Audio recommendation strategy determination method and apparatus, and electronic device

Abstract

The invention provides an audio recommendation strategy determination method and apparatus, and an electronic device. The method comprises the steps of determining a target audio in a specified platform; determining a plurality of first users of a specified platform, pushing an evaluation task of the target audio to each first user, and obtaining an evaluation parameter of the target audio generated by each first user for the evaluation task; determining a quality evaluation parameter of the target audio based on the evaluation parameter; if the quality evaluation parameter meets a preset second condition, determining a plurality of second users in the specified platform based on user portraits of users of the specified platform; and obtaining user behavior data of the plurality of second users for the target audio, and determining a recommendation strategy of the target audio in the specified platform based on the user behavior data and the user portraits of the plurality of second users. According to the mode, the corresponding recommendation strategy can be formulated for the audio with less exposure and higher quality, the song listening experience of the user is improved, and the user retention rate of the music platform is improved.

Description

Technical Field

[0001] This disclosure relates to the field of audio processing technology, and more specifically, to a method, apparatus, and electronic device for determining audio recommendation strategies. Background Technology

[0002] In music platforms, user behavior data is typically used as a fundamental indicator to measure song quality and popularity. This data includes play counts, number of favorites, number of comments, and completion rate. Recommendation algorithms process this data to evaluate song quality and influence subsequent recommendation methods and traffic distribution decisions. However, processing this data is inherently lagging. Without song exposure and data, recommendation algorithms lack sufficient information to evaluate song quality and provide effective suggestions for recommendation methods. This can easily lead to missed opportunities to develop the potential of songs like these, resulting in a poor user listening experience and significant user churn. Summary of the Invention

[0003] In view of this, the purpose of this disclosure is to provide a method, apparatus and electronic device for determining audio recommendation strategies, so as to formulate corresponding recommendation strategies for audio with less exposure but higher quality, improve the user's listening experience and improve the user retention rate of music platforms.

[0004] In a first aspect, embodiments of this disclosure provide a method for determining an audio recommendation strategy. The method includes: identifying a target audio on a specified platform; ensuring the target audio's release time and playback volume on the specified platform meet preset first conditions; identifying multiple first users on the specified platform, pushing evaluation tasks for the target audio to each first user, and obtaining evaluation parameters of the target audio generated by each first user for the evaluation task; the evaluation parameters indicate the first user's subjective evaluation of the target audio's melody, emotion, and expression; determining quality evaluation parameters for the target audio based on the evaluation parameters; if the quality evaluation parameters meet preset second conditions, identifying multiple second users on the specified platform based on user profiles of the users on the specified platform; the user profiles are determined based on the user's basic attribute data and historical audio listening data on the specified platform; pushing the target audio to each second user, and obtaining user behavior data of the multiple second users for the target audio; and determining a recommendation strategy for the target audio on the specified platform based on the user behavior data and the user profiles of the multiple second users.

[0005] Secondly, embodiments of this disclosure provide an audio recommendation strategy determination apparatus, comprising: a target audio determination module, configured to determine target audio on a specified platform; wherein the release time and playback volume of the target audio on the specified platform meet a preset first condition; an evaluation parameter acquisition module, configured to determine multiple first users on the specified platform, push evaluation tasks of the target audio to each first user, and acquire evaluation parameters of the target audio generated by each first user for the evaluation task; the evaluation parameters are used to indicate: the subjective evaluation of the melody, emotion, and expression of the target audio by the first user; a quality evaluation parameter determination module, configured to determine quality evaluation parameters of the target audio based on the evaluation parameters; a second user determination module, configured to determine multiple second users on the specified platform based on user profiles of users on the specified platform if the quality evaluation parameters meet the preset second condition; the user profiles are determined based on the user's basic attribute data and historical audio listening data on the specified platform; and a recommendation strategy determination module, configured to push the target audio to each second user, acquire user behavior data of multiple second users for the target audio, and determine the recommendation strategy of the target audio on the specified platform based on the user behavior data and the user profiles of multiple second users.

[0006] Thirdly, embodiments of the present invention provide an electronic device, including a processor and a memory, wherein the memory stores machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement the above-described audio recommendation strategy determination method. Fourthly, embodiments of the present invention provide a machine-readable storage medium storing machine-executable instructions. When the machine-executable instructions are invoked and executed by a processor, the machine-executable instructions cause the processor to implement the aforementioned audio recommendation strategy determination method.

[0007] The embodiments of the present invention bring the following beneficial effects: The aforementioned method, apparatus, and electronic device for determining audio recommendation strategies involve: identifying target audio on a specified platform; identifying multiple first users on the specified platform; pushing evaluation tasks for the target audio to each first user and obtaining evaluation parameters generated by each first user for the target audio in response to the evaluation tasks; determining quality evaluation parameters for the target audio based on the evaluation parameters; if the quality evaluation parameters meet a preset second condition; identifying multiple second users on the specified platform based on user profiles of the users on the specified platform; pushing the target audio to each second user and obtaining user behavior data of the multiple second users regarding the target audio; and determining a recommendation strategy for the target audio on the specified platform based on the user behavior data and the user profiles of the multiple second users. This method can formulate corresponding recommendation strategies for audio with low exposure but high quality, improving the user's listening experience and increasing user retention rate on music platforms.

[0008] Other features and advantages of this disclosure will be set forth in the following description and will be apparent in part from the description or may be learned by practicing the disclosure. The objects and other advantages of this disclosure are realized and obtained through the structures particularly pointed out in the description, claims and drawings.

[0009] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the specific embodiments of this disclosure or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0011] Figure 1 A flowchart illustrating a method for determining an audio recommendation strategy provided in this embodiment of the disclosure; Figure 2 A schematic diagram of the structure of an audio recommendation strategy determination device provided in an embodiment of this disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0013] In music platforms, user behavior data is typically used as a fundamental indicator to measure song quality and popularity. This data usually includes song plays, favorites, comments, and completion rates. After being modeled by recommendation algorithms, this data is used for cold start decisions, recommendation ranking, and traffic distribution decisions.

[0014] Furthermore, the metrics calculated based on this data are inherently lagging indicators, making it difficult to effectively evaluate newly released songs with fewer than 1,000 plays. These songs can be termed "cold start" songs. A cold start typically refers to a newly released song that, due to low play counts and a lack of user behavior data, is difficult for recommendation algorithms to identify and distribute. Without exposure and data, recommendation algorithms lack sufficient information to make judgments, causing high-quality cold start songs to be "submerged" and missing potential development opportunities.

[0015] In some variety shows, the production team sets up a public jury, where live audience members score songs or contestants' performances. This format is widely used in the entertainment industry to enhance the interactivity and objectivity of programs. However, this mechanism is only suitable for small-scale offline scenarios, has a limited sample size, and is time-consuming and costly. It cannot be standardized and scaled up on online platforms, nor can it be combined with user profiles to determine potential growth potential.

[0016] Based on this, the present disclosure provides a method, apparatus, and electronic device for determining audio recommendation strategies, which can be applied to scenarios for determining audio recommendation strategies.

[0017] See Figure 1 First, we will introduce a method for determining an audio recommendation strategy provided by an embodiment of the present invention. This method includes the following steps: Step S102: Determine the target audio on the specified platform; the release time and play count of the target audio on the specified platform meet the preset first condition.

[0018] The target audio can be audio corresponding to vocal works, instrumental works, etc., that is, it can be a song sung by human voices or instrumental music. The target audio can be audio recorded by a recording device or audio synthesized by a computer. The specific settings can be made according to needs, and there are no restrictions here.

[0019] The target audio files mentioned above are typically "cold start" songs, meaning songs that have been released for some time but have low play counts. Correspondingly, the first condition mentioned above usually includes sub-conditions regarding release time and sub-conditions regarding play counts.

[0020] A time range can be preset. When the time difference between the release time of the target audio and the current time falls within this time range, the release time of the target audio can be considered to have met the sub-condition for the release time. This time range can be [10 days, 15 days], and can be set according to needs; there are no restrictions here.

[0021] The aforementioned play count typically refers to the number of plays received by the target audio after its release, and can be specified as daily play count, total play count, etc. Sub-conditions for play count can be set for total play count, such as limiting the total play count of the target audio to less than a certain total play count threshold; this sub-condition can also be set for daily play count, such as limiting the daily play count of the target audio to not exceed a certain daily play count threshold; this sub-condition can also be set for play count trends, such as limiting the daily increase in play count to less than or equal to a certain increase threshold. This sub-condition can also be set for multiple metrics such as total play count, daily play count, and play count trends. Specific settings can be configured according to needs and are not limited here.

[0022] When the release time and play count of an audio file on a specified platform simultaneously meet the first condition, that audio file can be designated as the target audio. The target audio can be selected from the audio files on the specified platform according to a preset time frequency; alternatively, after each audio file is released, play count data can be acquired in real time, and the audio file can be designated as the target audio file once the release time and play count meet the first condition. Specific restrictions can be implemented according to requirements.

[0023] S104, identify multiple first users on the designated platform, push the evaluation task of the target audio to each first user, and obtain the evaluation parameters of the target audio generated by each first user for the evaluation task; the evaluation parameters are used to indicate: the first user's subjective evaluation of the melody, emotion and expression of the target audio.

[0024] The aforementioned designated platforms typically have multiple registered users. Users can listen to, favorite, comment on, and share audio on these platforms. The platforms record user activity data, which can be called user behavior data. This user behavior data can be analyzed, and the results can reflect, to some extent, a user's ability to evaluate music, also known as "appreciation ability." Then, a predetermined number of users with high evaluation abilities are selected as the first user.

[0025] By analyzing user behavior data, we can determine the number, type, and rating of songs a user listens to, favorites, comments on, and shares. We can also perform semantic analysis on user comments to determine the degree of match between the comments and the song's content. The song ratings, the degree of match between the comments and the song's content, and the number of comments can be used as evaluation criteria for a user's evaluation ability. Then, based on preset weights and the actual values ​​of these criteria, we can determine the user's evaluation ability parameters.

[0026] Furthermore, in determining a user's evaluation ability, data such as the user's previous completion level of audio evaluation tasks and the consistency between the evaluation content and the subsequent development of the audio can also be considered. These data are assigned weights and used, along with the song ratings, the degree of match between comments and the song's content, and the number of comments, to calculate the user's evaluation ability parameters. Therefore, a set number of users with high evaluation abilities can be designated as the first users.

[0027] After identifying multiple first users, the evaluation task for the target audio needs to be pushed to each first user, and the evaluation parameters generated by each first user for the target audio in response to the evaluation task need to be obtained. The evaluation task can be sent to the client controlling each first user via a link or window. The evaluation task typically requires users to provide subjective evaluations of the melody, emotion, and expression of the target audio. This can be done through ratings, written descriptions, or a combination of both, depending on the specific requirements.

[0028] Step S106: Determine the quality evaluation parameters of the target audio based on the evaluation parameters.

[0029] Generally speaking, each of the aforementioned first users will generate evaluation parameters for the target audio. These evaluation parameters can be analyzed, summarized, and processed. The processing results are considered to reflect the quality of the audio to a certain extent, and the processing results are determined as the quality evaluation parameters of the target audio.

[0030] When the evaluation parameter is a score, statistical parameters such as sum, mean, variance, and median of multiple user-generated scores can be calculated. As mentioned above, the evaluation parameters are multi-dimensional, corresponding to the melody, emotion, and expression of the target audio. A corresponding scoring mechanism can be set for each dimension, making the evaluation parameters generated by users for each dimension clearer. When calculating the statistical parameters of the score, statistical parameters can be calculated for each dimension. Based on the weights corresponding to each dimension and the calculated statistical parameters, quality evaluation parameters that can be used to reflect the quality of the target audio can be obtained.

[0031] When the evaluation parameter is text, semantic analysis can be performed on the text. Specifically, for each dimension of the target audio, such as melody, emotion, and expression, corresponding semantic analysis results can be obtained. Then, based on the semantic analysis results, the various dimensions of the target audio can be quantified. For example, multiple words that can represent the melody effect can be pre-defined, such as "catchy," "gentle and soothing," "noisy," and "rhythmic," each with a corresponding score. The distance between the semantic analysis result corresponding to the melody and the semantic features of each word can be calculated. Then, for each word, based on the semantic analysis result corresponding to the melody and the semantic feature distance between the word and the word, the score of the target audio on the melody effect represented by that word is determined, and then the total score of the target audio on the melody is calculated. Finally, similar to the method used when the evaluation parameter is a score, quality evaluation parameters that can be used to reflect the quality of the target audio can be obtained based on the weights corresponding to each dimension and the calculated statistical parameters.

[0032] When the evaluation includes both scores and text content, the two methods mentioned above can be combined to obtain quality evaluation parameters that can be used to reflect the quality of the target audio.

[0033] Step S108: If the quality evaluation parameters meet the preset second condition, determine multiple second users in the specified platform based on the user profile of the user on the specified platform; the user profile is determined based on the user's basic attribute data and historical audio listening data on the specified platform.

[0034] The second condition mentioned above is typically used to limit the target audio to a high quality. When the quality evaluation parameter can be quantified, i.e. has a parameter value, a parameter threshold can be preset, and the second condition can be that the quality evaluation parameter is greater than or equal to that parameter threshold.

[0035] When acquiring target audio based on frequency, if multiple target audios are identified at once, the second condition mentioned above can further limit the quality of the target audio to the highest quality audio among those identified. For example, a pre-set percentage parameter, such as 10%, can be used, and the second condition is that the quality evaluation parameter must be within the top 10% of all acquired target audios. That is, when 10 target audios are acquired, the target audio with the highest quality evaluation parameter can be determined to satisfy the second condition.

[0036] The two second conditions mentioned above can also be combined. For example, the second condition could be that the quality evaluation parameter is within the top 10% of all target audio samples acquired in this study, and the quality evaluation parameter is greater than or equal to a preset parameter threshold. Alternatively, other methods can be used to set the second condition, which are not restricted here.

[0037] After determining that the target audio meets the second condition, some users can be made to listen to the target audio first, and based on the behavioral data of these users regarding the target audio, the recommendation strategy for the target audio can be further determined. These users who listen to the target audio first need to be representative, and the characteristics and representativeness of users can usually be determined based on the user portraits of the users.

[0038] A user portrait usually refers to abstracting each specific piece of information of a user into tags and using these tags to concretize the user image to obtain a user portrait. The user portrait can be determined based on the basic attribute data of the user and the historical audio listening data on the specified platform. The above-mentioned basic attribute data can include age, gender, region, etc. The above-mentioned historical audio listening data can include the names, types, listening times, listening durations, etc. of the songs listened to by the user. Users on the platform can be classified based on the similarity of their user portraits to obtain multiple user categories. Then, some users can be selected from the multiple user types as the second users.

[0039] Step S110: Push the target audio to each second user, obtain the user behavioral data of multiple second users regarding the target audio, and determine the recommendation strategy of the target audio on the specified platform based on the user behavioral data and the user portraits of the multiple second users.

[0040] The target audio can be pushed to each second user individually in the form of a system message, etc., or the target audio can be placed in a playlist, such as playlists like "Daily Recommendations" and "Guess You Like" and pushed to each second user. The second user can listen to the song according to their own preferences, or ignore the song, and can also comment on and forward the song, generating user behavioral data regarding the target audio. Then, for By analyzing the log data of the second users, the user behavioral data regarding the target audio can be obtained. The play volume, collection volume, comment volume, forward volume, etc. of the target audio can be statistically calculated based on the user behavioral data. Usually, corresponding thresholds can be set in advance for multiple indicators such as play volume, collection volume, comment volume, forward volume, etc., and then the indicators of the target audio are compared with the corresponding thresholds. And based on the comparison results, it is determined whether the target audio has the potential to be liked by most users. If so, the recommendation intensity of the target audio can be increased. Specifically, the recommendation strategy can be set as: Push the target audio to 80% of the users on the specified platform.

[0041] If the target audio does not have the potential to be liked by most users, the target user portrait of the users who relatively like the target audio can be determined based on the user portraits of the second users and the user behavioral data regarding the target audio. Then the recommendation strategy is set as: Push the target audio to users with a relatively high similarity between their user portraits and the target user portrait.

[0042] The aforementioned method for determining an audio recommendation strategy involves: identifying a target audio on a specified platform; identifying multiple first users on the specified platform; pushing the evaluation task of the target audio to each first user and obtaining the evaluation parameters generated by each first user for the target audio in response to the evaluation task; determining quality evaluation parameters for the target audio based on the evaluation parameters; if the quality evaluation parameters meet a preset second condition; identifying multiple second users on the specified platform based on the user profiles of the users on the specified platform; pushing the target audio to each second user and obtaining user behavior data of the multiple second users regarding the target audio; and determining the recommendation strategy for the target audio on the specified platform based on the user behavior data and the user profiles of the multiple second users. This method can formulate corresponding recommendation strategies for audio with low exposure but high quality, improving the user's listening experience and increasing the user retention rate of the music platform.

[0043] In one specific embodiment, the first condition mentioned above may be: the time difference between the release time of the target audio on the designated platform and the current time is within a preset duration range, and the cumulative number of plays of the target audio on the designated platform is less than or equal to a preset number of plays threshold.

[0044] The cumulative playback count of audio on a specified platform takes time; therefore, the target audio needs to have been published on that platform for a period exceeding a certain duration, such as 7 days. Since the recommendation strategy for audio should be implemented as early as possible, a duration range can be set, such as 7 to 90 days. In the actual implementation, an endpoint within this duration range can also be set.

[0045] The aforementioned playback threshold can be set based on experience, such as 10,000. The first condition can also include that the target audio's daily playback volume on the specified platform is less than or equal to a certain threshold. For example, audio with fewer than 1,000 cumulative plays within 90 days of its release can be identified as the target audio.

[0046] The following embodiments provide a specific implementation for determining multiple first users of a specified platform.

[0047] In practical applications, different audio evaluation tasks can be distributed multiple times to the same user. The user's evaluation ability parameters are then determined based on the consistency of their historical audio ratings, the stability of their audio preferences, and their historical audio evaluation task processing performance. The consistency of historical audio ratings refers to whether the user's rating of an audio track in an audio evaluation task is consistent with the stable rating of that audio track on the specified platform. Consistency typically means that the deviation between the user's rating and the stable rating of the audio track is less than a certain value or percentage. The stability of audio preferences refers to whether the user prefers to rate a particular type of audio track highly in audio evaluation tasks. The historical audio evaluation task processing performance can include the duration of audio listening in the audio evaluation task, whether all aspects of the target audio were evaluated, and the number and quality of the written evaluation. Quantified data can be obtained for historical audio rating consistency, audio preference stability, and historical audio evaluation task processing performance. Then, based on this data and their corresponding weights, the user's evaluation ability parameters can be calculated.

[0048] Furthermore, multiple first users can be determined based on the evaluation ability parameters of users on a specified platform. Generally, users with stronger evaluation abilities can be selected as first users. In specific implementations, users whose evaluation ability parameters are greater than or equal to a preset evaluation ability threshold can be selected as first users. Alternatively, the evaluation ability parameters of users on the specified platform can be sorted from largest to smallest to obtain an evaluation ability ranking; then, users whose ranking in this ranking is less than or equal to a specified ranking position can be selected as first users. User listening preferences can also be considered, combining listening preferences that match the target audio with the user's evaluation ability parameters to jointly determine the first users.

[0049] The following embodiments provide a specific implementation method for determining the quality evaluation parameters of a target audio based on evaluation parameters.

[0050] In practical applications, evaluation parameters can include multiple basic rating parameters for the target audio and a recommendation rating. For example, the basic rating parameters might include a first rating parameter for the frequency response of the target audio, a second rating parameter for the singing technique, and a third rating parameter for the lyrics. Generally, a higher rating indicates that the user considers the target audio to be of higher quality in that aspect. The recommendation rating typically refers to the degree to which a user would recommend the target audio to other users. Generally, a higher recommendation rating indicates a higher evaluation of the target audio's quality by the user.

[0051] In practical implementation, an initial score can be calculated for each first user based on multiple basic rating parameters and a recommendation score. For example, all scores can be summed to obtain the initial score, or the summed result can be normalized and used as the initial score; there are no restrictions here. Then, based on the initial score, the recommendation score, and the first user's evaluation ability parameters, the first user's initial quality evaluation parameters for the target audio are calculated. Since the recommendation score is relatively important, its weight can be increased; for example, the product of the initial score, the recommendation score, and the first user's evaluation ability parameters can be used as the first user's initial quality parameters for the target audio. Based on multiple initial quality evaluation parameters for the target audio from the first users, the quality evaluation parameters for the target audio are determined. Specifically, all initial quality evaluation parameters can be summed to obtain the quality evaluation parameters for the target audio, or the average of the summed results can be used as the quality evaluation parameters for the target audio.

[0052] In practical implementation, scoring parameters with short listening durations or scores identical to previous scores can be removed from music evaluation tasks and excluded from determining the quality evaluation parameters of the target audio. When evaluation parameters include text content, parameters with minimal text content or those that are clearly copied and pasted from other people's music reviews should also be removed to maintain the validity of the quality evaluation parameters.

[0053] After determining the quality evaluation parameters, it can be judged whether the quality evaluation parameters are greater than or equal to the preset quality evaluation threshold. If so, it is determined that the quality evaluation parameters meet the second condition.

[0054] The following embodiments provide user categories corresponding to user profiles of users on a specified platform, and specific implementation methods for determining multiple user sets.

[0055] In practical applications, user profiles on a specified platform can be clustered to determine multiple user categories. These categories may include "trendy female vocal fans," "core users of Cantonese folk songs," and "Generation Z indie rock enthusiasts," among others. Each user profile corresponds to a specific user category.

[0056] Multiple user sets can be determined based on user profiles and corresponding user categories on a specified platform. Each user set includes multiple users whose user profiles correspond to the same user category. Multiple users are then selected from each user set and designated as second users. In other words, for different user categories, some users from that user category are selected as second users.

[0057] The following embodiments provide a specific implementation method for determining the recommendation strategy of target audio on a specified platform based on user behavior data and user profiles of multiple second users.

[0058] After obtaining user behavior data from second users regarding the target video, it is necessary to determine the potential parameters of the target audio based on this user behavior data and user profiles of multiple second users. The user behavior data includes second users' ratings of the target audio, their interactive actions, and playback repetition rates. Second user interactive actions can include actions such as adding to favorites, commenting, and sharing.

[0059] First, the rating characteristics of the target audio can be determined based on the ratings of multiple second users. These rating characteristics should include at least one of the following: average score, median score, and high score rate. Statistical calculations can be performed on the target audio ratings to calculate the average score, median score, and high score rate. Then, based on the user sets to which the multiple second users belong and their ratings of the target audio, the audience reach of the target audio can be determined. Furthermore, rating characteristics can be statistically analyzed for different user sets, and then the user sets with higher average scores or higher high score rates can be identified as the target audio's audience set. Finally, the audience reach of the target audio can be determined based on the size of the audience set.

[0060] Furthermore, the conversion rate of the target audio can be determined based on user behavior data from multiple second users. This conversion rate is calculated based on the number of times the second user has saved or forwarded the target video, and can also be determined by combining user behavior data from other users after the second user's recommendation. Further, the potential parameters of the target audio can be calculated based on its audience reach, rating characteristics, conversion rate, and preset weight parameters. In practice, the values ​​of these three factors can be multiplied by the weight parameters, and then the three calculation results can be added or multiplied together. The final calculation result is then used as the potential parameters of the target audio. The aforementioned weight parameters can be set based on experience.

[0061] Then, based on the potential parameters of the target audio, a recommendation strategy for the target audio on a specified platform needs to be determined. In specific implementation, if the potential parameters of the target audio are greater than or equal to a preset first threshold, the target audio is pushed to a first specified number of first target users. If the potential parameters of the target audio are less than the first threshold but greater than or equal to a preset second threshold, the user characteristics corresponding to the target audio are determined, and the target audio is pushed to a second specified number of second target users based on these user characteristics. The user profiles of the second target users match the user characteristics. The second specified number is less than a second specified number. For example, when the value range of the potential parameters is [0, 100], the first threshold can be set to 60, and the second threshold can be set to 40.

[0062] After implementing a recommendation strategy on the target audio, the target new user behavior data can be determined based on the implementation duration and potential parameters of the recommendation strategy. The potential parameters and the implementation duration of the recommendation strategy can be used as the target new user behavior data. This target new user behavior data typically refers to the target number of plays, but the number of favorites and comments can also be unified into the play count dimension; for example, one favorite is equivalent to two plays, and one comment is equivalent to five plays, etc.

[0063] Then, the actual user behavior data of the target audio on the designated platform during the implementation period of the recommendation strategy is obtained. The deviation between the target new user behavior data and the actual user behavior data is determined. This deviation can be absolute or relative, and is not limited here. If the deviation is greater than or equal to a preset deviation threshold, the rating ability parameter corresponding to the first user can be updated, the matching strategy between the target audio and the target set corresponding to the second target user can be updated, or the weight parameters in the calculation process of the potential parameter can be updated. For example, the consistency of the first user's historical audio ratings can be updated, thereby updating the rating ability parameter corresponding to the first user. The above matching strategy update process can be to adjust the matching degree between the target audio and the second target user, or the matching degree calculation method, etc. The above weight parameters can be adjusted based on the actual user behavior data and audience coverage breadth, rating characteristics, and conversion rate, so that the potential parameter calculated based on the adjusted weight parameters matches the actual user behavior data.

[0064] The following embodiment provides a specific implementation process for determining an audio recommendation strategy.

[0065] This process addresses the issue of designated platforms being unable to effectively identify high-quality content and accurately allocate resources when faced with a large number of "cold start songs" with insufficient play counts. It comprises two core phases: quality diagnostic assessment (determining quality evaluation parameters) and growth potential prediction assessment (determining potential parameters), supplemented by multiple enhancement mechanisms to improve the stability, credibility, and scalability of the scoring. Specifically, it can be implemented in the following ways: 1. Newly released songs with fewer than 1,000 daily plays are automatically entered into the cold start evaluation pool as cold start songs.

[0066] 2. The task engine prioritizes songs based on genre and language, pushing them to users with high appreciation for music. The task engine can be set up on or off a specific platform; there are no restrictions here.

[0067] 3. After users complete the scoring, the system records the scores for each song across three dimensions and subjective comments, and completes standardized preprocessing; 4. The system calculates the total score. If the total score is greater than or equal to the specified score (e.g., 60), the system proceeds to the next stage of processing.

[0068] 5. The platform accurately distributes songs to various user profiles and records their preference matching and interaction feedback.

[0069] 6. The system generates a final growth potential report and outputs "priority group suggestions" and "recommendation level suggestions" to the recommendation algorithm.

[0070] 7. The system continuously tracks subsequent data performance, updates user profiles and prediction model weights, and achieves iterative optimization of strategies.

[0071] This process can be embedded into the platform's recommendation system, enabling integrated support from pre-playback quality control to subsequent resource distribution strategies.

[0072] Phase 1: High-Accuracy Quality Diagnosis Mechanism The core objective of this stage is to obtain the most objective and reliable preliminary quality judgment results (i.e., the aforementioned "quality evaluation parameters") in the early stages when there is a lack of behavioral data for the songs, by using structured scoring mechanisms, user profiles, and multi-factor weighted calculations.

[0073] The platform constructs a "user appreciation model" based on users' historical behavior. The main evaluation criteria of this model include: the consistency between historical ratings and the platform's recognized quality results, the stability of users' preferences for different styles and languages ​​(such as whether they only rate a certain type of music), the completion rate of listening and review tasks and the quality of subjective comments (whether they provide effective language descriptions), and the subsequent performance of songs they have participated in (such as whether songs they participated in rating have a higher probability of being upgraded).

[0074] Based on the above factors, the platform dynamically calculates the user's rating ability parameters (also known as "rating credibility") and prioritizes the distribution of high-weight rating tasks accordingly to ensure the structural stability of the data.

[0075] The formula for calculating the credibility of the rating is as follows:

[0076] Where σ: standard deviation of sample scores, μ: average sample score, n: effective sample size, and ε: a small value to prevent division by zero.

[0077] In addition, task distribution is also based on matching users' musical style and language preferences with song characteristics to determine the first user, minimizing the scoring noise from "users who are not in the target group".

[0078] Then, a multi-dimensional, structured score can be given to the song. For a given song, the first user needs to rate it based on the following dimensions: Melody score (1-5): Consider rhythm, melody fluency, and memorability; Singing ability score (1-5): measures pitch accuracy, rhythm control, and emotional expression; Lyrics score (1-5): Evaluates the imagery, writing style, completeness of content, and emotional delivery; Subjective recommendation rating (1-5): reflects users' tendency to actively share the song; Subjective written music review: You must provide a minimum of 50 words of your listening experience for qualitative analysis.

[0079] After obtaining the evaluation parameters from the first user, rating behaviors such as "scoring time too short (e.g., <10 seconds)," "frequently scoring the middle value," or "obvious copy-pasting" can be automatically eliminated to ensure data authenticity.

[0080] Each user's individual score, along with their individual weighting factor, determines the final contribution of that data point. The total score can be calculated using the following formula: Total Score = ((Melody Score + Singing Skill Score + Lyrics Score) / 3 + Total Score) / 2 × Subjective Recommendation Index × Evaluation Credibility Among them, through platform historical modeling verification, songs that are willing to be recommended are more likely to have dissemination potential, so the subjective recommendation index is used as a weighted amplification item; Evaluation credibility reflects the seriousness of the first user's rating and is an indicator formed by the platform's statistics on the stability of user rating behavior, including rating range, content matching degree and past deviation.

[0081] This approach ensures that the score not only reflects users' evaluation of the song's quality but also their confidence that the song is "acceptable to the market," thus establishing the initial quality control standard for the song during its cold start phase.

[0082] Phase Two: A Multidimensional Prediction Mechanism for Growth Potential If a song reaches the platform's set threshold in the first stage of scoring (e.g., a weighted total score of ≥60 points), it will enter the second stage of evaluation, which aims to identify its potential audience, dissemination capabilities, and expected return on investment, serving as important inputs for the recommendation algorithm.

[0083] In practical applications, a designated platform can build a profile for each user based on the following characteristics: Music genre / language preference: Analysis of their long-term music listening distribution; Mood preferences: Preference for upbeat / healing / folk / rock / electronic music, etc. Interactive behaviors: frequency of saving, sharing, and commenting; Content dissemination tendency: the popularity of content that is actively shared; Basic attributes: age, region, gender, etc.

[0084] Through machine learning clustering technology, a series of "typical user group packages" are formed, such as: "trendy female vocal fans", "core users of Cantonese folk songs", and "Generation Z independent rock enthusiasts".

[0085] Selected songs will be pushed to multiple typical groups through scoring tasks, targeted distribution, and chart exposure, with at least 100 feedback data collected from each group. Further, the following metrics can be calculated: Positive Response Rate: The percentage of respondents who gave a score of ≥4. Conversion rate: Whether the user saved, commented, or actively shared the content; Playback repetition rate: The proportion of replays represents long-term engagement; Comparison bias: The difference in rating distribution among different groups, used to determine whether a song is a "niche hit" or a "broadly popular masterpiece". Furthermore, a growth index can be calculated based on the above indicators: Growth Score = (Average Score × Audience Reach × High Score Rate × Referral Conversion Rate) × Time Weighted Here, the Growth Score is equivalent to the product of the aforementioned potential parameter and the implementation duration. The time-weighted average is equivalent to the implementation duration of the strategy.

[0086] Specifically, songs can be categorized based on their growth index: When the growth index is ≥60, prioritize the cold start of the hot-selling resource pool, that is, push the target audio to a large number of users on the designated platform; When the growth index is between 40 and 60, targeted recommendations can be made to precisely reach users who like this type of audio. If the growth index is less than 40, you can temporarily refrain from investing and wait for natural feedback.

[0087] The platform records the discrepancy between the growth assessment output and the actual performance of the songs. If the actual performance is significantly higher or lower than the prediction, the following will be retrospectively corrected: the weight parameters of the user rating model, the tag matching strategy of the recommendation distribution strategy, and a certain weight deviation in the growth index.

[0088] This closed-loop mechanism ensures continuous model optimization, adapting to the rapidly changing music market environment and user aesthetic trends.

[0089] This method assesses song quality during the cold start phase when there are no plays, addressing the delayed judgment issue of existing models. It effectively captures and filters content invisible to the algorithm through human feedback. Then, based on the stratification of user feedback and recommendation intentions, it accurately identifies the song's audience suitability, thus saving on ineffective traffic costs. Trial operation data shows that songs scoring 60 points or higher had an upgrade rate more than 6 times the platform average in subsequent play, and the scoring coverage increased tenfold in the past year, enhancing content mining capabilities and the diversity of the platform's content ecosystem.

[0090] The comparison results of audio recommendations using the above method and those not using the above method are shown in Table 1.

[0091] Table 1

[0092] For the above method embodiments, see Figure 2 An audio recommendation strategy determination apparatus is shown, the apparatus comprising: The target audio determination module 202 is used to determine the target audio in a specified platform; the release time and playback volume of the target audio on the specified platform meet the preset first condition. The evaluation parameter acquisition module 204 is used to identify multiple first users on a specified platform, push the evaluation task of the target audio to each first user, and acquire the evaluation parameters of the target audio generated by each first user for the evaluation task; the evaluation parameters are used to indicate: the first user's subjective evaluation of the melody, emotion and expression of the target audio. The quality evaluation parameter determination module 206 is used to determine the quality evaluation parameters of the target audio based on the evaluation parameters. The second user determination module 208 is used to determine multiple second users in the specified platform based on the user profile of the user on the specified platform if the quality evaluation parameters meet the preset second conditions; the user profile is determined based on the user's basic attribute data and historical audio listening data on the specified platform. The recommendation strategy determination module 210 is used to push the target audio to each second user and obtain user behavior data of multiple second users for the target audio. Based on the user behavior data and the user profiles of multiple second users, the recommendation strategy of the target audio on the specified platform is determined.

[0093] The aforementioned audio recommendation strategy determination device identifies a target audio on a specified platform; identifies multiple first users on the specified platform, pushes evaluation tasks for the target audio to each first user, and obtains evaluation parameters for the target audio generated by each first user for the evaluation task; determines quality evaluation parameters for the target audio based on the evaluation parameters; if the quality evaluation parameters meet a preset second condition, identifies multiple second users on the specified platform based on user profiles of the users on the specified platform; pushes the target audio to each second user, and obtains user behavior data of the multiple second users for the target audio; and determines a recommendation strategy for the target audio on the specified platform based on the user behavior data and the user profiles of the multiple second users. This method can formulate corresponding recommendation strategies for audio with low exposure but high quality, improving the user's listening experience and increasing the user retention rate of the music platform.

[0094] The first condition mentioned above includes: the time difference between the release time of the target audio on the designated platform and the current time is within a preset duration range, and the cumulative number of plays of the target audio on the designated platform is less than or equal to a preset play count threshold.

[0095] The aforementioned evaluation parameter acquisition module is also used to: determine the evaluation ability parameters of users on a specified platform; the evaluation ability parameters are determined based on the consistency of users' historical audio ratings, the stability of audio preferences, and the processing status of historical audio evaluation tasks; based on the evaluation ability parameters of users on a specified platform, determine multiple first users; the evaluation ability parameters of the first users are greater than or equal to a preset evaluation ability threshold, or the ranking of the evaluation ability parameters of the first users in the evaluation ability ranking of users on the specified platform is less than or equal to a specified ranking; in the evaluation ability ranking, the evaluation ability parameters of users on the specified platform are arranged from largest to smallest.

[0096] The aforementioned evaluation parameters include multiple basic scoring parameters and a recommendation score for the target audio. The multiple basic scoring parameters include a first scoring parameter for the frequency of the target audio, a second scoring parameter for the singing technique of the target audio, and a third evaluation parameter for the lyrics of the target audio. The quality evaluation parameter determination module is also used to: calculate an initial score for each first user based on the multiple basic scoring parameters and the recommendation score; calculate the first user's initial quality evaluation parameters for the target audio based on the initial score, the recommendation score, and the first user's evaluation ability parameters; and determine the quality evaluation parameters for the target audio based on the initial quality evaluation parameters of the multiple first users for the target audio.

[0097] The above-mentioned device further includes: a judgment module for judging whether the quality evaluation parameter is greater than or equal to a preset quality evaluation threshold; and a condition satisfaction judgment module for determining that, if so, the quality evaluation parameter satisfies a second condition.

[0098] The aforementioned user profiles have corresponding user categories; user categories are determined by clustering user images of users on a specified platform; the second user determination module is also used to: determine multiple user sets based on the user categories corresponding to the user profiles of users on a specified platform; each user set includes multiple users, and the user categories corresponding to the user profiles of the multiple users are the same; determine multiple users from each user set, and determine the users determined from each user set as the second user.

[0099] The aforementioned recommendation strategy determination module is also used to: determine the potential parameters of the target audio based on user behavior data and user profiles of multiple second users; and determine the recommendation strategy for the target audio on a specified platform based on the potential parameters of the target audio.

[0100] The aforementioned multiple second users are categorized into multiple user sets based on their user profiles. User categories are determined by clustering user screens on a specified platform. User behavior data includes second users' ratings of the target audio, interactive actions, and playback repetition rates. Interactive actions include at least one of the following: collection, commenting, and forwarding. The recommendation strategy determination module is also used to: determine the rating characteristics of the target audio based on the ratings of multiple second users; rating characteristics include at least one of the following: average score, median score, and high score rate; determine the audience coverage of the target audio based on the user sets to which the multiple second users belong and their ratings of the target audio; determine the conversion rate of the target audio based on the user behavior data of multiple second users; and calculate the potential parameters of the target audio based on the audience coverage, rating characteristics, conversion rate, and preset weight parameters.

[0101] The aforementioned recommendation strategy determination module is further configured to: if the potential parameter of the target audio is greater than or equal to a preset first threshold, push the target audio to a first specified number of first target users; if the potential parameter of the target audio is less than the first threshold but greater than or equal to a preset second threshold, determine the user characteristics corresponding to the target audio, and push the target audio to a second specified number of second target users based on the user characteristics; the user profile of the second target users matches the user characteristics; and the second specified number is less than the second specified number.

[0102] The aforementioned first user has corresponding evaluation ability parameters; the aforementioned device further includes: a target data determination module, used to determine the target new user behavior data corresponding to the target audio based on the implementation duration and potential parameters of the recommendation strategy; an actual data acquisition module, used to acquire the actual user behavior data of the target audio on a specified platform within the implementation duration of the recommendation strategy; a deviation determination module, used to determine the deviation between the target new user behavior data and the actual user behavior data; and an operation execution module, used to perform at least one of the following operations if the deviation is greater than or equal to a preset deviation threshold: updating the evaluation ability parameters corresponding to the first user, updating the matching strategy between the target audio and the target set corresponding to the second target user, and updating the weight parameters in the calculation process of the potential parameters.

[0103] This embodiment also provides an electronic device, including a processor and a memory. The memory stores machine-executable instructions that can be executed by the processor. The processor executes the machine-executable instructions to implement the aforementioned audio recommendation strategy determination method, for example: The process involves identifying a target audio on a designated platform; ensuring the target audio's release time and play count on the platform meet a preset first condition; identifying multiple first users on the designated platform, pushing evaluation tasks for the target audio to each first user, and obtaining evaluation parameters generated by each first user for the evaluation task; these evaluation parameters indicate the first users' subjective evaluation of the target audio's melody, emotion, and expression; determining quality evaluation parameters for the target audio based on these evaluation parameters; if the quality evaluation parameters meet a preset second condition, identifying multiple second users on the designated platform based on their user profiles; these user profiles are determined based on users' basic attribute data and historical audio listening data on the designated platform; pushing the target audio to each second user, and obtaining user behavior data from these second users regarding the target audio; and determining a recommendation strategy for the target audio on the designated platform based on the user behavior data and the user profiles of these second users.

[0104] The above methods can formulate corresponding recommendation strategies for audio with low exposure but high quality, thereby improving the user's listening experience and increasing the user retention rate of the music platform.

[0105] Optionally, the first condition mentioned above includes: the time difference between the release time of the target audio on the specified platform and the current time is within a preset duration range, and the cumulative number of plays of the target audio on the specified platform is less than or equal to a preset play count threshold.

[0106] Optionally, the above step of determining multiple first users of a specified platform includes: determining the evaluation ability parameters of users on the specified platform; determining the evaluation ability parameters based on the consistency of users' historical audio ratings, the stability of audio preferences, and the processing status of historical audio evaluation tasks; determining multiple first users based on the evaluation ability parameters of users on the specified platform; the evaluation ability parameters of the first users are greater than or equal to a preset evaluation ability threshold, or the ranking of the evaluation ability parameters of the first users in the evaluation ability ranking of users on the specified platform is less than or equal to a specified ranking; in the evaluation ability ranking, the evaluation ability parameters of users on the specified platform are arranged from largest to smallest.

[0107] Optionally, the aforementioned evaluation parameters include multiple basic scoring parameters and a recommendation score for the target audio; the multiple basic scoring parameters include a first scoring parameter for the frequency of the target audio, a second scoring parameter for the singing skills of the target audio, and a third evaluation parameter for the lyrics of the target audio; the step of determining the quality evaluation parameters of the target audio based on the evaluation parameters includes: for each first user, calculating an initial score based on the multiple basic scoring parameters and the recommendation score; calculating the first user's initial quality evaluation parameters for the target audio based on the initial score, the recommendation score, and the first user's evaluation ability parameters; and determining the quality evaluation parameters of the target audio based on the multiple first users' initial quality evaluation parameters for the target audio.

[0108] Optionally, the above method further includes: determining whether the quality evaluation parameter is greater than or equal to a preset quality evaluation threshold; if so, determining that the quality evaluation parameter meets the second condition.

[0109] Optionally, the aforementioned user profile has a corresponding user category; the user category is determined by clustering the user images of users on a specified platform; the step of determining multiple second users on a specified platform based on the user profiles of users on a specified platform includes: determining multiple user sets based on the user categories corresponding to the user profiles of users on a specified platform; each user set includes multiple users, and the user categories corresponding to the user profiles of the multiple users are the same; determining multiple users from each user set, and identifying the users determined from each user set as second users.

[0110] Optionally, the steps described above for determining the recommendation strategy for the target audio on a specified platform based on user behavior data and user profiles of multiple second users include: determining the potential parameters of the target audio based on user behavior data and user profiles of multiple second users; and determining the recommendation strategy for the target audio on the specified platform based on the potential parameters of the target audio.

[0111] Optionally, the aforementioned multiple second users are categorized into multiple user sets based on their user profiles; user categories are determined by clustering user screens on a specified platform; user behavior data includes second users' ratings of the target audio, interactive actions, and playback repetition rates; interactive actions include at least one of the following: favorites, comments, and forwards; the step of determining the potential parameters of the target audio based on user behavior data and the user profiles of multiple second users includes: determining the rating characteristics of the target audio based on the ratings of multiple second users; rating characteristics include at least one of the following: average score, median score, and high score rate; determining the audience coverage of the target audio based on the user sets to which the multiple second users belong and their ratings of the target audio; determining the conversion rate of the target audio based on the user behavior data of multiple second users; and calculating the potential parameters of the target audio based on the audience coverage, rating characteristics, conversion rate, and preset weight parameters.

[0112] Optionally, the step of determining the recommendation strategy for the target audio on a specified platform based on the potential parameters of the target audio includes: if the potential parameters of the target audio are greater than or equal to a preset first threshold, pushing the target audio to a first specified number of first target users; if the potential parameters of the target audio are less than the first threshold but greater than or equal to a preset second threshold, determining the user characteristics corresponding to the target audio, and pushing the target audio to a second specified number of second target users based on the user characteristics; the user profile of the second target users matches the user characteristics; and the second specified number is less than the second specified number.

[0113] Optionally, the first user has corresponding evaluation ability parameters; the method further includes: determining the target new user behavior data corresponding to the target audio based on the implementation duration and potential parameters of the recommendation strategy; obtaining the actual user behavior data of the target audio on the specified platform within the implementation duration of the recommendation strategy; determining the deviation between the target new user behavior data and the actual user behavior data; if the deviation is greater than or equal to a preset deviation threshold, performing at least one of the following operations: updating the evaluation ability parameters corresponding to the first user, updating the matching strategy between the target audio and the target set corresponding to the second target user, and updating the weight parameters in the calculation process of the potential parameters.

[0114] See Figure 3 As shown, the electronic device includes a processor 100 and a memory 101. The memory 101 stores machine-executable instructions that can be executed by the processor 100. The processor 100 executes the machine-executable instructions to implement the aforementioned audio recommendation strategy determination method.

[0115] Furthermore, Figure 3The electronic device shown also includes a bus 102 and a communication interface 103, with the processor 100, the communication interface 103 and the memory 101 connected via the bus 102.

[0116] The memory 101 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 103 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 102 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0117] The processor 100 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 100 or by instructions in software form. The processor 100 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor may be a microprocessor or any conventional processor. The methods disclosed in the embodiments of this disclosure can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 101, and the processor 100 reads the information in memory 101 and, in conjunction with its hardware, completes the method of the aforementioned embodiments.

[0118] This embodiment also provides a machine-readable storage medium storing machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions cause the processor to implement the above-mentioned audio recommendation strategy determination method.

[0119] This disclosure provides an audio recommendation strategy determination method, apparatus, and electronic device, including a computer-readable storage medium storing program code. The program code includes instructions that can be used to execute the methods described in the preceding method embodiments, for example: The process involves identifying a target audio on a designated platform; ensuring the target audio's release time and play count on the platform meet a preset first condition; identifying multiple first users on the designated platform, pushing evaluation tasks for the target audio to each first user, and obtaining evaluation parameters generated by each first user for the evaluation task; these evaluation parameters indicate the first users' subjective evaluation of the target audio's melody, emotion, and expression; determining quality evaluation parameters for the target audio based on these evaluation parameters; if the quality evaluation parameters meet a preset second condition, identifying multiple second users on the designated platform based on their user profiles; these user profiles are determined based on users' basic attribute data and historical audio listening data on the designated platform; pushing the target audio to each second user, and obtaining user behavior data from these second users regarding the target audio; and determining a recommendation strategy for the target audio on the designated platform based on the user behavior data and the user profiles of these second users.

[0120] The above methods can formulate corresponding recommendation strategies for audio with low exposure but high quality, thereby improving the user's listening experience and increasing the user retention rate of the music platform.

[0121] Optionally, the above audio classification model is trained in the following way: An initial model is established; a sample set is determined; the sample set includes the audio features of multiple sample audios and the labeled quality scores of the sample audios; the labeled quality scores of the sample audios in the subset corresponding to each quality level are within the range of the quality scores corresponding to the quality level; training data is determined from the sample set; the training data includes the audio features of the target sample audio and the labeled quality scores of the target sample audio; the target sample audio is input into the initial model, and the initial model outputs the probability parameters corresponding to the audio features of the target sample audio; based on the probability parameters corresponding to the audio features of the target sample audio and the labeled quality scores of the target sample audio, the loss value of the initial model is calculated; based on the loss value, the model parameters of the initial model are adjusted, and the step of determining training data from the sample set is continued until the loss value converges, and the trained initial model is determined as the audio classification model.

[0122] Optionally, the first condition mentioned above includes: the time difference between the release time of the target audio on the specified platform and the current time is within a preset duration range, and the cumulative number of plays of the target audio on the specified platform is less than or equal to a preset play count threshold.

[0123] Optionally, the above step of determining multiple first users of a specified platform includes: determining the evaluation ability parameters of users on the specified platform; determining the evaluation ability parameters based on the consistency of users' historical audio ratings, the stability of audio preferences, and the processing status of historical audio evaluation tasks; determining multiple first users based on the evaluation ability parameters of users on the specified platform; the evaluation ability parameters of the first users are greater than or equal to a preset evaluation ability threshold, or the ranking of the evaluation ability parameters of the first users in the evaluation ability ranking of users on the specified platform is less than or equal to a specified ranking; in the evaluation ability ranking, the evaluation ability parameters of users on the specified platform are arranged from largest to smallest.

[0124] Optionally, the aforementioned evaluation parameters include multiple basic scoring parameters and a recommendation score for the target audio; the multiple basic scoring parameters include a first scoring parameter for the frequency of the target audio, a second scoring parameter for the singing skills of the target audio, and a third evaluation parameter for the lyrics of the target audio; the step of determining the quality evaluation parameters of the target audio based on the evaluation parameters includes: for each first user, calculating an initial score based on the multiple basic scoring parameters and the recommendation score; calculating the first user's initial quality evaluation parameters for the target audio based on the initial score, the recommendation score, and the first user's evaluation ability parameters; and determining the quality evaluation parameters of the target audio based on the multiple first users' initial quality evaluation parameters for the target audio.

[0125] Optionally, the above method further includes: determining whether the quality evaluation parameter is greater than or equal to a preset quality evaluation threshold; if so, determining that the quality evaluation parameter meets the second condition.

[0126] Optionally, the aforementioned user profile has a corresponding user category; the user category is determined by clustering the user images of users on a specified platform; the step of determining multiple second users on a specified platform based on the user profiles of users on a specified platform includes: determining multiple user sets based on the user categories corresponding to the user profiles of users on a specified platform; each user set includes multiple users, and the user categories corresponding to the user profiles of the multiple users are the same; determining multiple users from each user set, and identifying the users determined from each user set as second users.

[0127] Optionally, the steps described above for determining the recommendation strategy for the target audio on a specified platform based on user behavior data and user profiles of multiple second users include: determining the potential parameters of the target audio based on user behavior data and user profiles of multiple second users; and determining the recommendation strategy for the target audio on the specified platform based on the potential parameters of the target audio.

[0128] Optionally, the aforementioned multiple second users are categorized into multiple user sets based on their user profiles; user categories are determined by clustering user screens on a specified platform; user behavior data includes second users' ratings of the target audio, interactive actions, and playback repetition rates; interactive actions include at least one of the following: favorites, comments, and forwards; the step of determining the potential parameters of the target audio based on user behavior data and the user profiles of multiple second users includes: determining the rating characteristics of the target audio based on the ratings of multiple second users; rating characteristics include at least one of the following: average score, median score, and high score rate; determining the audience coverage of the target audio based on the user sets to which the multiple second users belong and their ratings of the target audio; determining the conversion rate of the target audio based on the user behavior data of multiple second users; and calculating the potential parameters of the target audio based on the audience coverage, rating characteristics, conversion rate, and preset weight parameters.

[0129] Optionally, the step of determining the recommendation strategy for the target audio on a specified platform based on the potential parameters of the target audio includes: if the potential parameters of the target audio are greater than or equal to a preset first threshold, pushing the target audio to a first specified number of first target users; if the potential parameters of the target audio are less than the first threshold but greater than or equal to a preset second threshold, determining the user characteristics corresponding to the target audio, and pushing the target audio to a second specified number of second target users based on the user characteristics; the user profile of the second target users matches the user characteristics; and the second specified number is less than the second specified number.

[0130] Optionally, the first user has corresponding evaluation ability parameters; the method further includes: determining the target new user behavior data corresponding to the target audio based on the implementation duration and potential parameters of the recommendation strategy; obtaining the actual user behavior data of the target audio on the specified platform within the implementation duration of the recommendation strategy; determining the deviation between the target new user behavior data and the actual user behavior data; if the deviation is greater than or equal to a preset deviation threshold, performing at least one of the following operations: updating the evaluation ability parameters corresponding to the first user, updating the matching strategy between the target audio and the target set corresponding to the second target user, and updating the weight parameters in the calculation process of the potential parameters.

[0131] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and apparatus described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0132] Furthermore, in the description of the embodiments of this disclosure, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this disclosure based on the specific circumstances.

[0133] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of this disclosure, in essence, or the parts that contribute to the prior art, or parts of the technical solutions, can be embodied in the form of software products. These computer software products are stored in a storage medium and include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0134] In the description of this disclosure, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this disclosure and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this disclosure. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0135] Finally, it should be noted that the above embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the protection scope of the claims.

Claims

1. A method for determining an audio recommendation strategy, characterized in that, The method includes: Identify the target audio on a designated platform; the release time and playback volume of the target audio on the designated platform meet a preset first condition; A plurality of first users on the designated platform are identified, and the evaluation task of the target audio is pushed to each of the first users. The evaluation parameters of the target audio generated by each of the first users for the evaluation task are obtained. The evaluation parameters are used to indicate the subjective evaluation of the first user on the melody, emotion and expression of the target audio. Based on the evaluation parameters, determine the quality evaluation parameters of the target audio; If the quality evaluation parameters meet the preset second condition, multiple second users in the designated platform are determined based on the user profile of the user on the designated platform; the user profile is determined based on the user's basic attribute data and the historical audio listening data on the designated platform. The target audio is pushed to each of the second users, and user behavior data of the multiple second users in response to the target audio is obtained. Based on the user behavior data and the user profiles of the multiple second users, a recommendation strategy for the target audio on the designated platform is determined.

2. The method according to claim 1, characterized in that, The step of identifying multiple first users of the specified platform includes: Determine the evaluation ability parameters of users in the specified platform; the evaluation ability parameters are determined based on the consistency of users' historical audio ratings, the stability of their audio preferences, and the processing status of their historical audio evaluation tasks. Based on the evaluation ability parameters of users in the designated platform, multiple first users are determined; the evaluation ability parameters of the first users are greater than or equal to a preset evaluation ability threshold, or the ranking of the evaluation ability parameters of the first users in the evaluation ability ranking of users in the designated platform is less than or equal to a specified ranking; in the evaluation ability ranking, the evaluation ability parameters of users in the designated platform are arranged from largest to smallest.

3. The method according to claim 1, characterized in that, The evaluation parameters include multiple basic scoring parameters and a recommendation level score for the target audio; the multiple basic scoring parameters include a first scoring parameter for the frequency of the target audio, a second scoring parameter for the singing skills of the target audio, and a third evaluation parameter for the lyrics of the target audio. The step of determining the quality evaluation parameters of the target audio based on the evaluation parameters includes: For each of the first users, an initial score is calculated based on the multiple basic rating parameters and the recommendation level score; Based on the initial rating, the recommendation rating, and the first user's evaluation ability parameters, the first user's initial quality evaluation parameters for the target audio are calculated. The quality evaluation parameters of the target audio are determined based on the initial quality evaluation parameters of the multiple first users for the target audio.

4. The method according to claim 1, characterized in that, The user profile has a corresponding user category; the user category is determined by clustering user images of users on the specified platform. The step of determining multiple second users on the specified platform based on the user profiles of users on the specified platform includes: Based on the user categories corresponding to the user profiles of the users on the specified platform, multiple user sets are determined; each user set includes multiple users, and the user categories corresponding to the user profiles of the multiple users are the same. Multiple users are determined from each of the user sets, and the users determined from each of the user sets are designated as the second user.

5. The method according to claim 1, characterized in that, The step of determining the recommendation strategy for the target audio on the designated platform based on the user behavior data and the user profiles of the multiple second users includes: Based on the user behavior data and the user profiles of the multiple second users, the potential parameters of the target audio are determined; Based on the potential parameters of the target audio, a recommendation strategy for the target audio on the specified platform is determined.

6. The method according to claim 5, characterized in that, The multiple second users are divided into multiple user sets based on the user categories corresponding to the user profiles of the second users; the user categories are determined by clustering the user screens of the users on the specified platform; the user behavior data includes the second users' ratings of the target audio, interactive operations, and playback repetition rates; the interactive operations include at least one of the following: collection operation, comment operation, and forwarding operation; The step of determining the potential parameters of the target audio based on the user behavior data and the user profiles of the multiple second users includes: Based on the ratings given to the target audio by the multiple second users, the rating characteristics of the target audio are determined; the rating characteristics include at least one of the following: average score, median score, and high score rate; Based on the user sets to which the multiple second users belong and their ratings of the target audio, the audience coverage of the target audio is determined; Based on the user behavior data of the multiple second users, the conversion rate of the target audio is determined; Based on the target audio's audience reach, rating characteristics, conversion rate, and preset weight parameters, the potential parameters of the target audio are calculated.

7. The method according to claim 5, characterized in that, The step of determining a recommendation strategy for the target audio on the specified platform based on the potential parameters of the target audio includes: If the potential parameter of the target audio is greater than or equal to a preset first threshold, the target audio will be pushed to a first specified number of first target users; If the potential parameter of the target audio is less than the first threshold and greater than or equal to the preset second threshold, the user characteristics corresponding to the target audio are determined, and the target audio is pushed to a second specified number of second target users based on the user characteristics; the user profile of the second target users matches the user characteristics; the second specified number is less than the second specified number.

8. An audio recommendation strategy determination device, characterized in that, The device includes: The target audio determination module is used to determine the target audio in a specified platform; the release time and playback volume of the target audio on the specified platform meet a preset first condition. The evaluation parameter acquisition module is used to identify multiple first users on the designated platform, push the evaluation task of the target audio to each first user, and acquire the evaluation parameters of the target audio generated by each first user for the evaluation task; the evaluation parameters are used to indicate: the first user's subjective evaluation of the melody, emotion and expression of the target audio; A quality evaluation parameter determination module is used to determine the quality evaluation parameters of the target audio based on the evaluation parameters. The second user determination module is used to determine multiple second users on the designated platform based on the user profile of the user on the designated platform if the quality evaluation parameters meet the preset second conditions; the user profile is determined based on the user's basic attribute data and the historical audio listening data on the designated platform. The recommendation strategy determination module is used to push the target audio to each of the second users, and obtain user behavior data of the multiple second users for the target audio. Based on the user behavior data and the user profiles of the multiple second users, the module determines the recommendation strategy of the target audio on the designated platform.

9. An electronic device, characterized in that, The system includes a processor and a memory, the memory storing machine-executable instructions that can be executed by the processor, the processor executing the machine-executable instructions to implement the audio recommendation strategy determination method according to any one of claims 1-7.

10. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores machine-executable instructions, which, when invoked and executed by a processor, cause the processor to implement the audio recommendation strategy determination method according to any one of claims 1-7.

Images