the search strategy for the push content provided by the user end.

[0094] Among them, the retrieval strategy can be marked as a scene resource retrieval strategy.

[0095] In an embodiment of the present disclosure, keyword expansion can be performed on the demand key information of the scene information type to improve the retrieval accuracy of the scene resource retrieval strategy, wherein the demand key information can be algorithmically processed based on a preset keyword expansion algorithm, and then the expanded keywords of the demand key information can be obtained according to the results of the algorithm processing.

[0096] Correspondingly, a preset scenario extension keyword storage information table can also be obtained, wherein there is a mapping relationship between the scenario information and the keyword information in the information table. In this scenario, the requirement key information can be used as a retrieval keyword, and the mapping keyword of the requirement key information can be obtained from the storage information table according to the mapping relationship as an extended keyword of the requirement key information.

[0097] In the disclosed embodiment, each music resource in the music resource library has tag information. The style, atmosphere, emotion and other related information of the music resource can be obtained through the tag information of the music resource. In this scenario, the resource library composed of the tag information of each music resource can be marked as a tag resource library.

[0098] In the embodiment of the present disclosure, the key information of the music recommendation requirement is of the scene information type. In this scenario, based on the startup execution of the corresponding scene resource retrieval strategy, multiple tag information associated with the music recommendation requirement can be obtained from the tag resource library according to the key information of the requirement and the corresponding extended keywords as multiple scene resource tags for the music recommendation requirement.

[0099] For example, Figure 2 Taking the music recommendation requirement "I want to listen to songs suitable for i people" as an example, the key information of the music recommendation requirement is "suitable for i people", and the type of the key information is the scene information type. Then, the extended keywords of the key information can be obtained, such as soothing, sad, etc., and then, based on the key information and the extended keywords, the scene resource tags related to the music recommendation requirement "I want to listen to songs suitable for i people" are obtained from the tag resource library.

[0100] Optionally, resource metadata of each music resource in the music resource library is obtained, and based on the required key information, extended keywords and scene resource tags, multiple first candidate music associated with the scene resource tags are obtained through the resource metadata of each music resource to obtain a candidate music set.

[0101] In the disclosed embodiment, each music resource in the music resource library has metadata, which can be marked as resource metadata. For any music resource, music description information of the music resource can be obtained from the resource metadata of the music resource.

[0102] In this scenario, the resource metadata of each music resource in the music resource library can be obtained, and the combined data consisting of the required key information, extended keywords and scene resource tags can be used as search keywords. Multiple related resource metadata can be filtered out from the resource metadata of each music resource, and the music resources of each of the multiple resource metadata can be marked as the first candidate music, thereby obtaining multiple first candidate music for music recommendation needs, and then forming a candidate music set.

[0103] As another possible implementation, in response to the key information type to which the requirement key information belongs being a text keyword information type, the text keyword resource retrieval strategy is determined as the retrieval strategy for music recommendation requirements, and the requirement key information is determined as the resource retrieval keyword.

[0104] In an embodiment of the present disclosure, the music recommendation demand input by the user may carry clear text keywords. In this scenario, clear text keywords of music resources can be extracted from the music recommendation demand as demand key information, and the key information type to which the demand key information belongs can be marked as a text keyword information type.

[0105] For example, the user terminal can input the music recommendation demand "I want to listen to dynamic songs". The key information extracted from the music recommendation demand can be "dynamic". Then the key information of the "dynamic" demand in the music recommendation demand can be extracted, and it can be determined that the key information belongs to the text keyword information type.

[0106] Furthermore, the music resource retrieval strategy corresponding to the text keyword information type can be determined as the text keyword resource retrieval strategy, and the required key information can be determined as the retrieval keyword used when performing music resource retrieval and marked as the resource retrieval keyword.

[0107] Optionally, in response to the retrieval strategy being a text keyword resource retrieval strategy, resource metadata of each music resource in the music resource library is obtained, and based on the resource retrieval keyword, multiple second candidate music for music recommendation needs are obtained through the resource metadata of each music resource to obtain a candidate music set.

[0108] In the disclosed embodiment, when the retrieval strategy for music resources is a text keyword resource retrieval strategy, the server can search the resource metadata of each music resource in the music resource library based on the resource retrieval keyword, obtain multiple resource metadata associated with the resource retrieval keyword, and then determine the music resources of each of the multiple resource metadata as the second candidate music of each of the multiple resource metadata, thereby obtaining a candidate resource set consisting of multiple second candidate music.

[0109] As another possible implementation, in response to the key information type to which the required key information belongs being a fuzzy information type, the intended resource retrieval strategy is determined to be a retrieval strategy for music recommendation requirements.

[0110] In the embodiment of the present disclosure, the key information of the demand in the music recommendation demand input by the user may be fuzzy information, wherein the music recommendation demand may not carry scene information and / or text keyword information, and may not carry other clear demand information for music resource retrieval. In this scenario, it can be determined that the key information type of the key information of the demand in the music recommendation demand is a fuzzy information type.

[0111] Among them, the music resource retrieval strategy corresponding to the fuzzy information type can be obtained and marked as the intended resource retrieval strategy.

[0112] Optionally, in response to the retrieval strategy being an intention resource retrieval strategy, the demand intent information of the music recommendation demand is identified, and the estimated scenario vector of the music recommendation demand is obtained based on the demand intent information, and based on the estimated scenario vector and the demand intent information, a backup resource tag is obtained from the tag resource library.

[0113] In the embodiment of the present disclosure, when the retrieval strategy for music resources is an intention resource retrieval strategy, the server can perform intent analysis on the music recommendation requirements input by the user terminal, obtain the listening intention carried by the user terminal when inputting the music recommendation requirements through the analysis results, and retrieve the music resources based on the intent information, thereby obtaining multiple third candidate music for the music recommendation requirements.

[0114] Among them, according to the intention analysis algorithm in the relevant technology, the music recommendation needs input by the user end can be algorithmically analyzed, and then the user end's demand intention information for music resources can be obtained based on the results of the algorithm analysis.

[0115] Optionally, the demand intention information may be subjected to scene estimation according to a preset scene vector estimation algorithm, and the estimated scene vector may be marked as the estimated scene vector of the music recommendation demand.

[0116] It can be understood that since the demand intention information obtained from the intent analysis may deviate from the user's actual music needs, the possible listening scenarios of the user side can be estimated based on the demand intention information, and then music resources can be retrieved based on the estimated scenario vector and demand intention information to improve the retrieval accuracy.

[0117] Optionally, the scene vector can be estimated based on the demand intention information through a pre-acquired scene vector estimation model (embedding model). It can be understood that the scene vector estimation model is used to estimate the scene atmosphere when the user inputs the music recommendation demand based on the demand intention information, thereby obtaining the estimated scene vector of the music recommendation demand.

[0118] In this scenario, the combination of the estimated scene vector and the demand intention information can be used as a search keyword to obtain the associated resource tags from the tag resource library as a backup tag for music recommendation needs.

[0119] Optionally, the combined information of the estimated scenario vector and the demand intention information can be used as a search keyword to search in the label resource library to obtain the correlation parameters between the combined information of the estimated scenario vector and the demand intention information and each label resource. For any label resource, when the correlation parameters between the label resource and the combined information of the estimated scenario vector and the demand intention information meet the preset correlation conditions, the label resource can be determined as a fallback resource label.

[0120] As an example, for the scene information of "speeding", if the key information of the demand in the music recommendation demand input by the user is fuzzy information, the demand intention information in the key information of the demand is obtained, and the scene vector of the demand intention information is estimated through the scene vector prediction model to obtain the estimated scene vector corresponding to the "speeding" scene.

[0121] In this example, based on the combined information of the estimated scene vector and the key demand information corresponding to the "racing" scene, multiple resource tags such as "excitement" and "rock" associated with the "racing" scene can be obtained from the tag resource library as multiple backup resource tags for the music recommendation demand.

[0122] Optionally, resource metadata of each music resource in the music resource library is obtained, and based on the demand intent information, estimated scenario vector and backup resource label, multiple third candidate music for music recommendation needs are obtained through the resource metadata of each music resource to obtain a candidate music set.

[0123] In the disclosed embodiment, based on the combined information of demand intention information, estimated scenario vector and backup resource label, multiple related music resources can be obtained from the resource metadata of each music resource in the music resource library, and the multiple related music resources can be respectively marked as third candidate music, thereby obtaining a candidate music set consisting of multiple third candidate music.

[0124] In order to better understand the method for obtaining the candidate music set proposed above, we can combine Figure 5 ,like Figure 5 As shown, the extended keyword information of each scene can be based on Figure 5 The key-value pairs are stored as shown, Figure 5 As shown, the key column is the scene information, and the value column is the extended keyword information of each scene information, where Figure 5 The extended keyword information of the self-driving scene shown is Figure 5 Shows ethnic customs, travel, and, Figure 5 The extended keyword information of the racing scene shown is Figure 5 Shows fast-paced, heavy metal.

[0125] like Figure 5 As shown, the user can input music recommendation requirements in the form of voice, and Figure 5 The automatic speech recognition module shown performs speech recognition on the music recommendation demand to extract key demand information in the music recommendation demand.

[0126] Optionally, when the key information required belongs to the scenario information type, it can be understood in conjunction with the following:

[0127] like Figure 5 As shown, when the key information required is scene type information self-driving, you can use Figure 5 The keyword expansion module shown is based on the required key information from Figure 5 Table 1 shows the extended keyword information of self-driving scene information obtained, which includes ethnic style and travel.

[0128] Further, according to the scene information of self-driving, the extended keyword information of ethnic style and travel, the scene resource tag of folk songs is obtained from the tag resource library, and then according to the scene information of self-driving, the extended keyword information of ethnic style and travel and the scene resource tag of folk songs, through Figure 5 The recall module shown is from Figure 5 In the music resource library shown in Table 2, music resource 1 is recalled as the first candidate music for the music recommendation requirement.

[0129] Optionally, when the key requirement information is a text keyword information type, it can be understood in conjunction with the following:

[0130] like Figure 5As shown, when the key information required is the text keyword information type fast rhythm, heavy metal, the keyword resource tag of rock can be obtained from the tag resource library according to the text keyword information of fast rhythm, heavy metal, and then according to the text keyword information of fast rhythm, heavy metal and the keyword resource tag of rock, Figure 5 The recall module shown is from Figure 5 Table 2 shows that music resource 2 is recalled from the music resource library as the second candidate music for the music recommendation requirement.

[0131] Optionally, when the key information of the requirement is fuzzy, it can be understood in combination with the following:

[0132] like Figure 5 As shown, when the key information of the requirement is fuzzy information type, it can be Figure 5 The scene vector estimation module shown obtains demand intention information of the demand key information and performs scene vector estimation based on the demand intention information to obtain the estimated scene vector movement of the music recommendation demand.

[0133] Furthermore, the corresponding extended keyword information fast-paced and exciting is obtained according to the estimated scene vector of the movement, and the associated backup resource tag rock is obtained from the tag resource library according to the estimated scene information of the drag racing and the extended keyword information of fast-paced and exciting. Figure 5 The recall module shown is from Figure 5 In the music resource library shown in Table 2, music resource 4 is recalled as the third candidate music for the music recommendation requirement.

[0134] It should be noted that when searching in the music resource library based on scene resource tags, text keyword resource tags and backup resource tags, the Oak Bay retrieval system in the relevant technology can be used for implementation, or the vector retrieval (Facebook AI Similarity Search, Faiss) system can be used for implementation, and no specific limitation is made here.

[0135] Optionally, at least one group of a plurality of first candidate music, a plurality of second candidate music, and a plurality of third candidate music is obtained to obtain a candidate music set that matches the required key information.

[0136] In the disclosed embodiment, a plurality of first candidate music is obtained through a scene resource retrieval strategy, a plurality of second candidate music is obtained through a text keyword resource retrieval strategy, and a plurality of third candidate music is obtained through an intention resource retrieval strategy.

[0137] Optionally, the server may implement the above three strategies in parallel and recall the candidate music obtained by each of the three strategies, thereby obtaining three groups of candidate music, namely, multiple first candidate music, multiple second candidate music and multiple third candidate music, to form a candidate music set.

[0138] Optionally, the server may also execute the above three strategies in sequence based on a preset execution order, thereby obtaining any one of three groups of candidate music, namely, multiple first candidate music, multiple second candidate music, and multiple third candidate music, to form a candidate music set.

[0139] As an example, the execution order among the scene resource retrieval strategy, the text keyword resource retrieval strategy and the intention resource retrieval strategy is set as follows: the scene resource retrieval strategy is executed first, the text keyword resource retrieval strategy is executed when the scene resource retrieval strategy cannot be executed, and the intention resource retrieval strategy is executed when both the scene resource retrieval strategy and the text keyword retrieval strategy cannot be executed.

[0140] In this example, if a plurality of first candidate music pieces are obtained according to the scene resource retrieval strategy, a candidate music set is directly obtained based on the plurality of first candidate music pieces.

[0141] Furthermore, when the scene resource retrieval strategy cannot be executed, the text keyword resource retrieval strategy is executed. When multiple second candidate music is obtained according to the text keyword resource retrieval strategy, a candidate music set is obtained according to the multiple second candidate music.

[0142] Also, when the scene resource retrieval strategy and the text keyword resource retrieval strategy cannot be executed, the intention resource retrieval strategy is executed. When multiple third candidate music are obtained according to the intention resource retrieval strategy, a candidate music set is obtained based on the multiple third candidate music.

[0143] S303: Based on the candidate music set and the pre-acquired large model, a target music list that meets the music recommendation requirement is screened out from the candidate music set, and music description information of each target music in the target music list is generated.

[0144] In the disclosed embodiment, it is necessary to construct corresponding input prompt words according to the input requirements of the large model, wherein the model prompt words of the large model can be generated according to the candidate music set.

[0145] Among them, a prompt word construction template of the large model can be obtained, and according to the prompt word construction template, the candidate matching music set is filled into the prompt word construction template to generate the model prompt words of the large model.

[0146] In the disclosed embodiment, a corresponding prompt word construction template exists for the large model. The prompt word construction template can be obtained, and the filling position of the candidate music set in the prompt word construction template can be obtained. The candidate music set is filled into the prompt word construction template according to the filling position, and the prompt words obtained after filling are determined as the model prompt words of the large model.

[0147] Optionally, based on the model prompt words and the large model, an adaptation evaluation is performed on each candidate music in the candidate music set in the model prompt words to obtain a target music list that passes the adaptation evaluation from the candidate matching music set.

[0148] Among them, the model prompt words are input into the big model, and the big model obtains multiple target music that have passed the adaptation evaluation from the candidate music set carried by the model prompt words, and obtains the music popularity values ​​of the multiple target music respectively, and sorts the multiple target music by popularity in order from high to low to form a target music list for music recommendation needs.

[0149] In the disclosed embodiment, the model prompt words can be input into the big model, and multiple music resources that are adapted to the music recommendation requirements can be screened out as output results from the candidate music set carried in the model prompt words, and the multiple music resources in the output results can be determined as multiple target music to be pushed to the user end.

[0150] Optionally, the popularity of multiple target music may be different. In this scenario, the target music with higher popularity can be pushed to the user end first, wherein the music popularity value of each of the multiple target music can be obtained, and the multiple target music can be sorted in order from high to low, and the list consisting of the sorted multiple target music can be marked as the target music list pushed to the user end.

[0151] As an example, Figure 5 As shown, set Figure 5 The music resources 1, 2, 3 and 4 shown are multiple target music obtained by the server for music recommendation needs. Figure 5 The sorting module shown sorts music resource 1, music resource 2, music resource 3 and music resource 4 as target music in descending order based on the music popularity value, and then obtains a target music list consisting of the four music resources as target music.

[0152] In the embodiment of the present disclosure, when each target music in the target music list is pushed to the user terminal, the music description information of the target music can be pushed together, wherein the resource description information of each music resource in the music resource library can be obtained to obtain the target resource description information of each target music in the target music list. For any target music, the target resource description information of the target music and the music recommendation requirements are integrated into text information to generate the music description information of the target music.

[0153] In the embodiment of the present disclosure, when the server pushes target music to the user end, it can also push the music description information of the target music, such as Figure 2 As shown, when the server pushes music song 2 as the target music to the user, it also pushes the music description information of the target music: "Singer 2's classic voice and the gentle melody of this song can create a calming atmosphere, helping you concentrate and successfully complete "big things."

[0154] In this scenario, the resource metadata of each target music in the target music list may be obtained and marked as the target resource description information of each target music.

[0155] Among them, for any target music, input information that meets the input requirements of the big model and carries the target resource description information can be generated based on the target resource description information of the target music, and input into the big model. The big model obtains at least part of the information associated with the music recommendation needs input by the user end from the target resource description information carried in the input information. Furthermore, based on this at least part of the information and the music recommendation needs input by the user end, text information such as recommendation reasons and precautions is generated and integrated, and the integrated text information is output as the music description information of the target music.

[0156] It should be noted that the large model proposed in the embodiments of the present disclosure can be a large language model (LLM) or a large model constructed based on a technology that combines information retrieval and text generation (RAG). No specific limitation is made here.

[0157] The music push method proposed in the present invention determines a music retrieval strategy according to the key information type to which the key information of the demand belongs, so as to obtain a candidate music set from a music resource library, and obtain a target music list based on the candidate music set and the large model, thereby improving the flexibility of obtaining the candidate music set and the accuracy of music retrieval. In the scenario where the music resource library is updated, there is no need to train and optimize the model, which reduces the training cost of the model, improves the degree of adaptation between the target music list and the music recommendation demand input by the user end, and thus improves the accuracy of the target music push. The user end only needs to input the music recommendation demand to obtain the target music list from the music resource library. Compared with people's daily search for target music from the commonly used music application software, the music search range of the user end is expanded, the user end does not need to search for music by itself, and the time for the user end to obtain the target music list is shortened, the music search experience of the user end is optimized, and user stickiness is increased.

[0158] In the above embodiment, regarding the acquisition of the music resource library, the following can be combined: Figure 6 Further understanding, Figure 6 FIG. 1 is a flow chart of a music push method according to another embodiment of the present disclosure. Figure 6 As shown, the method includes:

[0159] S601, obtain the singer list, the singer introduction details of each singer in the singer list and the album information of each singer from the open source music information library, wherein for any singer, the singer's singer introduction details and album information are obtained based on the open source introduction information on each music application software.

[0160] In the embodiment of the present disclosure, singer information and song information can be obtained from the open source music resource library. As an example, Figure 7 As shown, you can Figure 7 The open source music application software shown is regarded as the open source music information library proposed in the embodiment of the present disclosure, wherein the open source music application software can be used to store music information. Figure 7 The singer list module shown is from Figure 7 The open source music application software shown obtains all singer information to form Figure 7 A list of singers is shown.

[0161] Further, through Figure 7 The singer details acquisition module shown obtains the singer's detailed information from the open source music application software, wherein the singer's detailed information may include the singer's personal basic information, personal experience, and awards and achievements in the music field.

[0162] And, through Figure 7 The singer album information module shown obtains the album information previously released by each singer from the open source music application software.

[0163] It should be noted that the open source music application software for obtaining the singer details of each singer in the singer list and the open source music application software for obtaining the singer album information of each singer can be the same application software or different application software, and no specific limitation is made here.

[0164] S602: Obtain a song list and song information of each song in the song list from an open source music information library, wherein the song information at least includes lyrics information, comment information, and song details information of the song.

[0165] As an example, Figure 7 As shown, Figure 7 The open source music application software shown can be regarded as an open source music resource library, which can be used to Figure 7 The song information (singer + song) module shown obtains a song list consisting of full song information from the open source music application software, as well as the singer, lyrics information, song comment information and song details information of each song in the song list.

[0166] The song details information may include creative concept information such as the composition and lyrics of the music, and may also include information such as the atmosphere and meaning of the song, which is not specifically limited here.

[0167] It should be noted that the open source music application software for obtaining lyrics information of each song, the open source music application software for obtaining comment information of each song, and the open source music application software for obtaining song details information of each song can be the same open source music application software or different open source music application software, and no specific limitation is made here.

[0168] S603, constructing a music resource library based on the singer list, the singer introduction details of each singer, the album information of each singer, the song list and the song information of each song.

[0169] In the disclosed embodiment, after obtaining the relevant information of each singer in the singer list and the relevant information of each song in the song list, a resource library can be constructed based on the obtained information to obtain a music resource library.

[0170] As an example, Figure 7 As shown, according to Figure 7 The song information (singer + song), the lyrics information of each song, the comment information of each song and the song details information of each song can be constructed to obtain Figure 7 The song information library is shown.

[0171] And, according to Figure 7 The singer list, singer details and album information of each singer can be constructed. Figure 7The singer information database is shown.

[0172] Further, according to Figure 7 The song information library and singer information library shown, as well as the index information required to build the resource library, are obtained Figure 7 The music resource library is shown.

[0173] The music push method proposed in the present disclosure is that the user terminal only needs to input the music recommendation requirements to obtain the target music list from the music resource library. Compared with people's daily search for target music from the usual music application software, the music search range of the user terminal is expanded, the user terminal does not need to search for music by itself, and the time for the user terminal to obtain the target music list is shortened.

[0174] To better understand the above embodiments, Figure 8 , Figure 8 FIG. 1 is a flow chart of a music push method according to another embodiment of the present disclosure. Figure 8 As shown, the method includes:

[0175] User-side input Figure 8 The input information shown and passed Figure 8 The pre-identification module shown identifies whether the demand carried in the input information of the user terminal is a music recommendation demand under the music vertical category, wherein the pre-identification module can be called Figure 8 The full vocabulary shown is used to identify the vertical category to which the requirements carried in the input information belong.

[0176] Optionally, when it is identified that the demand carried in the input information of the user terminal is a music recommendation demand, Figure 8 The knowledge retrieval module shown searches for music resources to obtain a candidate music set for music recommendation needs.

[0177] Among them, you can Figure 8 The knowledge retrieval module shown identifies the key information type of the key information in the music recommendation requirement, thereby determining the corresponding music resource retrieval strategy, and retrieves the candidate music set of the music recommendation requirement from the music resource library through the execution of the music resource retrieval strategy.

[0178] like Figure 8 As shown, it can be achieved through Figure 8 The Oak Bay search method shown searches from a metadata database composed of resource metadata of each music resource in the music resource library, thereby obtaining a candidate music set for music recommendation needs.

[0179] like Figure 8 As shown, you can also Figure 8The vector retrieval (Facebook AI Similarity Search, Faiss) method shown is from Figure 8 The vector library composed of the scene vectors shown is searched to obtain the estimated scene vector of the music recommendation demand, and then based on the estimated scene vector and related information such as the key information of the demand, a candidate music set of the music recommendation demand is obtained from the music resource library.

[0180] Optionally, the candidate music in the candidate music set can be reordered based on a preset sorting strategy, and the model prompt words of the large model can be obtained according to the sorted candidate music set. Further, the model prompt words can be input into the Figure 8 In the large model shown, the large model is used to perform adaptation evaluation on each candidate music in the candidate music set carried in the model prompt word, so as to screen out multiple target music that pass the adaptation evaluation from the candidate music set, and then obtain a target music list consisting of multiple target music.

[0181] In addition, the target resource description information of each target music in the target music list is obtained, and based on the target resource description information of each target music and the music recommendation requirements, input information that meets the input requirements of the big model is generated and input into the big model. The big model integrates the target resource description information and music recommendation requirements carried in the input information into text information, and then obtains the integrated text information output by the big model, and uses it as the music description information of each target music.

[0182] Furthermore, the target music list and the music description information of each target music are returned to the user end in sequence.

[0183] The music push method proposed in the present invention determines a music retrieval strategy according to the key information type to which the required key information belongs, so as to obtain a candidate music set from a music resource library, and obtains a target music list according to the candidate music set and a large model. The music retrieval strategy is determined according to the key information type to which the required key information belongs to obtain the candidate music set, thereby improving the flexibility of obtaining the candidate music set and the accuracy of music retrieval. In the scenario where the music resource library is updated, there is no need to train and optimize the model, which reduces the training cost of the model, improves the degree of adaptation between the target music list and the music recommendation requirements input by the user end, and thus improves the accuracy of target music push. The user end only needs to input the music recommendation requirements to obtain the target music list from the music resource library. Compared with people's daily search for target music from the commonly used music application software, the music search range of the user end is expanded, the user end does not need to search for music by itself, shortens the time for the user end to obtain the target music list, optimizes the music search experience of the user end, and increases user stickiness.

[0184] Corresponding to the music push methods proposed in the above-mentioned embodiments, an embodiment of the present disclosure further proposes a music push device. Since the music push device proposed in the embodiment of the present disclosure corresponds to the music push methods proposed in the above-mentioned embodiments, the implementation method of the above-mentioned music push method is also applicable to the music push device proposed in the embodiment of the present disclosure, and will not be described in detail in the following embodiments.

[0185] Figure 9 FIG. 1 is a structural diagram of a music push device according to an embodiment of the present disclosure. Figure 9 As shown, the music push device 900 includes a first acquisition module 91, a second acquisition module 92, a third acquisition module 93 and a push module 94, wherein:

[0186] The first acquisition module 91 is used to obtain the music recommendation demand of the user terminal and extract the key information of the demand in the music recommendation demand;

[0187] The second acquisition module 92 is used to obtain the key information type of the required key information to determine the music retrieval strategy for the music recommendation requirement, and obtain a candidate music set that matches the required key information from the music resource library according to the music retrieval strategy;

[0188] The third acquisition module 93 is used to filter out a target music list that meets the music recommendation requirements from the candidate music set based on the candidate music set and the pre-acquired large model, and generate music description information for each target music in the target music list;

[0189] The push module 94 is used to push the target music list and the music description information of each target music to the user terminal.

[0190] In the embodiment of the present disclosure, the second acquisition module 92 is also used to: in response to the key information type to which the demand key information belongs being the scene information type, determine that the scene resource retrieval strategy is the retrieval strategy for the music recommendation demand; in response to the retrieval strategy being the scene resource retrieval strategy, obtain the extended keyword of the demand key information; obtain the scene resource tag of the demand key information from the tag resource library based on the demand key information and the extended keyword; obtain the resource metadata of each music resource in the music resource library, and obtain multiple first candidate music for the music recommendation demand through the resource metadata of each music resource based on the demand key information, the extended keyword and the scene resource tag, so as to obtain a candidate music set.

[0191] In the embodiment of the present disclosure, the second acquisition module 92 is also used to: in response to the key information type to which the key information of the requirement is a text keyword information type, determine the text keyword resource retrieval strategy as the retrieval strategy for the music recommendation requirement, and determine the key information of the requirement as the resource retrieval keyword; in response to the retrieval strategy being the text keyword resource retrieval strategy, obtain the resource metadata of each music resource in the music resource library, and according to the resource retrieval keyword, obtain multiple second candidate music for the music recommendation requirement through the resource metadata of each music resource to obtain a candidate music set.

[0192] In the embodiment of the present disclosure, the second acquisition module 92 is also used to: in response to the key information type of the demand key information being a fuzzy information type, determine that the intention resource retrieval strategy is a retrieval strategy for music recommendation demand; in response to the retrieval strategy being the intention resource retrieval strategy, identify the demand intent information of the music recommendation demand, and obtain the estimated scenario vector of the music recommendation demand based on the demand intent information; obtain a backup resource tag from the tag resource library based on the estimated scenario vector and the demand intent information; obtain resource metadata of each music resource in the music resource library, and obtain multiple third candidate music for the music recommendation demand through the resource metadata of each music resource based on the demand intent information, the estimated scenario vector and the backup resource tag to obtain a candidate music set.

[0193] In the embodiment of the present disclosure, the third acquisition module 93 is further used to: acquire at least one group of multiple first candidate music, multiple second candidate music and multiple third candidate music to obtain a candidate music set that matches the required key information; generate model prompt words of the large model based on the candidate music set; and perform adaptation evaluation on each candidate music in the candidate music set in the model prompt words based on the model prompt words and the large model to obtain a target music list that passes the adaptation evaluation from the candidate matching music set.

[0194] In the disclosed embodiment, the third acquisition module 93 is further used to: obtain a prompt word construction template for the large model; and fill the candidate matching music set into the prompt word construction template according to the prompt word construction template to generate model prompt words for the large model.

[0195] In the embodiment of the present disclosure, the third acquisition module 93 is also used to: input the model prompt words into the big model, and obtain multiple target music that have passed the adaptation evaluation from the candidate music set carried in the model prompt words through the big model; obtain the music popularity value of each of the multiple target music, and sort the multiple target music by popularity in order from high to low to form a target music list for music recommendation needs.

[0196] In the disclosed embodiment, the third acquisition module 93 is also used to: obtain resource description information of each music resource in the music resource library to obtain target resource description information of each target music in the target music list; for any target music, integrate the target resource description information of the target music and the music recommendation requirements into text information to generate music description information of the target music.

[0197] In the embodiment of the present disclosure, the device also includes a construction module, which is used to: obtain a singer list, singer introduction details of each singer in the singer list and album information of each singer from an open source music information library, wherein for any singer, the singer's singer detailed introduction and album information are obtained based on the open source introduction information on each music application software; obtain a song list and song information of each song in the song list from the open source music information library, wherein the song information includes at least lyrics information, comment information and song details information of the song; construct a music resource library based on the singer list, singer introduction details of each singer, album information of each singer, song list and song information of each song.

[0198] In the embodiment of the present disclosure, the first acquisition module 91 is also used to: obtain input information from the user end and identify the input intention of the input information; in response to identifying that the input intention is a music recommendation intention, obtain the music recommendation demand of the user end; perform semantic recognition on the music recommendation demand, and extract key demand information in the music recommendation demand.

[0199] The music push device proposed in the present disclosure obtains the music recommendation demand of the user end, extracts the key information of the demand in the music recommendation demand, obtains the key information type to which the key information of the demand belongs, obtains a candidate music set from a music resource library, obtains a target music list adapted to the music recommendation demand from the candidate music set based on the candidate music set and the big model, and generates music description information of each target music, and further pushes the target music list and the music description information of each target music to the user end. In the present disclosure, a music retrieval strategy is determined according to the key information type to which the required key information belongs, so as to obtain a candidate music set from a music resource library, and a target music list is obtained according to the candidate music set and a large model. The music retrieval strategy is determined according to the key information type to which the required key information belongs to obtain the candidate music set, thereby improving the flexibility of obtaining the candidate music set and improving the accuracy of music retrieval. In the scenario where the music resource library is updated, there is no need to train and optimize the model, which reduces the training cost of the model, improves the degree of adaptation between the target music list and the music recommendation requirements input by the user end, and thus improves the accuracy of the target music push. The user end only needs to input the music recommendation requirements to obtain the target music list from the music resource library. Compared with people's daily search for target music from their usual music application software, the music search range of the user end is expanded, the user end does not need to search for music by itself, and the time it takes for the user end to obtain the target music list is shortened, the music search experience of the user end is optimized, and user stickiness is increased.

[0200] To achieve the above embodiments, the present disclosure also provides an electronic device, a computer-readable storage medium, and a computer program product.

[0201] Figure 10 is a block diagram of an electronic device 100 according to an embodiment of the present disclosure, as shown in FIG. Figure 10 As shown, the electronic device 100 includes a memory 11, a processor 12, and a computer program stored in the memory 11 and executable on the processor 12. When the processor 12 executes the program instructions, the music push method applicable to the server provided in the above embodiment is implemented.

[0202] The music push method proposed in the present invention obtains the music recommendation demand of the user end, extracts the key information of the demand in the music recommendation demand, obtains the key information type to which the key information of the demand belongs, obtains a candidate music set from a music resource library, obtains a target music list adapted to the music recommendation demand from the candidate music set based on the candidate music set and a large model, and generates music description information of each target music, and further pushes the target music list and the music description information of each target music to the user end. In the present disclosure, a music retrieval strategy is determined according to the key information type to which the required key information belongs, so as to obtain a candidate music set from a music resource library, and a target music list is obtained according to the candidate music set and a large model. The music retrieval strategy is determined according to the key information type to which the required key information belongs to obtain the candidate music set, thereby improving the flexibility of obtaining the candidate music set and improving the accuracy of music retrieval. In the scenario where the music resource library is updated, there is no need to train and optimize the model, which reduces the training cost of the model, improves the degree of adaptation between the target music list and the music recommendation requirements input by the user end, and thus improves the accuracy of the target music push. The user end only needs to input the music recommendation requirements to obtain the target music list from the music resource library. Compared with people's daily search for target music from their usual music application software, the music search range of the user end is expanded, the user end does not need to search for music by itself, and the time it takes for the user end to obtain the target music list is shortened, the music search experience of the user end is optimized, and user stickiness is increased.

[0203] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0204] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0205] The program code for implementing the method itself can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0206] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0207] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0208] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or grid browser through which a user can interact with embodiments of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by digital data communication (e.g., a communication grid) in any form or medium. Examples of communication grids include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain grid.

[0209] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication grid. The client-server relationship is established by computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, a host product within the cloud computing service system that addresses the management difficulties and poor business scalability of traditional physical hosts and VPS services ("Virtual Private Server," or simply "VPS"). The server may also be a server in a distributed system or a server integrated with blockchain.

[0210] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and features of different embodiments or examples, unless they are mutually inconsistent.

[0211] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0212] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present disclosure belong.

[0213] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or otherwise processing it in a suitable manner if necessary, and then storing it in a computer memory.

[0214] It should be understood that various parts of the present disclosure can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0215] Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above-mentioned embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0216] In addition, the functional units in the various embodiments of the present disclosure may be integrated into a single processing module, each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0217] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. A person of ordinary skill in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present disclosure.

[0218] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0219] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A music push method, characterized in that: The method comprises: Obtaining music recommendation requirements from a user and extracting key information from the music recommendation requirements; Obtaining the key information type to which the key information of the requirement belongs to determine a music retrieval strategy for the music recommendation requirement, and obtaining a candidate music set that matches the key information of the requirement from a music resource library according to the music retrieval strategy; Based on the candidate music set and the pre-acquired large model, a target music list adapted to the music recommendation requirement is screened out from the candidate music set, and music description information of each target music in the target music list is generated; The target music list and the music description information of each target music are pushed to the user terminal.

2. The method according to claim 1, characterized in that The step of obtaining the key information type to which the key information of the requirement belongs, determining a music retrieval strategy for the music recommendation requirement, and obtaining a candidate music set matching the key information of the requirement from a music resource library according to the music retrieval strategy includes: In response to the key information type to which the key information of the requirement belongs being a scene information type, determining the scene resource retrieval strategy as the retrieval strategy for the music recommendation requirement; In response to the retrieval strategy being the scene resource retrieval strategy, acquiring an extended keyword of the requirement key information; According to the key requirement information and the extended keyword, a scene resource tag of the key requirement information is obtained from a tag resource library; Obtain resource metadata of each music resource in the music resource library, and according to the requirement key information, the extended keyword and the scene resource tag, obtain multiple first candidate music of the music recommendation requirement through the resource metadata of each music resource to obtain the candidate music set.

3. The method according to claim 1, characterized in that The step of obtaining the key information type to which the key information of the requirement belongs, determining a music retrieval strategy for the music recommendation requirement, and obtaining a candidate music set matching the key information of the requirement from a music resource library according to the music retrieval strategy includes: In response to the key information type to which the demand key information belongs being a text keyword information type, determining a text keyword resource retrieval strategy as a retrieval strategy for the music recommendation demand, and determining the demand key information as a resource retrieval keyword; In response to the retrieval strategy being the text keyword resource retrieval strategy, resource metadata of each music resource in the music resource library is obtained, and based on the resource retrieval keyword, multiple second candidate music for the music recommendation needs are obtained through the resource metadata of each music resource to obtain the candidate music set.

4. The method according to claim 1, wherein The step of obtaining the key information type to which the key information of the requirement belongs, determining a music retrieval strategy for the music recommendation requirement, and obtaining a candidate music set matching the key information of the requirement from a music resource library according to the music retrieval strategy includes: In response to the key information type to which the key information of the requirement belongs being a fuzzy information type, determining that the intended resource retrieval strategy is a retrieval strategy for the music recommendation requirement; In response to the retrieval strategy being the intention resource retrieval strategy, identifying demand intention information of the music recommendation demand, and obtaining an estimated scenario vector of the music recommendation demand based on the demand intention information; Obtaining a backup resource tag from a tag resource library based on the estimated scenario vector and the demand intention information; Obtain resource metadata of each music resource in the music resource library, and according to the demand intention information, the estimated scenario vector and the backup resource tag, obtain multiple third candidate music for the music recommendation demand through the resource metadata of each music resource to obtain the candidate music set.

5. The method according to any one of claims 1 to 4, characterized in that The step of screening a target music list adapted to the music recommendation requirement from the candidate music set based on the candidate music set and the pre-acquired large model includes: Obtaining at least one group of a plurality of first candidate music, a plurality of second candidate music, and a plurality of third candidate music to obtain the candidate music set matching the key information of the requirement; Generating model prompt words of the large model according to the candidate music set; Based on the model prompt words and the large model, an adaptation evaluation is performed on each candidate music in the candidate music set in the model prompt words to obtain the target music list that passes the adaptation evaluation from the candidate matching music set.

6. The method according to claim 5, characterized in that Generating the model prompt words of the large model according to the candidate music set includes: Obtaining a prompt word construction template for the large model; According to the prompt word construction template, the candidate matching music set is filled into the prompt word construction template to generate the model prompt words of the large model.

7. The method according to claim 6, characterized in that The step of performing an adaptation evaluation on each candidate music in the candidate music set in the model prompt word according to the model prompt word and the large model, so as to obtain the target music list that passes the adaptation evaluation from the candidate matching music set, includes: Inputting the model prompt word into the large model, and obtaining, through the large model, a plurality of target music pieces that pass the adaptation evaluation from the candidate music set carried in the model prompt word; The music popularity value of each of the multiple target music is obtained, and the multiple target music are ranked based on popularity in descending order to form the target music list for the music recommendation requirement.

8. The method according to claim 1, characterized in that The step of generating the music description information of each target music in the target music list includes: Obtain resource description information of each music resource in the music resource library to obtain target resource description information of each target music in the target music list; For any target music, text information is integrated between the target resource description information of the target music and the music recommendation requirement to generate the music description information of the target music.

9. The method according to claim 1, characterized in that Before obtaining the key information type to which the required key information belongs and obtaining a candidate music set matching the required key information from a music resource library according to the key information type, the method includes: Obtaining a list of singers, singer introduction details of each singer in the list, and album information of each singer from an open source music information database, wherein for any singer, the singer introduction details and album information are obtained based on open source introduction information on various music application software; Obtaining a song list and song information of each song in the song list from the open source music information library, wherein the song information at least includes lyrics information, comment information, and song details information of the song; The music resource library is constructed based on the singer list, the singer introduction details of each singer, the album information of each singer, the song list and the song information of each song.

10. The method according to claim 1, characterized in that The obtaining of the user's music recommendation demand and extracting key demand information in the music recommendation demand include: Obtaining input information from the user terminal and identifying the input intention of the input information; In response to identifying that the input intention is a music recommendation intention, obtaining the music recommendation demand of the user terminal; Perform semantic recognition on the music recommendation demand and extract key demand information in the music recommendation demand.

11. A music push device, characterized in that: The device comprises: A first acquisition module is used to acquire music recommendation requirements of a user terminal and extract key information of the requirements; A second acquisition module is configured to acquire the key information type to which the key information of the requirement belongs, so as to determine a music retrieval strategy for the music recommendation requirement, and acquire a candidate music set matching the key information of the requirement from a music resource library according to the music retrieval strategy; a third acquisition module, configured to filter out a target music list that meets the music recommendation requirement from the candidate music set based on the candidate music set and the pre-acquired large model, and generate music description information for each target music in the target music list; The push module is used to push the target music list and the music description information of each target music to the user terminal.

12. An electronic device, characterized in that: include: processor; a memory for storing executable instructions for the processor; The processor is configured to execute instructions to implement the method according to any one of claims 1 to 10.

13. A computer-readable storage medium, when instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the method according to any one of claims 1 to 10.

Images