User feature acquisition method for content search, music search method and device
By segmenting users into groups with low activity levels and combining group and individual characteristics, the problem of insufficient behavioral data for low-activity users is solved, enabling more accurate representation of search preferences and content recommendations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TENCENT MUSIC ENTERTAINMENT TECH (SHENZHEN) CO LTD
- Filing Date
- 2023-07-20
- Publication Date
- 2026-06-05
AI Technical Summary
Existing technologies struggle to accurately represent the search preferences of less active users because their behavioral data is insufficient, resulting in inaccurate content search results.
By dividing inactive users into user groups with varying activity levels, the content preference characteristics of these groups are obtained. Combined with individual content preference characteristics, a trained group and individual feature extraction model is used to fuse these characteristics to obtain the user's search preference features.
It improves the accuracy of the search preference representation for inactive users, enabling it to provide them with content search results that better meet their needs.
Smart Images
Figure CN116894123B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and in particular to a method for obtaining user features for content search, a music search method, a computer device, and a storage medium. Background Technology
[0002] With the rapid development of digital content, the styles and genres of digital content, such as music and film, have become increasingly diverse. Each user may have different content preferences, and by combining these preferences with the user's search history, content search can provide tailored search results.
[0003] In response, current content search technologies rely heavily on user behavior data to obtain user search preferences, and the amount of data needs to reach a certain level. However, for users with low activity levels, the amount of their behavior data is relatively small, making it difficult for this technology to accurately represent their search preferences. Summary of the Invention
[0004] Therefore, it is necessary to provide a method for obtaining user features for content search, a music search method, a computer device, and a storage medium to address the aforementioned technical problems.
[0005] Firstly, this application provides a method for obtaining user features in content search. The method includes:
[0006] For a current user of a content application whose activity level has not reached the first activity level, determine the user group to which the current user belongs among multiple pre-divided user groups; each user group includes users with various activity levels of the content application;
[0007] Obtain the content preference characteristics of the user group to which the user belongs; the content preference characteristics are determined based on the behavioral data of users in the user group whose activity level reaches a second activity level in the content application; the second activity level is not lower than the first activity level;
[0008] Obtain the current user's personal content preference characteristics;
[0009] Based on the group content preference characteristics and the individual content preference characteristics, the search preference characteristics of the current user in the content application are obtained.
[0010] In one embodiment, before determining the user group to which the current user belongs among the pre-divided multiple user groups, the method further includes: selecting one or more user attribute features that do not include activity from a user attribute feature library; clustering each user of the content application based on the one or more user attribute features; and determining the multiple user groups based on the clustering results.
[0011] In one embodiment, determining the plurality of user groups based on the clustering results includes: obtaining a plurality of user groups to be evaluated based on the clustering results; determining whether each user group to be evaluated contains users with various activity levels based on the activity level of each user in each user group to be evaluated; if yes, then determining the plurality of user groups to be evaluated as the plurality of user groups; if no, then returning to the step of selecting one or more user attribute features that do not contain activity levels from the user attribute feature library.
[0012] In one embodiment, determining whether each user group to be evaluated contains users with various activity levels based on the activity level of each user in each user group to be evaluated includes: forming a first activity range based on the first activity level and forming a second activity range based on the second activity level; for each user group to be evaluated, determining the number of users falling into the first activity range and the number of users falling into the second activity range based on the activity level of each user in the user group to be evaluated; and determining whether the user group to be evaluated contains users with various activity levels based on the number of users falling into the first activity range and the number of users falling into the second activity range.
[0013] In one embodiment, determining whether the user group to be evaluated contains users with various activity levels based on the number of users falling into the first activity range and the number of users falling into the second activity range includes: if the number of users falling into the first activity range reaches a first user number threshold and the number of users falling into the second activity range reaches a second user number threshold, then it is determined that the user group to be evaluated contains users with various activity levels; if the number of users falling into the first activity range does not reach the first user number threshold or the number of users falling into the second activity range does not reach the second user number threshold, then it is determined that the user group to be evaluated does not contain users with various activity levels; wherein, the second user number threshold is not lower than the first user number threshold.
[0014] In one embodiment, the method further includes: for each pre-divided user group, obtaining behavioral data samples of sample users in the content application whose activity level reaches a second activity level, as behavioral data samples of the pre-divided user group; training a group feature extraction model to be trained using the behavioral data samples of each pre-divided user group to obtain a trained group feature extraction model; obtaining the group content preference features of the user group includes: inputting the behavioral data of users in the user group whose activity level reaches a second activity level in the content application into the trained group feature extraction model to obtain the group content preference features of the user group output by the trained group feature extraction model.
[0015] In one embodiment, the method further includes: acquiring behavioral data samples of sample users of the content application in the content application; using the behavioral data samples of sample users of the content application and the group content preference features of their respective user groups to train a personal feature extraction model and a feature fusion model to be trained, thereby obtaining a trained personal feature extraction model and a feature fusion model; acquiring the personal content preference features of the current user includes: inputting the behavioral data of the current user in the content application into the trained personal feature extraction model to acquire the personal content preference features of the current user output by the trained personal feature extraction model; obtaining the search preference features of the current user in the content application based on the group content preference features and the personal content preference features includes: inputting the group content preference features output by the trained group feature extraction model and the personal content preference features output by the trained personal feature extraction model into the trained feature fusion model to acquire the search preference features of the current user in the content application output by the trained feature fusion model.
[0016] Secondly, this application also provides a music search method. The method includes:
[0017] Get the music search keywords provided by the current user;
[0018] If the current user's activity level does not reach the first activity level, then the current user's search preference features are obtained according to the method of any of the above embodiments;
[0019] If the current user's activity level reaches the second activity level, then the current user's search preference characteristics are determined based on the current user's behavioral data, or the current user's search preference characteristics are obtained according to the method of any of the above embodiments; wherein, the second activity level is not lower than the first activity level;
[0020] Based on the music search keywords and search preference features, obtain music search results.
[0021] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:
[0022] For a current user whose activity level in a content application has not reached a first activity level, determine the user group to which the current user belongs among multiple pre-divided user groups; each user group includes users with various activity levels in the content application; obtain the group content preference characteristics of the user group to which the user belongs; the group content preference characteristics are determined based on the behavioral data of users in the user group whose activity level reaches a second activity level in the content application; the second activity level is not lower than the first activity level; obtain the current user's personal content preference characteristics; and obtain the current user's search preference characteristics in the content application based on the group content preference characteristics and the personal content preference characteristics.
[0023] Fourthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:
[0024] Obtain the music search keywords provided by the current user; if the current user's activity level does not reach a first activity level, obtain the current user's search preference features according to the method of any of the above embodiments; if the current user's activity level reaches a second activity level, determine the current user's search preference features based on the current user's behavioral data, or obtain the current user's search preference features according to the method of any of the above embodiments; wherein, the second activity level is not lower than the first activity level; obtain music search results based on the music search keywords and search preference features.
[0025] Fifthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:
[0026] For a current user whose activity level in a content application has not reached a first activity level, determine the user group to which the current user belongs among multiple pre-divided user groups; each user group includes users with various activity levels in the content application; obtain the group content preference characteristics of the user group to which the user belongs; the group content preference characteristics are determined based on the behavioral data of users in the user group whose activity level reaches a second activity level in the content application; the second activity level is not lower than the first activity level; obtain the current user's personal content preference characteristics; and obtain the current user's search preference characteristics in the content application based on the group content preference characteristics and the personal content preference characteristics.
[0027] Sixthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:
[0028] Obtain the music search keywords provided by the current user; if the current user's activity level does not reach a first activity level, obtain the current user's search preference features according to the method of any of the above embodiments; if the current user's activity level reaches a second activity level, determine the current user's search preference features based on the current user's behavioral data, or obtain the current user's search preference features according to the method of any of the above embodiments; wherein, the second activity level is not lower than the first activity level; obtain music search results based on the music search keywords and search preference features.
[0029] The aforementioned method for obtaining user characteristics for content search, music search method, computer equipment, and storage medium, for a current user of a content application whose activity level has not reached a first activity level, determines the user group to which the current user belongs among multiple pre-divided user groups. Each user group includes users with various activity levels in the content application. The group content preference characteristics of the user group to which the user belongs are obtained. These group content preference characteristics are determined based on the behavioral data of users in the user group whose activity level reaches a second activity level in the content application. The second activity level is not lower than the first activity level. The personal content preference characteristics of the current user are obtained. Based on the group content preference characteristics and the personal content preference characteristics, the search preference characteristics of the current user in the content application are obtained. This solution targets users whose activity level is below the first level. By determining the user group to which the user belongs among multiple pre-divided user groups, the solution obtains the content preference characteristics of the user group to which the user belongs. Then, based on the user's personal content preference characteristics and the content preference characteristics of the user group to which the user belongs, the solution characterizes the user's search preference characteristics. This allows the search preference characteristics of low-activity users to be characterized by the content preference characteristics represented by the behavioral data of high-activity users in the user group to which the user belongs, which has sufficient behavioral data. This approach takes into account both group preferences and personal preferences, and can at least improve the accuracy of the search preference representation of low-activity users. As a result, it can also provide content search results that are more suitable for users with lower activity levels. Attached Figure Description
[0030] Figure 1 This is a diagram illustrating the application environment of the relevant methods in the embodiments of this application;
[0031] Figure 2 This is a flowchart illustrating the user feature acquisition method for content search in this embodiment of the application.
[0032] Figure 3 This is a flowchart illustrating the steps for determining multiple user groups in an embodiment of this application;
[0033] Figure 4 This is a flowchart illustrating the further steps of determining multiple user groups in an embodiment of this application;
[0034] Figure 5 This is a flowchart illustrating the steps for determining active users in an embodiment of this application;
[0035] Figure 6 This is a flowchart illustrating the music search method in an embodiment of this application;
[0036] Figure 7 This is a diagram showing the internal structure of a computer device in an embodiment of this application. Detailed Implementation
[0037] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0038] The user feature acquisition method for content search and the music search method provided in this application embodiment can be applied to, for example... Figure 1 The application environment shown includes a terminal 110 and a server 120, with the terminal 110 communicating with the server 120 via a network. A data storage system can store the data that the server 120 needs to process. The data storage system can be integrated onto the server 120 or located in the cloud or on other network servers. The terminal 110 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. The server 120 can be implemented using a standalone server or a server cluster consisting of multiple servers.
[0039] Specifically, the user feature acquisition method for content search and the music search method provided in this application can be executed by terminal 110 or server 120, or by terminal 110 and server 120 in cooperation. As an example, regarding the user feature acquisition method for content search in this application, taking server 120 as an example, for users whose activity level has not reached the first activity level, server 120 determines the user group to which the user belongs among multiple pre-divided user groups. Server 120 obtains the group content preference features of the user group to which the user belongs, and server 120 obtains the user's personal content preference features. Server 120 obtains the user's search preference features based on the group content preference features and the personal content preference features. As an example of the music search method in this application, taking terminal 110 and server 120 as examples, terminal 110 can obtain music search keywords provided by the current user and then send the music search keywords to server 120. Server 120 can determine the current user's search preference characteristics using an appropriate method based on the current user's activity level. Specifically: if the current user's activity level does not reach the first activity level, the search preference characteristics of the current user are obtained according to the aforementioned user feature acquisition method for content search; if the current user's activity level reaches the second activity level, the search preference characteristics of the current user are determined based on the current user's behavioral data or according to the aforementioned user feature acquisition method for content search. Then, server 120 can obtain music search results based on the music search keywords and search preference characteristics, and the music search results can be fed back to terminal 110 for display.
[0040] The following description, in conjunction with various embodiments and corresponding figures, will describe the user feature acquisition method for content search and the music search method provided in this application.
[0041] In one embodiment, such as Figure 2 As shown, a method for obtaining user features for content search is provided. This method can be executed by server 120 and may include the following steps:
[0042] Step S201: For the current user of a content application whose activity level has not reached the first activity level, determine the user group to which the current user belongs among the pre-divided multiple user groups; each user group includes users with various activity levels of the content application.
[0043] Step S202: Obtain the content preference characteristics of the user group to which the user belongs; the content preference characteristics are determined based on the behavioral data of users in the user group whose activity level reaches the second activity level in the content application; the second activity level is not lower than the first activity level.
[0044] Steps S201 to S202 are steps in this application to extract group content preference features from the search preference features of current users of content applications whose activity level has not reached the first activity level. Here, current users of content applications whose activity level has not reached the first activity level can be considered low-activity users within the content application. The level of user activity in the content application can be quantified and measured based on information such as usage time and login frequency in content applications such as music applications. The specific activity calculation method can be determined according to relevant business needs. Therefore, a first activity level can be preset. In step S201, when the current user's activity level does not reach the first activity level, the user group to which the current user belongs among the pre-divided multiple user groups can be determined. Here, all users using the content application can be pre-divided into multiple user groups. Unbiased features unrelated to activity level (unbiased features refer to treating all users equally without any bias) can be selected to classify all users of the content application into multiple user groups, so that each user group contains users with various activity levels in the content application, specifically including users with high activity level, medium activity level, and low activity level. After multiple user groups are formed according to features, the user group to which the current user belongs can be determined among the pre-divided multiple user groups based on the corresponding features of the current user whose activity level does not reach the first activity level during the content search service stage.
[0045] After determining the user group to which a user belongs, the process proceeds to step S202, where the content preference characteristics of that user group are obtained. These group content preference characteristics refer to the overall content preference characteristics of the user group, rather than those of a specific individual. These characteristics can be determined based on the behavioral data of users within the user group who reach a second level of activity in the content application. The second level of activity is not lower than the first level of activity and can be the same. Users within the user group who reach the second level of activity can be categorized as medium-activity and high-activity users. In other words, the content preference characteristics of a user group can be determined based on the behavioral data of medium-activity and high-activity users within that user group regarding content in the content application, thereby identifying the common content preference characteristics of the user group to which the user belongs. In practical applications, behavioral data refers to user behavior data within the content application, which can include content usage behavior data, content operation behavior data, etc. Taking a music application as an example, content usage behavior data can include listening to music, and content operation behavior data can include searching for music. The content preference characteristics of the corresponding user group can be learned based on the behavioral data of users within each user group who reach the second level of activity in the content application.
[0046] In some embodiments, the method of this application may further include the following steps before step S202:
[0047] For each pre-divided user group, obtain behavioral data samples of sample users whose activity level reaches the second highest level in the content application, and use them as behavioral data samples of the pre-divided user group; use the behavioral data samples of each pre-divided user group to train the group feature extraction model to be trained, and obtain the trained group feature extraction model.
[0048] Therefore, step S202, obtaining the content preference characteristics of the user group, may specifically include:
[0049] The behavioral data of users with the second highest activity level in the user group are input into the trained group feature extraction model to obtain the group content preference features of the user group output by the trained group feature extraction model.
[0050] This embodiment further includes extracting the content preference features of user groups through a trained group feature extraction model. Specifically, for each pre-divided user group, behavioral data samples of users with activity levels reaching the second level (high activity level) in the content application are obtained as behavioral data samples of the pre-divided user group. Then, the behavioral data samples of each sample user in the content application of each pre-divided user group are input into the group feature extraction model to be trained for training. This allows the group feature extraction model to learn how to extract the content preference features of user groups during training, resulting in a trained group feature extraction model. After obtaining the trained group feature extraction model, the content preference features of each user group can be extracted without waiting to determine the user's group. Specifically, the behavioral data of users with activity levels reaching the second level in the content application of each user group, including the aforementioned user group, can be input into the trained group feature extraction model to obtain the content preference features of each user group, including the aforementioned user group, output by the trained group feature extraction model, thereby obtaining the content preference features of the user group. In this model, user behavior data {fea1,fea2,...} can be first mapped to embedded features {fea1,fea2,...} => {BEem1,BEem2,...} in the group feature extraction model, and then the group content preference feature Group is output based on this. embed =GroupModel(BehaviorEmb).
[0051] Step S203: Obtain the current user's personal content preference characteristics.
[0052] Step S204: Based on the group content preference characteristics and the individual content preference characteristics, obtain the current user's search preference characteristics in content applications.
[0053] In steps S203 and S204, step S203 is the step in the personal content preference feature extraction of the current user's search preference features for users whose activity level has not reached the first activity level in this embodiment, and step S204 is the feature fusion step. Specifically, in step S203, the user's personal content preference features can be obtained based on the user's behavioral data in the content application. embed In step S204, the group content preference feature can be... embed Personal content preference characteristics embed The search preferences of users are then fused to obtain their search preference features. As one implementation method, these group content preference features can be grouped according to a certain fusion coefficient α. embedPersonal content preference characteristics embed The user's search preference features are obtained by fusion. embed =αGroup embed +(1-α)Personal embed .
[0054] In some embodiments, the method of this application may further include the following steps before step S203:
[0055] Obtain behavioral data samples of sample users of the content application; use the behavioral data samples of sample users of the content application and the group content preference characteristics of their respective user groups to train the personal feature extraction model and feature fusion model to be trained, and obtain the trained personal feature extraction model and feature fusion model.
[0056] Therefore, step S203, obtaining the current user's personal content preference features, can specifically include: inputting the current user's behavioral data in the content application into a trained personal feature extraction model to obtain the current user's personal content preference features output by the trained personal feature extraction model. And step S204, obtaining the current user's search preference features in the content application based on the group content preference features and the personal content preference features, can specifically include: inputting the group content preference features output by the trained group feature extraction model and the personal content preference features output by the trained personal feature extraction model into a trained feature fusion model to obtain the current user's search preference features in the content application output by the trained feature fusion model.
[0057] This embodiment further includes extracting personal content preference features and performing subsequent feature fusion processing sequentially through a trained personal feature extraction model and a feature fusion model. Specifically, behavioral data samples of various sample users (including high, medium, and low activity levels) using the content application can be obtained. These behavioral data samples are then input into the personal feature extraction model to be trained. The model first maps the behavioral data samples of each sample user to an embedded feature BehaviorEmb, and then outputs the personal content preference feature Personal based on this. embed =PersonalModel(BehaviorEmb), then personal content preference features Personal embed Group content preference characteristics embed The data is transmitted to the feature fusion model, which learns the fusion coefficient α to incorporate the group content preference features. embed Personal content preference characteristicsembed The search preference features of the sample users are obtained by fusion. embed =αGroup embed +(1-α)Personal embed Thus, the personal feature extraction model and the feature fusion model to be trained are trained simultaneously, so that the personal feature extraction model learns how to extract personal content preference features during training, and the feature fusion model learns the fusion coefficient α during training, resulting in the trained personal feature extraction model and feature fusion model.
[0058] After obtaining the trained personal feature extraction model and feature fusion model, when providing content search services to users in a content application, including current users whose activity level has not reached the first level of activity, the model can extract personal content preference features for each user and fuse group and personal features to obtain the user's search preference features. Specifically, the behavioral data of current users whose activity level has not reached the first level of activity in the content application can be input into the trained personal feature extraction model to obtain the current user's personal content preference features output by the trained personal feature extraction model. embed Then, this personal content preference feature can be... embed Group content preference characteristics embed The data is transmitted to a trained feature fusion model, which outputs the user's search preference features based on the learned fusion coefficient α, thereby incorporating personal content preference features. embed Group content preference characteristics embed The search preference features of the current user are obtained by fusion. embed =αGroup embed +(1-α)Personal embed This allows for the accurate extraction of the current user's search preference characteristics.
[0059] The aforementioned method for obtaining user characteristics for content search, music search method, computer equipment, and storage medium, for a current user of a content application whose activity level has not reached a first activity level, determines the user group to which the current user belongs among multiple pre-divided user groups. Each user group includes users with various activity levels in the content application. The group content preference characteristics of the user group to which the user belongs are obtained. These group content preference characteristics are determined based on the behavioral data of users in the user group whose activity level reaches a second activity level in the content application. The second activity level is not lower than the first activity level. The personal content preference characteristics of the current user are obtained. Based on the group content preference characteristics and the personal content preference characteristics, the search preference characteristics of the current user in the content application are obtained. This solution targets users whose activity level is below the first level. By determining the user group to which the user belongs among multiple pre-divided user groups, the solution obtains the content preference characteristics of the user group to which the user belongs. Then, based on the user's personal content preference characteristics and the content preference characteristics of the user group to which the user belongs, the solution characterizes the user's search preference characteristics. This allows the search preference characteristics of low-activity users to be characterized by the content preference characteristics represented by the behavioral data of high-activity users in the user group to which the user belongs, which has sufficient behavioral data. This approach takes into account both group preferences and personal preferences, and can at least improve the accuracy of the search preference representation of low-activity users. As a result, it can also provide content search results that are more suitable for users with lower activity levels.
[0060] In one embodiment, such as Figure 3 As shown, the method of this application can determine the multiple user groups by means of the following steps before determining the user group to which the current user belongs in the pre-divided multiple user groups in step S201:
[0061] Step S301: Select one or more user attribute features that do not include activity level from the user attribute feature library.
[0062] In this step, one or more user attribute features can be selected from a user attribute feature library formed based on the user attributes of all users using the content application. These one or more user attribute features do not include activity level, making the selected user attribute features irrelevant to user activity. As an example, user attribute features may include age, gender, and city.
[0063] Step S302: Cluster the users of the content application based on one or more user attribute features.
[0064] Step S303: Based on the clustering results, identify multiple user groups.
[0065] Steps S302 to S303 mainly involve clustering users based on one or more user attribute features that do not include activity levels, and determining multiple user groups based on the clustering results. Specifically, one or more user attribute features that do not include activity levels can first be used to generate corresponding embedded features (fea). embed Then, based on the corresponding embedding feature fea embed Each user is clustered, and multiple user groups are determined based on the clustering results. The solution in this embodiment clusters each user based on one or more user attribute features that do not include activity level, and forms multiple user groups accordingly. This achieves the goal of simplifying the pre-stratification of users, saving costs, and ensuring that each user group includes users with various levels of activity (such as high, medium, and low activity levels).
[0066] Furthermore, in one embodiment, such as Figure 4 As shown, step S303 in the above embodiment, which determines multiple user groups based on the clustering results, may specifically include:
[0067] Step S401: Based on the clustering results, multiple user groups to be evaluated are obtained.
[0068] Step S402: Based on the activity level of each user in each user group to be evaluated, determine whether each user group to be evaluated contains users with various activity levels.
[0069] Step S4031: If so, then the multiple user groups to be evaluated are identified as multiple user groups.
[0070] Step S4032: If not, return to the step of selecting one or more user attribute features from the user attribute feature library that do not contain activity level.
[0071] This embodiment can evaluate the activity levels of users in each user group when multiple user groups are obtained based on clustering results, and thus ultimately obtain multiple user groups containing users with various activity levels. Specifically, in step S401, the multiple user groups obtained based on the clustering results are set as multiple user groups to be evaluated, and then the activity level of each user in each user group to be evaluated is obtained. Then, in step S402, based on the activity level of each user in each user group to be evaluated, it is determined whether each user group to be evaluated contains users with various activity levels, such as whether it contains users with high, medium, and low activity levels. In step S4031, if yes, the multiple user groups to be evaluated are determined as multiple user groups. In step S4032, if not, the process returns to step S301, which is the step of selecting one or more user attribute features from the user attribute feature library that do not contain activity levels. Thus, the selection, clustering, and activity evaluation of one or more user attribute features can be repeated until it is determined that each user group to be evaluated contains users with various activity levels. Among them, reselecting one or more user attribute features means re-performing step S301, re-clustering means re-performing steps S302 and S401, and re-performing activity evaluation means re-performing step S402. Through this process, multiple user groups that all contain users with various levels of activity can be automatically evaluated and finally determined.
[0072] Furthermore, in one embodiment, such as Figure 5 As shown, step S402, which determines whether each user group to be evaluated contains users with various activity levels based on the activity level of each user in each user group to be evaluated, may include:
[0073] Step S501: Form a first activity interval based on the first activity level, and form a second activity interval based on the second activity level.
[0074] In this step, the activity range below the first activity level can be set as the first activity range, and the activity range above or equal to the second activity level can be set as the second activity range. The second activity level is not lower than the first activity level, and the second activity level can be equal to the first activity level. That is, in practical applications, the activity range below the first activity level can be set as the first activity range to locate low-activity users, and the activity range above or equal to the first activity level can be set as the second activity range to locate high-activity users.
[0075] Step S502: For each user group to be evaluated, determine the number of users falling into the first activity range and the number of users falling into the second activity range based on the activity level of each user in the user group to be evaluated.
[0076] In this step, for each user group to be evaluated, we can count the number of users whose activity level falls into the first activity level range and the number of users whose activity level falls into the second activity level range.
[0077] Step S503: Based on the number of users falling into the first activity range and the number of users falling into the second activity range, determine whether the user group to be evaluated includes users with various activity levels.
[0078] In this step, for each user group to be evaluated, based on the number of users whose activity falls into the first activity range and the number of users whose activity falls into the second activity range, it is determined whether the corresponding user group to be evaluated contains users with various activity levels, so as to accurately assess the inclusion of users with various activity levels in each user group to be evaluated.
[0079] Furthermore, in some embodiments, determining whether the user group to be evaluated includes users with various activity levels based on the number of users falling into the first activity range and the number of users falling into the second activity range in step S503 may include:
[0080] If the number of users falling into the first activity range reaches the first user number threshold and the number of users falling into the second activity range reaches the second user number threshold, then it is determined that the user group to be evaluated includes users with all levels of activity. If the number of users falling into the first activity range does not reach the first user number threshold or the number of users falling into the second activity range does not reach the second user number threshold, then it is determined that the user group to be evaluated does not include users with all levels of activity. The second user number threshold is not lower than the first user number threshold.
[0081] In this embodiment, in order to evaluate the user group to be evaluated by statistically analyzing the number of users falling into the first activity range and the number of users falling into the second activity range, a threshold for the number of users is further added to improve the contribution of the multiple user groups to the accuracy of extracting the content preference features of the group. Specifically, for the first activity level range, the number of users falling into the first activity level range needs to reach a first user number threshold. For the second activity level range, the number of users falling into the second activity level range needs to reach a second user number threshold, and this second user number threshold is not lower than the first user number threshold. When the number of users falling into the first activity level range reaches the first user number threshold and the number of users falling into the second activity level range reaches the second user number threshold, it is determined that the user group to be evaluated includes users with various activity levels. If the number of users falling into the first activity level range does not reach the first user number threshold or the number of users falling into the second activity level range does not reach the second user number threshold, one or more user attribute features can be re-selected, clustered, and activity evaluation processed. This ensures that each user group formed in the end contains a certain number of high, medium, and low activity users, and also ensures that high and medium activity users have a high proportion in each user group. Therefore, in the stage of extracting group content preference features, the behavioral data of high and medium activity users in the user group can better represent the group content preference features of the user group.
[0082] In one embodiment, such as Figure 6 As shown, a music search method is provided, which can be executed by server 120, and the method may include the following steps:
[0083] Step S601: Obtain the music search keywords provided by the current user.
[0084] Step S6021: If the current user's activity level has not reached the first activity level, then obtain the current user's search preference features according to the method of any of the above embodiments.
[0085] Step S6022: If the current user's activity level reaches the second activity level, then determine the current user's search preference features based on the current user's behavior data, or obtain the current user's search preference features according to the method of any of the above embodiments.
[0086] Step S603: Obtain music search results based on music search keywords and search preference features.
[0087] The solution of this embodiment can apply the user feature acquisition method for content search of the present application to the field of music search. Among them, with the rapid development of digital music and mobile music, the styles of music have evolved into a variety of forms, and everyone's music preferences are also different. This embodiment can combine the user's music preferences in music search and finally give music search results that satisfy the user. Specifically, in this embodiment, in step S601, the music search keyword input by the current user using the search service of the music application can be obtained. The server 120 also determines whether the activity of the current user reaches the first activity level, that is, it can determine whether the current user is a low-activity user. If the current user is a low-activity user, then in step S6021, the search preference feature of the current user is obtained according to the user feature acquisition method for content search provided in the foregoing embodiment of the present application. If the server 120 determines that the activity of the current user reaches the second activity level (the second activity level is not lower than the first activity level), that is, the current user is a high or medium-activity user, then the server 120 can directly determine the search preference feature of the current user according to the behavior data of the current user, or obtain the search preference feature of the current user according to the user feature acquisition method for content search provided in the foregoing embodiment of the present application. Thus, this embodiment can obtain the search preference feature of the current user when the current user belongs to a low-activity user or a high or medium-activity user, and then in step S603, according to the music search keyword and search preference feature of the current user, obtain the music search result. This solution can take into account the interest expression of the group while emphasizing the expression of personal interests, provide music search results adapted to high-activity users and low-activity users, and improve the search satisfaction of users in the music field. For example, when the current user inputs the music search keyword "Map of Mountains and Rivers", because different users like different versions, such as versions of different singers, for high or medium-activity users, their search preference features can be learned based on their behavior data and music search results can be given accordingly. For low-activity users, the user group to which they belong can be determined through user attribute features such as age and city, and the version preferred by the user group in young first- and second-tier cities, such as the singer's version, can be learned and music search results can be given accordingly, achieving the interest transfer of the same user group, giving possible interest expressions of the user, taking into account the preferences of the group and the personal music preferences of the user, and improving the search satisfaction of the user.
[0088] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0089] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database can be used to store data such as preference characteristics. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external devices via a network connection. When executed by the processor, the computer program implements a user feature acquisition method for content search and a music search method.
[0090] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0091] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0092] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.
[0093] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0094] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.
[0095] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0096] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0097] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for obtaining user features in content search, characterized in that, The method includes: For a current user of a content application whose activity level has not reached the first activity level, determine the user group to which the current user belongs among multiple pre-divided user groups; each user group includes users of the content application with various activity levels; the multiple user groups are formed by classifying the users of the content application according to unbiased features that are unrelated to activity level; Obtain the content preference characteristics of the user group to which the user belongs; the content preference characteristics are determined based on the behavioral data of users in the user group whose activity level reaches a second activity level in the content application; the second activity level is not lower than the first activity level; Obtain the current user's personal content preference characteristics; Based on the group content preference characteristics and the individual content preference characteristics, the search preference characteristics of the current user in the content application are obtained; the step of obtaining the search preference characteristics of the current user in the content application based on the group content preference characteristics and the individual content preference characteristics includes: fusing the group content preference characteristics and the individual content preference characteristics according to a fusion coefficient to obtain the search preference characteristics of the current user in the content application.
2. The method according to claim 1, characterized in that, Before determining the user group to which the current user belongs among the pre-divided multiple user groups, the method further includes: Select one or more user attribute features from the user attribute feature library that do not include activity level; Cluster the users of the content application based on one or more user attribute features; Based on the clustering results, the multiple user groups are identified.
3. The method according to claim 2, characterized in that, The step of determining the multiple user groups based on the clustering results includes: Based on the clustering results, several user groups to be evaluated were obtained; Based on the activity level of each user in each user group to be evaluated, determine whether each user group to be evaluated contains users with various activity levels; If so, then the plurality of user groups to be evaluated are identified as the plurality of user groups; If not, return to the step of selecting one or more user attribute features from the user attribute feature library that do not include activity level.
4. The method according to claim 3, characterized in that, The step of determining whether each user group to be evaluated contains users with various activity levels based on the activity level of each user in each user group to be evaluated includes: A first activity range is formed based on the first activity level, and a second activity range is formed based on the second activity level. For each user group to be evaluated, the number of users falling into the first activity range and the number of users falling into the second activity range are determined based on the activity level of each user in the user group to be evaluated. Based on the number of users falling into the first activity range and the number of users falling into the second activity range, determine whether the user group to be evaluated includes users with various activity levels.
5. The method according to claim 4, characterized in that, The step of determining whether the user group to be evaluated includes users with various activity levels based on the number of users falling into the first activity range and the number of users falling into the second activity range includes: If the number of users falling into the first activity range reaches the first user number threshold and the number of users falling into the second activity range reaches the second user number threshold, then it is determined that the user group to be evaluated includes users with various activity levels. If the number of users falling into the first activity range does not reach the first user number threshold or the number of users falling into the second activity range does not reach the second user number threshold, then it is determined that the user group to be evaluated does not include users with various activity levels. Wherein, the second user number threshold is not lower than the first user number threshold.
6. The method according to claim 1, characterized in that, The method further includes: For each pre-divided user group, obtain behavioral data samples of sample users whose activity level reaches the second level in the content application, and use them as behavioral data samples of the pre-divided user group. The group feature extraction model to be trained is obtained by using behavioral data samples of each pre-divided user group. The process of obtaining the content preference characteristics of the user group to which the user belongs includes: The behavioral data of users in the user group whose activity level reaches the second level in the content application are input into the trained group feature extraction model to obtain the group content preference features of the user group output by the trained group feature extraction model.
7. The method according to claim 1 or 6, characterized in that, The method further includes: Obtain sample user behavior data of the content application; Using the behavioral data samples of the sample users of the content application and the group content preference characteristics of their respective user groups, the personal feature extraction model and feature fusion model to be trained are trained to obtain the trained personal feature extraction model and feature fusion model. The process of obtaining the current user's personal content preference characteristics includes: Input the current user's behavior data in the content application into a trained personal feature extraction model to obtain the current user's personal content preference features output by the trained personal feature extraction model; The step of obtaining the current user's search preference characteristics in the content application based on the group content preference characteristics and the individual content preference characteristics includes: The group content preference features output by the trained group feature extraction model and the individual content preference features output by the trained individual feature extraction model are input into the trained feature fusion model to obtain the current user's search preference features in the content application output by the trained feature fusion model.
8. A music search method, characterized in that, The method includes: Get the music search keywords provided by the current user; If the current user's activity level does not reach the first activity level, then the search preference features of the current user are obtained according to any one of claims 1 to 7; If the current user's activity level reaches a second activity level, then the current user's search preference characteristics are determined based on the current user's behavioral data, or the current user's search preference characteristics are obtained according to any one of claims 1 to 7; wherein, the second activity level is not lower than the first activity level; Based on the music search keywords and search preference features, obtain music search results.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1 to 7 or claim 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1 to 7 or claim 8.