CN115048504A - Information pushing method and device, computer equipment and computer readable storage medium - Google Patents

Abstract

The embodiment of the invention relates to the technical field of artificial intelligence, and discloses an information pushing method, which comprises the following steps: classifying the watched contents of all users to obtain a plurality of content classifications; calculating a first probability of the target user watching each content to be recommended according to the preference degree of the target user to each content classification and the similarity of the content to be recommended and the watched content of the target user; calculating a second probability that each content to be recommended is recommended to the target user according to comment information of the user category to which the target user belongs on the content to be recommended; according to the first probability and the second probability, obtaining the heat score of each content to be recommended; and recommending the content to be recommended to a target user according to the popularity score. Through the mode, the content recommendation accuracy is improved.

Description

Information push method, apparatus, computer device, and computer-readable storage medium

technical field

Embodiments of the present invention relate to the technical field of artificial intelligence, and in particular, to an information push method, apparatus, computer device, and computer-readable storage medium.

Background technique

At present, content push generally uses collaborative filtering to push related content by analyzing the similarity of user relationship content or content that has been watched by users' friends, or simply pushes content based on hot content.

In the process of implementing the embodiments of the present invention, the inventor found that the existing users did not comprehensively consider the user and the content, the relationship between the user and the user, and the user's validity evaluation of the watched content when analyzing the probability of the user viewing the content, resulting in information Recommendations are less accurate.

SUMMARY OF THE INVENTION

In view of the above problems, embodiments of the present invention provide an information push method, apparatus, computer device, and computer-readable storage medium, which are used to solve the technical problem of low information recommendation accuracy existing in the prior art.

According to an aspect of the embodiments of the present invention, there is provided a method for pushing information, the method comprising:

Classify the watched content of all users to obtain multiple content classifications;

Calculate the first probability that the target user views each of the to-be-recommended content according to the target user's preference for each of the content categories and the similarity between the to-be-recommended content and the target user's watched content; the to-be-recommended content The content is any content in the multiple content categories except the content that the target user has watched;

Calculate the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the content to be recommended by the user category to which the target user belongs;

obtaining a popularity score of each content to be recommended according to the first probability and the second probability;

The to-be-content is recommended to the target user according to the popularity score.

In an optional manner, classifying the watched content of all users to obtain multiple content classifications, including: selecting K pieces of content from all the watched content as classification centers; Correlation of the classification centers; according to the correlations, a correlation matrix between each watched content is obtained; iteratively update the classification centers and the correlation matrix until the iterative threshold is met, and multiple content classifications are obtained.

In an optional manner, according to the preference degree of the target user for each of the content categories and the similarity between the content to be recommended and the watched content of the target user, calculating the target user to watch each content category. Describe the first probability of the content to be recommended, including: determining each content that has been watched by the target user and the classification center of each of the content categories; the preference degree of the target user for each of the content categories.

In an optional manner, according to the preference degree of the target user for each of the content categories and the similarity between the content to be recommended and the watched content of the target user, it is calculated that the target user watched each of the to-be-recommended users. The first probability of recommending content includes: determining a first feature vector of the content viewed by the target user; determining a second feature vector of the content to be recommended; according to the first feature vector and the second feature vector , and determine the similarity between the content to be recommended and the watched content of the target user.

In an optional manner, calculating the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs, including : obtain the viewing information of all users; obtain the feature vector of each user according to the viewing information and the word vector model; perform clustering according to the feature vector to obtain the user category to which each user belongs.

In an optional manner, calculating the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs, including : Determine the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs, according to the comment information on the content to be recommended by the user category to which the target user belongs; Euclidean distance between other users; according to the Euclidean distance and the emotion classification, calculate the second probability that each of the to-be-recommended content is recommended to the target user.

In an optional manner, the emotion classification of the content to be recommended by other users in the user category to which the target user belongs is determined according to the comment information on the content to be recommended by the user category to which the target user belongs, Including: obtaining the comment information of the user category to which the target user belongs to the content to be recommended; inputting the comment information into a sentiment classification model to obtain the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs ; The emotion classification model is obtained by training according to emotion classification samples in advance.

According to another aspect of the embodiments of the present invention, an apparatus for pushing information is provided, including:

The content classification module is used to classify the watched content of all users to obtain multiple content classifications;

The first probability calculation module is configured to calculate the probability that the target user watches each of the to-be-recommended content according to the target user's preference for each of the content categories and the similarity between the to-be-recommended content and the target user's watched content; The first probability; the content to be recommended is any content in the multiple content categories except the content that the target user has watched;

A second probability calculation module, configured to calculate the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs;

a popularity score module, configured to obtain a popularity score of each content to be recommended according to the first probability and the second probability;

A recommendation module, configured to recommend the to-be-content to target users according to the popularity score.

According to another aspect of the embodiments of the present invention, a computer device is provided, including: a processor, a memory, a communication interface, and a communication bus, and the processor, the memory, and the communication interface communicate with each other through the communication bus. communication between;

The memory is used for storing at least one executable instruction, and the executable instruction enables the processor to perform the operations of the above-mentioned information pushing method.

According to yet another aspect of the embodiments of the present invention, a computer-readable storage medium is provided, where at least one executable instruction is stored in the storage medium, and when the executable instruction is executed on a computer device, the computer device executes the above-mentioned The operation of the information push method.

In this embodiment of the present invention, multiple content categories are obtained by classifying the watched content of all users, and then according to the user's preference for each of the content categories and the similarity between the content to be recommended and the watched content of the target user Calculate the first probability of the target user viewing each of the to-be-recommended content, and then calculate that each of the to-be-recommended content is recommended to the target user according to the comment information of the user category to which the target user belongs. The second probability of the target user is obtained, the popularity score of each content to be recommended is obtained according to the first probability and the second probability, and finally the content to be recommended is recommended to the target user according to the popularity score, which can effectively improve the The accuracy of the information recommendation.

The above description is only an overview of the technical solutions of the embodiments of the present invention. In order to understand the technical means of the embodiments of the present invention more clearly, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and The advantages can be more clearly understood, and the following specific embodiments of the present invention are given.

Description of drawings

The drawings are only used to illustrate the embodiments and are not considered to be limiting of the present invention. Also, the same components are denoted by the same reference numerals throughout the drawings. In the attached image:

1 shows a schematic flowchart of an information push method provided by an embodiment of the present invention;

FIG. 2 shows a schematic structural diagram of an information push device provided by an embodiment of the present invention;

FIG. 3 shows a schematic structural diagram of a computer device provided by an embodiment of the present invention.

Detailed ways

Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited by the embodiments set forth herein.

At present, content push is performed by analyzing the similarity of user relationship content or the content that has been watched by users' friends, using collaborative filtering to push related content, or simply pushing content based on hot content. When obtaining the feature vector of the content, the existing solution mainly generates the feature vector of the content through user ratings, and does not consider features such as content playback time and playback training. When acquiring the hot content, the number of hot words in the content is simply counted, and the content without hot words cannot be counted. Or the hot content is calculated according to the user's search behavior, which is also based on quantity statistics. Generating hot content based on quantity statistics cannot comprehensively reflect the hotness of the content, but can better reflect the hotness of the content through the user's viewing behavior. When analyzing the probability of users watching content, the user and content, the relationship between users and users, and the user's evaluation of the effectiveness of the watched content are not comprehensively considered.

FIG. 1 shows a flowchart of an information push method provided by an embodiment of the present invention, where the method is executed by a computer device. The computer device may be a computer, a tablet computer, a mobile phone, a watch, an audio and video playback device, a wearable device, etc., which is not specifically limited in this embodiment of the present invention. As shown in Figure 1, the method includes the following steps:

Step 110: Classify the watched content of all users to obtain multiple content classifications.

The embodiment of the present invention further obtains user behavior data in advance before classifying the watched content of all users, and obtains the watched content corresponding to all users after processing the user behavior data. Specifically, the user behavior log is sent to kafka through the business background, the user behavior data of the user behavior log in kafka (distributed streaming media platform) is collected through flume (log collection system), and the user behavior data is filtered and extracted. Among them, an interceptor can be set in flume to filter abnormal data in user behavior data, such as empty user name, abnormal field value, etc. After filtering, the data information is extracted through sparkstreaming (a real-time computing framework built on Spark). First, the data information of the user's viewing behavior is obtained, and then the user id, watched content id, and content playback time are extracted from the viewing behavior data. , content playback time and other viewing information, and finally store the obtained viewing information into the hive (data warehouse tool) database.

After obtaining the watched content of all users, the word vector model is used to convert the watched content into the feature vector V of the watched content. Categorize to get multiple content categories. Which includes the following steps:

a. Select K contents from all the watched contents as the classification center.

b. Calculate the correlation of all viewed content with each of the classification centers. Specifically, according to the selected classification center, the correlation between the i-th watched content c _i and each classification center v _j in the watched content of all users is calculated:

Among them, p _ij is the correlation between the i-th all users' watched content c _i and the j-th classification center v _j ; d( _ci ,v _j ) is the i-th watched content among all the users' watched content The Euclidean distance between c _i and the jth classification center v _j ; t is a variable and takes the value 0-k.

c. According to the correlation, a correlation matrix between the respective viewed contents is obtained. The correlation matrix is a k*n matrix in which the viewed content is the column and the classification center is the row, and n is the number of the watched content of all users.

d. Iteratively update the classification center and the correlation matrix until the iterative threshold is met, and obtain multiple content classifications.

After obtaining the correlation matrix, update the classification center, and the update method is:

Among them, p _ij represents the correlation between the i-th watched content and the j-th watched content as the classification center; d _j is the feature vector of the watched content j, n represents the number of watched content of all users; i represents The ith watched content.

Through the iterative update of the classification center, when the Euclidean distance between the classification center point obtained in the current time and the classification center obtained in the previous iteration is less than the threshold e, it is determined that the iteration is stopped, and k content classifications are obtained.

Step 120: Calculate a first probability that the target user views each of the to-be-recommended contents according to the target user's preference for each of the content categories and the similarity between the to-be-recommended content and the target user's watched content.

The to-be-recommended content is any content in the multiple content categories except the content that the target user has watched.

Wherein, after the k content categories are obtained, each watched content of the target user and the classification center of each content classification are determined; according to each watched content of the target user and the classification center of each content classification, calculate the The preference level of the target user for each of the content categories. Specifically, the calculation method can be:

Among them, I(u, c _i ) represents the preference degree of the target user u to the i-th content category c _i among the k content categories; s represents the total number of watched contents corresponding to the target user u; X _j represents the target user u’s ith content category j watched content.

In this embodiment of the present invention, calculating the similarity between the content to be recommended and each watched content of the target user u includes: determining a first feature vector of the content watched by the target user; determining a second feature vector of the content to be recommended feature vector; according to the first feature vector and the second feature vector, determine the similarity between the content to be recommended and the watched content of the target user. Specifically, the calculation formula can be:

Among them, W _ij represents the similarity of the content category of the i-th watched content and the j-th content to be recommended among the s watched contents of the target user u; V _im represents the i-th watched content of the target user u. Feature vector; V _im represents the feature vector of the jth content to be recommended. After obtaining the user preference degree and the similarity between the to-be-recommended content and the target user's watched content respectively, according to the target user's preference degree for each of the content categories and the to-be-recommended content and the target user's watched content The similarity of the content is calculated, and the first probability that the target user watches each of the to-be-recommended content is calculated. Specifically, the following methods are included:

Among them, p _uj1 is the first probability that the target user u watches the j-th content to be recommended; s is the number of watched content corresponding to the target user u; W _ij is the i-th watched content and the j-th content of the target user u The similarity of the content to be recommended. When the content to be recommended is the watched content of the target user u, P _uj1 =1, that is, the viewing probability is 1. c _j is the classification center of the content classification to which the jth content to be recommended belongs.

Step 130: Calculate a second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the content to be recommended by the user category to which the target user belongs.

Wherein, classifying users in advance in this embodiment of the present invention includes: obtaining viewing information of all users, obtaining feature vectors of each user according to the viewing information and a word vector model; clustering according to the feature vectors to obtain each The user category to which the user belongs. Specifically, the feature vector W of all users can be obtained by training word2Vec. After the user feature vector W is obtained, all feature vectors are clustered. Specifically, steps a to d similar to the above content classification can be used to calculate all users for group classification, and obtain c user categories.

After obtaining the user category to which the target user belongs, the sentiment classification of the to-be-recommended content by other users in the user category to which the target user belongs is determined according to the comment information of the user category to which the target user belongs to the to-be-recommended content. Calculate the Euclidean distance between the target user and other users in the user category, and calculate the second probability that each of the to-be-recommended content is recommended to the user according to the Euclidean distance and the sentiment classification. Wherein, determining the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs is determined according to the comment information on the content to be recommended by the user category to which the target user belongs, which is specifically implemented by: obtaining the content to be recommended. Comment information on the content to be recommended by the user category to which the user belongs; input the comment information into a sentiment classification model to obtain the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs; the sentiment classification model It is pre-trained based on sentiment classification samples.

其中，P(w₁,w₂)表示词语w₁及w₂同时出现的概率；P(w₁)表示词语w₁出现的概率；P(w2)表示词语w2出现的概率。通过该公式分别计算所有词语与选取的种子单词的相关性，计算得到与积极单词的相关性PP以及与消极单词的相关性PN。然后，计算PP与PN的差值(PP-PN)，如果差值为正，则该形容词为积极的，若为负则为消极的，从而从所有词语中挖掘得到所有的积极和消极形容词。子步骤c:通过子步骤b中得到的所有词语的词性，筛选出每句评论的名词与形容词，并将名词与形容词组成词对。子步骤d:将训练数据中去除特殊符号的样本评论信息通过BERT预训练模型进行预训练。BERT是一个多任务模型，它的任务是由两个自监督任务组成，即MLM和NSP。MLM是指在训练的时候随即从输入语料上mask掉一些单词，然后通过的上下文预测该单词，MLM任务中，将子步骤b和子步骤c中获取到的带有词性的形容词与名词-形容词词对进行mask，其余训练步骤与BERT模型的训练过程相同，从而得到训练好的情感分类模型。In the embodiment of the present invention, the process of training the emotion classification model is as follows: using a BERT model (an open source pre-training model for natural language understanding) to train the emotion classification model. In the pre-training process of BERT, the way of randomly erasing one or several words in the Masked LM task is modified to Mask-specific words, and Masked LM can be used to learn the representation of words in context. The calculation method is as follows: Sub-step a: First, some seed words need to be selected, and the part of speech (positive or negative) of the words should be marked. For example, positive words are like, very good, etc., and negative words are dislike, bad, embarrassing and so on. Sub-step b: Mining more emotional attribute words through the selected seed words. The mining method is as follows: First, perform word segmentation on all the comment content, use the open source word segmentation tool (Stanford CoreNLP) to perform word segmentation and obtain the part of speech of each word. After getting all the word sets and the part of speech of each word, get all the adjectives. Then, calculate the relevance of all adjectives to the selected seed words. Calculated as follows:

Among them, P(w ₁ , w ₂ ) represents the probability that the words w ₁ and w ₂ appear at the same time; P(w ₁ ) represents the probability that the word w ₁ appears; P(w2) represents the probability that the word w2 appears. Through this formula, the correlations between all words and the selected seed words are calculated respectively, and the correlation PP with positive words and the correlation PN with negative words are calculated. Then, the difference between PP and PN (PP-PN) is calculated, if the difference is positive, the adjective is positive, if it is negative, it is negative, so that all positive and negative adjectives are mined from all words. Sub-step c: filter out the nouns and adjectives of each comment based on the parts of speech of all words obtained in sub-step b, and form word pairs with the nouns and adjectives. Sub-step d: Pre-train the sample comment information with special symbols removed from the training data through the BERT pre-training model. BERT is a multi-task model whose task is composed of two self-supervised tasks, namely MLM and NSP. MLM refers to masking some words from the input corpus during training, and then predicting the word through the context. In the MLM task, the adjective with part of speech and the noun-adjective word obtained in sub-step b and sub-step c are Mask the pair, and the rest of the training steps are the same as the training process of the BERT model, so as to obtain a trained sentiment classification model.

Wherein, after the sentiment classification model is obtained, the sentiment classification of each user's comment content for each watched content (that is, the content to be recommended) is calculated according to the obtained sentiment classification model, where the sentiment classification as "positive" is the recommendation. , negative is not recommended.

After determining the sentiment classification, calculate the Euclidean distance u _ub between the target user u and other users b in the user category, and calculate the recommended content for each content to be recommended according to the Euclidean distance and the sentiment classification a second probability for the target user. The specific calculation formula can be expressed as:

Among them, p _uj2 is the second probability of recommending the j-th content to be recommended to the target user u; U _ub is the Euclidean distance between the target user u and the user b; r is the sentiment classification of the user b for the recommended content j, where, 1 for recommended, -1 for not recommended, 0 for no reviews. q is the number of users in the user category to which the target user u belongs.

Step 140: Obtain a popularity score of each content to be recommended according to the first probability and the second probability.

Wherein, after obtaining the first probability and the second probability, according to the following calculation formula, determine the target probability that user u watches the jth content to be recommended:

After the target probability is obtained, the popularity score of each content to be recommended is determined according to the target probability. The specific popularity score can be calculated according to the following formula:

Among them, Score _j is the popularity score of the jth content to be recommended, and C is the total number of all users.

Step 150: Recommend the to-be-content to target users according to the popularity score.

After the popularity score of each content to be recommended is obtained, the content to be recommended with high popularity can be determined according to the popularity score, and the content to be recommended with high popularity is recommended to the corresponding target user u. Among them, the content to be recommended can be sorted according to the popularity score, and the content to be recommended in the top TOPN is taken and recommended to the corresponding target users.

In this embodiment of the present invention, a plurality of content categories are obtained by classifying the watched content of all users, and then according to the preference degree of the target user for each content classification and the similarity between the content to be recommended and the watched content of the target user , calculate the first probability of the target user viewing each of the to-be-recommended content, and then calculate that each of the to-be-recommended content is recommended to the The second probability of the target user. According to the first probability and the second probability, the popularity score of each content to be recommended is obtained, and finally the content to be recommended is recommended to the target user according to the popularity score, which can effectively improve the information Recommended accuracy.

FIG. 2 shows a schematic structural diagram of an information pushing apparatus provided by an embodiment of the present invention. As shown in Figure 2, the device 200 includes:

The content classification module 210 is used to classify the watched content of all users to obtain a plurality of content classifications;

The first probability calculation module 220 is configured to calculate the target user's viewing of each of the to-be-recommended content according to the target user's preference for each of the content categories and the similarity between the to-be-recommended content and the target user's watched content The first probability of ; the to-be-recommended content is any content in the multiple content categories except the content that the target user has watched;

A second probability calculation module 230, configured to calculate a second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the content to be recommended by the user category to which the target user belongs;

A popularity score module 240, configured to obtain a popularity score of each content to be recommended according to the first probability and the second probability;

The recommendation module 250 is configured to recommend the to-be-content to target users according to the popularity score.

In an optional manner, classifying the watched content of all users to obtain multiple content classifications, including: selecting K pieces of content from all the watched content as classification centers; Correlation of the classification centers; according to the correlations, a correlation matrix between each watched content is obtained; iteratively update the classification centers and the correlation matrix until the iterative threshold is met, and multiple content classifications are obtained.

In an optional manner, according to the preference degree of the target user for each of the content categories and the similarity between the content to be recommended and the watched content of the target user, calculating the target user to watch each content category. Describe the first probability of the content to be recommended, including: determining each content that has been watched by the target user and the classification center of each of the content categories; the preference degree of the target user for each of the content categories.

In an optional manner, according to the preference degree of the target user for each of the content categories and the similarity between the content to be recommended and the watched content of the target user, it is calculated that the target user watched each of the to-be-recommended users. The first probability of recommending content includes: determining a first feature vector of the content viewed by the target user; determining a second feature vector of the content to be recommended; according to the first feature vector and the second feature vector , and determine the similarity between the content to be recommended and the watched content of the target user.

In an optional manner, calculating the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs, including : obtain the viewing information of all users; obtain the feature vector of each user according to the viewing information and the word vector model; perform clustering according to the feature vector to obtain the user category to which each user belongs.

In an optional manner, calculating the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs, including : Determine the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs, according to the comment information on the content to be recommended by the user category to which the target user belongs; Euclidean distance between other users; according to the Euclidean distance and the emotion classification, calculate the second probability that each of the to-be-recommended content is recommended to the target user.

In an optional manner, the emotion classification of the content to be recommended by other users in the user category to which the target user belongs is determined according to the comment information on the content to be recommended by the user category to which the target user belongs, Including: obtaining the comment information of the user category to which the target user belongs to the content to be recommended; inputting the comment information into a sentiment classification model to obtain the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs ; The emotion classification model is obtained by training according to emotion classification samples in advance.

The specific working process of the information pushing apparatus according to the embodiment of the present invention is substantially the same as the specific flow steps of the above-mentioned method embodiments, and will not be repeated here.

In this embodiment of the present invention, a plurality of content categories are obtained by classifying the watched content of all users, and then according to the preference degree of the target user for each content classification and the similarity between the content to be recommended and the watched content of the target user , calculate the first probability of the target user viewing each of the to-be-recommended content, and then calculate that each of the to-be-recommended content is recommended to the The second probability of the target user. According to the first probability and the second probability, the popularity score of each content to be recommended is obtained, and finally the content to be recommended is recommended to the target user according to the popularity score, which can effectively improve the information Recommended accuracy.

FIG. 3 shows a schematic structural diagram of a computer device provided by an embodiment of the present invention. The specific embodiment of the present invention does not limit the specific implementation of the computer device.

As shown in FIG. 3 , the computer equipment may include: a processor (processor) 402 , a communications interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

The processor 402 , the communication interface 404 , and the memory 406 communicate with each other through the communication bus 408 . The communication interface 404 is used for communicating with network elements of other devices such as clients or other servers. The processor 402 is configured to execute the program 410, and specifically may execute the relevant steps in the above-mentioned embodiments of the method for pushing information.

Specifically, program 410 may include program code, which includes computer-executable instructions.

The processor 402 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included in the computer equipment may be the same type of processors, such as one or more CPUs; or may be different types of processors, such as one or more CPUs and one or more ASICs.

The memory 406 is used to store the program 410 . Memory 406 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

The program 410 can be specifically called by the processor 402 to make the computer device perform the following operations:

Classify the watched content of all users to obtain multiple content classifications;

Calculate the first probability that the target user views each of the to-be-recommended content according to the target user's preference for each of the content categories and the similarity between the to-be-recommended content and the user's watched content; the to-be-recommended content any one content in the plurality of content categories except the content that the target user has watched;

Calculate the second probability that each of the to-be-recommended content is recommended to the user according to the comment information on the to-be-recommended content of the user category to which the target user belongs;

obtaining a popularity score of each content to be recommended according to the first probability and the second probability;

The to-be-content is recommended to the target user according to the popularity score.

In an optional manner, classifying the watched content of all users to obtain multiple content classifications, including: selecting K pieces of content from all the watched content as classification centers; Correlation of the classification centers; according to the correlations, a correlation matrix between each watched content is obtained; iteratively update the classification centers and the correlation matrix until the iterative threshold is met, and multiple content classifications are obtained.

In an optional manner, calculating the target user's viewing of each said The first probability of the content to be recommended includes: determining each content viewed by the target user and the classification center of each content category; The preference level of the target user for each of the content categories.

In an optional manner, according to the preference degree of the target user for each of the content categories and the similarity between the content to be recommended and the watched content of the target user, it is calculated that the target user watched each of the to-be-recommended users. The first probability of recommending content includes: determining a first feature vector of the content viewed by the target user; determining a second feature vector of the content to be recommended; according to the first feature vector and the second feature vector , and determine the similarity between the content to be recommended and the watched content of the target user.

In an optional manner, calculating the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs, including : obtain the viewing information of all users; obtain the feature vector of each user according to the viewing information and the word vector model; perform clustering according to the feature vector to obtain the user category to which each user belongs.

In an optional manner, calculating the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs, including : Determine the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs, according to the comment information on the content to be recommended by the user category to which the target user belongs; Euclidean distance between users; according to the Euclidean distance and the sentiment classification, calculate the second probability that each of the to-be-recommended content is recommended to the target user.

In an optional manner, the emotion classification of the content to be recommended by other users in the user category to which the target user belongs is determined according to the comment information on the content to be recommended by the user category to which the target user belongs, Including: obtaining the comment information of the user category to which the target user belongs to the content to be recommended; inputting the comment information into a sentiment classification model to obtain the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs ; The emotion classification model is obtained by training according to emotion classification samples in advance.

In this embodiment of the present invention, a plurality of content categories are obtained by classifying the watched content of all users, and then according to the preference degree of the target user for each content classification and the similarity between the content to be recommended and the watched content of the target user , calculate the first probability of the target user viewing each of the to-be-recommended content, and then calculate that each of the to-be-recommended content is recommended to the The second probability of the target user. According to the first probability and the second probability, the popularity score of each content to be recommended is obtained, and finally the content to be recommended is recommended to the target user according to the popularity score, which can effectively improve the information Recommended accuracy.

An embodiment of the present invention provides a computer-readable storage medium, where the storage medium stores at least one executable instruction, and when the executable instruction is executed on a computer device, causes the computer device to execute any of the above method embodiments. Information push method.

Specifically, the executable instructions can be used to cause the computer device to perform the following operations:

Classify the watched content of all users to obtain multiple content classifications;

Calculate the first probability that the target user views each of the to-be-recommended content according to the target user's preference for each of the content categories and the similarity between the to-be-recommended content and the target user's watched content; the to-be-recommended content The content is any content in the multiple content categories except the content that the target user has watched;

Calculate the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the content to be recommended by the user category to which the target user belongs;

obtaining a popularity score of each content to be recommended according to the first probability and the second probability;

The to-be-content is recommended to the target user according to the popularity score.

In an optional manner, classifying the watched content of all users to obtain multiple content classifications, including: selecting K pieces of content from all the watched content as classification centers; Correlation of the classification centers; according to the correlations, a correlation matrix between each watched content is obtained; iteratively update the classification centers and the correlation matrix until the iterative threshold is met, and multiple content classifications are obtained.

In an optional manner, according to the preference of the target user for each of the content categories and the similarity between the content to be recommended and the watched content of the target user, calculating the user to watch each of the content The first probability of the content to be recommended includes: determining each content viewed by the target user and the classification center of each content category; The preference level of the target user for each of the content categories.

In an optional manner, according to the target user's preference for each of the content categories and the similarity between the content to be recommended and the user's watched content, it is calculated that the target user watches each of the to-be-recommended content The first probability of the content includes: determining the first feature vector of the watched content of the target user; determining the second feature vector of the content to be recommended; according to the first feature vector and the second feature vector, The similarity between the content to be recommended and the watched content of the target user is determined.

In an optional manner, calculating the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs, including : obtain the viewing information of all users; obtain the feature vector of each user according to the viewing information and the word vector model; perform clustering according to the feature vector to obtain the user category to which each user belongs.

In an optional manner, calculating the second probability that each of the to-be-recommended content is recommended to the target user according to the comment information on the to-be-recommended content according to the user category to which the target user belongs, including : Determine the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs, according to the comment information on the content to be recommended by the user category to which the target user belongs; Euclidean distance between other users; according to the Euclidean distance and the emotion classification, calculate the second probability that each of the to-be-recommended content is recommended to the target user.

In an optional manner, the emotion classification of the content to be recommended by other users in the user category to which the target user belongs is determined according to the comment information on the content to be recommended by the user category to which the target user belongs, Including: obtaining the comment information of the user category to which the target user belongs to the content to be recommended; inputting the comment information into a sentiment classification model to obtain the sentiment classification of the content to be recommended by other users in the user category to which the target user belongs ; The emotion classification model is obtained by training according to emotion classification samples in advance.

In this embodiment of the present invention, a plurality of content categories are obtained by classifying the watched content of all users, and then according to the preference degree of the target user for each content classification and the similarity between the content to be recommended and the watched content of the target user , calculate the first probability of the target user viewing each of the to-be-recommended content, and then calculate that each of the to-be-recommended content is recommended to the The second probability of the target user. According to the first probability and the second probability, the popularity score of each content to be recommended is obtained, and finally the content to be recommended is recommended to the target user according to the popularity score, which can effectively improve the information Recommended accuracy.

An embodiment of the present invention provides an information push apparatus, which is used for executing the above information push method.

An embodiment of the present invention provides a computer program, and the computer program can be invoked by a processor to cause a computer device to execute the information pushing method in any of the foregoing method embodiments.

An embodiment of the present invention provides a computer program product. The computer program product includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are run on a computer, the computer is made to execute any of the above. The information push method in the method embodiment.

The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used with teaching based on this. The structure required to construct such a system is apparent from the above description. Furthermore, embodiments of the present invention are not directed to any particular programming language. It is to be understood that various programming languages may be used to implement the inventions described herein, and that the descriptions of specific languages above are intended to disclose the best mode for carrying out the invention.

In the description provided herein, numerous specific details are set forth. It will be understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it is to be understood that, in the above description of exemplary embodiments of the invention, various features of the embodiments of the invention are sometimes grouped together into a single implementation in order to simplify the invention and to aid in the understanding of one or more of the various aspects of the invention. examples, figures, or descriptions thereof. This disclosure, however, should not be construed as reflecting an intention that the invention as claimed requires more features than are expressly recited in each claim.

Those skilled in the art can understand that the modules in the device in the embodiment can be adaptively changed and arranged in one or more devices different from the embodiment. The modules or units or components in the embodiments may be combined into one module or unit or component, and they may be divided into multiple sub-modules or sub-units or sub-assemblies. All features disclosed in this specification (including accompanying claims, abstract and drawings) and any method so disclosed may be employed in any combination, unless at least some of such features and/or procedures or elements are mutually exclusive. All processes or units of equipment are combined. Each feature disclosed in this specification (including accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

It should be noted that the above-described embodiments illustrate rather than limit the invention, and that alternative embodiments may be devised by those skilled in the art without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. do not denote any order. These words can be interpreted as names. The steps in the above embodiments should not be construed as limitations on the execution order unless otherwise specified.

Claims (10)

1. An information pushing method, characterized in that the method comprises:

classifying the watched contents of all users to obtain a plurality of content classifications;

calculating a first probability of the target user watching each content to be recommended according to the preference degree of the target user to each content classification and the similarity between the content to be recommended and the content watched by the target user; the content to be recommended is any one of the content categories except the content watched by the target user;

calculating a second probability that each content to be recommended is recommended to the target user according to comment information of the user category to which the target user belongs on the content to be recommended;

according to the first probability and the second probability, obtaining the heat score of each content to be recommended;

and recommending the content to be recommended to a target user according to the popularity score.

2. The method of claim 1, wherein the classifying the viewed content of all users results in a plurality of content classifications, comprising:

selecting K contents from all watched contents as a classification center;

calculating the correlation of all the viewed contents with each classification center;

obtaining a correlation matrix among the watched contents according to the correlation;

and iteratively updating the classification center and the correlation matrix until an iteration threshold is met, and obtaining a plurality of content classifications.

3. The method according to claim 1, wherein the calculating a first probability that the target user watches each content to be recommended according to the preference degree of the target user for each content category and the similarity between the content to be recommended and the watched content of the target user comprises:

determining each viewed content of a target user and a classification center of each content classification;

and calculating the preference degree of the target user to each content classification according to each watched content of the target user and the classification center of each content classification.

4. The method according to any one of claims 1 to 3, wherein calculating the first probability that the target user watches each of the contents to be recommended according to the preference degree of the target user for each of the content categories and the similarity between the contents to be recommended and the watched contents of the target user comprises:

determining a first feature vector of viewed content of the target user;

determining a second feature vector of the content to be recommended;

and determining the similarity between the content to be recommended and the watched content of the target user according to the first feature vector and the second feature vector.

5. The method according to any one of claims 1 to 3, wherein the calculating of the second probability that each content to be recommended is recommended to the target user according to the comment information of the user category to which the target user belongs for the content to be recommended comprises:

acquiring the watching information of all users;

obtaining a feature vector of each user according to the viewing information and the word vector model;

and clustering according to the characteristic vectors to obtain the user category to which each user belongs.

6. The method according to claim 5, wherein the calculating a second probability that each content to be recommended is recommended to the target user according to comment information of the content to be recommended of the user category to which the target user belongs comprises:

according to the comment information of the user category to which the target user belongs to the content to be recommended, determining the emotion classification of other users in the user category to which the target user belongs to the content to be recommended;

calculating Euclidean distances between the target user and other users in the user category to which the target user belongs;

and calculating a second probability that each content to be recommended is recommended to the user according to the Euclidean distance and the emotion classification.

7. The method according to claim 6, wherein the determining the emotional classification of the content to be recommended by other users in the user category to which the target user belongs according to the comment information of the user category to which the target user belongs to the content to be recommended comprises:

obtaining comment information of the user category to which the target user belongs to the content to be recommended;

inputting the comment information into an emotion classification model to obtain emotion classification of other users in the user category to which the target user belongs on the content to be recommended; the emotion classification model is obtained by training according to emotion classification samples in advance.

8. An information pushing apparatus, characterized in that the apparatus comprises:

the content classification module is used for classifying the watched contents of all the users to obtain a plurality of content classifications;

the first probability calculation module is used for calculating first probabilities of the target users for watching the contents to be recommended according to the preference degrees of the target users for the contents in each category and the similarity between the contents to be recommended and the watched contents of the target users; the content to be recommended is any one of the content categories except the content watched by the target user;

the second probability calculation module is used for calculating a second probability that each content to be recommended is recommended to the target user according to comment information of the user category to which the target user belongs on the content to be recommended;

the popularity scoring module is used for obtaining popularity scores of the contents to be recommended according to the first probability and the second probability;

and the recommending module is used for recommending the content to be recommended to the target user according to the popularity score.

9. A computer device, comprising: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, and the executable instruction causes the processor to execute the operation of the information pushing method according to any one of claims 1-7.

10. A computer-readable storage medium, having at least one executable instruction stored therein, which when executed on a computer device, causes the computer device to perform the operations of the information push method according to any one of claims 1 to 7.

Images