Data processing method and device, computer readable medium and electronic equipment

Abstract

The application discloses a data processing method and device, a computer readable medium and an electronic device. The method comprises: obtaining an original data set of a target object on a multimedia platform, the original data set being a set of multimedia data; performing clustering processing on the multimedia data in the original data set based on a plurality of preset cluster numbers to obtain clustering results corresponding to the preset cluster numbers; determining a target cluster number according to the clustering results corresponding to the preset cluster numbers, and obtaining a plurality of multimedia clusters according to the clustering results corresponding to the target cluster number; and screening the plurality of multimedia clusters according to the multimedia data aggregation degree corresponding to each multimedia cluster to obtain a target multimedia cluster of the target object. The multimedia data in the target multimedia cluster obtained by the application is more consistent with the interest points of the target object, thereby improving the judgment accuracy of the interest points of the target object.

Description

Technical Field

[0001] This application belongs to the field of Internet and computer technology, and specifically relates to a data processing method, apparatus, computer-readable medium, and electronic device. Background Technology

[0002] Currently, when determining an audience's interests, the system first obtains access data for the pushed information, such as page views and clicks. Then, it simply checks if the access data meets certain criteria, such as whether the page views or clicks exceed a threshold. If the criteria are met, the pushed information is determined to correspond to the audience's interests. This method of information determination is relatively simple, but its accuracy in identifying audience interests is low and needs further improvement.

[0003] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] The purpose of this application is to provide a data processing method, apparatus, computer-readable medium, and electronic device to solve the problems of simple and low-precision object interest point determination methods in related technologies.

[0005] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0006] According to one aspect of the embodiments of this application, a data processing method is provided, including:

[0007] Obtain the original dataset of the target object on the multimedia platform. The original dataset is a collection of multimedia data, which represents the content data accessed by the target object on the multimedia platform.

[0008] The multimedia data in the original dataset are clustered based on multiple preset cluster numbers to obtain clustering results corresponding to each preset cluster number.

[0009] The number of target clusters is determined based on the clustering results corresponding to the number of preset clusters, and multiple multimedia clusters are obtained based on the clustering results corresponding to the number of target clusters. The multimedia clusters represent a set of multimedia data with similar content features.

[0010] The target multimedia clusters of the target object are obtained by filtering the multiple multimedia clusters based on the degree of aggregation of multimedia data corresponding to each multimedia cluster.

[0011] According to one aspect of the embodiments of this application, a data processing apparatus is provided, comprising:

[0012] The raw data acquisition module is used to acquire the raw dataset of the target object on the multimedia platform. The raw dataset is a collection of multimedia data, which represents the content data accessed by the target object on the multimedia platform.

[0013] The clustering module is used to perform clustering processing on the multimedia data in the original dataset based on multiple preset cluster numbers, and obtain the clustering results corresponding to each preset cluster number.

[0014] The multimedia cluster determination module is used to determine the number of target clusters based on the clustering results corresponding to the number of preset clusters, and to obtain multiple multimedia clusters based on the clustering results corresponding to the number of target clusters. The multimedia clusters represent a set of multimedia data with similar content features.

[0015] The filtering module is used to filter the multiple multimedia clusters according to the degree of aggregation of multimedia data corresponding to each multimedia cluster, so as to obtain the target multimedia cluster of the target object.

[0016] In one embodiment of this application, the clustering module includes:

[0017] The feature extraction unit is used to extract features from the multimedia data in the original dataset to obtain multiple multimedia content features;

[0018] The clustering unit is used to perform clustering processing on the multimedia data in the original dataset based on multiple preset cluster numbers and according to the multiple multimedia content features, so as to obtain the clustering results corresponding to each preset cluster number.

[0019] In one embodiment of this application, the clustering result corresponding to the preset number of clusters includes multiple feature clusters, and each feature cluster includes multiple multimedia content features; the clustering unit is further used for:

[0020] The clustering parameters corresponding to the preset number of clusters are determined based on the distance from each multimedia content feature to the cluster center of the feature cluster in which the multimedia content feature is located;

[0021] A clustering evaluation curve is generated based on the number of each preset cluster and the clustering parameters corresponding to each preset cluster, and the number of preset clusters corresponding to the inflection point in the clustering evaluation curve is taken as the target number of clusters.

[0022] In one embodiment of this application, the feature extraction unit includes:

[0023] A text information acquisition unit is used to acquire text description information of the multimedia data;

[0024] The word segmentation unit is used to segment the text description information to obtain multiple words corresponding to the text description information.

[0025] The vector mapping unit is used to perform vector mapping processing on each word corresponding to the text description information to obtain multiple word vectors corresponding to the description information of the multimedia information.

[0026] The vector fusion unit is used to perform vector fusion on multiple word vectors corresponding to the text description information to obtain the multimedia content features corresponding to the multimedia data.

[0027] In one embodiment of this application, the word vector is a dense word vector; the vector mapping unit is specifically used for:

[0028] The vocabulary is mapped to a low-dimensional vector to obtain sparse word vectors.

[0029] High-dimensional vector mapping is performed on the sparse word vectors of the vocabulary to obtain the dense word vectors of the vocabulary.

[0030] In one embodiment of this application, the screening module includes:

[0031] The distance determination unit is used to determine the distance from the multimedia content feature corresponding to the multimedia data to the center of the multimedia cluster, wherein the multimedia content feature is obtained by feature extraction of the multimedia data.

[0032] The clustering degree determination unit is used to determine the multimedia data clustering degree of the multimedia cluster based on the sum of the squares of the distances from all multimedia content features to the cluster center;

[0033] The first filtering unit is used to select the multimedia cluster as the target multimedia cluster of the target object when the aggregation degree of the multimedia data is greater than a first preset threshold.

[0034] In one embodiment of this application, the screening module further includes:

[0035] The target object behavior parameter acquisition unit is used to mark multimedia clusters with a multimedia data aggregation degree greater than a first preset threshold as the first multimedia cluster, and to acquire the target object behavior parameters of the multimedia data in the first multimedia cluster. The target object behavior parameters are generated based on the target object accessing the multimedia data.

[0036] The first evaluation parameter determination unit is used to perform statistical processing on the target object behavior parameters of all multimedia data in the first multimedia cluster to obtain the first evaluation parameters of the first multimedia cluster based on the target object.

[0037] The second filtering unit is used to select the first multimedia cluster as the target multimedia cluster of the target object when the first evaluation parameter meets the first preset condition.

[0038] In one embodiment of this application, the first evaluation parameter meeting the first preset condition includes:

[0039] Obtain the non-target object behavior parameters of multimedia data in the first multimedia cluster, and perform statistical processing on the non-target object behavior parameters of all multimedia data in the first multimedia cluster to obtain the second evaluation parameters of the first multimedia cluster based on non-target objects; wherein, the non-target object is all accessed objects of the first multimedia cluster except for the target object, and the non-target object behavior parameters are generated based on the access of the multimedia data by the non-target object;

[0040] When the ratio of the first evaluation parameter to the second evaluation parameter is greater than the second preset threshold, it is determined that the first evaluation parameter meets the first preset condition.

[0041] In one embodiment of this application, the screening module further includes:

[0042] The second multimedia cluster determination unit is used to mark the first multimedia cluster as the second multimedia cluster when the ratio of the first evaluation parameter to the second evaluation parameter is greater than the second preset threshold.

[0043] The behavior parameter acquisition unit is used to acquire a first object behavior parameter generated based on the access of a target object and a second object behavior parameter generated based on the access of a non-target object within a preset time period for multimedia data in the second multimedia cluster; the generation time of the first object behavior parameter and the second object behavior parameter is after the generation time of the target object behavior parameter.

[0044] The evaluation parameter determination unit is used to determine a third evaluation parameter of the second multimedia cluster based on the target object within a preset time period based on the first object behavior parameter corresponding to the second multimedia cluster, and to determine a fourth evaluation parameter of the second multimedia cluster based on the non-target object within a preset time period based on the second object behavior parameter corresponding to the second multimedia cluster.

[0045] The third screening unit is used to select the second multimedia cluster as the target multimedia cluster of the target object when the third evaluation parameter and the fourth evaluation parameter meet the second preset conditions.

[0046] In one embodiment of this application, the preset duration includes multiple time periods, the third evaluation parameter of the second multimedia cluster includes the third evaluation parameter of the second multimedia cluster in each time period, and the fourth evaluation parameter of the second multimedia cluster includes the fourth evaluation parameter of the second multimedia cluster in each time period; the third filtering unit is specifically used for:

[0047] Determine the ratio of the third evaluation parameter to the fourth evaluation parameter for each time period in the second multimedia cluster;

[0048] When the number of time periods with a ratio greater than the second preset threshold is greater than a preset number, the second multimedia cluster is taken as the target multimedia cluster of the target object.

[0049] In one embodiment of this application, the apparatus further includes:

[0050] The data filtering module is used to filter the multimedia data in the original dataset to remove dirty data from the original dataset and obtain the target dataset.

[0051] Accordingly, the clustering module is specifically used for:

[0052] Clustering is performed on the multimedia data in the target dataset.

[0053] In one embodiment of this application, the data filtering module is specifically used for:

[0054] Obtain at least one target object behavior parameter of the multimedia data, the target object behavior parameter being generated based on the target object's access to the multimedia data;

[0055] When the behavior parameters of at least one target object meet the third preset condition, the multimedia data will be used as the multimedia data in the target dataset.

[0056] When the behavior parameters of at least one target object do not meet the third preset condition, the multimedia data is removed as dirty data.

[0057] According to one aspect of the embodiments of this application, a computer-readable medium is provided, on which a computer program is stored, which, when executed by a processor, implements the data processing method as described in the above technical solutions.

[0058] According to one aspect of the embodiments of this application, an electronic device is provided, the electronic device comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform a data processing method as described above by executing the executable instructions.

[0059] According to one aspect of the embodiments of this application, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the data processing method as described in the above technical solutions.

[0060] In the technical solution provided in this application, multimedia clusters are obtained by clustering the original dataset, and target multimedia clusters are obtained by filtering the multimedia clusters according to the degree of aggregation of multimedia data. This makes the multimedia data in the target multimedia clusters more closely match the interest points of the target object, thereby improving the accuracy of the judgment of the interest points of the target object. Furthermore, the technical solution of this application can quickly and accurately construct the interest point dataset of the target object, providing data support for subsequent model development, reducing the difficulty of model development, and improving the efficiency of model development.

[0061] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0062] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0063] Figure 1 An exemplary system architecture block diagram illustrating the application of the technical solution of this application is shown schematically.

[0064] Figure 2 A flowchart illustrating a data processing method provided in one embodiment of this application is shown schematically.

[0065] Figure 3 A flowchart illustrating a clustering process provided in one embodiment of this application is shown schematically.

[0066] Figure 4 A flowchart illustrating feature extraction of multimedia data according to an embodiment of this application is shown schematically.

[0067] Figure 5 A schematic diagram of a clustering evaluation curve provided in one embodiment of this application is shown.

[0068] Figure 6 A flowchart illustrating a data processing method provided in one embodiment of this application is shown schematically.

[0069] Figure 7 A schematic block diagram of the data processing apparatus provided in an embodiment of this application is shown.

[0070] Figure 8 A schematic diagram of a computer system architecture suitable for implementing the embodiments of this application is shown. Detailed Implementation

[0071] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.

[0072] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0073] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0074] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0075] It is understood that in the specific implementation of this application, data related to object information is involved. When the above embodiments of this application are applied to specific products or technologies, permission or consent from the object is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0076] Figure 1 An exemplary system architecture block diagram illustrating the application of the technical solution of this application is shown schematically.

[0077] like Figure 1As shown, the system architecture 100 may include terminal devices 110, a network 120, and a server 130. Terminal devices 110 may include smartphones, tablets, laptops, smart voice interaction devices, smart home appliances, in-vehicle terminals, etc. Server 130 may be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. Network 120 may be a communication medium of various connection types capable of providing a communication link between terminal devices 110 and server 130, such as a wired communication link or a wireless communication link.

[0078] Depending on the implementation requirements, the system architecture in this application embodiment can have any number of terminal devices, networks, and servers. For example, server 130 can be a server group composed of multiple server devices. In addition, the technical solutions provided in this application embodiment can be applied to terminal device 110, or to server 130, or can be implemented jointly by terminal device 110 and server 130. This application does not impose any special limitations on this.

[0079] For example, terminal device 110 collects the original dataset of the target object on a multimedia platform. This original dataset is a collection of multimedia data, which represents the content data accessed by the target object on the multimedia platform. Terminal device 110 sends the original dataset to server 130. Server 130 performs clustering processing on the multimedia data in the original dataset based on multiple preset cluster numbers, obtaining clustering results corresponding to each preset cluster number. Based on the clustering results corresponding to each preset cluster number, server 130 determines the target cluster number and obtains multiple multimedia clusters based on the clustering results corresponding to the target cluster number. Each multimedia cluster represents a set of multimedia data with similar content features. Finally, server 130 filters the multiple multimedia clusters based on the degree of aggregation of multimedia data corresponding to each multimedia cluster, obtaining the target multimedia cluster for the target object. The target multimedia cluster represents the set of multimedia data that the object is interested in, that is, the set of multimedia data corresponding to the object's points of interest.

[0080] In one embodiment of this application, the data processing method provided in this embodiment is executed by server 130. Server 130 is used to receive the original dataset uploaded by terminal device 110, and perform clustering and filtering processing based on the original dataset to obtain target multimedia clusters. Accordingly, the data processing device is generally located in server 130. However, it is readily understood by those skilled in the art that the data processing method provided in this embodiment can also be executed by terminal device 110, and accordingly, the data processing device can also be located in terminal device 110. This exemplary embodiment does not impose any special limitations on this. For example, in one exemplary embodiment, terminal device 110 is used to collect the original dataset and perform clustering and filtering processing based on the original dataset to obtain target multimedia clusters.

[0081] The data processing method provided in this application will be described in detail below with reference to specific implementation methods.

[0082] Figure 2 The flowchart illustrating one embodiment of the data processing method provided in this application is shown schematically. This method can be executed by a terminal device, such as... Figure 1 Device 110 shown can also be executed by a server, such as... Figure 1 Server 130 is shown. (As shown) Figure 2 As shown, the data processing method provided in this application embodiment includes steps 210 to 240, as detailed below:

[0083] Step 210: Obtain the original dataset of the target object on the multimedia platform. The original dataset is a collection of multimedia data, which represents the content data accessed by the target object on the multimedia platform.

[0084] Specifically, a multimedia platform is a content sharing platform that provides multimedia data. Multimedia data refers to various content data that can be disseminated through the network, such as videos, articles, and music. Users can access multimedia content and interact with it through the multimedia platform, such as clicking on videos, watching videos, liking videos, and forwarding videos.

[0085] The multimedia data in the original dataset refers to the multimedia data corresponding to the content data accessed by the target object in the multimedia platform. The original dataset is a collection of multimedia data, that is, the original dataset includes multiple multimedia data, for example, the original dataset includes video data 1, video data 2, video data 3, etc.

[0086] In the embodiments of this application, the target object refers to the object whose point of interest needs to be determined. The target object can be a single object or a collection of objects of a certain type, such as female objects, male objects, objects aged between 20 and 30 years old, etc.

[0087] In one embodiment of this application, the multimedia data in the original dataset corresponds to the multimedia data accessed by a target object on a multimedia platform within a set time period, such as multimedia data from January 1, 2021 to March 31, 2021. Therefore, when acquiring multimedia data, it is necessary to filter the multimedia data in the database based on the time period in which the data is acquired and the target object. For example, firstly, multimedia data is initially acquired from the database based on the time period in which the data is acquired, and then the initially acquired multimedia data is filtered for target objects to obtain the original dataset. For example, if the target object is a female object, then multimedia data whose access objects include female objects is filtered from the initially acquired multimedia data to constitute the original dataset.

[0088] In one embodiment of this application, after obtaining the original dataset, the original dataset can be filtered to remove dirty data and obtain a cleaner target dataset. Dirty data includes data whose parameters do not meet the conditions, abnormal data, etc.

[0089] In one embodiment of this application, the process of filtering the original dataset specifically includes: obtaining at least one target object behavior parameter of the multimedia data, the target object behavior parameter being generated based on the target object accessing the multimedia data; when at least one target object behavior parameter meets a third preset condition, the multimedia data is used as multimedia data in the target dataset; when at least one target object behavior parameter does not meet the third preset condition, the multimedia data is removed as dirty data.

[0090] The target object behavior parameters of multimedia data are data generated based on the target object's access to the multimedia data. For example, the target object behavior parameters can be the number of objects accessing the data, the number of clicks, the browsing time, and the click-through rate (CTR). The click-through rate of multimedia data refers to the click-through rate of the multimedia data, which is the ratio between the actual number of clicks and the actual number of impressions (or exposures) of the multimedia data.

[0091] Multimedia data is filtered based on at least one target object behavior parameter. If at least one target object behavior parameter of the multimedia data does not meet a third preset condition, the multimedia data is removed as dirty data. For example, click volume and click pass rate can be selected to filter multimedia data. If the click volume of the multimedia data exceeds a click threshold and the click pass rate of the multimedia data exceeds a pass rate threshold, the multimedia data is included in the target dataset; otherwise, if the click volume of the multimedia data is lower than the click threshold or the click pass rate of the multimedia data is lower than the pass rate threshold, the multimedia data is removed as dirty data. Exposure volume or other parameters of the multimedia data can also be used for filtering, and this application embodiment does not impose any limitations.

[0092] Step 220: Perform clustering processing on the multimedia data in the original dataset based on multiple preset cluster numbers to obtain the clustering results corresponding to each preset cluster number.

[0093] Specifically, clustering refers to classifying multimedia data in an original dataset. The clustering result divides the multimedia data in the original dataset into multiple categories, each containing multiple multimedia data that are similar in a certain aspect (also referred to as multimedia clusters). The number of categories in the clustering result is the number of clusters set during clustering. In the embodiments of this application, clustering is performed according to multiple preset numbers of clusters. For each preset number of clusters, the clustering result is that the multimedia data is divided into the preset number of clusters.

[0094] In one embodiment of this application, such as Figure 3 As shown, the steps for clustering the multimedia data in the original dataset include steps 310 to 320, as detailed below:

[0095] Step 310: Extract features from the multimedia data in the original dataset to obtain multiple multimedia content features.

[0096] Specifically, feature extraction of multimedia data involves extracting multimedia content features that can represent the multimedia data. For example, if the multimedia data is video data, a convolutional neural network can be used to extract features from the video data to obtain video content features that can represent the video data.

[0097] In one embodiment of this application, such as Figure 4 As shown, the steps for feature extraction from multimedia data include steps 410 to 440, as detailed below:

[0098] Step 410: Obtain the text description information of the multimedia data.

[0099] Specifically, the text description information of multimedia data refers to the descriptive information of multimedia data in text form. For example, the text description information can be the title of the multimedia data, a brief introduction of the multimedia data content, etc.

[0100] Step 420: Perform word segmentation on the text description information to obtain multiple words corresponding to the text description information.

[0101] Specifically, word segmentation is the process of dividing a text sequence into individual words. Word segmentation includes Chinese word segmentation and English word segmentation. English word segmentation is relatively simple, generally achieved by recognizing punctuation marks and spaces. Chinese word segmentation has various methods, including: string matching-based segmentation methods, understanding-based segmentation methods, and statistical segmentation methods.

[0102] String matching-based word segmentation methods match text descriptions against entries in a pre-defined, sufficiently large machine dictionary according to certain rules. If a matching string is found in the machine dictionary, a successful match is achieved, and a word in the text description is identified. Matching rules include forward matching, backward matching, minimum segmentation, simple word segmentation, and methods combining word segmentation and part-of-speech tagging. String matching-based word segmentation methods are relatively simple and fast.

[0103] Understanding-based word segmentation methods perform syntactic and semantic analysis simultaneously with word segmentation, utilizing syntactic and semantic information to address ambiguities arising during the segmentation process. Due to the complexity of Chinese syntax and semantics, this type of word segmentation method is also quite complex.

[0104] Statistical word segmentation is a method that utilizes machine learning models for word segmentation. This method first trains the statistical machine learning model using a large number of segmented text samples. The segmented text samples are then used to segment the descriptive information within the text. This approach, which uses machine learning models to segment descriptive information, achieves higher accuracy.

[0105] Step 430: Perform vector mapping processing on each word corresponding to the text description information to obtain multiple word vectors corresponding to the multimedia information description information.

[0106] Specifically, vector mapping of words involves mapping words represented in the form of Chinese characters or English characters to word vectors represented in the form of real numbers according to certain rules. For example, if a Chinese character is mapped to the number 1, then the number 1 is the word vector of that Chinese character.

[0107] In one embodiment of this application, the process of vector mapping of words can be as follows: first, perform low-dimensional vector mapping on the words to obtain sparse word vectors of the words; then, perform high-dimensional vector mapping on the sparse word vectors of the words to obtain dense word vectors of the words, which are the word vectors of the words.

[0108] Specifically, firstly, one-hot encoding is used to transform the vocabulary into sparse word vectors. One-hot encoding, also known as single-hot encoding or one-bit effective encoding, represents each word or character with either 0 or 1, thus resulting in sparse word vectors. Then, the Word2Vec model is used to transform the sparse word vectors into multi-dimensional dense word vectors, which are the final word vectors of the vocabulary.

[0109] Step 440: Perform vector fusion on multiple word vectors corresponding to the text description information to obtain the multimedia content features corresponding to the multimedia data.

[0110] Specifically, vector fusion of multiple word vectors means combining multiple word vectors into one vector. The fused vector is the vector that can represent the text description information, which is the multimedia content feature corresponding to the multimedia data.

[0111] In one embodiment of this application, vector fusion can be achieved by weighted averaging of multiple word vectors, that is, multiplying each word vector by its weight and then summing the results to obtain a fused vector. The weight of each word vector can be determined based on the importance of the corresponding word in the text description information; the higher the importance, the greater the weight.

[0112] Continue to refer to Figure 3 Step 320: Based on multiple preset cluster numbers, cluster the multimedia data in the original dataset according to multiple multimedia content features to obtain the clustering results corresponding to each preset cluster number.

[0113] Specifically, after extracting the multimedia content features from the multimedia data, the multimedia content features are clustered based on a preset number of clusters. The clustering results corresponding to this preset number of clusters include multiple feature clusters. Then, the multimedia content features in each feature cluster are mapped to the corresponding multimedia data, thus realizing the classification of the multimedia data.

[0114] Continue to refer to Figure 2 Step 230: Determine the number of target clusters based on the clustering results corresponding to the number of preset clusters, and obtain multiple multimedia clusters based on the clustering results corresponding to the number of target clusters. Multimedia clusters represent a collection of multimedia data with similar content features.

[0115] Specifically, a multimedia cluster represents a collection of multimedia data with similar content characteristics. In other words, the multimedia data within a multimedia cluster belong to the same type, and this type of multimedia data is equivalent to a point of interest (POI). This POI may be a POI of the target object, so the multimedia cluster is also equivalent to a set of candidate POI data for the target object. For example, if a multimedia cluster contains only video data about beauty and makeup, then "beauty and makeup" may be a POI of the target object.

[0116] When starting clustering, it is uncertain how many clusters are needed to achieve the best clustering effect. Therefore, this application sets multiple preset cluster numbers, performs clustering once based on each preset cluster number, analyzes the quality of the clustering results corresponding to each preset cluster number, and selects one preset cluster number as the target cluster number. The clustering result corresponding to the target cluster number is then the required multiple multimedia clusters.

[0117] In one embodiment of this application, the actual data processed in the clustering process is multimedia content features. Therefore, the direct clustering result corresponding to a preset number of clusters includes multiple feature clusters, and each feature cluster includes multiple multimedia content features. The process of determining the target cluster data is as follows: determine the clustering parameters corresponding to the preset number of clusters based on the distance from each multimedia content feature to the cluster center of the feature cluster in which the multimedia content feature is located; generate a clustering evaluation curve based on each preset number of clusters and the clustering parameters corresponding to each preset number of clusters; and take the preset number of clusters corresponding to the inflection point in the clustering evaluation curve as the target number of clusters.

[0118] Specifically, clustering based on a predetermined number of clusters yields multiple feature clusters. Each feature cluster includes multiple multimedia content features and a defined cluster center. Within a feature cluster, the distance from each multimedia content feature to the cluster center indicates the clustering effectiveness. Closer distances indicate better clustering, while wider distances indicate poorer clustering. The sum of the squares of the distances from each multimedia content feature to the cluster center is denoted as the cluster parameter. This cluster parameter evaluates the clustering effectiveness and reflects the dispersion of multimedia content features (or multimedia data) within the cluster (i.e., their distance from the cluster center). The sum of the cluster parameters for all feature clusters corresponding to a predetermined number of clusters is denoted as the clustering parameter for that predetermined number of clusters, used to evaluate the clustering effectiveness for that predetermined number of clusters.

[0119] Optionally, the cluster parameters of the feature clusters can also be other feature data that can reflect the clustering effect, such as the sum of the variances between each multimedia content feature and the cluster center, the silhouette coefficient of the feature cluster, etc.

[0120] Different preset numbers of clusters can produce different clustering results; that is, different preset numbers of clusters correspond to different clustering parameters. After performing clustering based on each preset number of clusters, the clustering parameters corresponding to each preset number of clusters can be obtained. By plotting the preset number of clusters on the x-axis and the evaluation parameters of the preset number of clusters on the y-axis, a clustering evaluation curve can be drawn, such as... Figure 5 As shown. Once the clustering effect stabilizes, the evaluation parameters will not change significantly. Therefore, the preset number of clusters corresponding to the inflection point in the clustering evaluation curve can be used as the target number of clusters. After a certain point, even if the value of the preset number of clusters is increased, the change in clustering parameters is very small. Such a point is called the inflection point. That is, the difference between the clustering parameters corresponding to the (k-1)th preset number of clusters and the clustering parameters corresponding to the kth preset number of clusters is greater than a preset threshold, while the difference between the clustering parameters corresponding to the kth preset number of clusters and the clustering parameters corresponding to the (k+1)th preset number of clusters is less than a preset threshold. Then, the kth preset number of clusters is the preset number of clusters corresponding to the inflection point. Figure 5 The preset number of clusters is K points.

[0121] In one embodiment of this application, clustering can be performed in ascending order of the number of preset clusters. When the difference between the clustering parameter corresponding to the (k-1)th preset cluster number and the clustering parameter corresponding to the kth preset cluster number is greater than a preset threshold, and the difference between the clustering parameter corresponding to the kth preset cluster number and the clustering parameter corresponding to the (k+1)th preset cluster number is less than a preset threshold, then the kth preset cluster number is the target cluster number (k starts from 1). At this point, clustering can be stopped, and clustering processing is no longer required for the (k+2)th and subsequent preset cluster numbers.

[0122] Generally speaking, the larger the preset number of clusters, the smaller the clustering parameter corresponding to the preset number of clusters. Therefore, in one embodiment of this application, the preset number of clusters when the clustering parameter is less than the set value can be used as the target number of clusters.

[0123] Continue to refer to Figure 2 Step 240: Based on the degree of aggregation of multimedia data corresponding to each multimedia cluster, filter multiple multimedia clusters to obtain the target multimedia cluster of the target object.

[0124] Specifically, the degree of multimedia data clustering reflects the clustering effect of multimedia clusters. A higher degree of clustering indicates a better clustering effect, while a lower degree indicates a worse effect. If a multimedia cluster has a poor clustering effect, it means that the multimedia data in that cluster does not represent the target audience's interests. Therefore, multiple multimedia clusters are filtered based on their clustering degree, and the clusters with good clustering effects are selected as target multimedia clusters. These target multimedia clusters more accurately reflect the target audience's interests.

[0125] In the technical solution provided in this application, multimedia clusters are obtained by clustering the original dataset, and target multimedia clusters are obtained by filtering the multimedia clusters according to the degree of aggregation of multimedia data. This makes the multimedia data in the target multimedia clusters more closely match the interest points of the target object, thereby improving the accuracy of the judgment of the interest points of the target object. Furthermore, the technical solution of this application can quickly and accurately construct the interest point dataset of the target object, providing data support for subsequent model development, reducing the difficulty of model development, and improving the efficiency of model development.

[0126] In one embodiment of this application, within the feature cluster corresponding to the multimedia cluster, the sum of the squares of the distances from all multimedia content features to the cluster center (i.e., the cluster parameter of the multimedia cluster) reflects the dispersion of the multimedia data. Dispersion and clustering are two opposite concepts. Therefore, the negative of the sum of the squares of the distances from all multimedia content features to the cluster center (i.e., the negative of the cluster parameter of the multimedia cluster) can be used as the degree of multimedia data clustering. Thus, the smaller the cluster parameter of the multimedia cluster, the greater the degree of multimedia data clustering; conversely, the larger the cluster parameter, the smaller the degree of multimedia data clustering. When the degree of multimedia data clustering exceeds a first preset threshold, the corresponding multimedia cluster is the target multimedia cluster of the target object.

[0127] In one embodiment of this application, the target multimedia cluster can also be determined directly by the cluster parameter of the multimedia cluster. That is, when the cluster parameter of the multimedia cluster is less than a certain threshold, the corresponding multimedia cluster is the target multimedia cluster of the target object.

[0128] In one embodiment of this application, after filtering based on the degree of multimedia data aggregation, the method may further include: marking multimedia clusters with a degree of multimedia data aggregation greater than a first preset threshold as first multimedia clusters; obtaining target object behavior parameters of multimedia data in the first multimedia cluster, wherein the target object behavior parameters are generated based on the target object accessing multimedia data; performing statistical processing on the target object behavior parameters of all multimedia data in the first multimedia cluster to obtain a first evaluation parameter of the first multimedia cluster based on the target object; and when the first evaluation parameter meets the first preset condition, using the first multimedia cluster as the target multimedia cluster of the target object.

[0129] Specifically, multimedia clusters with a data aggregation degree greater than a first preset threshold are selected and marked as the first multimedia cluster. Then, the selected first multimedia clusters are further filtered according to the target object behavior parameters of the multimedia data in the first multimedia cluster to obtain the target multimedia clusters that meet the conditions.

[0130] Statistical processing is performed on the target object behavior parameters of all multimedia data in the first multimedia cluster to obtain the first evaluation parameter of the first multimedia cluster based on the target object. The target object behavior parameters for statistical processing can be any data that reflects the behavior of the target object towards the multimedia data, such as the number of clicks, click pass rate, number of accessed objects, etc. of the multimedia data. The first evaluation parameter can be the average, weighted average, standard deviation, median, etc. of the target object behavior parameters of all multimedia data. For example, the first multimedia cluster includes 5 multimedia data, and the target object behavior parameters corresponding to these 5 multimedia data are C. 11 C 12 C 13 C 14 C 15 Then the first multimedia cluster is based on the first evaluation parameter S1 of the target object = (C 11 +C 12 +C 13 +C 14 +C 15 ) / 5.

[0131] In all access objects of multimedia data, excluding the target object, the remaining objects are non-target objects. The behavioral data generated by non-target objects are denoted as non-target object behavioral parameters. Since the goal is to obtain the target multimedia cluster of the target object, it is necessary to reflect the difference between the target object and non-target objects. Therefore, when determining whether the first evaluation parameter meets the conditions, the non-target object behavioral parameters of the multimedia data in the first multimedia cluster are first obtained, and statistical processing is performed to obtain the second evaluation parameter of the first multimedia cluster based on the non-target object. Then, the target multimedia cluster is determined based on the magnitude of the first evaluation parameter and the second evaluation parameter. For example, the first multimedia cluster includes 5 multimedia data, and the target object behavioral parameters corresponding to these 5 multimedia data are C... 21 C 22 C 23 C 24 C 25 Then the first multimedia cluster is based on the second evaluation parameter S2 of the non-target object, which is S2 = (C 21 +C 22 +C 23 +C 24 +C 25 ) / 5.

[0132] When the first evaluation parameter is greater than the second evaluation parameter, or the ratio of the first evaluation parameter to the second evaluation parameter is greater than the second preset threshold, it indicates that the multimedia data in the first multimedia cluster is more easily accessed by the target object, meaning that the multimedia data in the first multimedia cluster better matches the target object's interests. Therefore, the first multimedia cluster can be used as the target multimedia cluster for the target object. Conversely, if the first evaluation parameter is less than the second evaluation parameter, or the ratio of the first evaluation parameter to the second evaluation parameter is less than the second preset threshold, it indicates that the multimedia data in the first multimedia cluster better matches the interests of non-target objects and is not suitable as the target multimedia cluster for the target object.

[0133] In one embodiment of this application, considering a situation where multimedia data matches both the interests of the target object and those of non-target objects, the first multimedia cluster with a first evaluation parameter greater than a certain threshold can be directly used as the target multimedia cluster of the target object. Alternatively, the evaluation parameters of the first multimedia cluster based on all accessed objects can be calculated, and then the first multimedia cluster whose ratio of the first evaluation parameter to the evaluation parameters based on all accessed objects is less than a second preset threshold can be used as the target multimedia cluster of the target object.

[0134] In one embodiment of this application, the filtering based on the first evaluation parameter may further include: when the ratio of the first evaluation parameter to the second evaluation parameter is greater than a second preset threshold, marking the first multimedia cluster as the second multimedia cluster; obtaining a first object behavior parameter generated by access to a target object and a second object behavior parameter generated by access to a non-target object within a preset time period for multimedia data in the second multimedia cluster; determining a third evaluation parameter of the second multimedia cluster based on the target object within a preset time period based on the first object behavior parameter corresponding to the second multimedia cluster, and determining a fourth evaluation parameter of the second multimedia cluster based on a non-target object within a preset time period based on the second object behavior parameter corresponding to the second multimedia cluster; when the third evaluation parameter and the fourth evaluation parameter meet the second preset condition, designating the second multimedia cluster as the target multimedia cluster of the target object.

[0135] Specifically, a first multimedia cluster whose ratio of the first evaluation parameter to the second evaluation parameter is greater than a second preset threshold is marked as a second multimedia cluster. Then, the object behavior parameters generated by the multimedia data in the second multimedia cluster within a preset time period are obtained to determine whether the second multimedia cluster can be used as a target multimedia cluster. This is equivalent to determining whether the object behavior parameters generated by the multimedia data in the second multimedia cluster within a preset time period are consistent with the target object behavior parameters of the multimedia data.

[0136] The multimedia data in the second multimedia cluster is obtained by clustering and filtering the original dataset. The multimedia data in the original dataset corresponds to the target object within a set time period. Therefore, the target object behavior parameters corresponding to the multimedia data are naturally also the object behavior parameters within the set time period. In this embodiment, the preset duration is after the set time period corresponding to the original dataset; that is, the generation time of the first and second object behavior parameters is after the generation time of the target object behavior parameters. Furthermore, the preset duration can also be specified to be shorter than the set time period. For example, if the set time period is from January 1, 2021 to March 31, 2021, and the preset duration is from April 1, 2021 to April 7, 2021, then the target object behavior parameter is the object behavior data of the multimedia data based on the target object within the set time period from January 1, 2021 to March 31, 2021. The first object behavior parameter is the object behavior data based on the target object within the preset duration from April 1, 2021 to April 7, 2021. The second object behavior parameter is the object behavior data based on non-target objects within the preset duration from April 1, 2021 to April 7, 2021.

[0137] The behavioral parameters of the first object within a set time period are statistically analyzed to obtain a third evaluation parameter based on the target object; the behavioral parameters of the second object within the set time period are statistically analyzed to obtain a fourth evaluation parameter based on the non-target object. If the ratio of the third evaluation parameter to the fourth evaluation parameter is greater than a second preset threshold, it indicates that within the set time period, the multimedia data in the second multimedia cluster still better matches the target object's interests. Therefore, it can be determined that the behavior of the target object and the non-target object within the set time period is consistent with the behavior of the target object and the non-target object within the set time period. Thus, it can be considered that the third evaluation parameter and the fourth evaluation parameter meet the second preset condition, and the second multimedia cluster can be used as the target multimedia cluster of the target object. If the ratio of the third evaluation parameter to the fourth evaluation parameter is less than the second preset threshold, it is considered that within the set time period, the target object's interests have changed, and the second multimedia cluster is not suitable as the target multimedia cluster of the target object.

[0138] In one embodiment of this application, a preset duration can be divided into multiple time periods. For each time period, a third evaluation parameter can be calculated based on the first object behavior parameter of that time period, and a fourth evaluation parameter can be calculated based on the second object behavior parameter of that time period. This allows for the determination of the ratio of the third evaluation parameter to the fourth evaluation parameter within each time period. When the number of time periods with a ratio greater than a second preset threshold exceeds a preset number, the behavior of the target object within the set duration is considered consistent with its behavior within the set time period, thus allowing the corresponding second multimedia category to be designated as the target multimedia category of the target object. For example, if the preset duration is from April 1, 2021 to April 7, 2021, it can be divided into 7 time periods, each lasting one day. The third evaluation parameter is calculated based on the first object behavior parameter for each day, and the fourth evaluation parameter is calculated based on the second object behavior parameter for each day, thereby determining whether the ratio of the third evaluation parameter to the fourth evaluation parameter for each day exceeds a second preset threshold. If the number of days with a ratio greater than the second preset threshold reaches a preset number (e.g., 5 days), the corresponding second multimedia category can be designated as the target multimedia category of the target object.

[0139] In one embodiment of this application, after obtaining the target multimedia cluster, the interest points of the target object can be determined based on the multimedia data in the target multimedia cluster. For example, if the multimedia data in the target multimedia cluster are all videos about beauty, then one interest point of the target object can be determined to be beauty. In some cases, a target multimedia cluster can also correspond to multiple interest point tags. Based on multiple target multimedia clusters, an interest point dataset of the target object can be constructed, and then this interest point dataset can be used to recommend information to the target object. For example, an information recommendation model can be trained using the interest point dataset, and then the trained information recommendation model can predict the information that the target object is interested in and recommend that information to the target object.

[0140] The following example uses multimedia data (video data) and the target object (female object) to illustrate the specific process of the data processing method in a specific embodiment of this application.

[0141] like Figure 6 As shown, the data processing method of this application includes:

[0142] S601. Obtain the original dataset src_fdata of female objects within a specified time period (January 1, 2021 to March 31, 2021) from the dataset. The original dataset src_fdata includes multiple video data.

[0143] S602. Filter the video data in the original dataset src_fdata according to the video exposure threshold e_thresh, the object click threshold c_thresh, and the video CTR threshold ctr_thresh. Remove video data with video exposure below the video exposure threshold e_thresh, video clicks below the object click threshold c_thresh, and video CTR below the video CTR threshold ctr_thresh as dirty data. The remaining video data forms a relatively clean target dataset fdata={vt_1,vt_2,...,vt_n}, where vt_i (1≤i≤n) represents a video data, and the video data is represented in the form of a video title, that is, vt_i represents the video title.

[0144] S603. First, the video title vt_i is segmented to obtain multiple words corresponding to the video title vt_i; then, each word is mapped to a vector using the pre-trained word vector model Word2Vec to obtain the word vector corresponding to the word; finally, all word vectors are weighted and summed according to the weights {w_1,w_2,...,w_m} of each word to obtain the content feature vt_v_i of the video title vt_i, which constitutes the feature dataset fdata_v={vt_v_1,vt_v_2,...,vt_v_n}.

[0145] S604. Perform K-means clustering on the feature dataset fdata_v={vt_v_1, vt_v_2, ..., vt_v_n} to obtain K video clusters fcluster={candidate_fc_1, candidate_fc_2, ..., candidate_fc_K}, where candidate_fc_j (1≤j≤K) represents a video cluster, equivalent to the candidate interest point cluster of female objects. During the clustering process, the number of clusters is obtained using the inflection point method; details can be found in step 320 above and will not be repeated here.

[0146] S605. Filter the K video clusters fcluster={candidate_fc_1, candidate_fc_2, ...,candidate_fc_K}, including the following three aspects:

[0147] First, video clusters are filtered based on the point of interest purity threshold sq_thresh. The sum of the squares of the distances from all data points in a video cluster candidate_fc_j to the cluster center is taken as the purity of the video cluster candidate_fc_j. Video clusters candidate_fc_j with a purity sq_j higher than the point of interest purity threshold sq_thresh are taken as the first video cluster.

[0148] Second, calculate the video CTR of the first video cluster based on female subjects within a set time period (January 1, 2021 to March 31, 2021), and obtain the video CTR of the first video cluster based on male subjects within the same set time period (January 1, 2021 to March 31, 2021). When the ratio of the video CTR based on female subjects to the video CTR based on male subjects is higher than 1.3, the first video cluster is marked as the second video cluster. Alternatively, obtain the video CTR of the first video cluster based on all subjects within the set time period (January 1, 2021 to March 31, 2021), and when the ratio of the video CTR based on female subjects to the video CTR based on all subjects is higher than 1.3, the first video cluster is marked as the second video cluster.

[0149] Third, obtain the video data of the second video cluster within a set time period (April 1, 2021 to April 7, 2021) based on female objects and video CTRs based on difficult objects. If the ratio of the video CTR based on female objects to the video CTR based on difficult objects within this set time period is higher than 1.3, then this second video cluster is taken as the target video cluster, resulting in the set of female object interest points: female_instersting={fi_1, fi_2, ..., fi_m}, where fi_k (1≤k≤m) represents a target video cluster. Alternatively, calculate the ratio between the video CTR based on female objects and the video CTR based on difficult objects for each day within the set time period. If the number of days with a ratio greater than 1.3 reaches 6, then this second video cluster is taken as the target video cluster.

[0150] It should be noted that although the steps of the method in this application are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0151] The following describes an apparatus embodiment of this application, which can be used to execute the data processing method described in the above embodiments of this application. Figure 7 A schematic block diagram of the data processing apparatus provided in an embodiment of this application is shown. Figure 7 As shown, the data processing apparatus includes:

[0152] The raw data acquisition module 710 is used to acquire the raw dataset of the target object on the multimedia platform. The raw dataset is a collection of multimedia data, which represents the content data accessed by the target object on the multimedia platform.

[0153] Clustering module 720 is used to perform clustering processing on the multimedia data in the original dataset based on multiple preset cluster numbers to obtain clustering results corresponding to each preset cluster number.

[0154] The multimedia cluster determination module 730 is used to determine the number of target clusters based on the clustering results corresponding to the number of preset clusters, and to obtain multiple multimedia clusters based on the clustering results corresponding to the number of target clusters. The multimedia clusters represent a set of multimedia data with similar content features.

[0155] The filtering module 740 is used to filter the multiple multimedia clusters according to the degree of aggregation of multimedia data corresponding to each multimedia cluster, so as to obtain the target multimedia cluster of the target object.

[0156] In one embodiment of this application, the clustering module 720 includes:

[0157] The feature extraction unit is used to extract features from the multimedia data in the original dataset to obtain multiple multimedia content features;

[0158] The clustering unit is used to perform clustering processing on the multimedia data in the original dataset based on multiple preset cluster numbers and according to the multiple multimedia content features, so as to obtain the clustering results corresponding to each preset cluster number.

[0159] In one embodiment of this application, the clustering result corresponding to the preset number of clusters includes multiple feature clusters, and each feature cluster includes multiple multimedia content features; the multimedia cluster determination module 730 is used for:

[0160] The clustering parameters corresponding to the preset number of clusters are determined based on the distance from each multimedia content feature to the cluster center of the feature cluster in which the multimedia content feature is located;

[0161] A clustering evaluation curve is generated based on the number of each preset cluster and the clustering parameters corresponding to each preset cluster, and the number of preset clusters corresponding to the inflection point in the clustering evaluation curve is taken as the target number of clusters.

[0162] In one embodiment of this application, the feature extraction unit includes:

[0163] A text information acquisition unit is used to acquire text description information of the multimedia data;

[0164] The word segmentation unit is used to segment the text description information to obtain multiple words corresponding to the text description information.

[0165] The vector mapping unit is used to perform vector mapping processing on each word corresponding to the text description information to obtain multiple word vectors corresponding to the description information of the multimedia information.

[0166] The vector fusion unit is used to perform vector fusion on multiple word vectors corresponding to the text description information to obtain the multimedia content features corresponding to the multimedia data.

[0167] In one embodiment of this application, the word vector is a dense word vector; the vector mapping unit is specifically used for:

[0168] The vocabulary is mapped to a low-dimensional vector to obtain sparse word vectors.

[0169] High-dimensional vector mapping is performed on the sparse word vectors of the vocabulary to obtain the dense word vectors of the vocabulary.

[0170] In one embodiment of this application, the screening module 740 includes:

[0171] The distance determination unit is used to determine the distance from the multimedia content feature corresponding to the multimedia data to the center of the multimedia cluster, wherein the multimedia content feature is obtained by feature extraction of the multimedia data.

[0172] The clustering degree determination unit is used to determine the multimedia data clustering degree of the multimedia cluster based on the sum of the squares of the distances from all multimedia content features to the cluster center;

[0173] The first filtering unit is used to select the multimedia cluster as the target multimedia cluster of the target object when the aggregation degree of the multimedia data is greater than a first preset threshold.

[0174] In one embodiment of this application, the screening module 740 further includes:

[0175] The target object behavior parameter acquisition unit is used to mark multimedia clusters with a multimedia data aggregation degree greater than a first preset threshold as the first multimedia cluster, and to acquire the target object behavior parameters of the multimedia data in the first multimedia cluster. The target object behavior parameters are generated based on the target object accessing the multimedia data.

[0176] The first evaluation parameter determination unit is used to perform statistical processing on the target object behavior parameters of all multimedia data in the first multimedia cluster to obtain the first evaluation parameters of the first multimedia cluster based on the target object.

[0177] The second filtering unit is used to select the first multimedia cluster as the target multimedia cluster of the target object when the first evaluation parameter meets the first preset condition.

[0178] In one embodiment of this application, the first evaluation parameter meeting the first preset condition includes:

[0179] Obtain the non-target object behavior parameters of multimedia data in the first multimedia cluster, and perform statistical processing on the non-target object behavior parameters of all multimedia data in the first multimedia cluster to obtain the second evaluation parameters of the first multimedia cluster based on non-target objects; wherein, the non-target object is all accessed objects of the first multimedia cluster except for the target object, and the non-target object behavior parameters are generated based on the access of the multimedia data by the non-target object;

[0180] When the ratio of the first evaluation parameter to the second evaluation parameter is greater than the second preset threshold, it is determined that the first evaluation parameter meets the first preset condition.

[0181] In one embodiment of this application, the screening module 740 further includes:

[0182] The second multimedia cluster determination unit is used to mark the first multimedia cluster as the second multimedia cluster when the ratio of the first evaluation parameter to the second evaluation parameter is greater than the second preset threshold.

[0183] The behavior parameter acquisition unit is used to acquire a first object behavior parameter generated based on the access of a target object and a second object behavior parameter generated based on the access of a non-target object within a preset time period for multimedia data in the second multimedia cluster; the generation time of the first object behavior parameter and the second object behavior parameter is after the generation time of the target object behavior parameter.

[0184] The evaluation parameter determination unit is used to determine a third evaluation parameter of the second multimedia cluster based on the target object within a preset time period based on the first object behavior parameter corresponding to the second multimedia cluster, and to determine a fourth evaluation parameter of the second multimedia cluster based on the non-target object within a preset time period based on the second object behavior parameter corresponding to the second multimedia cluster.

[0185] The third screening unit is used to select the second multimedia cluster as the target multimedia cluster of the target object when the third evaluation parameter and the fourth evaluation parameter meet the second preset conditions.

[0186] In one embodiment of this application, the preset duration includes multiple time periods, the third evaluation parameter of the second multimedia cluster includes the third evaluation parameter of the second multimedia cluster in each time period, and the fourth evaluation parameter of the second multimedia cluster includes the fourth evaluation parameter of the second multimedia cluster in each time period; the third filtering unit is specifically used for:

[0187] Determine the ratio of the third evaluation parameter to the fourth evaluation parameter for each time period in the second multimedia cluster;

[0188] When the number of time periods with a ratio greater than the second preset threshold is greater than a preset number, the second multimedia cluster is taken as the target multimedia cluster of the target object.

[0189] In one embodiment of this application, the apparatus further includes:

[0190] The data filtering module is used to filter the multimedia data in the original dataset to remove dirty data from the original dataset and obtain the target dataset.

[0191] Accordingly, clustering module 720 is specifically used for:

[0192] Clustering is performed on the multimedia data in the target dataset.

[0193] In one embodiment of this application, the data filtering module is specifically used for:

[0194] Obtain at least one target object behavior parameter of the multimedia data, the target object behavior parameter being generated based on the target object's access to the multimedia data;

[0195] When the behavior parameters of at least one target object meet the third preset condition, the multimedia data will be used as the multimedia data in the target dataset.

[0196] When the behavior parameters of at least one target object do not meet the third preset condition, the multimedia data is removed as dirty data.

[0197] The specific details of the data processing apparatus provided in the various embodiments of this application have been described in detail in the corresponding method embodiments, and will not be repeated here.

[0198] Figure 8 A schematic block diagram of a computer system architecture for implementing an electronic device according to embodiments of the present application is shown.

[0199] It should be noted that, Figure 8 The computer system 800 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0200] like Figure 8 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 802 or programs loaded from storage section 808 into random access memory (RAM). The RAM 803 also stores various programs and data required for system operation. The CPU 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output interface 805 (I / O interface) is also connected to the bus 804.

[0201] The following components are connected to the input / output interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a local area network card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0202] Specifically, according to embodiments of this application, the processes described in the various method flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by central processing unit 801, it performs various functions defined in the system of this application.

[0203] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such transmitted data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0204] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0205] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0206] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the methods according to the embodiments of this application.

[0207] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0208] The data collection and processing plan outlined in this application must be implemented in strict accordance with the requirements of relevant national laws and regulations, obtaining the informed consent or separate consent of the data subject (or having a legal basis as stipulated by the relevant national laws and regulations), and conducting subsequent data use and processing within the scope authorized by laws and regulations and the data subject.

[0209] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A data processing method, characterized in that, include: Obtain the original dataset of the target object on the multimedia platform. The original dataset is a collection of multimedia data, which represents the content data accessed by the target object on the multimedia platform. The target object is an object whose points of interest need to be determined, and the target object is a collection of objects of a certain type. The multimedia platform is a content sharing platform that provides multimedia data. Obtain the text description information of the multimedia data; The text description information is segmented to obtain multiple words corresponding to the text description information; Vector mapping is performed on each word corresponding to the text description information to obtain multiple word vectors corresponding to the multimedia information description information. Vector fusion is performed on multiple word vectors corresponding to the text description information to obtain the multimedia content features corresponding to the multimedia data; Based on multiple preset cluster numbers, the multimedia data in the original dataset is clustered according to multiple multimedia content features to obtain clustering results corresponding to each preset cluster number. The number of target clusters is determined based on the clustering results corresponding to the number of preset clusters, and multiple multimedia clusters are obtained based on the clustering results corresponding to the number of target clusters. The multimedia clusters represent a set of multimedia data with similar content features. In the multimedia cluster, the distance from the multimedia content feature corresponding to the multimedia data to the cluster center is determined; The degree of multimedia data aggregation of the multimedia cluster is determined by the sum of the squares of the distances from all multimedia content features to the cluster center. Multimedia clusters with a multimedia data aggregation degree greater than a first preset threshold are marked as the first multimedia cluster. Target object behavior parameters of the multimedia data in the first multimedia cluster are obtained. These target object behavior parameters are generated based on the target object's access to the multimedia data. Statistical processing is performed on the target object behavior parameters of all multimedia data in the first multimedia cluster to obtain the first evaluation parameters of the first multimedia cluster based on the target object; Obtain non-target object behavior parameters of multimedia data in the first multimedia cluster, and perform statistical processing on the non-target object behavior parameters of all multimedia data in the first multimedia cluster to obtain a second evaluation parameter of the first multimedia cluster based on non-target objects; wherein, the non-target object is all access objects of the first multimedia cluster other than the target object, and the non-target object behavior parameter is generated based on the non-target object accessing the multimedia data; when the ratio of the first evaluation parameter to the second evaluation parameter is greater than a second preset threshold, determine that the first evaluation parameter meets the first preset condition, and designate the first multimedia cluster as the target multimedia cluster of the target object; or Calculate the evaluation parameters of the first multimedia cluster based on all accessed objects; and select the first multimedia cluster whose ratio of the first evaluation parameter to the evaluation parameters based on all accessed objects is less than a second preset threshold as the target multimedia cluster of the target object. The interest points of the target object are determined based on the multimedia data in the target multimedia cluster, and the interest points are interest point tags.

2. The data processing method according to claim 1, characterized in that, The clustering results corresponding to the preset number of clusters include multiple feature clusters, and each feature cluster includes multiple multimedia content features; The number of target clusters is determined based on the clustering results corresponding to the number of preset clusters, including: The clustering parameters corresponding to the preset number of clusters are determined based on the distance from each multimedia content feature to the cluster center of the feature cluster in which the multimedia content feature is located; A clustering evaluation curve is generated based on the number of each preset cluster and the clustering parameters corresponding to each preset cluster, and the number of preset clusters corresponding to the inflection point in the clustering evaluation curve is taken as the target number of clusters.

3. The data processing method according to claim 1, characterized in that, The word vectors are dense word vectors; vector mapping is performed on each word corresponding to the text description information to obtain multiple word vectors corresponding to the description information of the multimedia information, including: The vocabulary is mapped to a low-dimensional vector to obtain sparse word vectors. High-dimensional vector mapping is performed on the sparse word vectors of the vocabulary to obtain the dense word vectors of the vocabulary.

4. The data processing method according to claim 1, characterized in that, The method further includes: When the ratio of the first evaluation parameter to the second evaluation parameter is greater than the second preset threshold, the first multimedia cluster is marked as the second multimedia cluster; The system acquires a first object behavior parameter generated based on access to a target object and a second object behavior parameter generated based on access to a non-target object within a preset time period for multimedia data in the second multimedia cluster; the first object behavior parameter and the second object behavior parameter are generated after the target object behavior parameter is generated. The third evaluation parameter of the second multimedia cluster based on the target object within a preset time period is determined based on the first object behavior parameter corresponding to the second multimedia cluster, and the fourth evaluation parameter of the second multimedia cluster based on the non-target object within a preset time period is determined based on the second object behavior parameter corresponding to the second multimedia cluster. When the third evaluation parameter and the fourth evaluation parameter meet the second preset condition, the second multimedia cluster is taken as the target multimedia cluster of the target object.

5. The data processing method according to claim 4, characterized in that, The preset duration includes multiple time periods, the third evaluation parameter of the second multimedia cluster includes the third evaluation parameter of the second multimedia cluster in each time period, and the fourth evaluation parameter of the second multimedia cluster includes the fourth evaluation parameter of the second multimedia cluster in each time period. When the third evaluation parameter and the fourth evaluation parameter meet the second preset condition, the second multimedia cluster is taken as the target multimedia cluster of the target object, including: Determine the ratio of the third evaluation parameter to the fourth evaluation parameter for each time period in the second multimedia cluster; When the number of time periods with a ratio greater than the second preset threshold is greater than a preset number, the second multimedia cluster is taken as the target multimedia cluster of the target object.

6. The data processing method according to claim 1, characterized in that, Before performing clustering processing on the multimedia data in the original dataset, the method further includes: The multimedia data in the original dataset is filtered to remove dirty data and obtain the target dataset. Accordingly, the multimedia data in the original dataset is subjected to clustering processing, including: Clustering is performed on the multimedia data in the target dataset.

7. The data processing method according to claim 6, characterized in that, Filtering the multimedia data in the original dataset includes: Obtain at least one target object behavior parameter of the multimedia data, the target object behavior parameter being generated based on the target object's access to the multimedia data; When the behavior parameters of at least one target object meet the third preset condition, the multimedia data will be used as the multimedia data in the target dataset. When the behavior parameters of at least one target object do not meet the third preset condition, the multimedia data is removed as dirty data.

8. A data processing apparatus, characterized in that, include: The raw data acquisition module is used to acquire the raw dataset of the target object on the multimedia platform. The raw dataset is a collection of multimedia data, which represents the content data accessed by the target object on the multimedia platform. The target object is an object whose points of interest need to be determined, and the target object is a collection of objects of a certain type. The multimedia platform is a content sharing platform that provides multimedia data. The clustering module is used to perform clustering processing on the multimedia data in the original dataset based on multiple preset cluster numbers, and obtain the clustering results corresponding to each preset cluster number. The clustering module includes: a feature extraction unit, used to extract features from the multimedia data in the original dataset to obtain multiple multimedia content features; and a clustering unit, used to perform clustering processing on the multimedia data in the original dataset based on the multiple multimedia content features according to multiple preset cluster numbers to obtain clustering results corresponding to each preset cluster number. The feature extraction unit includes: a text information acquisition unit for acquiring text description information of the multimedia data; a word segmentation unit for segmenting the text description information to obtain multiple words corresponding to the text description information; a vector mapping unit for performing vector mapping processing on each word corresponding to the text description information to obtain multiple word vectors corresponding to the description information of the multimedia data; and a vector fusion unit for performing vector fusion on the multiple word vectors corresponding to the text description information to obtain multimedia content features corresponding to the multimedia data. The multimedia cluster determination module is used to determine the number of target clusters based on the clustering results corresponding to the number of preset clusters, and to obtain multiple multimedia clusters based on the clustering results corresponding to the number of target clusters. The multimedia clusters represent a set of multimedia data with similar content features. The filtering module is used to filter the multiple multimedia clusters according to the degree of aggregation of multimedia data corresponding to each multimedia cluster, so as to obtain the target multimedia cluster of the target object; The filtering module includes: a distance determination unit, used to determine the distance from the multimedia content features corresponding to the multimedia data to the cluster center in the multimedia cluster; a clustering degree determination unit, used to determine the multimedia data clustering degree of the multimedia cluster based on the sum of the squares of the distances from all multimedia content features to the cluster center; and a first filtering unit, used to select the multimedia cluster as the target multimedia cluster of the target object when the multimedia data clustering degree is greater than a first preset threshold. The filtering module further includes: a target object behavior parameter acquisition unit, used to mark multimedia clusters with a multimedia data aggregation degree greater than a first preset threshold as a first multimedia cluster, and acquire target object behavior parameters of multimedia data in the first multimedia cluster, wherein the target object behavior parameters are generated based on the target object accessing the multimedia data; a first evaluation parameter determination unit, used to perform statistical processing on the target object behavior parameters of all multimedia data in the first multimedia cluster to obtain a first evaluation parameter of the first multimedia cluster based on the target object; and a second filtering unit, used to select the first multimedia cluster as the target multimedia cluster of the target object when the first evaluation parameter meets a first preset condition. The first evaluation parameter meeting the first preset condition includes: obtaining non-target object behavior parameters of multimedia data in the first multimedia cluster, and performing statistical processing on the non-target object behavior parameters of all multimedia data in the first multimedia cluster to obtain a second evaluation parameter of the first multimedia cluster based on non-target objects; wherein, the non-target object is all access objects of the first multimedia cluster other than the target object, and the non-target object behavior parameters are generated based on the non-target objects accessing the multimedia data; when the ratio of the first evaluation parameter to the second evaluation parameter is greater than a second preset threshold, it is determined that the first evaluation parameter meets the first preset condition; The device is further configured to: calculate the evaluation parameters of the first multimedia cluster based on all accessed objects; and designate the first multimedia cluster whose ratio of the first evaluation parameters to the evaluation parameters based on all accessed objects is less than a second preset threshold as the target multimedia cluster of the target object; After obtaining the target multimedia cluster, the device is further configured to: determine the interest points of the target object based on the multimedia data in the target multimedia cluster, wherein the interest points are interest point tags.

9. A computer-readable medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the data processing method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor executes the executable instructions to cause the electronic device to perform the data processing method according to any one of claims 1 to 7.

11. A computer program product, characterized in that, The computer program product includes computer instructions stored in a computer-readable storage medium, a processor of a computer device reading the computer instructions from the computer-readable storage medium, and the processor executing the computer instructions to cause the computer device to perform the data processing method as described in any one of claims 1 to 7.

Images