On-demand content clustering method and device, electronic equipment and storage medium

Abstract

The invention discloses an on-demand content clustering method, an on-demand content clustering device, electronic equipment and a storage medium, aiming at data characteristics in an IPTV (Internet Protocol Television) service environment, and optimizing a media resource clustering algorithm cooperating with user behaviors. And in combination with the media asset content and the user data, the clustering accuracy can be effectively improved. In addition, large model feedback information is innovatively introduced to correct the IPTV recommendation model. By constructing a two-tuple task and inquiring whether the two large models are similar or not, the weight or parameter of the recommendation model can be adjusted according to the feedback of the large models, so that the clustering effect is further optimized. The IPTV multi-dimensional fine granularity classification method is remarkably innovated and improved, has the advantages and effects of efficient resource utilization, multi-dimensional classification, vectorization strategy innovation, efficient model training and optimization, promotion of business innovation development and the like, and can be widely applied to the technical field of data processing.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, electronic device and storage medium for clustering on-demand content. Background Art

[0002] With the rapid development of mobile Internet technology, although smart phones and tablets have played an important role in entertainment consumption, large-screen TVs still maintain an irreplaceable position in the field of film and television entertainment due to their immersive viewing experience. As an important part of large-screen TV services, IPTV transmits TV signals through the Internet protocol, providing users with rich film and television resources and high-quality viewing experience, and is one of the current mainstream viewing methods.

[0003] However, compared with social software and Internet applications that have massive user feedback data, IPTV is relatively backward in the development of content classification functions. The granularity of classification is relatively coarse, and it is not as detailed as the classification of social software Internet platforms, which is not conducive to the platform's recommendation and search functions. This is mainly because IPTV's media data is unique and extremely dependent on the information of the media itself (such as title, type, cast, etc.). At the same time, because large-screen content is subject to the influence of broadcast control and cannot be as interactive as social software Internet platforms, IPTV lacks sufficient user behavior data (such as ratings of viewing users and comments on programs, etc.) to support advanced clustering algorithms. Traditional clustering methods, limited by the richness of data, are difficult to achieve the clustering effect of advanced algorithms in the Internet field. Summary of the invention

[0004] The present invention provides a method, device, electronic device and storage medium for clustering on-demand content, aiming to solve one of the technical problems in the related art at least to a certain extent. The embodiments of the present invention can effectively realize clustering of on-demand content of IPTV.

[0005] On the one hand, an embodiment of the present invention provides a method for clustering on-demand content, including:

[0006] Obtain all media content of the network TV, and obtain the user attributes corresponding to each media content based on the broadcast data of the media content;

[0007] Performing a first vectorization process on the media asset content to obtain a media asset vector corresponding to each media asset content; performing a second vectorization process on the user attributes to obtain a user vector corresponding to each user attribute;

[0008] Content data is obtained by sorting the corresponding media asset content and user attributes; the corresponding media asset vectors and user vectors are combined and spliced ​​to obtain vector data corresponding to each content data;

[0009] Input the vector data into a preset classification model for classification processing to obtain the probability distribution corresponding to the vector data;

[0010] The classification model includes a convolutional layer, a pooling layer, a fully connected layer, and an output layer; the probability distribution includes the category probability of each category, the sum of all category probabilities in the probability distribution is 1, and the largest category probability is the predicted category of the vector data;

[0011] Based on the probability distribution processing, the information entropy corresponding to each content data is obtained, and the information entropy is used to construct a content pair according to the content data; the content pair includes two content data;

[0012] The content pair is input into a preset large model for classification and discrimination to obtain a discrimination result; the discrimination result indicates whether the two content data in the content pair are of the same type;

[0013] The prediction accuracy of the classification model is obtained based on the discrimination results of all content pairs and the probability distribution of all vector data;

[0014] When the prediction accuracy is less than the preset threshold, the loss function is used to feedback and adjust the model parameters of the classification model, and the step of inputting the vector data into the preset classification model for classification processing is returned to obtain the probability distribution corresponding to the vector data, until the prediction accuracy is greater than or equal to the preset threshold, and the adjusted classification model is used as the target model;

[0015] The target model is used to classify the media content of the network television set into types, so that the network television set can perform content clustering on the media content according to the result of the type classification.

[0016] Optionally, the media asset content includes a content name and description, a publication year, and a cast and crew list; performing a first vectorization process on the media asset content to obtain a media asset vector corresponding to each media asset content includes the following steps:

[0017] Input the content name and description into the pre-trained language representation model to obtain a first vectorization result;

[0018] Determine the era range according to the publication year, and obtain a second vectorization result corresponding to the era range based on a preset first mapping table; wherein the first mapping table includes a vector category result corresponding to each era range;

[0019] Binary encode all the cast and crew members in the cast and crew list to obtain a third vectorization result;

[0020] The media asset vector includes a first vectorization result, a second vectorization result and a third vectorization result.

[0021] Optionally, the user attributes include gender parameters, age parameters, region parameters and user habit data; performing a second vectorization process on the user attributes to obtain a user vector corresponding to each user attribute includes the following steps:

[0022] Obtaining a fourth vector quantization result corresponding to the gender parameter based on a preset second mapping table; wherein the second mapping table includes vector gender results corresponding to different gender parameters;

[0023] Binary encode the age parameter to obtain the fifth quantization result;

[0024] Determine the straight-line distance between the region and the target region according to the region parameter, convert the straight-line distance into a binary number, and obtain a sixth vectorization result;

[0025] Inputting the user habit data into the pre-trained language representation model for processing to obtain the seventh vectorization result; the user habit data includes viewing rate, viewing time and viewing times;

[0026] The user vector includes a fourth vectorization result, a fifth vectorization result, a sixth vectorization result and a seventh vectorization result.

[0027] Optionally, the user attribute includes the number of times the user has viewed the media content; combining and concatenating the corresponding media vectors and user vectors includes the following steps:

[0028] Get the binary representation of the number of views;

[0029] Based on the preset length, the media vector, user vector and binary representation are filled with the value 0;

[0030] The padded media vector, user vector and binary representation are sequentially concatenated.

[0031] Optionally, inputting the vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data includes the following steps:

[0032] Extract features from vector data through a convolutional layer to obtain feature data;

[0033] The feature data is reduced in dimension through the pooling layer to obtain feature information;

[0034] The feature information is integrated and mapped through multiple fully connected layers to obtain the classification results;

[0035] Among them, a preset classification function is embedded in the output layer;

[0036] The classification result is converted into probability distribution through the classification function of the output layer.

[0037] Optionally, obtaining information entropy corresponding to each content data based on probability distribution processing includes the following steps:

[0038] The information entropy corresponding to the content data is obtained by performing a preset accumulation operation based on all category probabilities in the probability distribution;

[0039] Among them, the expression of information entropy is:

[0040]

[0041] In the formula, H(X) represents the information entropy corresponding to the content data X; n represents the total number of categories; i represents the i-th category; p(x i ) represents the category probability of the i-th category.

[0042] Optionally, constructing content pairs according to content data using information entropy includes the following steps:

[0043] Based on all content data, construct all initial content pairs in groups of two to obtain all combinations;

[0044] Determine the entropy value of each initial content pair according to the sum of information entropies corresponding to two content data in the initial content pair;

[0045] The preset number of initial contents with the largest entropy value is used as the content pairs finally constructed;

[0046] The content pairs include content pairs of the same type and content pairs of different types.

[0047] On the other hand, an embodiment of the present invention provides a device for clustering on-demand content, including:

[0048] The first module is used to obtain all media content of the network TV and obtain the user attributes corresponding to each media content based on the broadcast data of the media content;

[0049] The second module is used to perform a first vectorization process on the media asset content to obtain a media asset vector corresponding to each media asset content; and perform a second vectorization process on the user attributes to obtain a user vector corresponding to each user attribute;

[0050] The third module is used to obtain content data according to the corresponding media content and user attributes; combine and splice the corresponding media vectors and user vectors to obtain vector data corresponding to each content data;

[0051] The fourth module is used to input the vector data into a preset classification model for classification processing to obtain the probability distribution corresponding to the vector data;

[0052] The classification model includes a convolutional layer, a pooling layer, a fully connected layer, and an output layer; the probability distribution includes the category probability of each category, the sum of all category probabilities in the probability distribution is 1, and the largest category probability is the predicted category of the vector data;

[0053] The fifth module is used to obtain information entropy corresponding to each content data based on probability distribution processing, and construct a content pair according to the content data using the information entropy; the content pair includes two content data;

[0054] The sixth module is used to classify and distinguish the content pair into a preset large model to obtain a distinction result; the distinction result indicates whether the two content data in the content pair are of the same type;

[0055] The seventh module is used to obtain the prediction accuracy of the classification model based on the discrimination results of all content pairs and the probability distribution of all vector data;

[0056] The eighth module is used to use the loss function to feedback and adjust the model parameters of the classification model when the prediction accuracy is less than the preset threshold, return to execute the step of inputting the vector data into the preset classification model for classification processing, and obtain the probability distribution corresponding to the vector data, until the prediction accuracy is greater than or equal to the preset threshold, and use the adjusted classification model as the target model;

[0057] The target model is used to classify the media content of the network television set into types, so that the network television set can perform content clustering on the media content according to the result of the type classification.

[0058] On the other hand, an embodiment of the present invention provides an electronic device, including: a processor and a memory; the memory is used to store programs; the processor executes the program to implement the above-mentioned on-demand content clustering method.

[0059] On the other hand, an embodiment of the present invention provides a computer storage medium, in which a program executable by a processor is stored. When the program executable by the processor is executed by the processor, it is used to implement the above-mentioned on-demand content clustering method.

[0060] The embodiment of the present invention obtains all the media content of the network TV, obtains the user attributes corresponding to each media content based on the broadcast data of the media content; performs a first vectorization process on the media content to obtain a media vector corresponding to each media content; performs a second vectorization process on the user attributes to obtain a user vector corresponding to each user attribute; obtains content data according to the corresponding media content and user attributes; combines and splices the corresponding media vectors and user vectors to obtain vector data corresponding to each content data; inputs the vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data; wherein the classification model includes a convolutional layer, a pooling layer, a fully connected layer and an output layer; the probability distribution includes a category probability for each category, the sum of all category probabilities in the probability distribution is 1, and the largest category probability is the predicted category of the vector data; based on the probability distribution processing, The information entropy corresponding to each content data is used to construct a content pair according to the content data; the content pair includes two content data; the content pair is input into a preset large model for classification and discrimination to obtain a discrimination result; the discrimination result represents whether the two content data in the content pair are of the same type; the prediction accuracy of the classification model is obtained based on the discrimination results of all content pairs and the probability distribution of all vector data; when the prediction accuracy is less than the preset threshold, the model parameters of the classification model are feedback-adjusted using the loss function, and the step of inputting the vector data into the preset classification model for classification processing is returned to execute, and the probability distribution corresponding to the vector data is obtained until the prediction accuracy is greater than or equal to the preset threshold, and the adjusted classification model is used as the target model; wherein the target model is used to classify the media content of the network TV set by type, so that the network TV set performs content clustering on the media content according to the result of the type classification. The present invention is aimed at the media content and limited user attributes unique to the network TV set, and makes full use of the existing data resources without significantly increasing the data collection cost, and realizes more accurate film and television content classification through the auxiliary classification model of the result of the large model, thereby promoting the innovative development of IPTV services under the background of big data and artificial intelligence technology. The embodiments of the present invention can effectively implement clustering of on-demand content. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The accompanying drawings are used to provide a further understanding of the technical solution of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the technical solution of the present invention and do not constitute a limitation on the technical solution of the present invention.

[0062] Figure 1 A schematic diagram of an implementation environment for clustering on-demand content provided by an embodiment of the present invention;

[0063] Figure 2 A schematic diagram of a flow chart of a method for clustering on-demand content provided by an embodiment of the present invention;

[0064] Figure 3 A schematic diagram of an expansion process of a first vectorization process provided by an embodiment of the present invention;

[0065] Figure 4 A schematic diagram of an expansion process of a second vectorization process provided by an embodiment of the present invention;

[0066] Figure 5 A schematic diagram of the expansion process of step S300 provided in an embodiment of the present invention;

[0067] Figure 6 A schematic diagram of the overall process of a method for clustering on-demand content provided by an embodiment of the present invention;

[0068] Figure 7 A schematic diagram of the application of a large model provided in an embodiment of the present invention to determine the same type of media assets;

[0069] Figure 8 A schematic diagram of the application of a large model provided in an embodiment of the present invention to determine different types of media assets;

[0070] Fig. 9 A schematic diagram of the structure of a device for clustering on-demand content provided by an embodiment of the present invention;

[0071] Fig.10 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0073] It should be noted that, although the functional modules are divided in the system schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowchart. The terms "first / S100", "second / S200", etc. in the specification, claims and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0074] Reference to "embodiments" in the present invention means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in the present invention may be combined with other embodiments.

[0075] To facilitate the understanding of the technical solution of the present invention, the following is an explanation of the specific names of the technical features that may appear in the technical solution of the present invention:

[0076] IPTV (Interactive Personality TV): Internet TV uses TVs, personal computers and handheld devices as display terminals, accesses broadband networks through set-top boxes or computers, and realizes digital TV, time-shifted TV, interactive TV and other services. The emergence of Internet TV has brought people a new way of watching TV. It has changed the previous passive TV viewing mode and realized on-demand viewing of TV and stop watching at any time.

[0077] BERT (Bidirectional Encoder Representation from Transformers): BERT is a pre-trained language representation model. Its core principle is to use the masked language model (MLM) to pre-train the bidirectional language model to generate deep bidirectional language representation.

[0078] like Figure 1 FIG. 1 is a schematic diagram of an implementation environment provided by an embodiment of the present invention. Figure 1 The implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be connected to a network wirelessly or wired to complete data transmission and exchange.

[0079] Server 101 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0080] In addition, the server 101 can also be a node server in the blockchain network. Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm.

[0081] The terminal 102 may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal 102 and the server 101 may be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiment of the present invention.

[0082] Based on the example Figure 1In the implementation environment shown, an embodiment of the present invention provides a method for clustering on-demand content. The following is an example of the method for clustering on-demand content being applied in the terminal 102 . It can be understood that the method for clustering on-demand content can also be applied in the server 101 .

[0083] Reference Figure 2 , Figure 2 The present invention provides a flowchart of a method for clustering on-demand content applied to a terminal according to an embodiment of the present invention. The execution subject of the method for clustering on-demand content can be any of the aforementioned computer devices (including a server or a terminal). Figure 2 , the method comprises the following steps:

[0084] S100, obtaining all media content of the network TV, and obtaining user attributes corresponding to each media content based on the broadcast data of the media content;

[0085] For example, in some specific implementations:

[0086] 1. Acquisition of Internet TV media content:

[0087] The media content of online TV sets comes from a wide range of sources, and may include digitized versions of traditional TV programs, online self-produced dramas, short videos, documentaries, movies, variety shows, etc.

[0088] These media resources are usually uploaded to the online TV platform by content providers for users to watch. Content providers may include TV stations, film and television production companies, individual creators, etc.

[0089] 2. Collection of playback data:

[0090] Playback data refers to various data generated when users watch media content, such as viewing time, number of views, playback position, number of pauses, number of fast-forward and fast-rewind times, etc.

[0091] This data can be collected in real time through the background system of the Internet TV and used to analyze users' viewing behavior and preferences.

[0092] 3. Obtaining user attributes:

[0093] Based on the playback data, data analysis technology can be used to infer the attributes of users, which may include age, gender, region, occupation, hobbies, etc.

[0094] For example, by analyzing the types of programs that users watch during a specific time period, we can infer the users' age and interests; by analyzing the sources of the programs that users watch (such as domestic TV stations, foreign TV stations, online self-made programs, etc.), we can infer the users' geographical and cultural background.

[0095] Acquiring user attributes through the broadcast data of online TV media content is an effective data analysis method, which helps the platform to have a deeper understanding of user needs and behavioral characteristics, thereby providing more personalized services and content recommendations.

[0096] S200, performing a first vectorization process on the media asset content to obtain a media asset vector corresponding to each media asset content; performing a second vectorization process on the user attributes to obtain a user vector corresponding to each user attribute;

[0097] It should be noted that the media asset content includes the content name and description, the year of publication and the cast and crew list; in some embodiments, such as Figure 3 As shown, performing a first vectorization process on the media asset content to obtain a media asset vector corresponding to each media asset content may include the following steps: S211, inputting the content name and description into a pre-trained language representation model for processing to obtain a first vectorization result; S212, determining the era range according to the year of publication, and obtaining a second vectorization result corresponding to the era range based on a preset first mapping table; wherein the first mapping table includes vector category results corresponding to each era range; S213, performing binary encoding on all cast and crew members in the cast and crew content to obtain a third vectorization result; wherein the media asset vector includes the first vectorization result, the second vectorization result and the third vectorization result.

[0098] For example, in some specific implementations, the vectorization of media asset content may be implemented as follows:

[0099] Media assets are mainly judged by the following attributes: content name, director, actor, release time, publishing area, description, etc. Different vectorization strategies need to be adopted for the above different attributes.

[0100] 1. Content name and description use the pre-trained BERT model.

[0101] 2. The length of the publication year is 4 digits, but it can be divided according to the years, with every 10 years being divided into a stage. For example, before the 1990s is divided into one category, from 1990 to 2000 is divided into one category, from 2000 to 2010 is divided into one category, from 2010 to 2020 is divided into one category, and from 2020 to the present is divided into one category.

[0102] 3. Directors and actors are just names and do not carry semantic information. First, extract all the name information and encode all the cast and crew members in binary to represent the personnel. If the length is not enough, fill it with 0.

[0103] It should be noted that user attributes include gender parameters, age parameters, region parameters and user habit data; in some embodiments, such as Figure 4As shown, performing a second vectorization process on user attributes to obtain a user vector corresponding to each user attribute may include the following steps: S221, obtaining a fourth vectorization result corresponding to the gender parameter based on a preset second mapping table; wherein the second mapping table includes vector gender results corresponding to different gender parameters; S222, binary encoding the age parameter to obtain a fifth vectorization result; S223, determining the straight-line distance between the corresponding region and the target region based on the regional parameter, converting the straight-line distance into a binary number, and obtaining a sixth vectorization result; S224, inputting the user habit data into a pre-trained language representation model for processing to obtain a seventh vectorization result; the user habit data includes viewing rate, viewing time and viewing times; wherein the user vector includes the fourth vectorization result, the fifth vectorization result, the sixth vectorization result and the seventh vectorization result.

[0104] Exemplarily, in some specific implementations, the initial vectorization rules for user-related attributes are as follows: the gender parameter uses an all-0 or all-1 vector to represent male and female, age is represented by binary, and if the length is insufficient, it is filled with 0, the region uses a binary number representing the straight-line distance from Beijing, and the user habit data uses the BERT model word vector model. The above is just a simple example of the user's attributes. In fact, the user can be described in more dimensions according to the business, such as viewing rate, viewing time, and number of views, etc.

[0105] S300, obtaining content data according to the corresponding media asset content and user attributes; combining and splicing the corresponding media asset vectors and user vectors to obtain vector data corresponding to each content data;

[0106] It should be noted that the user attributes include the number of times the user has viewed the media content; in some embodiments, such as Figure 5 As shown, combining and splicing the corresponding media vectors and user vectors may include the following steps: S301, obtaining the binary representation corresponding to the number of views; S302, based on a preset length, filling the media vector, user vector and binary representation with the value 0; S303, sequentially splicing the filled media vector, user vector and binary representation.

[0107] For example, in some specific implementations, the media asset vector and the user vector are combined, and the concatenation method is: media asset vector + user vector of the media asset content + binary representation of the number of times the user has watched the media asset, and the length is set according to the media asset content vector and the user vector, and the insufficient length is filled with 0. In other words, the media asset content vector is now not only described based on the media asset content itself, but also incorporates the user's behavior information.

[0108] S400, inputting the vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data;

[0109] The classification model includes a convolutional layer, a pooling layer, a fully connected layer, and an output layer; the probability distribution includes the category probability of each category, the sum of all category probabilities in the probability distribution is 1, and the largest category probability is the predicted category of the vector data;

[0110] It should be noted that, in some embodiments, inputting vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data may include the following steps: extracting features from the vector data through a convolutional layer to obtain feature data; performing dimensionality reduction processing on the feature data through a pooling layer to obtain feature information; integrating and mapping the feature information through multiple fully connected layers to obtain a classification result; wherein a preset classification function is embedded in the output layer; and converting the classification result into a probability distribution through the classification function of the output layer.

[0111] For example, in some specific implementations, the present invention may use a convolutional neural network (CNN) as the main classification model architecture, and CNN can effectively extract features of input data. The vectorized content attribute data is used as the input of CNN, and features of different levels are gradually extracted through multiple convolutional layers and pooling layers.

[0112] The main layers are as follows:

[0113] Convolution layer: Use convolution kernels of different sizes to capture features of different scales. For example, for the vectors of content names and descriptions, smaller convolution kernels can be used to extract local semantic features. For binary-encoded director and actor information, larger convolution kernels can be used to capture the overall combination of personnel features.

[0114] Pooling layer: Use maximum pooling or average pooling operations to reduce feature dimensions while retaining important feature information.

[0115] Fully connected layer: After the convolutional layer and the pooling layer, several fully connected layers are connected to further integrate and map the extracted features to obtain the final classification result.

[0116] In the output layer of the model, the softmax function can be used as the classification function. The softmax function converts the output value of the model into a probability distribution, so that each output value is between 0 and 1, and the sum of all output values ​​is 1. In this way, for a given on-demand content, the model can output the probability of it belonging to each category, and the category with the highest probability is the predicted classification of the content.

[0117] S500, obtaining information entropy corresponding to each content data based on probability distribution processing, and constructing a content pair according to the content data using the information entropy; the content pair includes two content data;

[0118] It should be noted that, in some embodiments, obtaining the information entropy corresponding to each content data based on probability distribution processing may include the following steps:

[0119] The information entropy corresponding to the content data is obtained by performing a preset accumulation operation based on all category probabilities in the probability distribution;

[0120] Among them, the expression of information entropy is:

[0121]

[0122] In the formula, H(X) represents the information entropy corresponding to the content data X; n represents the total number of categories; i represents the i-th category; p(x i ) represents the category probability of the i-th category.

[0123] It should be noted that, in some embodiments, using information entropy to construct content pairs based on content data may include the following steps: constructing initial content pairs of all combinations in groups of two based on all content data; determining the entropy value of each initial content pair based on the sum of the information entropies corresponding to the two content data in the initial content pair; taking a preset number of initial contents with the largest entropy value as the final constructed content pairs; wherein the content pairs include content pairs of the same type and content pairs of different types.

[0124] For example, in some specific implementations, if content pairs are randomly constructed to query the large model, the efficiency will be very low because many content pairs may not contain information. In order to efficiently utilize the large model, content pairs can be constructed based on information entropy.

[0125] The formula for information entropy is:

[0126]

[0127] In the formula, H(X) represents the information entropy corresponding to the content data X; n represents the total number of categories; i represents the i-th category; p(x i ) represents the category probability of the i-th category.

[0128] When constructing content pairs, content pairs with larger entropy values ​​are used as input into the large model.

[0129] S600, classify and identify the content into a preset large model to obtain a identification result;

[0130] The discrimination result indicates whether the two content data in the content pair are of the same type;

[0131] For example, in some specific implementations, the content pairs can be sorted according to a preset text template and input into the big model. The present invention introduces the big model to provide feedback to the small model. The present invention proposes to use the feedback information of the big model to continuously correct the media asset similarity calculation results of the above-mentioned small model. This is mainly achieved through the following three stages:

[0132] The first stage is the task of similar media asset identification, which aims to optimize the small model's ability to judge similar media assets. The information of two media assets <media asset 1, media asset 2 similar to media asset 1> is input into the large model, allowing the large model to judge whether the two media assets are of the same type. The following is a specific example:

[0133] [Please answer, "Content "Palace Drama 1", User A attributes: 20 years old, female, Shandong, active time period: 8 pm to 1 am, User A has watched 100 times, ..." and "Content "Palace Drama 2", User B attributes: 23 years old, female, Shandong, active time period: 8 pm to 1 am, User B has watched 100 times..." can these two media assets be classified as similar content? You can output 1, but you cannot output 0. ]

[0134] The second stage is the task of distinguishing dissimilar tasks, <media asset 1, media asset 3 that is dissimilar to media asset 1>, the purpose of which is to optimize the small model's ability to judge dissimilar media assets. The following is a specific example:

[0135] [Please answer, are the two media assets "Content "Palace Drama", User A's attributes: 20 years old, female, Shandong, active time period: 8pm to 1am, User A has watched 100 times,..." and "Content "War Drama", User C's attributes: 73 years old, female, Shandong, active time period: 7pm to 9pm, User C has watched 30 times" the same type of media assets? You can output 1, but you cannot output 0. ]

[0136] Input the questions of the above tasks to the big model, which will answer the above inquiries based on its own perspective or in combination with some information on the Internet. After getting the answer, the result will be returned to the small model, so that the small model can continuously correct the model optimization effect. Each round can be set to distinguish 100 pairs of similar content and 100 pairs of different content. When the error between the output result of the model and the judgment result of the big model after N consecutive rounds is less than 5%, the training can be stopped, and the recommendation model is assumed to have been trained. N depends on the size of the training set.

[0137] S700, obtaining the prediction accuracy of the classification model based on the discrimination results of all content pairs and the probability distribution of all vector data;

[0138] For example, in some specific application scenarios, for example, the large model determines that the content pair constructed by the content data corresponding to the first media asset and the content data corresponding to the second media asset are the same type of content, but the classification model predicts different classifications for the first media asset and the second media asset. In this case, the classification results of the classification model for the first media asset and the second media asset can be marked as inaccurate, and then all the results can be counted to achieve statistics on the prediction accuracy.

[0139] S800, when the prediction accuracy is less than the preset threshold, the model parameters of the classification model are adjusted by feedback using the loss function, and the step of inputting the vector data into the preset classification model for classification processing is returned to obtain the probability distribution corresponding to the vector data, until the prediction accuracy is greater than or equal to the preset threshold, and the adjusted classification model is used as the target model;

[0140] The target model is used to classify the media content of the network television set into types, so that the network television set can perform content clustering on the media content according to the result of the type classification.

[0141] For example, in some specific implementations, the aforementioned questions constructed based on content pairs can be input into the big model, and the big model will answer the above inquiries based on its own perspective or in combination with some information on the Internet. After getting the answer, the result is returned to the small model, so that the small model can continuously correct the model optimization effect. Each round can be set to distinguish 100 pairs of similar content and 100 pairs of different content. When the error between the output result of the model and the judgment result of the big model after N consecutive rounds is less than 5%, the training can be stopped, and the default recommendation model has been trained. N depends on the size of the training set.

[0142] In some specific application scenarios, the loss function can adopt the cross entropy loss function (Cross Entropy Loss). The cross entropy loss function is often used in classification problems to measure the difference between the probability distribution predicted by the model and the actual label. For the classification task of IPTV on-demand content, each content has a specific category label, and the output of the model is the predicted probability of each category. The cross entropy loss function can prompt the model to continuously adjust parameters so that the predicted probability distribution is as close as possible to the actual category label distribution.

[0143] In order to explain the principle of the technical solution of the present invention in detail, the overall process of the present invention is described below in combination with some specific embodiments. It is easy to understand that the following is an explanation of the technical principle of the present invention and cannot be regarded as a limitation of the present invention.

[0144] First of all, it should be explained that, in view of the problems of the existing technology, how to design and implement a method that can adapt to the characteristics of IPTV data, effectively improve the granularity of detailed classification and increase the dimension of classification under limited user data resources, and better support the development of IPTV large-screen film and television services has become an urgent problem to be solved in the current development of IPTV services. It is in this context that the present invention is proposed, aiming to provide a more efficient and accurate media resource clustering solution for the IPTV platform through combining diverse media resource data with user behavior data, so as to better support the search and recommendation functions of IPTV.

[0145] At present, the development of big models is very rapid. Considering the excellent effects that big models have shown in various fields, the present invention does first consider the use of big models. However, the historical behavior data of users cannot be fully transmitted to the big model for analysis, and the services currently provided by the big model to the outside world are all implemented using APIs. Even if the historical behavior of users is input into the big model, it will exceed the token limit of the big model. Therefore, the big model cannot directly serve as the "decision maker" in the recommendation. Therefore, the present invention considers borrowing the capabilities of the big model, participating in the training of the small model, and assisting in optimizing the training of the small model.

[0146] like Figure 6 As shown, the technical solution of the present invention can be implemented as follows:

[0147] Media asset data is a very rich data resource. The first step of the present invention is to vectorize the media asset content. Media assets mainly have the following attributes: content name, director, actor, release time, publishing area, description, etc. Different vectorization strategies need to be adopted for the above different attributes.

[0148] 1. Content name and description use the pre-trained BERT model.

[0149] 2. The length of the publication year is 4 digits, but it can be divided according to the years, with every 10 years being divided into a stage. For example, before the 1990s is divided into one category, from 1990 to 2000 is divided into one category, from 2000 to 2010 is divided into one category, from 2010 to 2020 is divided into one category, and from 2020 to the present is divided into one category.

[0150] 3. Directors and actors are just names and do not carry semantic information. First, extract all the name information and encode all the cast and crew members in binary to represent the personnel. If the length is not enough, fill it with 0.

[0151] The initial vectorization rules for user-related attributes are as follows: the gender parameter uses a vector of all 0s or all 1s to represent male and female, age is represented by binary, and 0 is filled if the length is insufficient, the region uses a binary number representing the straight-line distance from Beijing, and the user habit data uses the BERT model word vector model. The above is just a simple example of user attributes. In fact, users can be described in more dimensions based on the business, such as viewing rate, viewing time, and number of views, etc.

[0152] Next, the media vector and the user vector are combined in the following way: media vector + user vector of the media content + binary representation of the number of times the user has watched the media content. The length is set according to the media content vector and the user vector, and the insufficient length is filled with 0. In other words, the media content vector is now not only described based on the media content itself, but also incorporates the user's behavior information.

[0153] Next is the network training phase, which is implemented using a convolutional neural network (CNN). As the main classification model architecture, CNN can effectively extract the features of input data. The vectorized content attribute data is used as the input of CNN, and features of different levels are gradually extracted through multiple convolutional layers and pooling layers.

[0154] The main layers are as follows:

[0155] Convolution layer: Use convolution kernels of different sizes to capture features of different scales. For example, for the vectors of content names and descriptions, smaller convolution kernels can be used to extract local semantic features. For binary-encoded director and actor information, larger convolution kernels can be used to capture the overall combination of personnel features.

[0156] Pooling layer: Use maximum pooling or average pooling operations to reduce feature dimensions while retaining important feature information.

[0157] Fully connected layer: After the convolutional layer and the pooling layer, several fully connected layers are connected to further integrate and map the extracted features to obtain the final classification result.

[0158] In the output layer of the model, the softmax function can be used as the classification function. The softmax function converts the output value of the model into a probability distribution, so that each output value is between 0 and 1, and the sum of all output values ​​is 1. In this way, for a given on-demand content, the model can output the probability of it belonging to each category, and the category with the highest probability is the predicted classification of the content.

[0159] The loss function can be a cross entropy loss function. The cross entropy loss function is often used in classification problems to measure the difference between the probability distribution predicted by the model and the actual label. For the classification task of IPTV on-demand content, each content has a specific category label, and the output of the model is the predicted probability of each category. The cross entropy loss function can prompt the model to continuously adjust parameters so that the predicted probability distribution is as close as possible to the actual category label distribution.

[0160] The present invention introduces a large model to provide feedback to the small model. The present invention proposes to use the feedback information of the large model to continuously correct the media asset similarity calculation results of the above small model. This is mainly achieved through the following three stages:

[0161] The first stage is the task of similar media asset identification, which aims to optimize the small model's ability to judge similar media assets. The information of two media assets <media asset 1, media asset 2 similar to media asset 1> is input to the large model, allowing the large model to judge whether the two media assets are of the same type. Figure 7 As shown, the following is a specific example:

[0162] [Please answer, "Content "Palace Drama 1", User A attributes: 20 years old, female, Shandong, active time period: 8 pm to 1 am, User A has watched 100 times, ..." and "Content "Palace Drama 2", User B attributes: 23 years old, female, Shandong, active time period: 8 pm to 1 am, User B has watched 100 times..." can these two media assets be classified as similar content? You can output 1, but you cannot output 0. ]

[0163] The second stage is the task of distinguishing dissimilar tasks, <media asset 1, media asset 3 that is dissimilar to media asset 1>, the purpose of which is to optimize the small model’s ability to judge dissimilar media assets. Figure 8 As shown, the following is a specific example:

[0164] [Please answer, "Content "Palace Drama x", User A attributes: 20 years old, female, Shandong, active time period: 8pm to 1am, User A has watched 100 times, ..." and "Content "War Drama y", User C attributes: 73 years old, female, Shandong, active time period: 7pm to 9pm, User C has watched 30 times" are the same type of media assets? You can output 1, but you cannot output 0. ]

[0165] If we randomly construct content pairs to query the large model, the efficiency will be very low, because many content pairs may not contain information (for example, content 1 and content 3 belong to completely different categories). In order to efficiently use the large model, we construct it based on information entropy.

[0166] The formula for information entropy is:

[0167]

[0168] Where: H(X) represents the information entropy corresponding to the content data X; n represents the total number of categories; i represents the i-th category; p(x i ) represents the category probability of the i-th category.

[0169] When constructing content pairs, content pairs with larger entropy values ​​are used as input into the large model.

[0170] Input the questions of the above tasks to the big model, which will answer the above questions based on its own perspective or combined with some information on the Internet. After getting the answer, the result will be returned to the small model, so that the small model can continuously correct the model optimization effect. Each round is set to distinguish 100 pairs of similar content and 100 pairs of different content. When the error between the output result of the model and the judgment result of the big model after N consecutive rounds is less than 5%, the training can be stopped, and the recommendation model is assumed to have been trained. N depends on the size of the training set.

[0171] In some specific implementations, the model training process of the embodiment of the present invention can be implemented as follows:

[0172] Step 1: Data preprocessing:

[0173] First, the media data and user data are structured.

[0174] Step 2: Attribute vectorization:

[0175] Based on the preprocessed data, media content data and user data are vectorized.

[0176] Step 3: Convolutional neural network training for clustering:

[0177] The data in step 2 is used as the input of the convolutional neural network, and the convolutional neural network performs convolution to extract features, clustering, etc.

[0178] Step 4: Build content pairs and input into the big model to get feedback:

[0179] Phase 1: Construct a binary task <media asset 1, media asset 2 that is similar to media asset 1>, ask the big model whether the two are of the same type, and adjust the weights or parameters of the trained clustering model based on the feedback from the big model.

[0180] Phase 2: Construct a binary task <media asset 1, media asset 3 which is not similar to media asset 1 and is of a different category>, which is the same as above, but media asset 3 is not of the same category as media asset 1.

[0181] Step 6: Iterate the above steps:

[0182] Repeat the above steps until the accuracy reaches 95% after N rounds.

[0183] In summary, the main purpose of the present invention is to design and implement a multi-dimensional and fine-grained classification method for IPTV in an IPTV service environment, targeting its unique media data and limited user behavior data, which is not limited to classification based on age, such as what the elderly like and what children like; based on regional classification, northerners like and southerners like; based on active time: those who like staying up late and those who like getting up early, etc. The algorithm aims to improve the labeling ability of IPTV media assets, so that users can search more finely when searching, or recommend more suitable content to users when recommending, thereby improving the click-through rate of the content. Through the implementation of the present invention, it is expected that, without significantly increasing the cost of data collection, the existing data resources can be fully utilized to achieve more accurate classification of film and television content, and promote the innovative development of IPTV services in the context of big data and artificial intelligence technology.

[0184] Specifically, the present invention optimizes the media clustering algorithm for collaborative user behavior based on the data characteristics in the IPTV service environment. Combining media content and user data can effectively improve the accuracy of clustering. In addition, the present invention innovatively introduces large model feedback information to correct the IPTV recommendation model. By constructing a binary task and asking the large model whether the two are similar, the present invention can adjust the weight or parameters of the recommendation model according to the feedback of the large model, thereby further optimizing the clustering effect.

[0185] Compared with the prior art, the present invention has at least the following beneficial effects:

[0186] 1. Efficiently utilize the combination of large and small models:

[0187] The present invention cleverly integrates the capabilities of the large model into the training of the small model, and optimizes the classification effect of the small model through the feedback of the large model on the similarity of media assets. This approach not only avoids the high cost and resource consumption when directly using the large model for real-time recommendation, but also fully utilizes the advantages of the large model in feature extraction and semantic understanding.

[0188] Compared with traditional methods, the solution of the present invention significantly improves the accuracy and refinement of classification while maintaining low data collection costs.

[0189] 2. Multi-dimensional and fine-grained classification of IPTV content based on user behavior:

[0190] The present invention breaks through the limitation of traditional classification based on media asset content itself, and realizes fine-grained classification based on multiple dimensions such as media asset content and user behavior data.

[0191] This multi-dimensional classification method can more accurately capture the subtle differences between media content and user needs, thereby improving the accuracy of the recommendation system and user satisfaction.

[0192] 3. Innovation of vectorization strategy:

[0193] For media asset data and user data with different attributes, the present invention adopts different vectorization strategies, such as using a pre-trained BERT model to process content names and descriptions, and performing binary encoding on directors and actors.

[0194] These innovative vectorization strategies not only improve the representation ability of the data, but also provide high-quality input for subsequent convolutional neural network training.

[0195] 4. Efficient model training and optimization:

[0196] The present invention adopts convolutional neural network (CNN) as the main classification model architecture, gradually extracts features through multiple convolutional layers and pooling layers, and realizes accurate classification of media asset content.

[0197] During the model training process, the present invention introduces the concept of information entropy and constructs high-information content to discriminate the input to the large model, thereby improving the efficiency of model training.

[0198] By continuously correcting the media asset similarity calculation results of the small model, the present invention ultimately achieves a high degree of consistency with the discrimination results of the large model, ensuring the accuracy and reliability of the recommendation system.

[0199] 5. Promote the innovative development of IPTV services:

[0200] The implementation of the present invention will significantly improve the user experience and competitiveness of IPTV services. Through more accurate classification and recommendation of film and television content, users can find their favorite content more quickly, and improve viewing satisfaction and loyalty.

[0201] At the same time, the present invention also provides strong technical support and demonstration effect for the innovative development of IPTV services under the background of big data and artificial intelligence technology.

[0202] In summary, the present invention has achieved significant innovation and improvement in the IPTV multi-dimensional fine-grained classification method, and has the advantages and effects of efficient resource utilization, multi-dimensional classification, vectorization strategy innovation, efficient model training and optimization, and promotion of business innovation and development.

[0203] On the other hand, Fig. 9 As shown, an embodiment of the present invention provides a device 900 for clustering on-demand content, including:

[0204] The first module 901 is used to obtain all media content of the network TV, and obtain user attributes corresponding to each media content based on the broadcast data of the media content;

[0205] The second module 902 is used to perform a first vectorization process on the media asset content to obtain a media asset vector corresponding to each media asset content; perform a second vectorization process on the user attributes to obtain a user vector corresponding to each user attribute;

[0206] The third module 903 is used to obtain content data according to the corresponding media asset content and user attributes; combine and splice the corresponding media asset vectors and user vectors to obtain vector data corresponding to each content data;

[0207] The fourth module 904 is used to input the vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data;

[0208] The classification model includes a convolutional layer, a pooling layer, a fully connected layer, and an output layer; the probability distribution includes the category probability of each category, the sum of all category probabilities in the probability distribution is 1, and the largest category probability is the predicted category of the vector data;

[0209] The fifth module 905 is used to obtain information entropy corresponding to each content data based on probability distribution processing, and construct a content pair according to the content data using the information entropy; the content pair includes two content data;

[0210] The sixth module 906 is used to classify and distinguish the content pair into a preset macro model to obtain a distinction result; the distinction result indicates whether the two content data in the content pair are of the same type;

[0211] The seventh module 907 is used to obtain the prediction accuracy of the classification model based on the discrimination results of all content pairs and the probability distribution of all vector data;

[0212] The eighth module 908 is used to use the loss function to feedback and adjust the model parameters of the classification model when the prediction accuracy is less than the preset threshold, return to the step of inputting the vector data into the preset classification model for classification processing, and obtain the probability distribution corresponding to the vector data, until the prediction accuracy is greater than or equal to the preset threshold, and use the adjusted classification model as the target model;

[0213] The target model is used to classify the media content of the network television set into types, so that the network television set can perform content clustering on the media content according to the result of the type classification.

[0214] The contents of the method embodiments of the present invention are all applicable to the device embodiments. The functions specifically implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0215] On the other hand, an embodiment of the present invention further provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above sensitive information method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0216] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0217] like Fig.10 As shown, Fig.10 A specific example of the hardware structure of an electronic device 1000 according to an embodiment is shown. The electronic device 1000 includes:

[0218] The processor 1001 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention.

[0219] The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1002, and the processor 1001 calls and executes the network node population optimization method of the embodiment of the present invention;

[0220] Input / output interface 1003, used to implement information input and output;

[0221] The communication interface 1004 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0222] A bus 1005 , which transmits information between various components of the device (e.g., the processor 1001 , the memory 1002 , the input / output interface 1003 , and the communication interface 1004 );

[0223] The processor 1001 , the memory 1002 , the input / output interface 1003 and the communication interface 1004 are connected to each other in communication within the device via the bus 1005 .

[0224] The electronic device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0225] The contents of the method embodiments of the present invention are all applicable to the electronic device embodiments. The functions specifically implemented by the electronic device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0226] Another aspect of an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the above method.

[0227] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (Compact Disc Read to Only Memory, CD to ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than computer readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0228] The contents of the method embodiments of the present invention are all applicable to the computer-readable storage medium embodiments. The functions specifically implemented by the computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0229] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the above method.

[0230] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0231] It should be noted that, although several modules of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0232] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD to ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation of the present invention.

[0233] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operation and logic flow presented by the present invention. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part described as a larger operation is performed independently.

[0234] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of various functional modules in the device disclosed in the present invention, the actual implementation of the module will be understood within the conventional skills of engineers. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0235] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0236] The logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be embodied in any computer-readable medium for use by an instruction execution device, apparatus or device (such as a computer-based device, a device including a processor, or other device that can fetch instructions from an instruction execution device, apparatus or device and execute the instructions), or in conjunction with such instruction execution device, apparatus or device. For purposes of this specification, "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution device, apparatus or device, or in conjunction with such instruction execution device, apparatus or device.

[0237] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0238] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution device. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0239] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0240] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0241] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A method for clustering on-demand content, characterized in that: The following steps are involved: Acquire all media content of the network TV, and obtain user attributes corresponding to each media content based on the broadcast data of the media content; Performing a first vectorization process on the media asset content to obtain a media asset vector corresponding to each of the media asset content; performing a second vectorization process on the user attributes to obtain a user vector corresponding to each of the user attributes; Arrange the corresponding media asset content and the user attributes to obtain content data; combine and splice the corresponding media asset vectors and the user vectors to obtain vector data corresponding to each content data; Inputting the vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data; The classification model includes a convolutional layer, a pooling layer, a fully connected layer and an output layer; the probability distribution includes a category probability of each category, the sum of all the category probabilities in the probability distribution is 1, and the largest category probability is the predicted category of the vector data; Based on the probability distribution processing, information entropy corresponding to each content data is obtained, and a content pair is constructed according to the content data using the information entropy; the content pair includes two content data; The content pair is input into a preset macro model for classification and discrimination to obtain a discrimination result; the discrimination result indicates whether the two content data in the content pair are of the same type; Obtaining the prediction accuracy of the classification model based on the discrimination results of all the content pairs and the probability distribution of all the vector data; When the prediction accuracy is less than a preset threshold, the model parameters of the classification model are feedback-adjusted using a loss function, and the step of inputting the vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data is returned to the step of obtaining a probability distribution corresponding to the vector data until the prediction accuracy is greater than or equal to the preset threshold, and the adjusted classification model is used as the target model; The target model is used to classify the media content of the network television set into types, so that the network television set can perform content clustering on the media content according to the result of the type classification.

2. The method for clustering on-demand content according to claim 1, characterized in that: The media asset content includes a content name and description, a publication year, and a cast and crew list; performing a first vectorization process on the media asset content to obtain a media asset vector corresponding to each of the media asset contents includes the following steps: Inputting the content name and description into a pre-trained language representation model for processing to obtain a first vectorization result; Determine the era range of the publication year according to the publication year, and obtain a second vectorization result corresponding to the era range based on a preset first mapping table; wherein the first mapping table includes a vector category result corresponding to each era range; Binary-code all the cast and crew members in the cast and crew list to obtain a third vectorization result; The media asset vector includes the first vectorization result, the second vectorization result and the third vectorization result.

3. The method for clustering on-demand content according to claim 1, characterized in that: The user attributes include gender parameters, age parameters, region parameters and user habit data; the second vectorization processing of the user attributes to obtain a user vector corresponding to each of the user attributes includes the following steps: Obtaining a fourth vector quantization result corresponding to the gender parameter based on a preset second mapping table; wherein the second mapping table includes vector gender results corresponding to different gender parameters; Binary encoding the age parameter to obtain a fifth quantization result; Determine the straight-line distance between the region and the target region according to the region parameter, convert the straight-line distance into a binary number, and obtain a sixth quantization result; Inputting the user habit data into a pre-trained language representation model for processing to obtain a seventh vectorization result; the user habit data includes a viewing rate, a viewing duration, and a viewing number; The user vector includes the fourth vectorization result, the fifth vectorization result, the sixth vectorization result and the seventh vectorization result.

4. The method for clustering on-demand content according to claim 1, characterized in that: The user attribute includes the number of times the user has viewed the media content; and combining and splicing the corresponding media vector and the user vector includes the following steps: Obtaining a binary representation corresponding to the number of views; Based on a preset length, padding the media asset vector, the user vector and the binary representation with a value of 0; The media asset vector, the user vector and the binary representation after the filling process are sequentially concatenated.

5. The method for clustering on-demand content according to claim 1, characterized in that: The step of inputting the vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data comprises the following steps: Performing feature extraction on the vector data through the convolution layer to obtain feature data; Performing dimensionality reduction processing on the feature data through the pooling layer to obtain feature information; Integrate and map the feature information through multiple fully connected layers to obtain a classification result; Wherein, a preset classification function is embedded in the output layer; The classification result is converted into the probability distribution through the classification function of the output layer.

6. The method for clustering on-demand content according to claim 1, characterized in that: The obtaining of information entropy corresponding to each content data based on the probability distribution processing comprises the following steps: Performing a preset accumulation operation based on all the category probabilities in the probability distribution to obtain the information entropy corresponding to the content data; Wherein, the expression of the information entropy is: In the formula, H(X) represents the information entropy corresponding to the content data X; n represents the total number of categories; i represents the i-th category; p(x i ) represents the category probability of the i-th category.

7. The method for clustering on-demand content according to claim 1, characterized in that: The step of constructing a content pair according to the content data using the information entropy comprises the following steps: According to all the content data, constructing all combinations of initial content pairs in groups of two; Determine the entropy value of each of the initial content pairs according to the sum of the information entropies corresponding to the two content data in the initial content pair; Taking the preset number of initial contents with the largest entropy values ​​as the content pairs finally constructed; The content pairs include content pairs of the same type and content pairs of different types.

8. A device for clustering on-demand content, characterized in that: include: The first module is used to obtain all media content of the network TV, and obtain the user attribute corresponding to each media content based on the broadcast data of the media content; The second module is used to perform a first vectorization process on the media asset content to obtain a media asset vector corresponding to each of the media asset content; and perform a second vectorization process on the user attributes to obtain a user vector corresponding to each of the user attributes; The third module is used to obtain content data according to the corresponding relationship between the media asset content and the user attributes; combine and splice the corresponding relationship between the media asset vector and the user vector to obtain vector data corresponding to each content data; The fourth module is used to input the vector data into a preset classification model for classification processing to obtain a probability distribution corresponding to the vector data; The classification model includes a convolutional layer, a pooling layer, a fully connected layer and an output layer; the probability distribution includes a category probability of each category, the sum of all the category probabilities in the probability distribution is 1, and the largest category probability is the predicted category of the vector data; A fifth module is used to obtain information entropy corresponding to each content data based on the probability distribution processing, and construct a content pair according to the content data using the information entropy; the content pair includes two content data; The sixth module is used to classify and distinguish the content pair into a preset large model to obtain a distinction result; the distinction result indicates whether the two content data in the content pair are of the same type; A seventh module is used to obtain the prediction accuracy of the classification model based on the discrimination results of all the content pairs and the probability distribution of all the vector data; The eighth module is used for, when the prediction accuracy is less than a preset threshold, using a loss function to perform feedback adjustment on the model parameters of the classification model, returning to the step of inputting the vector data into the preset classification model for classification processing to obtain the probability distribution corresponding to the vector data, until the prediction accuracy is greater than or equal to the preset threshold, and using the adjusted classification model as the target model; The target model is used to classify the media content of the network television set into types, so that the network television set can perform content clustering on the media content according to the result of the type classification.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 7.

10. A computer storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to implement the method according to any one of claims 1 to 7 when executed by the processor.

Images