CN112000821B - Multimedia information pushing method, device, server and storage medium - Google Patents

Multimedia information pushing method, device, server and storage medium Download PDF

Info

Publication number
CN112000821B
CN112000821B CN202010848523.7A CN202010848523A CN112000821B CN 112000821 B CN112000821 B CN 112000821B CN 202010848523 A CN202010848523 A CN 202010848523A CN 112000821 B CN112000821 B CN 112000821B
Authority
CN
China
Prior art keywords
information
multimedia information
ordering
account
model
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.)
Active
Application number
CN202010848523.7A
Other languages
Chinese (zh)
Other versions
CN112000821A (en
Inventor
张志伟
林靖
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 CN202010848523.7A priority Critical patent/CN112000821B/en
Publication of CN112000821A publication Critical patent/CN112000821A/en
Application granted granted Critical
Publication of CN112000821B publication Critical patent/CN112000821B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
    • G06F16/43—Querying
    • G06F16/438—Presentation of query results
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
    • G06F16/43—Querying
    • G06F16/435—Filtering based on additional data, e.g. user or group profiles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
    • G06F16/44—Browsing; Visualisation therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Databases & Information Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The disclosure relates to a multimedia information pushing method, a device, a server and a storage medium, wherein the method comprises the following steps: acquiring candidate multimedia information corresponding to an account to be pushed; obtaining sequencing labels of each candidate multimedia information through a target sequencing model; the target ordering model is obtained by fusing tree models selected from at least two information ordering models and is used for outputting ordering labels of the multimedia information; according to the sorting labels, sorting the candidate multimedia information to obtain sorted candidate multimedia information; pushing the sorted candidate multimedia information to an account to be pushed. By adopting the method, the target ordering model for ordering the candidate multimedia information is obtained by fusing the tree models selected from at least two information ordering models, and the redundancy of the model structure does not occur, so that the overall operation efficiency of the target ordering model is higher, and the pushing efficiency of the multimedia information is further improved.

Description

Multimedia information pushing method, device, server and storage medium
Technical Field
The disclosure relates to the field of computer technology, and in particular, to a method, a device, a server and a storage medium for pushing multimedia information.
Background
With the development of computer technology, various applications for browsing multimedia information are endless, and more users choose to browse multimedia information, such as video, pictures, etc., through the applications; in order to realize accurate pushing of multimedia information, the information obtained by recall generally needs to be sequenced.
In the related art, in order to integrate the advantages of different information sorting models in information sorting, the present multimedia information pushing method generally trains to obtain another information sorting model (the model structure of the information sorting model includes the model structure of the trained information sorting model) on the basis of one information sorting model, continues to train to obtain another information sorting model on the basis of the information sorting model obtained by training, and so on until the last information sorting model is obtained by training, and uses the last information sorting model as a final fusion model, sorts the multimedia information through the fusion model, and pushes the sorted multimedia information to a corresponding account; however, if the number of the information ordering models is large, the model structure of the fusion model obtained by gradual accumulation training is more and more complex, redundancy is easy to occur, and the ordering processing time of the multimedia information is long, so that the pushing efficiency of the multimedia information is low.
Disclosure of Invention
The disclosure provides a method, a device, a server and a storage medium for pushing multimedia information, so as to at least solve the problem of low pushing efficiency of the multimedia information in the related technology. The technical scheme of the present disclosure is as follows:
according to a first aspect of an embodiment of the present disclosure, there is provided a multimedia information pushing method, including:
acquiring candidate multimedia information corresponding to an account to be pushed;
obtaining the sequencing labels of the candidate multimedia information through a target sequencing model; the target sorting models are obtained by fusing tree models selected from at least two information sorting models, sorting labels used for outputting multimedia information are obtained, and the number of the tree models selected from each information sorting model is determined according to the information sorting accuracy of each information sorting model;
according to the sorting labels, sorting the candidate multimedia information to obtain sorted candidate multimedia information;
pushing the sequenced candidate multimedia information to the account to be pushed.
In an exemplary embodiment, the obtaining candidate multimedia information corresponding to the account to be pushed includes:
Receiving an information pushing request of the account to be pushed; the information pushing request comprises key information input by triggering of the account to be pushed;
and recalling the multimedia information with the information matching degree between the key information and the information base higher than the preset matching degree as candidate multimedia information corresponding to the account to be pushed.
In an exemplary embodiment, the obtaining candidate multimedia information corresponding to the account to be pushed further includes:
acquiring account characteristics of an account to be pushed;
acquiring a multimedia information category conforming to the account characteristics;
and screening out the multimedia information which accords with the account characteristics from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as candidate multimedia information corresponding to the account to be pushed.
In an exemplary embodiment, the screening the multimedia information corresponding to the account feature from the multimedia information corresponding to the multimedia information category includes:
acquiring information characteristics of the multimedia information corresponding to the multimedia information category;
acquiring feature similarity between the account features and each information feature;
and screening out the multimedia information with the feature similarity larger than the preset similarity from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as the multimedia information conforming to the account feature.
In an exemplary embodiment, the pushing the ranked candidate multimedia information to the account to be pushed includes:
sequentially screening out a preset number of candidate multimedia information from the sequenced candidate multimedia information, and taking the candidate multimedia information as target multimedia information corresponding to the account to be pushed;
pushing the target multimedia information to the account to be pushed according to a preset pushing frequency.
In an exemplary embodiment, the target ranking model is fused by:
acquiring information ordering accuracy of at least two information ordering models, and determining the number of tree models selected from each information ordering model according to the information ordering accuracy;
according to a preset selection sequence, correspondingly selecting tree models corresponding to the number of the tree models from the tree models contained in each information ordering model, and obtaining the tree models correspondingly selected from each information ordering model;
and carrying out fusion processing on the tree models correspondingly selected from the information ordering models to obtain the target ordering model.
In an exemplary embodiment, the obtaining the information ordering accuracy of at least two information ordering models includes:
Collecting a test sample set; the test sample set comprises sample multimedia information and actual sequencing labels corresponding to the sample multimedia information;
respectively inputting the sample multimedia information into each information sequencing model to obtain a prediction sequencing label of each information sequencing model on the sample multimedia information;
and respectively acquiring errors between the prediction ordering labels of the information ordering models on the sample multimedia information and the actual ordering labels of the sample multimedia information, and acquiring the information ordering accuracy of the information ordering models according to the errors.
According to a second aspect of the embodiments of the present disclosure, there is provided a multimedia information pushing apparatus, including:
the multimedia message obtaining unit is configured to obtain candidate multimedia message corresponding to the account to be pushed;
a ranking tag determining unit configured to perform ranking of each of the candidate multimedia message information by the target ranking model; the target sorting models are obtained by fusing tree models selected from at least two information sorting models, sorting labels used for outputting multimedia information are obtained, and the number of the tree models selected from each information sorting model is determined according to the information sorting accuracy of each information sorting model;
The multimedia information ordering unit is configured to order the candidate multimedia information according to the ordering label to obtain ordered candidate multimedia information;
and the multimedia information pushing unit is configured to push the sorted candidate multimedia information to the account to be pushed.
In an exemplary embodiment, the multimedia information obtaining unit is further configured to perform receiving an information push request of the account to be pushed; the information pushing request comprises key information input by triggering of the account to be pushed; and recalling the multimedia information with the information matching degree between the key information and the information base higher than the preset matching degree as candidate multimedia information corresponding to the account to be pushed.
In an exemplary embodiment, the multimedia information obtaining unit is further configured to perform obtaining account characteristics of an account to be pushed; acquiring a multimedia information category conforming to the account characteristics; and screening out the multimedia information which accords with the account characteristics from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as candidate multimedia information corresponding to the account to be pushed.
In an exemplary embodiment, the multimedia information obtaining unit is further configured to perform obtaining an information feature of the multimedia information corresponding to the multimedia information category; acquiring feature similarity between the account features and each information feature; and screening out the multimedia information with the feature similarity larger than the preset similarity from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as the multimedia information conforming to the account feature.
In an exemplary embodiment, the multimedia information pushing unit is further configured to perform sequentially screening a preset number of candidate multimedia information from the sorted candidate multimedia information, as target multimedia information corresponding to the account to be pushed; pushing the target multimedia information to the account to be pushed according to a preset pushing frequency.
In an exemplary embodiment, the apparatus further includes a model fusion unit configured to perform obtaining information ordering accuracy of at least two information ordering models, and determine the number of tree models selected from each of the information ordering models according to the information ordering accuracy; according to a preset selection sequence, correspondingly selecting tree models corresponding to the number of the tree models from the tree models contained in each information ordering model, and obtaining the tree models correspondingly selected from each information ordering model; and carrying out fusion processing on the tree models correspondingly selected from the information ordering models to obtain the target ordering model.
In an exemplary embodiment, the model fusion unit is further configured to perform collecting a set of test samples; the test sample set comprises sample multimedia information and actual sequencing labels corresponding to the sample multimedia information; respectively inputting the sample multimedia information into each information sequencing model to obtain a prediction sequencing label of each information sequencing model on the sample multimedia information; and respectively acquiring errors between the prediction ordering labels of the information ordering models on the sample multimedia information and the actual ordering labels of the sample multimedia information, and acquiring the information ordering accuracy of the information ordering models according to the errors.
According to a third aspect of embodiments of the present disclosure, there is provided 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 multimedia information pushing method as described in any of the embodiments of the first aspect.
According to a fourth aspect of embodiments of the present disclosure, there is provided a storage medium comprising: the instructions in the storage medium, when executed by a processor of a server, enable the server to perform the multimedia information pushing method described in any one of the embodiments 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 at least one processor of a device reads and executes the computer program, causing the device to perform the multimedia information pushing method as 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:
acquiring candidate multimedia information corresponding to an account to be pushed; then, sorting labels of each candidate multimedia information are obtained through a target sorting model; the target ordering model is obtained by fusing tree models selected from at least two information ordering models and is used for outputting ordering labels of multimedia information, and the number of the tree models selected from each information ordering model is determined according to the information ordering accuracy of each information ordering model; finally, according to the sorting labels, sorting the candidate multimedia information to obtain sorted candidate multimedia information; pushing the sorted candidate multimedia information to an account to be pushed; the aim of sorting the candidate multimedia information according to the target sorting model obtained by fusing the tree models selected from the at least two information sorting models is fulfilled, and the overall operation efficiency of the target sorting model is higher because the model structure of the target sorting model obtained by fusing the tree models selected from the at least two information sorting models is not redundant, so that the sorting efficiency of the multimedia information is improved, and the pushing efficiency of the multimedia information is further improved; meanwhile, the defect that the pushing efficiency of the multimedia information is low due to long sequencing processing time of the multimedia information due to the fact that redundancy is easy to occur in a model structure of a fusion model obtained through gradual accumulation training is avoided.
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 disclosure and together with the description, serve to explain the principles of the disclosure and do not constitute an undue limitation on the disclosure.
Fig. 1 is an application environment diagram illustrating a multimedia information pushing method according to an exemplary embodiment.
Fig. 2 is a flowchart illustrating a multimedia information pushing method according to an exemplary embodiment.
FIG. 3 is a flowchart illustrating the fusion steps of a target ranking model according to an exemplary embodiment.
Fig. 4 is a flowchart illustrating another multimedia information pushing method according to an exemplary embodiment.
Fig. 5 is a block diagram illustrating a multimedia information pushing apparatus according to an exemplary embodiment.
Fig. 6 is an internal structural diagram of a server shown according to an exemplary embodiment.
Detailed Description
In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions of 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 foregoing figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.
The multimedia information pushing method provided by the disclosure can be applied to an application environment as shown in fig. 1. The terminal 110 interacts with the server 120 through a network, and the server 120 obtains candidate multimedia information corresponding to an account to be pushed logged in the terminal 110; obtaining sequencing labels of each candidate information through a target sequencing model; the target ordering model is obtained by fusing tree models selected from at least two information ordering models and is used for outputting ordering labels of multimedia information, and the number of the tree models selected from each information ordering model is determined according to the information ordering accuracy of each information ordering model; according to the sorting labels, sorting the candidate multimedia information to obtain sorted candidate multimedia information; pushing the sequenced candidate multimedia information to a terminal 110 corresponding to the account to be pushed, and displaying the sequenced candidate multimedia information by the terminal 110 through a terminal interface. The terminal 110 may be, but not limited to, various smartphones, tablet computers, notebook computers, etc., and the server 120 may be implemented by a stand-alone server or a server cluster formed by a plurality of servers.
Fig. 2 is a flowchart illustrating a multimedia information pushing method according to an exemplary embodiment, and as shown in fig. 2, the multimedia information pushing method is used in the server shown in fig. 1, and includes the following steps:
in step S210, candidate multimedia information corresponding to the account to be pushed is obtained.
The account refers to a registered account of an application program in the terminal, such as a registered account of a short video application program, a registered account of a video browsing program, and the like. The account to be pushed refers to an authorized account which needs to be processed and analyzed, and specifically refers to a pushing object of candidate multimedia information; in an actual scene, the account to be pushed can refer to a push object of a video or a push object of a picture.
The candidate multimedia information refers to multimedia information which is preliminarily screened from mass multimedia information and accords with an account to be pushed, such as a favorite video, a favorite picture and the like of the account to be pushed; in an actual scenario, the candidate multimedia information may refer to multimedia information such as video and pictures recalled by recall algorithms such as a hot spot recall algorithm, a theme recall algorithm, a collaborative recall algorithm, a related recall algorithm, an environment recall algorithm, and the like.
Specifically, the server receives a multimedia information pushing request of an account to be pushed, and screens multimedia information conforming to the account to be pushed from a local database or a network according to the multimedia information pushing request of the account to be pushed, and the multimedia information is used as candidate multimedia information corresponding to the account to be pushed. Therefore, the candidate multimedia information corresponding to the account to be pushed is acquired, so that subsequent sorting processing of the candidate multimedia information corresponding to the account to be pushed is facilitated, and all the multimedia information is not required to be sorted, so that the sorting processing time of the multimedia information is reduced, and the efficiency of subsequent pushing of the multimedia information is further improved.
For example, in a video search scene, a terminal responds to a video search operation of a currently logged account to be pushed on a terminal interface to obtain keywords input by the account to be pushed, such as AAA; generating an information pushing request according to keywords input by an account to be pushed, and sending the information pushing request to a corresponding server; the server analyzes the received information pushing request to obtain keywords input by the account to be pushed, and screens out videos conforming to the account to be pushed from an information base storing a plurality of videos to serve as candidate multimedia information corresponding to the account to be pushed.
Further, the server may further obtain an account feature code of the account to be pushed and an information feature code of each information in the information base, determine feature similarity between the account feature code and each information feature code, and screen out multimedia information with feature similarity greater than a preset similarity (for example, 0.6) from each multimedia information in the information base, as candidate multimedia information corresponding to the account to be pushed. The account features refer to attribute features of an account to be pushed, such as gender, age, region, occupation, interests and the like, and the information features refer to attribute features of information, such as information content, information authors, information labels and the like; the account feature codes refer to low-dimensional feature vectors which are subjected to compression coding and used for representing low-level semantics of account features, the information feature codes refer to low-dimensional feature vectors which are subjected to compression coding and used for representing low-level semantics of information features, and the low-dimensional feature vectors and the information feature codes are obtained through model learning generated by feature codes trained in advance.
In step S220, ranking tags of each candidate multimedia information are obtained through the target ranking model; the target sorting model is obtained by fusing tree models selected from at least two information sorting models and used for outputting sorting labels of multimedia information, and the number of the tree models selected from each information sorting model is determined according to the information sorting accuracy of each information sorting model.
The ranking label refers to a ranking basis of candidate multimedia information, and may be CTR (Click-Through-Rate), relevance, etc.; in an actual scene, the higher the CTR or the correlation degree of the candidate multimedia information, the more front the ranking position of the candidate multimedia information.
The information sorting model is used for sorting the multimedia information and comprises decision forest models of a plurality of tree models, such as a model obtained by training based on manually marked data, a model obtained by training based on posterior behavior data and the like; the model obtained through training based on the manually marked data has a better effect on correlation, and the model obtained through training based on posterior behavior data has a better effect on an online index.
The tree model is a model with a tree branch structure based on feature space division, such as a GBDT (Gradient Boost Decision Tree, gradient lifting iterative decision tree) model, a XGBoost (eXtreme Gradient Boosting, extreme gradient lifting) tree model, a LightGBM (Light Gradient Boosting Machine, lightweight gradient lifting) tree model, and the like.
The information sorting accuracy is used for measuring the accuracy of sorting labels output by the information sorting models, and the information sorting accuracy corresponding to different information sorting models is different; if the information sorting accuracy of the information sorting model is higher, the number of tree models selected from the information sorting model is larger.
The target ordering model is obtained by fusing tree models selected from at least two information ordering models and is used for ordering the multimedia information so as to output an ordering label of the multimedia information; it should be noted that, the at least two information ordering models refer to two or more information ordering models, and the disclosure is not limited in particular.
Specifically, the server inputs candidate multimedia information into a target sorting model, and outputs initial sorting labels of the candidate multimedia information through each information sorting model in the target sorting model; screening out initial sequencing labels with the largest proportion from the initial sequencing labels of the candidate multimedia information output by each information sequencing model, and taking the initial sequencing labels as sequencing labels of the candidate multimedia information; with reference to the method, ranking tags of each candidate multimedia message can be obtained. Therefore, the target ordering model for ordering the candidate multimedia information is obtained by fusing tree models selected from at least two information ordering models, and the model structure of the target ordering model is not redundant, so that the overall operation efficiency of the target ordering model is higher, the ordering processing of the candidate multimedia information is faster, and the subsequent pushing efficiency of the multimedia information is improved.
For example, assuming that there are 6 information ranking models, if the initial ranking labels of the candidate multimedia information output by each information ranking model are A1, A2, A3, A1, A2, A1, respectively, the initial ranking label is the largest duty ratio of A1, which indicates that the final ranking label of the candidate multimedia information is A1.
Further, the server can also input the candidate multimedia information into a target sorting model, and output an initial sorting label of the candidate multimedia information through each information sorting model in the target sorting model; acquiring weights corresponding to the information ordering accuracy of each information ordering model, and corresponding to the weights of the initial ordering labels of the candidate multimedia information output by each information ordering model; weighting the initial sequencing labels of the candidate multimedia information output by each information sequencing model according to the weight of the initial sequencing labels of the candidate multimedia information output by each information sequencing model to obtain the sequencing labels of the candidate multimedia information; with reference to the method, ranking tags of each candidate multimedia message can be obtained.
For example, if the initial ranking tag is the initial click rate, the server may perform weighting processing on the initial click rate of the candidate multimedia information output by each information ranking model according to the weight of the initial click rate of the candidate multimedia information output by each information ranking model, so as to obtain the target click rate of the candidate multimedia information, and use the target click rate as the ranking tag of the candidate multimedia information.
In step S230, each candidate multimedia information is ranked according to the ranking tag, so as to obtain ranked candidate multimedia information.
For example, the server sorts the candidate multimedia information according to the order of the click rate or the correlation degree from high to low, so as to obtain the sorted candidate multimedia information. Therefore, the efficiency of outputting the sorting labels through the target sorting model is higher, so that the sorting time of the candidate multimedia information is greatly shortened, the sorting efficiency of the multimedia information is improved, and the subsequent pushing efficiency of the multimedia information is further improved.
In step S240, the sorted candidate multimedia information is pushed to the account to be pushed.
Specifically, the server pushes the sequenced candidate multimedia information to a corresponding account to be pushed; thus, the accurate pushing of the multimedia information is facilitated.
In the multimedia information pushing method, candidate multimedia information corresponding to an account to be pushed is obtained; then, sorting labels of each candidate multimedia information are obtained through a target sorting model; the target ordering model is obtained by fusing tree models selected from at least two information ordering models and is used for outputting ordering labels of multimedia information, and the number of the tree models selected from each information ordering model is determined according to the information ordering accuracy of each information ordering model; finally, according to the sorting labels, sorting the candidate multimedia information to obtain sorted candidate multimedia information; pushing the sorted candidate multimedia information to an account to be pushed; the aim of sorting the candidate multimedia information according to the target sorting model obtained by fusing the tree models selected from the at least two information sorting models is fulfilled, and the overall operation efficiency of the target sorting model is higher because the model structure of the target sorting model obtained by fusing the tree models selected from the at least two information sorting models is not redundant, so that the sorting efficiency of the multimedia information is improved, and the pushing efficiency of the multimedia information is further improved; meanwhile, the defect that the pushing efficiency of the multimedia information is low due to long sequencing processing time of the multimedia information due to the fact that redundancy is easy to occur in a model structure of a fusion model obtained through gradual accumulation training is avoided.
In an exemplary embodiment, in step S210, obtaining candidate multimedia information corresponding to an account to be pushed includes: receiving an information pushing request of an account to be pushed; the information pushing request comprises key information of triggering input of an account to be pushed; and recalling the multimedia information with the information matching degree between the key information and the information library being higher than the preset matching degree as candidate multimedia information corresponding to the account to be pushed.
The information matching degree between the multimedia information and the key information is used for measuring the correlation degree between the multimedia information and the key information; the higher the information matching degree, the higher the correlation degree.
For example, in a video search scene, a terminal responds to a video search operation of a currently logged account to be pushed on a terminal interface to obtain keywords input by the account to be pushed, such as AAA; generating an information pushing request according to keywords input by an account to be pushed, and sending the information pushing request to a corresponding server; the server analyzes the received information pushing request to obtain keywords which are triggered and input by the account to be pushed; and respectively counting the information matching degree between the keywords and each video in the information base, and screening out videos with the information matching degree higher than the preset matching degree from each video in the information base, wherein the videos are used as candidate multimedia information corresponding to the account to be pushed.
According to the technical scheme provided by the embodiment of the disclosure, the candidate multimedia information corresponding to the account to be pushed is acquired, so that the subsequent sorting processing of the candidate multimedia information corresponding to the account to be pushed is facilitated, and the sorting processing of all the multimedia information is not needed, thereby reducing the sorting processing time of the multimedia information and further improving the sorting processing efficiency of the multimedia information.
In an exemplary embodiment, in step S210, obtaining candidate multimedia information corresponding to an account to be pushed further includes: acquiring account characteristics of an account to be pushed; acquiring multimedia information categories conforming to account characteristics; and screening out the multimedia information which accords with the account characteristics from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as candidate multimedia information corresponding to the account to be pushed.
Wherein, the multimedia information category refers to the information type of the multimedia information; for example, if the multimedia information is video, the information type may be a fun video, a variety video, a food video, a travel video, or the like.
Specifically, the server acquires account characteristics of an account to be pushed and category characteristics of a multimedia information category, wherein the category characteristics refer to attribute characteristics of the multimedia information category, such as category contents; then, the server calculates the feature similarity between the account feature and each multimedia information category, and screens out the multimedia information category with the feature similarity larger than a preset threshold value from each multimedia information category as the multimedia information category conforming to the account feature; and acquiring information characteristics of the multimedia information corresponding to the multimedia information category, screening the multimedia information with the information characteristics matched with the account characteristics from the multimedia information corresponding to the multimedia information category, taking the multimedia information as the multimedia information conforming to the account characteristics, and taking the multimedia information as candidate multimedia information corresponding to the account to be pushed.
According to the technical scheme provided by the embodiment of the disclosure, the multimedia information category which accords with the account characteristics of the account to be pushed is acquired firstly, and then the multimedia information which accords with the account characteristics is screened out from the multimedia information corresponding to the multimedia information category and is used as the candidate multimedia information corresponding to the account to be pushed, so that the determined candidate multimedia information is more accurate, and the determination accuracy of the candidate multimedia information is improved.
In an exemplary embodiment, screening multimedia information conforming to account features from multimedia information corresponding to a multimedia information category includes: acquiring information characteristics of multimedia information corresponding to the multimedia information category; acquiring feature similarity between account features and information features; and screening out the multimedia information with the feature similarity larger than the preset similarity from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as the multimedia information conforming to the account feature.
The feature similarity between the account features and the information features is used for measuring the correlation between the account features and the information features; in general, the greater the feature similarity, the greater the correlation between account features and information features.
Specifically, the server inputs the multimedia information corresponding to the multimedia information category into a pre-trained information feature extraction model, and performs information feature extraction processing on the multimedia information corresponding to the multimedia information category through the pre-trained information feature extraction model to obtain the information feature of the multimedia information corresponding to the multimedia information category; according to a preset feature similarity statistics instruction, counting feature similarity between account features and information features, and correspondingly serving as the feature similarity of each piece of multimedia information; if the feature similarity of the multimedia information is larger than the preset similarity, taking the multimedia information as the multimedia information conforming to the account feature; thus, the multimedia information conforming to the account characteristics can be screened from the multimedia information corresponding to the multimedia information category.
According to the technical scheme provided by the embodiment of the disclosure, the multimedia information with the feature similarity between the information features and the account features being larger than the preset similarity is screened out from the multimedia information corresponding to the multimedia information types and used as the multimedia information conforming to the account features, so that the purpose of determining the multimedia information conforming to the account features according to the feature similarity between the information features of the multimedia information and the account features of the account to be pushed is achieved, and the accuracy of the determined multimedia information is improved.
In an exemplary embodiment, in step S240, pushing the sorted candidate multimedia information to the account to be pushed includes: sequentially screening out a preset number of candidate multimedia information from the sequenced candidate multimedia information, and taking the candidate multimedia information as target multimedia information corresponding to an account to be pushed; and pushing the target multimedia information to the account to be pushed according to the preset pushing frequency.
The preset pushing frequency refers to the pushing frequency of the target multimedia information, for example, 5 pieces of target multimedia information are pushed at a time.
For example, the server screens out the first 20 candidate multimedia information from the sorted candidate multimedia information, and the first 20 candidate multimedia information is used as target multimedia information corresponding to the account to be pushed; pushing the target multimedia information to a terminal corresponding to an account to be pushed according to a preset pushing frequency, and displaying the received target multimedia information through a terminal interface.
According to the technical scheme provided by the embodiment of the disclosure, the preset number of candidate multimedia information is screened, and then the screened candidate multimedia information is pushed to the corresponding account to be pushed, so that the defect that the information pushing accuracy is low due to the fact that all candidate multimedia information is pushed to the account to be pushed is avoided, and the accurate pushing of the multimedia information is realized.
In an exemplary embodiment, as shown in FIG. 3, the target ranking model is fused by:
in step S310, information ordering accuracy of at least two information ordering models is acquired, and the number of tree models selected from the respective information ordering models is determined according to the information ordering accuracy.
Specifically, the server acquires information ordering accuracy of at least two information ordering models from a local database, and determines the number proportion of tree models selected from each information ordering model according to the information ordering accuracy; and determining the number of the tree models selected from each information ordering model according to the preset number of the tree models and the proportion of the number of the tree models selected from each information ordering model.
In step S320, according to a preset selection order, tree models corresponding to the number of the tree models are correspondingly selected from the tree models included in each information ordering model, so as to obtain tree models correspondingly selected from each information ordering model.
The preset selection sequence refers to a selection sequence of the tree models in the information ordering model, and specifically refers to a interception sequence of the tree models in the information ordering model; in an actual scene, the tree models corresponding to the number of the selected tree models are selected from the tree models contained in the information ordering model, namely, the tree models corresponding to the number of the previous tree models are intercepted from the tree models contained in the information ordering model according to the sequence from left to right.
In step S330, a fusion process is performed on the tree model selected correspondingly from the information ranking models, so as to obtain a target ranking model.
Specifically, the server acquires a preset model fusion instruction, and performs fusion processing on a tree model correspondingly selected from the information ordering models according to the preset model fusion instruction to obtain a fused model serving as a target ordering model.
For example, it is assumed that there are two information ranking models, namely, an information ranking model A and an information ranking model B, the number of tree models included in the information ranking model A is I, the number of tree models included in the information ranking model B is J, and the number of tree models selected from the information ranking model A is tree A The number of tree models selected from the information ordering model B is tree B Next, from among the I tree models included in the information ordering model A, the front tree is truncated from left to right A A tree model, denoted as i; from J tree models included in the information ordering model B, a front tree is cut from left to right B A tree model, denoted j; the tree models correspondingly intercepted from the information ordering models A and B are fused to obtain a target ordering model tree fusion ＝i∪j。
According to the technical scheme provided by the embodiment of the disclosure, according to the information sorting accuracy of the information sorting models, a corresponding number of tree models are correspondingly selected from the tree models contained in each information sorting model, and then the tree models correspondingly selected from each information sorting model are subjected to fusion processing, so that the purpose of dynamic fusion of the tree models is realized, the problem that redundancy easily occurs in a model structure of a fusion model obtained by gradual accumulation training, the sorting processing time of multimedia information is long, and the defect of low subsequent multimedia information pushing efficiency is avoided.
In an exemplary embodiment, the step S330 includes performing a fusion process on the tree model correspondingly selected from the information ranking models to obtain the target ranking model, which specifically includes the following steps: respectively carrying out fusion processing on the tree models correspondingly selected from the information sorting models to obtain new information sorting models; the information ordering accuracy of each new information ordering model corresponds to the information ordering accuracy of each information ordering model; performing fusion processing on each new information ordering model to obtain a target ordering model; in the target sorting model, the weight of the sorting label output by each new information sorting model is matched with the information sorting accuracy of each new information sorting model.
The information sorting accuracy of each new information sorting model has a one-to-one correspondence with the information sorting accuracy of each information sorting model, and the information sorting accuracy of each information sorting model can be correspondingly used as the information sorting accuracy of each new information sorting model.
The matching of the weight of the sorting label output by each new information sorting model with the information sorting accuracy of each new information sorting model means that the weight of the sorting label output by each new information sorting model has a one-to-one correspondence with the information sorting accuracy of each new information sorting model, for example, the higher the information sorting accuracy of the new information sorting model is, the higher the weight of the sorting label output by the new information sorting model is; the sum of the weights of the ranking tags output by the new information ranking models is equal to 1.
Specifically, the server respectively performs fusion processing on the tree models correspondingly selected from the information ordering models according to a preset model fusion instruction to obtain new information ordering models, and the information ordering accuracy of the information ordering models is correspondingly used as the information ordering accuracy of the new information ordering models; then, according to a preset model fusion instruction, continuing fusion processing on each new information sequencing model to obtain a target sequencing model; and simultaneously, according to the corresponding relation between the information ordering accuracy and the weights, determining the weights of the ordering labels output by each new information ordering model based on the information ordering accuracy of each new information ordering model.
According to the technical scheme provided by the embodiment of the disclosure, as the target ordering model for ordering the candidate information is obtained by fusing the tree models selected from at least two information ordering models, redundancy does not occur in the model structure, so that the overall operation efficiency of the target ordering model is higher, the ordering processing of the information is faster, and the subsequent information pushing efficiency is improved.
In an exemplary embodiment, the step S310 includes the following steps of: collecting a test sample set; the test sample set comprises sample multimedia information and actual sequencing labels corresponding to the sample multimedia information; respectively inputting the sample multimedia information into each information sequencing model to obtain a prediction sequencing label of each information sequencing model on the sample multimedia information; and respectively acquiring errors between the prediction ordering labels of the information ordering models on the sample multimedia information and the actual ordering labels of the sample multimedia information, and acquiring the information ordering accuracy of the information ordering models according to the errors.
The sample multimedia information is multimedia information marked with an actual sorting label, and specifically may be a picture, a video, etc. marked with an actual click rate or an actual relevance.
The error between the prediction ordering label of the sample multimedia information and the actual ordering label of the sample multimedia information is used for measuring the deviation degree between the prediction ordering label of the sample multimedia information and the actual ordering label of the sample multimedia information; the larger the error, the lower the accuracy of the information ordering representing the information ordering model.
Specifically, the server acquires errors between the prediction ordering labels and the actual ordering labels of the multimedia information of each sample by the information ordering model, and adds the errors between the prediction ordering labels and the actual ordering labels of the multimedia information of each sample by the information ordering model to obtain target errors of the information ordering model on a test sample set; obtaining the information ordering accuracy of the information ordering model according to the corresponding relation between the target error and the information ordering accuracy; with reference to the method, the information ordering accuracy of each information ordering model can be obtained.
For example, it is assumed that there are two information ordering models, namely, an information ordering model A and an information ordering model B, and on a given test sample set, the positive order pair number pos corresponding to the two models can be obtained A 、pos B And further obtaining the information ordering accuracy of the two models. If the model A and the model B are classified, the prediction accuracy acc corresponding to the model A and the model B can be obtained on a given test sample set A 、acc B And further obtaining the information ordering accuracy of the two models. If the regression model A and the regression model B are adopted, on a given test sample set, the root mean square error rmse corresponding to the two models can be obtained A 、rmse B And further obtaining the information ordering accuracy of the two models. It should be noted that, according to the category of the model, an evaluation index of the model on a given test sample set is obtained, so as to obtain the information ordering accuracy of the model.
According to the technical scheme provided by the embodiment of the disclosure, the information ordering accuracy of at least two information ordering models is obtained, so that the number of the tree models selected from each information ordering model is determined according to the information ordering accuracy of the information ordering models, and further the tree models corresponding to the number of each tree model are selected from each information ordering model, and the dynamic fusion of the tree models is facilitated.
In an exemplary embodiment, the step S310 determines the number of tree models selected from the respective information sorting models according to the information sorting accuracy, and specifically includes the following steps: normalizing the information sorting accuracy of each information sorting model to obtain the number proportion of tree models selected from each information sorting model; and obtaining the number of the tree models selected from each information ordering model according to the preset number of the tree models and the proportion of the number of the tree models selected from each information ordering model.
The preset number of tree models refers to the total number of tree models that need dynamic fusion, and can be adjusted according to actual situations, and is not limited herein.
Specifically, the server obtains products between the preset number of tree models and the proportion of the number of tree models selected from each information ordering model respectively as the number of tree models selected from each information ordering model.
For example, it is assumed that there are two information ranking models, namely, an information ranking model A and an information ranking model B, and the corresponding information ranking accuracy is an indicator A 、indicator B The tree model quantity ratios of the information sorting model A and the information sorting model B are respectively as follows:
if the number of the preset tree models is N, the number of the tree models selected from the information sorting model A and the information sorting model B is respectively as follows:
wherein,representation of the taking frac A X N integer, ">Representation of the taking frac B An integer of x N.
According to the technical scheme provided by the embodiment of the disclosure, the number of the tree models selected from each information ordering model is determined according to the information ordering accuracy, so that the number of the tree models selected from each information ordering model is accurately determined, the overall operation efficiency of the target ordering model obtained by subsequent fusion is higher, and the ordering processing efficiency of the multimedia information is further improved.
Fig. 4 is a flowchart illustrating another multimedia information pushing method according to an exemplary embodiment, as shown in fig. 4, for use in the server shown in fig. 1, comprising the steps of:
in step S410, information ordering accuracy of at least two information ordering models is acquired, and the number of tree models selected from the respective information ordering models is determined according to the information ordering accuracy.
In step S420, according to a preset selection order, tree models corresponding to the number of the tree models are correspondingly selected from the tree models included in each information ordering model, so as to obtain tree models correspondingly selected from each information ordering model.
In step S430, a fusion process is performed on the tree model selected correspondingly from the respective information ranking models, so as to obtain a target ranking model.
In step S440, an information push request of an account to be pushed is received; the information pushing request comprises key information of triggering input of an account to be pushed; and recalling the multimedia information with the information matching degree between the key information and the information library being higher than the preset matching degree as candidate multimedia information corresponding to the account to be pushed.
In step S450, ranking tags of each candidate multimedia information are obtained through the target ranking model.
In step S460, the candidate multimedia information is ranked according to the ranking tag, so as to obtain ranked candidate multimedia information.
In step S470, a preset number of candidate multimedia information is sequentially screened out from the sorted candidate multimedia information, and the candidate multimedia information is used as target multimedia information corresponding to the account to be pushed; and pushing the target multimedia information to the account to be pushed according to the preset pushing frequency.
In the multimedia information pushing method, candidate multimedia information corresponding to an account to be pushed is obtained; then, sorting labels of each candidate multimedia information are obtained through a target sorting model; the target ordering model is obtained by fusing tree models selected from at least two information ordering models and is used for outputting ordering labels of multimedia information, and the number of the tree models selected from each information ordering model is determined according to the information ordering accuracy of each information ordering model; finally, according to the sorting labels, sorting the candidate multimedia information to obtain sorted candidate multimedia information; pushing the sorted candidate multimedia information to an account to be pushed; the aim of sorting the candidate multimedia information according to the target sorting model obtained by fusing the tree models selected from the at least two information sorting models is fulfilled, and the overall operation efficiency of the target sorting model is higher because the model structure of the target sorting model obtained by fusing the tree models selected from the at least two information sorting models is not redundant, so that the sorting efficiency of the multimedia information is improved, and the pushing efficiency of the multimedia information is further improved; meanwhile, the defect that the pushing efficiency of the multimedia information is low due to long sequencing processing time of the multimedia information due to the fact that redundancy is easy to occur in a model structure of a fusion model obtained through gradual accumulation training is avoided.
It should be understood that, although the steps in the flowcharts of fig. 2-4 are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in fig. 2-4 may include multiple steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the steps or stages are performed necessarily performed in sequence, but may be performed alternately or alternately with at least a portion of the steps or stages in other steps or other steps.
Fig. 5 is a block diagram illustrating a multimedia information pushing apparatus according to an exemplary embodiment. Referring to fig. 5, the apparatus includes a multimedia information acquisition unit 510, a ranking tag determination unit 520, a multimedia information ranking unit 530, and a multimedia information pushing unit 540.
The multimedia information obtaining unit 510 is configured to perform obtaining candidate multimedia information corresponding to an account to be pushed.
A ranking tag determining unit 520 configured to perform ranking of each candidate multimedia information by the target ranking model; the target sorting model is obtained by fusing tree models selected from at least two information sorting models and used for outputting sorting labels of multimedia information, and the number of the tree models selected from each information sorting model is determined according to the information sorting accuracy of each information sorting model.
The multimedia information sorting unit 530 is configured to sort the candidate multimedia information according to the sorting labels, so as to obtain sorted candidate multimedia information.
The multimedia information pushing unit 540 is configured to perform pushing the sorted candidate pair media information to the account to be pushed.
In an exemplary embodiment, the multimedia information obtaining unit 510 is further configured to perform receiving an information push request of an account to be pushed; the information pushing request comprises key information of triggering input of an account to be pushed; and recalling the multimedia information with the information matching degree between the key information and the information library being higher than the preset matching degree as candidate multimedia information corresponding to the account to be pushed.
In an exemplary embodiment, the multimedia information obtaining unit 510 is further configured to perform obtaining account characteristics of an account to be pushed; acquiring multimedia information categories conforming to account characteristics; and screening out the multimedia information which accords with the account characteristics from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as candidate multimedia information corresponding to the account to be pushed.
In an exemplary embodiment, the multimedia information obtaining unit 510 is further configured to perform obtaining information features of the multimedia information corresponding to the multimedia information category; acquiring feature similarity between account features and information features; and screening out the multimedia information with the feature similarity larger than the preset similarity from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as the multimedia information conforming to the account feature.
In an exemplary embodiment, the multimedia information pushing unit 540 is further configured to perform sequentially screening out a preset number of candidate multimedia information from the sorted candidate multimedia information, as target multimedia information corresponding to the account to be pushed; and pushing the target multimedia information to the account to be pushed according to the preset pushing frequency.
In an exemplary embodiment, the multimedia information pushing apparatus of the present disclosure further includes a model fusion unit configured to perform obtaining information ordering accuracy of at least two information ordering models, and determine the number of tree models selected from the respective information ordering models according to the information ordering accuracy; according to a preset selection sequence, correspondingly selecting tree models corresponding to the number of each tree model from the tree models contained in each information ordering model, and obtaining the tree models correspondingly selected from each information ordering model; and carrying out fusion processing on the tree models correspondingly selected from the information sorting models to obtain the target sorting model.
In an exemplary embodiment, the model fusion unit is further configured to perform collecting a set of test samples; the test sample set comprises sample multimedia information and actual sequencing labels corresponding to the sample multimedia information; respectively inputting the sample multimedia information into each information sequencing model to obtain a prediction sequencing label of each information sequencing model on the sample multimedia information; and respectively acquiring errors between the prediction ordering labels of the information ordering models on the sample multimedia information and the actual ordering labels of the sample multimedia information, and acquiring the information ordering accuracy of the information ordering models according to the errors.
The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.
Fig. 6 is a block diagram illustrating an apparatus 600 for performing the above-described multimedia information pushing method according to an exemplary embodiment. For example, device 600 may be a server. Referring to fig. 6, device 600 includes a processing component 620 that further includes one or more processors and memory resources represented by memory 622 for storing instructions, such as applications, executable by processing component 620. The application program stored in memory 622 may include one or more modules each corresponding to a set of instructions. Further, the processing component 620 is configured to execute instructions to perform the multimedia information pushing method described above.
The device 600 may also include a power component 624 configured to perform power management of the device 600, a wired or wireless network interface 626 configured to connect the device 600 to a network, and an input/output (I/O) interface 628. The device 600 may operate based on an operating system stored in memory 622, such as Windows Server, mac OS XTM, unixTM, linuxTM, freeBSDTM, or the like.
In an exemplary embodiment, a storage medium is also provided, such as memory 622, including instructions executable by a processor of device 600 to perform the above-described method. The storage medium may be a non-transitory computer readable storage medium, for example, a ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.
In an exemplary embodiment, there is also provided a computer program product comprising a computer program stored in a readable storage medium, from which at least one processor of the device reads and executes the computer program, causing the device to perform the multimedia information pushing method in any of the embodiments of the disclosure.
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 disclosure is intended to cover any adaptations, uses, or adaptations of the disclosure following the general 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 is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims (16)

1. A multimedia information pushing method, comprising:
acquiring candidate multimedia information corresponding to an account to be pushed;
obtaining the sequencing labels of the candidate multimedia information through a target sequencing model; the target sorting models are obtained by fusing tree models selected from at least two information sorting models, sorting labels used for outputting multimedia information are obtained by fusing the target sorting models according to the number of the tree models selected from each information sorting model and the ratio of the number of the tree models selected from each information sorting model; the number proportion of the tree models selected from the information ordering models is obtained by normalizing the information ordering accuracy of the information ordering models;
According to the sorting labels, sorting the candidate multimedia information to obtain sorted candidate multimedia information;
pushing the sequenced candidate multimedia information to the account to be pushed.
2. The method for pushing multimedia information according to claim 1, wherein the obtaining candidate multimedia information corresponding to an account to be pushed includes:
receiving an information pushing request of the account to be pushed; the information pushing request comprises key information input by triggering of the account to be pushed;
and recalling the multimedia information with the information matching degree between the key information and the information base higher than the preset matching degree as candidate multimedia information corresponding to the account to be pushed.
3. The method for pushing multimedia information according to claim 1, wherein the obtaining candidate multimedia information corresponding to an account to be pushed further comprises:
acquiring account characteristics of an account to be pushed;
acquiring a multimedia information category conforming to the account characteristics;
and screening out the multimedia information which accords with the account characteristics from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as candidate multimedia information corresponding to the account to be pushed.
4. The method for pushing multimedia information according to claim 3, wherein the screening multimedia information conforming to the account feature from the multimedia information corresponding to the multimedia information category includes:
acquiring information characteristics of the multimedia information corresponding to the multimedia information category;
acquiring feature similarity between the account features and each information feature;
and screening out the multimedia information with the feature similarity larger than the preset similarity from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as the multimedia information conforming to the account feature.
5. The method for pushing the sequenced candidate multimedia information to the account to be pushed according to claim 1, comprising:
sequentially screening out a preset number of candidate multimedia information from the sequenced candidate multimedia information, and taking the candidate multimedia information as target multimedia information corresponding to the account to be pushed;
pushing the target multimedia information to the account to be pushed according to a preset pushing frequency.
6. The method for pushing multimedia information according to any one of claims 1 to 5, wherein the target ranking model is obtained by fusing:
Acquiring information ordering accuracy of at least two information ordering models, and determining the number of tree models selected from each information ordering model according to the information ordering accuracy;
according to a preset selection sequence, correspondingly selecting tree models corresponding to the number of the tree models from the tree models contained in each information ordering model, and obtaining the tree models correspondingly selected from each information ordering model;
and carrying out fusion processing on the tree models correspondingly selected from the information ordering models to obtain the target ordering model.
7. The method according to claim 6, wherein the obtaining the information ordering accuracy of at least two information ordering models comprises:
collecting a test sample set; the test sample set comprises sample multimedia information and actual sequencing labels corresponding to the sample multimedia information;
respectively inputting the sample multimedia information into each information sequencing model to obtain a prediction sequencing label of each information sequencing model on the sample multimedia information;
and respectively acquiring errors between the prediction ordering labels of the information ordering models on the sample multimedia information and the actual ordering labels of the sample multimedia information, and acquiring the information ordering accuracy of the information ordering models according to the errors.
8. A multimedia information pushing apparatus, comprising:
the multimedia information acquisition unit is configured to acquire candidate multimedia information corresponding to an account to be pushed;
a ranking tag determining unit configured to perform ranking of each of the candidate multimedia information by the target ranking model; the target sorting models are obtained by fusing tree models selected from at least two information sorting models, sorting labels used for outputting multimedia information are obtained by fusing the target sorting models according to the number of the tree models selected from each information sorting model and the ratio of the number of the tree models selected from each information sorting model; the number proportion of the tree models selected from the information ordering models is obtained by normalizing the information ordering accuracy of the information ordering models;
the multimedia information ordering unit is configured to order the candidate multimedia information according to the ordering label to obtain ordered candidate multimedia information;
and the multimedia information pushing unit is configured to push the sorted candidate multimedia information to the account to be pushed.
9. The multimedia information pushing apparatus according to claim 8, wherein the multimedia information obtaining unit is further configured to perform receiving an information pushing request of the account to be pushed; the information pushing request comprises key information input by triggering of the account to be pushed; and recalling the multimedia information with the information matching degree between the key information and the information base higher than the preset matching degree as candidate multimedia information corresponding to the account to be pushed.
10. The multimedia information pushing apparatus according to claim 8, wherein the multimedia information obtaining unit is further configured to perform obtaining account characteristics of an account to be pushed; acquiring a multimedia information category conforming to the account characteristics; and screening out the multimedia information which accords with the account characteristics from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as candidate multimedia information corresponding to the account to be pushed.
11. The multimedia information pushing apparatus according to claim 10, wherein the multimedia information obtaining unit is further configured to perform obtaining information features of the multimedia information corresponding to the multimedia information category; acquiring feature similarity between the account features and each information feature; and screening out the multimedia information with the feature similarity larger than the preset similarity from the multimedia information corresponding to the multimedia information category, and taking the multimedia information as the multimedia information conforming to the account feature.
12. The multimedia information pushing device according to claim 8, wherein the multimedia information pushing unit is further configured to perform sequentially screening a preset number of candidate multimedia information from the sorted candidate multimedia information as target multimedia information corresponding to the account to be pushed; pushing the target multimedia information to the account to be pushed according to a preset pushing frequency.
13. The multimedia information pushing apparatus according to any one of claims 8 to 12, further comprising a model fusion unit configured to perform obtaining an information ranking accuracy of at least two information ranking models, determining a number of tree models selected from each of the information ranking models according to the information ranking accuracy; according to a preset selection sequence, correspondingly selecting tree models corresponding to the number of the tree models from the tree models contained in each information ordering model, and obtaining the tree models correspondingly selected from each information ordering model; and carrying out fusion processing on the tree models correspondingly selected from the information ordering models to obtain the target ordering model.
14. The multimedia information pushing device of claim 13, wherein the model fusion unit is further configured to perform collecting a set of test samples; the test sample set comprises sample multimedia information and actual sequencing labels corresponding to the sample multimedia information; respectively inputting the sample multimedia information into each information sequencing model to obtain a prediction sequencing label of each information sequencing model on the sample multimedia information; and respectively acquiring errors between the prediction ordering labels of the information ordering models on the sample multimedia information and the actual ordering labels of the sample multimedia information, and acquiring the information ordering accuracy of the information ordering models according to the errors.
15. 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 multimedia information pushing method of any of claims 1 to 7.
16. A storage medium, which when executed by a processor of a server, enables the server to perform the multimedia information pushing method of any one of claims 1 to 7.
CN202010848523.7A 2020-08-21 2020-08-21 Multimedia information pushing method, device, server and storage medium Active CN112000821B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010848523.7A CN112000821B (en) 2020-08-21 2020-08-21 Multimedia information pushing method, device, server and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010848523.7A CN112000821B (en) 2020-08-21 2020-08-21 Multimedia information pushing method, device, server and storage medium

Publications (2)

Publication Number Publication Date
CN112000821A CN112000821A (en) 2020-11-27
CN112000821B true CN112000821B (en) 2024-03-26

Family

ID=73474028

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010848523.7A Active CN112000821B (en) 2020-08-21 2020-08-21 Multimedia information pushing method, device, server and storage medium

Country Status (1)

Country Link
CN (1) CN112000821B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110263235A (en) * 2019-06-05 2019-09-20 深圳市腾讯计算机系统有限公司 Information pushes object updating method, device and computer equipment
CN110457534A (en) * 2019-07-30 2019-11-15 深圳市腾讯网域计算机网络有限公司 A kind of data processing method based on artificial intelligence, device, terminal and medium
CN111339327A (en) * 2020-02-20 2020-06-26 北京达佳互联信息技术有限公司 Work recommendation method and device, server and storage medium

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
RU2666336C1 (en) * 2017-08-01 2018-09-06 Общество С Ограниченной Ответственностью "Яндекс" Method and system for recommendation of media-objects

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110263235A (en) * 2019-06-05 2019-09-20 深圳市腾讯计算机系统有限公司 Information pushes object updating method, device and computer equipment
CN110457534A (en) * 2019-07-30 2019-11-15 深圳市腾讯网域计算机网络有限公司 A kind of data processing method based on artificial intelligence, device, terminal and medium
CN111339327A (en) * 2020-02-20 2020-06-26 北京达佳互联信息技术有限公司 Work recommendation method and device, server and storage medium

Also Published As

Publication number Publication date
CN112000821A (en) 2020-11-27

Similar Documents

Publication Publication Date Title
CN110837579B (en) Video classification method, apparatus, computer and readable storage medium
CN110909182B (en) Multimedia resource searching method, device, computer equipment and storage medium
CN110321422B (en) Method for training model on line, pushing method, device and equipment
CN111125435B (en) Video tag determination method and device and computer equipment
CN112364204B (en) Video searching method, device, computer equipment and storage medium
CN107918657B (en) Data source matching method and device
CN111783712A (en) Video processing method, device, equipment and medium
CN111708909B (en) Video tag adding method and device, electronic equipment and computer readable storage medium
CN112749330B (en) Information pushing method, device, computer equipment and storage medium
CN111429161B (en) Feature extraction method, feature extraction device, storage medium and electronic equipment
CN111984824A (en) Multi-mode-based video recommendation method
CN111159563A (en) Method, device and equipment for determining user interest point information and storage medium
CN115712780A (en) Information pushing method and device based on cloud computing and big data
CN112364184A (en) Method, device, server and storage medium for ordering multimedia data
CN113886711B (en) Content recommendation method, device, server and storage medium
CN113220974B (en) Click rate prediction model training and search recall method, device, equipment and medium
CN112115354B (en) Information processing method, device, server and storage medium
CN112000821B (en) Multimedia information pushing method, device, server and storage medium
CN114282119B (en) Scientific and technological information resource retrieval method and system based on heterogeneous information network
CN117216362A (en) Content recommendation method, device, apparatus, medium and program product
CN115269998A (en) Information recommendation method and device, electronic equipment and storage medium
CN104965853B (en) The recommendation of polymeric type application, the multi-party mthods, systems and devices for recommending source polymerization
CN111651643A (en) Processing method of candidate content and related equipment
CN112711717A (en) Room product searching method and device and electronic equipment
CN112000888B (en) Information pushing method, device, server 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
GR01 Patent grant
GR01 Patent grant