CN113674008A - Directional label recommendation method, device, server and storage medium - Google Patents

Directional label recommendation method, device, server and storage medium Download PDF

Info

Publication number
CN113674008A
CN113674008A CN202010406273.1A CN202010406273A CN113674008A CN 113674008 A CN113674008 A CN 113674008A CN 202010406273 A CN202010406273 A CN 202010406273A CN 113674008 A CN113674008 A CN 113674008A
Authority
CN
China
Prior art keywords
account
user
tag
feature vector
sample
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010406273.1A
Other languages
Chinese (zh)
Inventor
何攀
高小平
秦烁
徐奇
王建明
郑秋野
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Dajia Internet Information Technology Co Ltd
Original Assignee
Beijing Dajia Internet Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Dajia Internet Information Technology Co Ltd filed Critical Beijing Dajia Internet Information Technology Co Ltd
Priority to CN202010406273.1A priority Critical patent/CN113674008A/en
Publication of CN113674008A publication Critical patent/CN113674008A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0242—Determining effectiveness of advertisements
    • G06Q30/0244—Optimization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0251—Targeted advertisements

Landscapes

  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Finance (AREA)
  • Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The disclosure relates to a directional tag recommendation method, a directional tag recommendation device, an electronic device and a storage medium. The method comprises the following steps: acquiring account characteristic vectors of content publishing accounts and label characteristic vectors corresponding to user labels; the account feature vector is obtained by fusing multi-attribute features based on the content publishing account; the label feature vector is obtained based on multi-attribute feature fusion of a user account corresponding to the user label; determining vector similarity between the account feature vector and each of the tag feature vectors; determining a target user label in each user label according to the vector similarity; and recommending the target user tag to the content publishing account so that the content publishing account directs the content to be published to the user account corresponding to the target user tag. By adopting the method, the advertiser can efficiently select the targeted crowd label suitable for the advertisement to be delivered.

Description

Directional label recommendation method, device, server and storage medium
Technical Field
The present disclosure relates to the field of internet technologies, and in particular, to a method and an apparatus for recommending a directional tag, a server, and a storage medium.
Background
Internet advertisement refers to commercial advertisement that directly or indirectly promotes goods or services through internet media such as websites, web pages, internet applications, and the like. Compared with the traditional four-large-transmission media advertisement and the outdoor advertisement which is prepared with the blue-green characters, the internet advertisement has the unique advantage and is an important part for implementing the modern marketing media strategy.
In the prior art, in order to enable an advertisement to be delivered to a target user accurately, an advertiser needs to select a suitable targeted crowd label for the advertisement to be delivered on an advertisement data management platform according to an advertisement attribute of the advertisement to be delivered so as to accurately deliver the advertisement to be delivered to the target crowd.
However, the advertisement data management platform often provides a wide variety of crowd tags for the advertiser to select, which makes the advertiser unable to efficiently select a targeted crowd tag suitable for the advertisement to be delivered.
Disclosure of Invention
The present disclosure provides a method, an apparatus, a server, and a storage medium for recommending a targeted tag, so as to at least solve the problem in the related art that an advertiser cannot efficiently select a targeted crowd tag suitable for an advertisement to be delivered. The technical scheme of the disclosure is as follows:
according to a first aspect of the embodiments of the present disclosure, there is provided a directional tag recommendation method, including:
acquiring account characteristic vectors of content publishing accounts and label characteristic vectors corresponding to user labels; the account feature vector is obtained by fusing multi-attribute features based on the content publishing account; the label feature vector is obtained based on multi-attribute feature fusion of a user account corresponding to the user label;
determining vector similarity between the account feature vector and each of the tag feature vectors;
determining a target user label in each user label according to the vector similarity;
and recommending the target user tag to the content publishing account so that the content publishing account directs the content to be published to the user account corresponding to the target user tag.
In one possible implementation manner, the obtaining an account feature vector of a content publishing account includes:
acquiring the multi-attribute characteristics of the content publishing account; the multi-attribute feature of the content publishing account is a feature used for representing a plurality of user attribute information of the content publishing account;
and fusing the multi-attribute features of the content publishing account through a pre-trained feature vector generation model to generate an account feature vector of the content publishing account.
In a possible implementation manner, the obtaining of the tag feature vector corresponding to each user tag includes:
acquiring the multi-attribute characteristics of the user account corresponding to the user tag; the multi-attribute feature of the user account is a feature of a plurality of user attribute information used for characterizing the user account;
and fusing the multi-attribute features of the user account corresponding to the user label through a pre-trained feature vector generation model to generate a label group feature vector of the user label.
In one possible implementation, the method further includes:
acquiring training samples, wherein each training sample comprises a multi-attribute feature of a sample content release account, a multi-attribute feature of a user account corresponding to a sample user label, and a content release result between the sample content release account and the user account corresponding to the sample user label;
training a feature vector generation model to be trained based on the multi-attribute feature of the sample content publishing account, the multi-attribute feature of the user account corresponding to the sample user label and the content delivery result;
and when the trained feature vector generation model meets the preset training condition, obtaining the pre-trained feature vector generation model.
In a possible implementation manner, the training a feature vector generation model to be trained based on the multi-attribute feature of the sample content publishing account, the multi-attribute feature of the user account corresponding to the sample user tag, and the content delivery result includes:
based on the multi-attribute features of the sample content publishing account and the multi-attribute features of the user account corresponding to the sample user label, obtaining an account feature vector of the sample content publishing account and a label feature vector of the sample user label through the feature vector generation model to be trained;
determining a content delivery prediction result between the sample content publishing account and the user account corresponding to the sample user label according to the account feature vector of the sample content publishing account and the label feature vector of the sample user label;
determining a loss function of the feature vector generation model to be trained according to the difference between the content delivery prediction result and the content delivery result;
adjusting the model parameters of the feature vector generation model to be trained according to the loss function;
and retraining the feature vector generation model after the model parameters are adjusted until the trained feature vector generation model meets the preset training conditions.
In one possible implementation manner, the determining, according to the account feature vector of the sample content publishing account and the tag feature vector of the sample user tag, a content delivery prediction result between the sample content publishing account and the user account corresponding to the sample user tag includes:
obtaining a vector inner product between the account characteristic vector of the sample content publishing account and the label characteristic vector of the sample user label to obtain a vector inner product result;
and determining an activation function value corresponding to the vector inner product result through a preset activation function, and taking the activation function value as the content delivery prediction result.
In one possible implementation manner, the determining, according to the vector similarity, a target user tag in each of the user tags includes:
sequencing the plurality of user tags according to the vector similarity;
selecting N user tags from the sorted user tags as the target user tags, wherein the minimum value of the vector similarity of the selected target user tags is larger than the maximum value of the vector similarity of the unselected target user tags; wherein N is a positive integer greater than or equal to 1.
According to a second aspect of the embodiments of the present disclosure, there is provided a directional tag recommendation apparatus, including:
the acquisition unit is configured to acquire the account feature vector of the content publishing account and the tag feature vector corresponding to each user tag; the account feature vector is obtained by fusing multi-attribute features based on the content publishing account; the label feature vector is obtained based on multi-attribute feature fusion of a user account corresponding to the user label; (ii) a
A determining unit configured to perform determining a vector similarity between the account feature vector and each of the tag feature vectors;
a recall unit configured to perform determining a target user tag among the user tags according to the vector similarity;
and the recommending unit is configured to recommend the target user tag to the content publishing account so that the content publishing account directs the content to be published to the user account corresponding to the target user tag.
In a possible implementation manner, the obtaining unit is specifically configured to perform obtaining the multi-attribute feature of the content distribution account; the multi-attribute feature of the content publishing account is a feature used for representing a plurality of user attribute information of the content publishing account; and fusing the multi-attribute features of the content publishing account through a pre-trained feature vector generation model to generate an account feature vector of the content publishing account.
In a possible implementation manner, the obtaining unit is specifically configured to perform obtaining a multi-attribute feature of a user account corresponding to the user tag; the multi-attribute feature of the user account is a feature of a plurality of user attribute information used for characterizing the user account; and fusing the multi-attribute features of the user account corresponding to the user label through a pre-trained feature vector generation model to generate a label group feature vector of the user label.
In one possible implementation manner, the directional tag recommendation apparatus further includes:
the system comprises a sample obtaining unit and a sample obtaining unit, wherein the sample obtaining unit is configured to obtain training samples, and each training sample comprises a multi-attribute feature of a sample content publishing account, a multi-attribute feature of a user account corresponding to a sample user label, and a content release result between the sample content publishing account and the user account corresponding to the sample user label;
the training unit is configured to execute training of a feature vector generation model to be trained on the basis of the multi-attribute feature of the sample content publishing account, the multi-attribute feature of the user account corresponding to the sample user label and the content delivery result;
and the model determining unit is configured to execute the step of obtaining the pre-trained feature vector generation model when the trained feature vector generation model meets the preset training condition.
In a possible implementation manner, the training unit is specifically configured to execute an account feature vector of the sample content publishing account and a tag feature vector of the sample user tag by using the feature vector generation model to be trained based on the multi-attribute feature of the sample content publishing account and the multi-attribute feature of the user account corresponding to the sample user tag; determining a content delivery prediction result between the sample content publishing account and the user account corresponding to the sample user label according to the account feature vector of the sample content publishing account and the label feature vector of the sample user label; determining a loss function of the feature vector generation model to be trained according to the difference between the content delivery prediction result and the content delivery result; adjusting the model parameters of the feature vector generation model to be trained according to the loss function; and retraining the feature vector generation model after the model parameters are adjusted until the trained feature vector generation model meets the preset training conditions.
In a possible implementation manner, the training unit is specifically configured to perform obtaining of an inner vector product between an account feature vector of the sample content publishing account and a tag feature vector of the sample user tag, so as to obtain an inner vector product result; and determining an activation function value corresponding to the vector inner product result through a preset activation function, and taking the activation function value as the content delivery prediction result.
In a possible implementation manner, the recall unit is specifically configured to perform sorting of the plurality of user tags according to the vector similarity; selecting N user tags from the sorted user tags as the target user tags, wherein the minimum value of the vector similarity of the selected target user tags is larger than the maximum value of the vector similarity of the unselected target user tags; wherein N is a positive integer greater than or equal to 1.
According to a third aspect of the embodiments of the present disclosure, there is provided an electronic device, including a memory and a processor, where the memory stores a computer program, and the processor implements the directional tag recommendation method according to the first aspect or any one of the possible implementations of the first aspect when executing the computer program.
According to a fourth aspect of embodiments of the present disclosure, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, implements a directional tag recommendation method according to the first aspect or any one of the possible implementations of the first aspect.
According to a fifth aspect of embodiments of the present disclosure, there is provided a computer program product comprising a computer program stored in a readable storage medium, from which the at least one processor of the apparatus reads and executes the computer program, such that the apparatus performs the directional tag recommendation method described in any one of the embodiments of the first aspect.
The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects: the method comprises the steps of obtaining multi-attribute feature fusion based on a content publishing account; the method comprises the steps of calculating a plurality of label feature vectors corresponding to the user labels, recommending target user labels for targeted release of contents to be released for the content release accounts by determining the vector similarity between the account feature vectors and the label feature vectors based on the label feature vectors obtained by fusing the multi-attribute features of the user accounts corresponding to the user labels, and accordingly providing accurate guidance for the content release accounts such as advertisers in the selection process of various crowd labels, and facilitating the advertisers to efficiently select the targeted crowd labels suitable for the advertisements in the crowd labels provided by the advertisement data management platform.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure and are not to be construed as limiting the disclosure.
FIG. 1 is a diagram illustrating an application environment for a method of directional tag recommendation, according to an example embodiment.
FIG. 2 is a flow diagram illustrating a method for directional tag recommendation in accordance with an exemplary embodiment.
Fig. 3 is a network structure diagram illustrating a feature vector generation model according to an exemplary embodiment.
FIG. 4 is a recall diagram of a user tag shown in accordance with an exemplary embodiment.
FIG. 5 is a flow diagram illustrating a method for directional tag recommendation in accordance with an exemplary embodiment.
FIG. 6 is a block diagram illustrating a directional tag recommendation apparatus according to an example embodiment.
Fig. 7 is an internal block diagram of a server according to an example embodiment.
Detailed Description
In order to make the technical solutions of the present disclosure better understood by those of ordinary skill in the art, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.
It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein are capable of operation in sequences other than those illustrated or otherwise described herein. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
The directional tag recommendation method provided by the present disclosure can be applied to the application environment shown in fig. 1. Wherein the tag recommendation server 110 communicates with the client 120 through a network. The tag recommendation server 110 obtains an account feature vector of a content distribution account on the client 120 and tag feature vectors corresponding to tags of each user; the account feature vector is obtained by fusing multi-attribute features based on the content publishing account; the label feature vector is obtained based on fusion of multi-attribute features of a user account corresponding to the user label; the tag recommendation server 110 determines the vector similarity between the account feature vector and each tag feature vector; the tag recommendation server 110 determines a target user tag in each user tag according to the vector similarity; the tag recommendation server 110 recommends the target user tag to the content publishing account on the client 120, so that the content publishing account directs the content to be published to the user account corresponding to the target user tag. In practical applications, the tag recommendation server 110 may be implemented by a separate server or a server cluster composed of a plurality of servers. The client 120 may be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices
Fig. 2 is a flowchart illustrating a directional tag recommendation method according to an exemplary embodiment, which is used in the tag recommendation server 110 of fig. 1, as shown in fig. 2, and includes the following steps.
In step S210, an account feature vector of the content distribution account and tag feature vectors corresponding to the user tags are obtained; the account feature vector is obtained by fusing multi-attribute features based on the content publishing account; the label feature vector is obtained based on fusion of multi-attribute features of the user account corresponding to the user label.
The content publishing account may refer to an account for publishing content to be published.
The content to be published can be multimedia content such as advertisements, videos, pictures, music, articles and the like. For example, when the content to be published is an advertisement, the content publishing account may also be named an advertiser account.
The account feature vector may refer to a feature vector for characterizing a content distribution account. In practical application, the account feature vector can be obtained based on fusion of multi-attribute features of the content publishing account. Specifically, for example, the content publishing account is an advertiser account, the account feature vector corresponding to the advertiser account may be obtained by performing fusion processing on a plurality of advertiser attribute features of the advertiser, such as industry information of the advertiser, tag usage history information (e.g., frequency of use of a crowd tag, overlap ratio of use of the historical crowd tag), advertisement information to be published of the advertiser (corresponding to an advertisement creative, a creative title, a creative tag, and the like).
The user tag may be a tag used for the content publishing account to select so that the content platform can directionally promote the content to be published of the content publishing account, for example, a crowd tag.
The tag feature vector may refer to a feature vector for characterizing a tag of a user. In practical application, the account feature vector may be obtained by fusing multi-attribute features of the user account corresponding to the user tag. Specifically, the account feature vector of the user tag may be obtained by fusing a plurality of user attribute features, such as gender, age, region, business interest, and the like, of the user corresponding to the user tag.
In a specific implementation, taking a content to be published as an advertisement as an example, when an advertiser needs to target the advertisement to be published to a corresponding target user, the advertiser needs to select a corresponding crowd tag for the advertisement to be published, and at this time, the tag recommendation server 110 obtains an account feature vector of a content publishing account, that is, an ad embedding of an account feature vector of an advertiser account, and tag embedding of a tag feature vector corresponding to each user tag. In practical applications, the account feature vector and the tag feature vector may be 64-dimensional vectors.
In step S220, a vector similarity between the account feature vector and each tag feature vector is determined.
In a specific implementation, after the tag recommendation server 110 obtains the account feature vector of the content distribution account and the tag feature vectors corresponding to the tags of the users, the tag recommendation server 110 may calculate the vector similarity between the account feature vector and each of the tag feature vectors. Specifically, the tag recommendation server 110 may use a neighborhood search engine to calculate cosine similarity between the account feature vector and each tag feature vector on line, further determine a corresponding cosine similarity distance, and use the cosine similarity distance to represent the vector similarity between the account feature vector and the tag feature vector.
In practical applications, the vector Similarity between the account feature vector a and the tag feature vector B may be expressed as:
Figure BDA0002491427590000081
wherein the content of the first and second substances,Aiis each component of the account feature vector a; b isiAre the components of the label feature vector B. In practice, n may be equal to 64.
In step S230, a target user tag is determined among the user tags according to the vector similarity.
In a specific implementation, the tag recommendation server 110 may recall tags according to the vector similarity corresponding to each user tag, and determine a target user tag in each user tag.
In step S240, the target user tag is recommended to the content publishing account, so that the content publishing account directs the content to be published to the user account corresponding to the target user tag.
In a specific implementation, when the content to be published is an advertisement to be published, after the tag recommendation server 110 determines that a target user tag is determined in each user tag, the tag recommendation server 110 takes the target user tag as a tag recommendation result, and sends the tag recommendation result to a user terminal of an advertiser account, so that the user terminal displays the tag recommendation result, an advertiser can select the target user tag, a corresponding crowd tag is printed on the advertisement to be published, and the advertisement platform can accurately deliver the advertisement to be published of the advertiser to a target client corresponding to the advertiser according to the crowd tag, thereby realizing accurate targeted delivery of the advertisement.
In the directional tag recommendation method, the directional tag recommendation method is obtained by acquiring multi-attribute feature fusion based on a content release account; the method comprises the steps of calculating a plurality of label feature vectors corresponding to the user labels, recommending target user labels for targeted release of contents to be released for the content release accounts by determining the vector similarity between the account feature vectors and the label feature vectors based on the label feature vectors obtained by fusing the multi-attribute features of the user accounts corresponding to the user labels, and accordingly providing accurate guidance for the content release accounts such as advertisers in the selection process of various crowd labels, and facilitating the advertisers to efficiently select the targeted crowd labels suitable for the advertisements in the crowd labels provided by the advertisement data management platform.
In an exemplary embodiment, obtaining an account feature vector for a content publication account includes: acquiring the multi-attribute characteristics of a content publishing account; the multi-attribute characteristic of the content publishing account is a characteristic used for representing a plurality of user attribute information of the content publishing account; and fusing the multi-attribute features of the content publishing account through a pre-trained feature vector generation model to generate an account feature vector of the content publishing account.
The multi-attribute feature of the content publishing account is a feature of a plurality of user attribute information used for representing the content publishing account. Specifically, for example, the content publishing account is an advertiser account, and the multi-attribute feature of the advertiser account may refer to features of a plurality of advertiser attribute information such as industry information of the advertiser, tag usage history information (e.g., frequency of usage of the crowd tag, degree of overlap of usage of the historical crowd tag), advertisement information to be published of the advertiser (corresponding to an advertisement creative, a creative title, a creative tag, and the like).
In a specific implementation, taking the content publishing account as an advertiser account as an example, the tag recommendation server 110 may specifically include, in the process of obtaining the account feature vector of the content publishing account: the tag recommendation server 110 may first process, into discrete variables, characteristics corresponding to a plurality of advertiser attribute information, such as industry information of an advertiser, tag usage history information (e.g., frequency of usage of a crowd tag, overlap ratio of usage of a historical crowd tag), advertisement information to be published of an advertiser (corresponding to an advertisement creative, a creative title, a creative tag, etc.), and the like. Then, expressing the discrete features by using an initialized random vector as the multi-attribute features of the content publishing account; then, inputting the multi-attribute features of the content publishing account into the pre-trained feature vector generation model, fusing the multi-attribute features of the content publishing account through the pre-trained feature vector generation model, and generating the account feature vector of the content publishing account, namely adv embedding.
According to the technical scheme of the embodiment, the multi-attribute characteristics of a plurality of user attribute information used for representing the content publishing account are obtained; and the multi-attribute characteristics of the content publishing account are fused through a pre-trained characteristic vector generation model, so that an account characteristic vector capable of accurately expressing the corresponding characteristics of the content publishing account is generated.
In an exemplary embodiment, obtaining the tag feature vector corresponding to each user tag includes: acquiring multi-attribute characteristics of a user account corresponding to a user tag; the multi-attribute feature of the user account is a feature of a plurality of user attribute information used for representing the user account; and fusing the multi-attribute features of the user account corresponding to the user label through the pre-trained feature vector generation model to generate a label group feature vector of the user label.
The multi-attribute feature of the user account is a feature of a plurality of user attribute information used for characterizing the user account. Specifically, the multi-attribute feature of the user account may refer to features of a plurality of user attribute information, such as gender, age, region, business interest, and the like, of the user corresponding to the user tag.
In a specific implementation, in the process of obtaining the tag feature vector corresponding to each user tag, the tag recommendation server 110 may specifically include: the tag recommendation server 110 may first calculate, by using the offline crowd insight imaging system, characteristics corresponding to a plurality of user attribute information, such as gender, age, region, business interest, and the like, of a user corresponding to a user tag; the above features are then processed into discrete variables. Then, expressing the discrete features by using an initialized random vector as the multi-attribute features of the user account corresponding to the user label; then, inputting the multi-attribute features of the user account corresponding to the user label into a pre-trained feature vector generation model, fusing the multi-attribute features of the user account corresponding to the user label through the pre-trained feature vector generation model, and generating a tag embedding feature vector of the user label.
According to the technical scheme of the embodiment, the multi-attribute characteristics of a plurality of user attribute information used for representing the content publishing account are obtained; and the multi-attribute characteristics of the content publishing account are fused through a pre-trained characteristic vector generation model, so that an account characteristic vector capable of accurately expressing the corresponding characteristics of the content publishing account is generated.
In an exemplary embodiment, further comprising: acquiring training samples, wherein each training sample comprises a multi-attribute feature of a sample content publishing account, a multi-attribute feature of a user account corresponding to a sample user label, and a content release result between the sample content publishing account and the user account corresponding to the sample user label; training a feature vector generation model to be trained based on the multi-attribute features of the sample content publishing account, the multi-attribute features of the user account corresponding to the sample user label and the content delivery result; and when the trained feature vector generation model meets the preset training condition, obtaining a pre-trained feature vector generation model.
In a specific implementation, before the tag recommendation server 110 obtains the account feature vector of the content distribution account and the tag feature vector corresponding to each user tag, the tag recommendation server 110 further needs to train the feature vector generation model, which may specifically include: the label recommendation server 110 obtains training samples; each training sample comprises the multi-attribute characteristics of the sample content publishing account, the multi-attribute characteristics of the user account corresponding to the sample user label, and the content release result between the sample content publishing account and the user account corresponding to the sample user label. Specifically, for example, when the content publishing account is the advertiser account, the tag recommendation server 110 may select some advertisement plans and advertisement creatives with good delivery effects (e.g., CTR (Click-Through-Rate) and CVR (Conversion Rate) are high) as positive samples, randomly select some advertisement plans and advertisement creatives as negative samples, perform feature extraction processing, and determine the multi-attribute features of the sample content publishing account, the multi-attribute features of the user accounts corresponding to the sample user tags, and the content delivery results between the sample content publishing account and the user accounts corresponding to the sample user tags.
In practical application, the content delivery result may refer to whether the sample user tag is selected when the advertisement is targeted by the sample content publishing account; if yes, recording the content delivery result as 1; if yes, the content delivery result may be recorded as 0.
Then, the tag recommendation server 110 may train the feature vector generation model to be trained based on the multi-attribute feature of the sample content publishing account, the multi-attribute feature of the user account corresponding to the sample user tag, and the content delivery result; and when the trained feature vector generation model meets the preset training condition, obtaining a pre-trained feature vector generation model.
In practical application, the preset training condition may be that the number of times of training the model satisfies a preset number threshold, the loss of the feature vector generation model is less than a preset threshold, and the like.
According to the technical scheme of the embodiment, the feature vector generation model to be trained can be accurately trained based on the multi-attribute features of the sample content publishing account, the multi-attribute features of the user account corresponding to the sample user tags and the content delivery result, so that the account feature vector of the content publishing account and the tag feature vector corresponding to each user tag can be accurately generated by the pre-trained feature vector generation model.
In an exemplary embodiment, training the feature vector generation model to be trained based on the multi-attribute features of the sample content publishing account, the multi-attribute features of the user account corresponding to the sample user tag, and the content delivery result includes: based on the multi-attribute features of the sample content publishing account and the multi-attribute features of the user account corresponding to the sample user label, generating a model through the feature vector to be trained to obtain an account feature vector of the sample content publishing account and a label feature vector of the sample user label; determining a content delivery prediction result between the sample content publishing account and the user account corresponding to the sample user label according to the account characteristic vector of the sample content publishing account and the label characteristic vector of the sample user label; determining a loss function of the feature vector generation model to be trained according to the difference between the content delivery prediction result and the content delivery result; according to the loss function, adjusting the model parameters of the feature vector generation model to be trained; and retraining the feature vector generation model after the model parameters are adjusted until the trained feature vector generation model meets the preset training conditions.
In a specific implementation, in the process of training the feature vector generation model to be trained based on the multi-attribute feature of the sample content publishing account, the multi-attribute feature of the user account corresponding to the sample user tag, and the content delivery result, the tag recommendation server 110 may generate the model based on the multi-attribute feature of the sample content publishing account and the multi-attribute feature of the user account corresponding to the sample user tag, and obtain the account feature vector of the sample content publishing account and the tag feature vector of the sample user tag through the feature vector to be trained. Specifically, the feature vector generation model to be trained may fuse the multi-attribute features of the sample content publishing account to generate an account feature vector of the sample content publishing account. The feature vector generation model to be trained can fuse the multi-attribute features of the user accounts corresponding to the sample user tags to generate tag feature vectors of the sample user tags.
Then, the tag recommendation server 110 determines a content delivery prediction result between the sample content publishing account and the user account corresponding to the sample user tag according to the account feature vector of the sample content publishing account and the tag feature vector of the sample user tag; i.e., the predicted probability value that the sample content publication account selects the sample user label when targeting the advertisement.
And determining a Log Loss function of the feature vector generation model to be trained at the label recommendation server 110 according to the difference between the content delivery prediction result and the content delivery result. Wherein the loss function ltCan be expressed as:
lt(wt)＝-ytlogpt-(1-yt)log(1-pt)
wherein, wtIs a parameter of the model; p is a radical oftDelivering a prediction result for the content; y istE {0,1} is the label of the content delivery result, i.e., the sample.
Finally, the label recommendation server 110 solves the gradient of the loss function according to the loss function by adopting a random gradient descent method, and then adjusts the model parameters of the feature vector generation model to be trained layer by layer; and the feature vector generation model after the model parameters are adjusted is retrained until the trained feature vector generation model meets the preset training conditions.
According to the technical scheme of the embodiment, the loss function of the feature vector generation model to be trained is determined according to the difference between the content delivery prediction result and the content delivery result obtained through calculation, so that the model parameters of the feature vector generation model to be trained can be accurately adjusted, the pre-trained feature vector generation model obtained through training can accurately generate the account feature vector of the content publishing account and the tag feature vector corresponding to each user tag.
To facilitate understanding of those skilled in the art, fig. 3 exemplarily provides a network structure diagram of a feature vector generation model; as shown in fig. 3, wherein the feature vector generation model may include a first subnetwork 310 and a second subnetwork 320; wherein, the hidden layers of the first sub-network 310 and the second sub-network 320 are three layers; the first sub-network 310 is configured to fuse multi-attribute features of the input content distribution accounts, for example, advertiser-related features, and generate an account feature vector of the content distribution accounts. The second sub-network 320 is configured to fuse multi-attribute features of the user account corresponding to the input user tag, for example, relevant features of the crowd tag, and generate a tag group feature vector of the user tag.
In an exemplary embodiment, determining a content delivery prediction result between a sample content publishing account and a user account corresponding to a sample user tag according to an account feature vector of the sample content publishing account and a tag feature vector of the sample user tag includes: obtaining a vector inner product between an account characteristic vector of a sample content publishing account and a label characteristic vector of a sample user label to obtain a vector inner product result; and determining an activation function value corresponding to the vector inner product result through a preset activation function, and using the activation function value as a content delivery prediction result.
In a specific implementation, in the process of determining the content delivery prediction result between the sample content publishing account and the user account corresponding to the sample user tag according to the account feature vector of the sample content publishing account and the tag feature vector of the sample user tag, the tag recommendation server 110 may obtain a vector inner product between the account feature vector of the sample content publishing account and the tag feature vector of the sample user tag to obtain a vector inner product result. Then, the tag recommendation server 110 calculates an activation function value corresponding to the product result in the vector through a preset activation function, and the activation function value is used as a content delivery prediction result. In practical applications, the content delivery prediction result may refer to a tag recommendation probability. The activation function may be a sigmoid function.
In practical applications, the vector inner product result distance can be expressed as:
Figure BDA0002491427590000121
wherein, CiIssuing each component of an account feature vector C of the account for the sample content; diThe components of the label feature vector D for the sample user label.
According to the technical scheme of the embodiment, a vector inner product result is obtained by obtaining a vector inner product between an account characteristic vector of a sample content publishing account and a label characteristic vector of a sample user label; and determining an activation function value corresponding to the vector inner product result through a preset activation function, and accurately calculating a content delivery prediction result.
In an exemplary embodiment, determining a target user tag among the user tags according to the vector similarity includes: sequencing the plurality of user tags according to the vector similarity; and selecting N user tags from the sorted user tags as target user tags, wherein the minimum value of the vector similarity of the selected target user tags is larger than the maximum value of the vector similarity of the unselected target user tags.
Wherein N is a positive integer greater than or equal to 1.
In a specific implementation, in the process that the tag recommendation server 110 determines the target user tag in each user tag according to the vector similarity, the tag recommendation server 110 may sort the plurality of user tags according to the vector similarity to obtain a plurality of sorted user tags; then, the tag recommendation server 110 selects N user tags as target user tags from the sorted user tags. The minimum value of the vector similarity of the selected target user label is larger than the maximum value of the vector similarity of the unselected target user labels. Specifically, the tag recommendation server 110 may rank the plurality of user tags in order of the vector similarity from large to small; then, the tag recommendation server 110 selects the top N user tags as target user tags from the sorted user tags. Wherein N may be 10.
To facilitate understanding by those skilled in the art, FIG. 4 provides a recall schematic of a user tag.
According to the technical scheme of the embodiment, a plurality of user tags are sequenced according to the vector similarity; and selecting N user tags as target user tags from the sorted user tags, so that the target user tags can be selectively and accurately recommended to the content publishing account.
Fig. 5 is a flowchart illustrating a directional tag recommendation method according to an exemplary embodiment, which is used in the tag recommendation server 110 of fig. 1, as shown in fig. 5, and includes the following steps. Step S510, acquiring the multi-attribute characteristics of the content publishing account; the multi-attribute feature of the content publishing account is a feature of a plurality of user attribute information used for characterizing the content publishing account. Step S520, fusing the multi-attribute features of the content publishing account through a pre-trained feature vector generation model to generate an account feature vector of the content publishing account. Step S530, acquiring the multi-attribute characteristics of the user account corresponding to the user tag; the multi-attribute feature of the user account is a feature of a plurality of user attribute information used for characterizing the user account. And S540, fusing the multi-attribute features of the user account corresponding to the user label through a pre-trained feature vector generation model to generate a label group feature vector of the user label. Step S550, determining vector similarity between the account feature vector and each of the tag feature vectors. And step S560, sequencing the plurality of user tags according to the vector similarity. Step S570, selecting N user tags from the sorted user tags as the target user tags, wherein the minimum value of the vector similarity of the selected target user tags is larger than the maximum value of the vector similarity of the unselected target user tags; wherein N is a positive integer greater than or equal to 1. Step S580, recommending the target user tag to the content publishing account, so that the content publishing account directs the content to be published to the user account corresponding to the target user tag. The specific limitations of the above steps may refer to the upper specific limitations on a directional tag recommendation method, which are not described herein again.
It should be understood that although the steps in the flowcharts of fig. 2 and 5 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2 and 5 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed in turn or alternately with other steps or at least some of the other steps.
FIG. 6 is a block diagram illustrating a directional tag recommendation apparatus according to an example embodiment. Referring to fig. X, the apparatus includes:
the acquiring unit 610 is configured to perform acquiring account feature vectors of the content distribution accounts and tag feature vectors corresponding to the user tags; the account feature vector is obtained by fusing multi-attribute features based on the content publishing account; the label feature vector is obtained based on multi-attribute feature fusion of a user account corresponding to the user label;
a determining unit 620 configured to perform determining a vector similarity between the account feature vector and each of the tag feature vectors;
a recall unit 630 configured to perform determining a target user tag among the user tags according to the vector similarity;
the recommending unit 640 is configured to recommend the target user tag to the content publishing account, so that the content publishing account directs the content to be published to the user account corresponding to the target user tag.
In an exemplary embodiment, the obtaining unit 610 is specifically configured to perform obtaining the multi-attribute feature of the content distribution account; the multi-attribute feature of the content publishing account is a feature used for representing a plurality of user attribute information of the content publishing account; and fusing the multi-attribute features of the content publishing account through a pre-trained feature vector generation model to generate an account feature vector of the content publishing account.
In an exemplary embodiment, the obtaining unit 610 is specifically configured to perform obtaining the multi-attribute feature of the user account corresponding to the user tag; the multi-attribute feature of the user account is a feature of a plurality of user attribute information used for characterizing the user account; and fusing the multi-attribute features of the user account corresponding to the user label through a pre-trained feature vector generation model to generate a label group feature vector of the user label.
In an exemplary embodiment, the directional tag recommendation apparatus further includes:
the system comprises a sample obtaining unit and a sample obtaining unit, wherein the sample obtaining unit is configured to obtain training samples, and each training sample comprises a multi-attribute feature of a sample content publishing account, a multi-attribute feature of a user account corresponding to a sample user label, and a content release result between the sample content publishing account and the user account corresponding to the sample user label;
the training unit is configured to execute training of a feature vector generation model to be trained on the basis of the multi-attribute feature of the sample content publishing account, the multi-attribute feature of the user account corresponding to the sample user label and the content delivery result;
and the model determining unit is configured to execute the step of obtaining the pre-trained feature vector generation model when the trained feature vector generation model meets the preset training condition.
In an exemplary embodiment, the training unit is specifically configured to execute generating a model through the feature vector to be trained based on the multi-attribute feature of the sample content publishing account and the multi-attribute feature of the user account corresponding to the sample user tag to obtain an account feature vector of the sample content publishing account and a tag feature vector of the sample user tag; determining a content delivery prediction result between the sample content publishing account and the user account corresponding to the sample user label according to the account feature vector of the sample content publishing account and the label feature vector of the sample user label; determining a loss function of the feature vector generation model to be trained according to the difference between the content delivery prediction result and the content delivery result; adjusting the model parameters of the feature vector generation model to be trained according to the loss function; and retraining the feature vector generation model after the model parameters are adjusted until the trained feature vector generation model meets the preset training conditions.
In an exemplary embodiment, the training unit is specifically configured to perform obtaining of an inner vector product between an account feature vector of the sample content publishing account and a tag feature vector of the sample user tag, to obtain an inner vector product result; and determining an activation function value corresponding to the vector inner product result through a preset activation function, and taking the activation function value as the content delivery prediction result.
In an exemplary embodiment, the recalling unit 630 is specifically configured to perform sorting the plurality of user tags according to the vector similarity; selecting N user tags from the sorted user tags as the target user tags, wherein the minimum value of the vector similarity of the selected target user tags is larger than the maximum value of the vector similarity of the unselected target user tags; wherein N is a positive integer greater than or equal to 1.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
Fig. 7 is a block diagram illustrating an apparatus 700 for performing a directed tag recommendation method according to an example embodiment. For example, device 700 may be a server. Referring to fig. 7, device 700 includes a processing component 720 that further includes one or more processors, and memory resources, represented by memory 722, for storing instructions, such as applications, that are executable by processing component 720. The application programs stored in memory 722 may include one or more modules that each correspond to a set of instructions. Further, the processing component 720 is configured to execute instructions to perform the directional tag recommendation method described above.
The device 700 may also include a power component 724 configured to perform power management for the device 700, a wired or wireless network interface 726 configured to connect the device 700 to a network, and an input/output (I/O) interface 728. The device 700 may operate based on an operating system stored in memory 722, such as Window77 over, MacO7XTM, UnixTM, LinuxTM, FreeB7DTM, or the like.
In an exemplary embodiment, a storage medium comprising instructions, such as memory 722 comprising instructions, executable by a processor of device 700 to perform the above-described method is also provided. The storage medium may be a non-transitory computer readable storage medium, which may be, for example, a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims (10)

1. A method for recommending a directional tag, comprising:
acquiring account characteristic vectors of content publishing accounts and label characteristic vectors corresponding to user labels; the account feature vector is obtained by fusing multi-attribute features based on the content publishing account; the label feature vector is obtained based on multi-attribute feature fusion of a user account corresponding to the user label;
determining vector similarity between the account feature vector and each of the tag feature vectors;
determining a target user label in each user label according to the vector similarity;
and recommending the target user tag to the content publishing account so that the content publishing account directs the content to be published to the user account corresponding to the target user tag.
2. The method of claim 1, wherein the obtaining the account feature vector of the content distribution account comprises:
acquiring the multi-attribute characteristics of the content publishing account; the multi-attribute feature of the content publishing account is a feature used for representing a plurality of user attribute information of the content publishing account;
and fusing the multi-attribute features of the content publishing account through a pre-trained feature vector generation model to generate an account feature vector of the content publishing account.
3. The method of claim 1, wherein the obtaining of the tag feature vector corresponding to each user tag comprises:
acquiring the multi-attribute characteristics of the user account corresponding to the user tag; the multi-attribute feature of the user account is a feature of a plurality of user attribute information used for characterizing the user account;
and fusing the multi-attribute features of the user account corresponding to the user label through a pre-trained feature vector generation model to generate a label group feature vector of the user label.
4. A directional tag recommendation method according to any one of claims 2 or 3, further comprising:
acquiring training samples, wherein each training sample comprises a multi-attribute feature of a sample content release account, a multi-attribute feature of a user account corresponding to a sample user label, and a content release result between the sample content release account and the user account corresponding to the sample user label;
training a feature vector generation model to be trained based on the multi-attribute feature of the sample content publishing account, the multi-attribute feature of the user account corresponding to the sample user label and the content delivery result;
and when the trained feature vector generation model meets the preset training condition, obtaining the pre-trained feature vector generation model.
5. The method according to claim 4, wherein the training of the feature vector generation model to be trained based on the multi-attribute feature of the sample content publishing account, the multi-attribute feature of the user account corresponding to the sample user tag, and the content delivery result comprises:
based on the multi-attribute features of the sample content publishing account and the multi-attribute features of the user account corresponding to the sample user label, obtaining an account feature vector of the sample content publishing account and a label feature vector of the sample user label through the feature vector generation model to be trained;
determining a content delivery prediction result between the sample content publishing account and the user account corresponding to the sample user label according to the account feature vector of the sample content publishing account and the label feature vector of the sample user label;
determining a loss function of the feature vector generation model to be trained according to the difference between the content delivery prediction result and the content delivery result;
adjusting the model parameters of the feature vector generation model to be trained according to the loss function;
and retraining the feature vector generation model after the model parameters are adjusted until the trained feature vector generation model meets the preset training conditions.
6. The method of claim 5, wherein the determining the content delivery prediction result between the sample content distribution account and the user account corresponding to the sample user tag according to the account feature vector of the sample content distribution account and the tag feature vector of the sample user tag comprises:
obtaining a vector inner product between the account characteristic vector of the sample content publishing account and the label characteristic vector of the sample user label to obtain a vector inner product result;
and determining an activation function value corresponding to the vector inner product result through a preset activation function, and taking the activation function value as the content delivery prediction result.
7. The method of claim 1, wherein the determining a target user tag among the user tags according to the vector similarity comprises:
sequencing the plurality of user tags according to the vector similarity;
selecting N user tags from the sorted user tags as the target user tags, wherein the minimum value of the vector similarity of the selected target user tags is larger than the maximum value of the vector similarity of the unselected target user tags; wherein N is a positive integer greater than or equal to 1.
8. A directional tag recommendation apparatus, comprising:
the acquisition unit is configured to acquire the account feature vector of the content publishing account and the tag feature vector corresponding to each user tag; the account feature vector is obtained by fusing multi-attribute features based on the content publishing account; the label feature vector is obtained based on multi-attribute feature fusion of a user account corresponding to the user label;
a determining unit configured to perform determining a vector similarity between the account feature vector and each of the tag feature vectors;
a recall unit configured to perform determining a target user tag among the user tags according to the vector similarity;
and the recommending unit is configured to recommend the target user tag to the content publishing account so that the content publishing account directs the content to be published to the user account corresponding to the target user tag.
9. A server, comprising:
a processor;
a memory for storing the processor-executable instructions;
wherein the processor is configured to execute the instructions to implement the directional tag recommendation method of any of claims 1-7.
10. A storage medium having instructions that, when executed by a processor of a server, enable the server to perform a directed tag recommendation method as claimed in any one of claims 1 to 7.
CN202010406273.1A 2020-05-14 2020-05-14 Directional label recommendation method, device, server and storage medium Pending CN113674008A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010406273.1A CN113674008A (en) 2020-05-14 2020-05-14 Directional label recommendation method, device, server and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010406273.1A CN113674008A (en) 2020-05-14 2020-05-14 Directional label recommendation method, device, server and storage medium

Publications (1)

Publication Number Publication Date
CN113674008A true CN113674008A (en) 2021-11-19

Family

ID=78537170

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010406273.1A Pending CN113674008A (en) 2020-05-14 2020-05-14 Directional label recommendation method, device, server and storage medium

Country Status (1)

Country Link
CN (1) CN113674008A (en)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109345302A (en) * 2018-09-27 2019-02-15 腾讯科技（深圳）有限公司 Machine learning model training method, device, storage medium and computer equipment
US20190163500A1 (en) * 2017-11-28 2019-05-30 Intuit Inc. Method and apparatus for providing personalized self-help experience
CN110162191A (en) * 2019-04-03 2019-08-23 腾讯科技（深圳）有限公司 A kind of expression recommended method, device and storage medium
CN110232152A (en) * 2019-05-27 2019-09-13 腾讯科技（深圳）有限公司 Content recommendation method, device, server and storage medium
CN110263265A (en) * 2019-04-10 2019-09-20 腾讯科技（深圳）有限公司 User tag generation method, device, storage medium and computer equipment
CN110309405A (en) * 2018-03-08 2019-10-08 腾讯科技（深圳）有限公司 A kind of item recommendation method, device and storage medium
CN110348907A (en) * 2019-07-12 2019-10-18 深圳市腾讯计算机系统有限公司 A kind of orientation method and device of advertisement crowd

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190163500A1 (en) * 2017-11-28 2019-05-30 Intuit Inc. Method and apparatus for providing personalized self-help experience
CN110309405A (en) * 2018-03-08 2019-10-08 腾讯科技（深圳）有限公司 A kind of item recommendation method, device and storage medium
CN109345302A (en) * 2018-09-27 2019-02-15 腾讯科技（深圳）有限公司 Machine learning model training method, device, storage medium and computer equipment
CN110162191A (en) * 2019-04-03 2019-08-23 腾讯科技（深圳）有限公司 A kind of expression recommended method, device and storage medium
CN110263265A (en) * 2019-04-10 2019-09-20 腾讯科技（深圳）有限公司 User tag generation method, device, storage medium and computer equipment
CN110232152A (en) * 2019-05-27 2019-09-13 腾讯科技（深圳）有限公司 Content recommendation method, device, server and storage medium
CN110348907A (en) * 2019-07-12 2019-10-18 深圳市腾讯计算机系统有限公司 A kind of orientation method and device of advertisement crowd

Similar Documents

Publication Publication Date Title
US10846617B2 (en) Context-aware recommendation system for analysts
US10755303B2 (en) Generating actionable suggestions for improving user engagement with online advertisements
CN111242748B (en) Method, apparatus, and storage medium for recommending items to a user
JP7130560B2 (en) Optimizing dynamic creatives to deliver content effectively
US20110082824A1 (en) Method for selecting an optimal classification protocol for classifying one or more targets
US20130263181A1 (en) Systems and methods for defining video advertising channels
CN108509457A (en) A kind of recommendation method and apparatus of video data
US20160283481A1 (en) Method and apparatus for combining text search and recommendation engines
KR102451099B1 (en) Purchsase inducement system and method through inferencing user's needs based on artificial intelligence
CN110209931B (en) Media content pushing method and device, storage medium and electronic device
CN111008335B (en) Information processing method, device, equipment and storage medium
US10380209B2 (en) Systems and methods of providing recommendations of content items
CN113379449B (en) Multimedia resource recall method and device, electronic equipment and storage medium
US11188579B2 (en) Personalized dynamic content via content tagging and transfer learning
CN113742567A (en) Multimedia resource recommendation method and device, electronic equipment and storage medium
US11615444B2 (en) Recommending that an entity in an online system create content describing an item associated with a topic having at least a threshold value of a performance metric and to add a tag describing the item to the content
CN111967924A (en) Commodity recommendation method, commodity recommendation device, computer device, and medium
US20190050890A1 (en) Video dotting placement analysis system, analysis method and storage medium
CN111552835A (en) File recommendation method and device and server
US20180239790A1 (en) Provision device, provision method and non-transitory computer readable storage medium
Allaymoun et al. Business intelligence model to analyze social network advertising
US20230316106A1 (en) Method and apparatus for training content recommendation model, device, and storage medium
CN112115354A (en) Information processing method, information processing apparatus, server, and storage medium
WO2014093621A2 (en) Proposing objects to a user to efficiently discover demographics from item ratings
CN110717094A (en) Information recommendation method and device, computer equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination