CN111859220A - Method and device for displaying information - Google Patents

Method and device for displaying information Download PDF

Info

Publication number
CN111859220A
CN111859220A CN202010721543.8A CN202010721543A CN111859220A CN 111859220 A CN111859220 A CN 111859220A CN 202010721543 A CN202010721543 A CN 202010721543A CN 111859220 A CN111859220 A CN 111859220A
Authority
CN
China
Prior art keywords
information
target object
representation
expression
data index
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
CN202010721543.8A
Other languages
Chinese (zh)
Inventor
刘乾超
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 ByteDance Network Technology Co Ltd
Original Assignee
Beijing ByteDance Network 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 ByteDance Network Technology Co Ltd filed Critical Beijing ByteDance Network Technology Co Ltd
Priority to CN202010721543.8A priority Critical patent/CN111859220A/en
Publication of CN111859220A publication Critical patent/CN111859220A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/957Browsing optimisation, e.g. caching or content distillation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/22Matching criteria, e.g. proximity measures
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/084Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Databases & Information Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The embodiment of the disclosure discloses a method and a device for displaying information. One embodiment of the method comprises: acquiring an information set for expression of a target object, wherein the information for expression in the information set for expression is used for expressing the target object; predicting data index values of the information for expression in the information set for expression according to the target object and the information for expression by using a pre-trained prediction model, wherein the prediction model is a double-tower model; selecting the presentation information and the presentation from the presentation information set based on the corresponding data index values. The embodiment realizes the pre-screening of the information for representation of the target object so as to show the information for representation in a targeted manner.

Description

Method and device for displaying information
Technical Field
The embodiment of the disclosure relates to the technical field of computers, in particular to a method and a device for displaying information.
Background
With the rapid development of the mobile internet, browsing information through a mobile terminal is one of the current main ways for users to obtain information. For many objects such as products, services, advertisements, etc., various information (such as advertisements, product documents, advertisement documents, content titles, etc.) representing the objects are usually designed, and the information is presented at the user terminal to make the user know the related content of the objects.
Generally, information representing the above objects is generally developed and designed by a professional designer according to the characteristics and the like of the objects. Different information displayed at the user terminal and representing the above objects may generate different display effects (e.g., different numbers of users interacting with the information, etc.).
Disclosure of Invention
The embodiment of the disclosure provides a method and a device for displaying information.
In a first aspect, an embodiment of the present disclosure provides a method for presenting information, the method including: acquiring an information set for expression of a target object, wherein the information for expression in the information set for expression is used for expressing the target object; predicting data index values of the information for expression in the information set for expression according to the target object and the information for expression by using a pre-trained prediction model, wherein the prediction model is a double-tower model; selecting the presentation information and the presentation from the presentation information set based on the corresponding data index values.
In some embodiments, the predicting the data index value of the information for representation according to the target object and the information for representation by using the pre-trained prediction model includes: the target object and the information for representation are encoded by using a prediction model, and a data index value of the information for representation is calculated from the encoding result corresponding to each of the target object and the information for representation.
In some embodiments, the calculating the data index value of the presentation information according to the encoding results corresponding to the target object and the presentation information respectively includes: the similarity of the coding results corresponding to the target object and the representation information is determined, and the data index value of the representation information is calculated according to the similarity.
In some embodiments, the predictive model is trained based on rank learning.
In some embodiments, the prediction model comprises a first coding network for coding the target object, wherein the first coding network comprises an input layer, a feature embedding representation layer, a self-attention layer, an output layer, and a second coding network for coding the active information, wherein the second coding network comprises an input layer, a feature embedding representation layer, a self-attention layer, an output layer.
In some embodiments, the encoding the target object and the representation information using the prediction model respectively includes: encoding the target object according to the attribute information of the target object by using the prediction model, wherein the attribute information of the target object includes but is not limited to: the display content information of the target object, the user group information of the target object, the type information of the target object and the owner information of the target object.
In some embodiments, the selecting the presentation information from the presentation information set and presenting according to the corresponding data index values includes: and selecting the information for expression from the information set for expression and displaying the information at the user terminal according to the data index values respectively corresponding to the data index values and the user behavior information corresponding to the user terminal.
In a second aspect, an embodiment of the present disclosure provides an apparatus for presenting information, the apparatus including: an acquisition unit configured to acquire a set of information for expression of a target object, wherein information for expression in the set of information for expression is used to express the target object; a prediction unit configured to predict a data index value of expression information from a target object and the expression information using a pre-trained prediction model for the expression information in the expression information set, wherein the prediction model is a two-tower model; and a presentation unit configured to select the presentation information from the presentation information set and present the presentation information according to the respective corresponding data index values.
In some embodiments, the above-mentioned prediction unit is further configured to encode the target object and the information for representation respectively by using a prediction model, and to calculate the data index value of the information for representation according to the encoding result corresponding to the target object and the information for representation respectively.
In some embodiments, the prediction unit is further configured to determine similarities of the encoding results corresponding to the target object and the representation information, respectively, and to calculate the data index value of the representation information according to the similarities.
In some embodiments, the predictive model is trained based on rank learning.
In some embodiments, the prediction model includes a first coding network for coding the target object, wherein the first coding network includes an input layer, a feature embedding representation layer, a self-attention layer, and an output layer, and a second coding network for coding the useful information, wherein the second coding network includes an input layer, a feature embedding representation layer, a self-attention layer, and an output layer.
In some embodiments, the prediction unit is further configured to encode the target object according to attribute information of the target object by using a prediction model, wherein the attribute information of the target object includes, but is not limited to: the display content information of the target object, the user group information of the target object, the type information of the target object and the owner information of the target object.
In some embodiments, the presentation unit is further configured to select the presentation information from the presentation information set and present the presentation information at the user terminal according to the corresponding data index value and the user behavior information corresponding to the user terminal.
In a third aspect, an embodiment of the present disclosure provides a server, including: one or more processors; storage means for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method as described in any implementation of the first aspect.
In a fourth aspect, embodiments of the present disclosure provide a computer-readable medium on which a computer program is stored, which computer program, when executed by a processor, implements the method as described in any of the implementations of the first aspect.
According to the method and the device for displaying the information, the data index values corresponding to the representation information of the target object are predicted by using the double-tower model, so that the display effect of the representation information can be pre-estimated before the representation information of the target object is displayed, and the representation information can be flexibly selected for displaying according to the predicted data index values, so that high-quality display of the representation information of the target object is realized.
Drawings
Other features, objects and advantages of the disclosure will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, made with reference to the accompanying drawings in which:
FIG. 1 is an exemplary system architecture diagram in which one embodiment of the present disclosure may be applied;
FIG. 2 is a flow diagram of one embodiment of a method for presenting information in accordance with the present disclosure;
FIG. 3 is a network architecture diagram of a predictive model in an embodiment in accordance with the present disclosure;
FIG. 4 is a schematic diagram of one application scenario of a method for presenting information in accordance with an embodiment of the present disclosure;
FIG. 5 is a flow diagram of yet another embodiment of a method for presenting information in accordance with the present disclosure;
FIG. 6 is a schematic block diagram illustrating one embodiment of an apparatus for presenting information in accordance with the present disclosure;
FIG. 7 is a schematic structural diagram of an electronic device suitable for use in implementing embodiments of the present disclosure.
Detailed Description
The present disclosure is described in further detail below with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the relevant invention and not restrictive of the invention. It should be noted that, for convenience of description, only the portions related to the related invention are shown in the drawings.
It should be noted that, in the present disclosure, the embodiments and features of the embodiments may be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
Fig. 1 illustrates an exemplary architecture 100 to which embodiments of the disclosed method for presenting information or apparatus for presenting information may be applied.
As shown in fig. 1, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
The terminal devices 101, 102, 103 interact with a server 105 via a network 104 to receive or send messages or the like. Various client applications may be installed on the terminal devices 101, 102, 103. For example, browser-like applications, search-like applications, shopping-like applications, instant messaging-like applications, social platform software, information flow-like applications, and the like.
The terminal apparatuses 101, 102, and 103 may be hardware or software. When the terminal devices 101, 102, 103 are hardware, they may be various electronic devices including, but not limited to, smart phones, tablet computers, e-book readers, laptop portable computers, desktop computers, and the like. When the terminal apparatuses 101, 102, 103 are software, they can be installed in the electronic apparatuses listed above. It may be implemented as multiple pieces of software or software modules (e.g., multiple pieces of software or software modules to provide distributed services) or as a single piece of software or software module. And is not particularly limited herein.
The server 105 may be a server that provides various services, such as a backend server that provides support for client applications installed on the terminal devices 101, 102, 103. The server 105 may predict data index values corresponding to the pieces of performance information in the set of performance information of the target object, select pieces of performance information from the set of performance information according to the prediction results, and push the pieces of performance information to the terminal devices 101, 102, and 103, and the terminal devices 101, 102, and 103 may display the pieces of received pushed information.
It should be noted that the method for presenting information provided by the embodiment of the present disclosure is generally performed by the server 105, and accordingly, the apparatus for presenting information is generally disposed in the server 105.
In some cases, the terminal apparatuses 101, 102, and 103 may predict data index values corresponding to the respective pieces of performance information in the set of performance information of the target object, and select and present the performance information from the set of performance information according to the prediction result. In this case, the method for presenting information may be executed by the terminal apparatuses 101, 102, and 103, and accordingly, the apparatus for presenting information may be provided in the terminal apparatuses 101, 102, and 103. At this point, the exemplary system architecture 100 may not have the server 105 and the network 104.
The server 105 may be hardware or software. When the server 105 is hardware, it may be implemented as a distributed server cluster composed of a plurality of servers, or may be implemented as a single server. When the server 105 is software, it may be implemented as multiple pieces of software or software modules (e.g., multiple pieces of software or software modules used to provide distributed services), or as a single piece of software or software module. And is not particularly limited herein.
It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
With continued reference to FIG. 2, a flow 200 of one embodiment of a method for presenting information in accordance with the present disclosure is shown. The method for presenting information comprises the following steps:
in step 201, an information set for expression of a target object is acquired.
In the present embodiment, the target object may be various objects. For example, the target object may be various physical objects (e.g., various items, etc.). As another example, the target object may also be various objects that do not have an entity (e.g., various types of services, advertisements, promos, etc.).
The presentation information in the presentation information set may be used to present the target object. The presentation information may be various types of information. For example, the information for presentation may be text, image, audio-video, or the like.
The information for representing the target object may be different depending on the target object. For example, when the target object is an article, the information for expression may be an advertisement, a poster, a product document, an advertisement word, or the like of the article. For example, when the target object is an advertisement, the information for expression may be an advertisement word, a document, a title, or the like of the advertisement.
In this embodiment, the execution subject of the method for presenting information (e.g., the server 105 shown in fig. 1) may obtain the pre-stored information set for representation of the target object from the local, or may obtain the information set for representation of the target object from another storage device.
For example, the execution body may acquire the presentation information set of the target object from a terminal used by the owner of the target object. Wherein the owner of the target object may be a developer, a manufacturer, an owner, a propagator, etc. of the target object.
Step 202, for the information for expression in the information set for expression, a data index value of the information for expression is predicted from the target object and the information for expression using a prediction model trained in advance.
In this embodiment, the pre-trained predictive model may be a two-tower model. The double tower model structurally distinguishes two input sources explicitly. The two input sources of the prediction model correspond to the target object and the information for representation of the target object, respectively, and the output of the prediction model may be a data index value of the information for representation.
Wherein one input source of the predictive model may input information characterizing the target object and another input source of the predictive model may input information characterizing the target object.
The information for representing the target object can be flexibly set according to the actual application scene and the application requirement. The information used to characterize the target object may also vary from target object to target object. For example, the information for characterizing the target object may be attribute information of the target object. Wherein the attribute indicated by the attribute information of the target object may be specified in advance by a technician. For example, when the target object is a broadcast video, the information for representing the target object may be the content of the target object itself, that is, the broadcast video, or the information for representing the target object may be information such as a keyword extracted from the broadcast video.
The data index corresponding to the data index value output by the prediction model can be specified in advance by technicians according to actual application requirements. The data index value of the information for performance may be used to measure the performance of the information for performance in the aspect indicated by the data index. As an example, the data index value may be one or more of a click rate, a browsing volume, a conversion rate, and the like.
In this embodiment, the prediction model may be trained by a technician in advance based on a machine learning method. For example, the prediction model can be trained by the following steps:
step one, training data is obtained.
In this step, the training data may include a preset object information set, an information set for performance that is shown in history of an object indicated by each object information in the preset object information set, and a data index value of each information for performance in the information set for performance corresponding to the object indicated by each object information. Wherein the object information may be used to characterize the object. The training data may be obtained from a designated data platform or database, or the like.
And step two, training the initial model by using the acquired training data based on a machine learning method, and determining the trained initial model as a prediction model.
In this step, the initial model may be various double tower models. For example, the initial model may be an existing open source, some two-tower model that has been trained or untrained. Such as DSSM (Deep Structured semantic models), etc. For another example, the initial model may be a two-tower model built by a technician using a deep learning framework such as Keras, tensrflow, or the like.
The initial model may then be trained using the training data. Specifically, the object information in the object information set and the representation information in the representation information set shown in the corresponding history may be used as the input of the initial model, the data index value corresponding to the input representation information may be used as the expected output of the initial model, and the parameters of each network layer of the initial model may be continuously adjusted by using algorithms such as gradient descent and back propagation based on a preset loss function, so as to train the initial model and obtain a trained prediction model. The loss function can be designed by a technician in advance according to the actual application requirements.
In some optional implementation manners of this embodiment, for the information for representation in the information set for representation of the target object, the target object and the information for representation may be encoded (Embedding) by using a prediction model, and an encoding result of the target object and an encoding result of the information for representation may be obtained. Then, the prediction model may calculate the data index value of the expression information based on the obtained encoding result of the target object and the encoding result of the expression information.
Alternatively, the prediction model may comprise a first coding network for coding the target object and a second coding network for coding the activity information.
The first coding network may include an input layer, a Feature Embedding (Feature Embedding) representation layer, a self-attention layer (self-attention layer), and an output layer. The second coding network may also include an input layer, a feature embedding representation layer, a self-attention layer, and an output layer. The feature embedding representation layer can be used for converting the information input by the input layer into a feature representation with a preset size. The self-attention layer can capture the internal relevance of the input information and quickly extract the important features of the input information. The output layer may output a vector of information characterizing the input.
The first coding network and the second coding network may be various existing networks for information coding, or may be coding networks obtained by adjusting various existing networks for information coding. For example, the first coding network and the second coding network may be constructed based on word2vec, DNN (Deep Neural Networks), CNN (Convolutional Neural Networks), LSTM (Long Short-Term Memory network), and the like.
In this case, the prediction model may further include an output network for calculating a data index value of the presentation information from the coding results corresponding to the target object and the presentation information, respectively.
Therefore, the two input sources can be structurally distinguished by the double-tower model to respectively encode the target object and the information for representation, so that the calculation amount for distinguishing the boundaries of different input sources is saved, the complexity and the calculation amount of the prediction model are reduced, and the pre-storage speed of the prediction model is further improved.
With continued reference to fig. 3, fig. 3 is a schematic diagram 300 of the network structure of the predictive model of the present embodiment. The network structure of the prediction model illustrated in fig. 3 is only an example, and should not bring any limitation to the network structure of the pre-stored model in the embodiment of the present disclosure.
As shown in fig. 3, the prediction model may include two input sources, a first encoding network 301 and a second encoding network 302, respectively. One of the coding networks is used for coding the target object and generating a vector for representing the target object, and the other coding network is used for coding the information for representing the target object and generating a vector for representing the information for representing the target object. Each encoding network may include an input layer, a feature embedding representation layer, a self-attention layer, and an output layer.
The predictive model may also include a data index output layer. The data index output layer is configured to calculate an inner product of a vector characterizing the target object and a vector of information for representing the target object, which are output by the first encoding network 301 and the second encoding network 302, respectively, to output a predicted data index value of the information for representing the target object.
It should be noted that, for convenience of describing the two coding networks, the two coding networks are named as a first coding network and a second coding network, respectively, and those skilled in the art should understand that the first or the second does not constitute a specific limitation on the coding networks.
In this embodiment, when the target object is encoded, an encoding result of the target object may be obtained according to information used for characterizing the target object.
Alternatively, the information for characterizing the target object may be attribute information of the target object. At this time, the prediction model may encode the target object according to the attribute information of the target object. Wherein, the attribute information of the target object includes but is not limited to: the display content information of the target object, the user group information of the target object, the type information of the target object and the owner information of the target object.
The display content information of the target object may refer to the display content corresponding to the target object. For example, when the target object is an article, the display content information of the article may be an image, a video, or the like for displaying the article. For example, when the target object is an advertisement video, the display content information of the advertisement video may be the advertisement video itself or a partial image extracted from the advertisement video.
The user population information of the target object may indicate a user population to which the target object is directed. It should be understood that the user group to which the target object is directed may be different according to different application requirements. The user group information may be various information that can be used to characterize a user group. For example, the attribute information may be common to the users in the user group.
The type information of the target object may indicate a type to which the target object belongs. It should be understood that different types may be set by different dividing methods according to different application requirements in practice. As an example, the type of the target object may be determined according to a domain to which the target object belongs.
The owner information of the target object may indicate the owner of the target object. The owner of the target object may be the developer, manufacturer, owner, propagator, etc. of the target object.
In particular, the prediction model may encode information for characterizing the target object based on various existing encoding techniques (e.g., one-hot-only encoding, etc.). For example, when the target object is characterized using N attribute information of the target object, the N attribute information of the target object may be mapped into one N-dimensional vector as an encoding result of the target object. Wherein each dimension in the N-dimensional vector may represent one attribute information of the target object.
In this way, the target object can be encoded using one or more items of attribute information of the display content information, the user group information, the genre information, and the owner information of the target object, and the data index value of the information for representing the target object can be predicted using these items of attribute information.
Accordingly, when encoding the information for expression of the target object, the prediction model can encode the information for expression based on various conventional encoding techniques. It should be understood that the encoding technique used may vary depending on the information used for presentation.
For example, when the information for presentation is audio/video, the prediction model may encode the information for presentation based on an existing encoding technique for audio/video (such as various audio/video feature extraction techniques). For example, when the presentation information is a text such as an advertisement word, a pattern, or a title of an advertisement, the prediction model may encode the presentation information by using various encoding techniques for the text (for example, various word vector representation methods).
Alternatively, after encoding the target object and the information for expression, respectively, the prediction model may calculate the degree of similarity between the encoding result of the target object and the encoding result of the information for expression, and calculate the data index value of the information for expression based on the calculated degree of similarity.
The similarity between the encoding result of the target object and the encoding result of the information for expression may be calculated using various similarity calculation methods (e.g., euclidean distance, cosine distance, parameterized similarity testimony, etc.). The correspondence between the similarity between the coding result of the design target object and the coding result of the presentation information and the data index value of the presentation information can also be flexibly used according to actual application requirements. For example, the data index value representing the current information may be represented directly using the similarity between the encoding result of the target object and the encoding result of the current information. In this case, the data index value representing the useful information is represented by using the similarity between the encoding result of the target object and the encoding result of the useful information, and the calculation process of the data index value can be simplified, so that the complexity of the prediction model can be reduced, and the prediction efficiency of the prediction model can be improved.
Alternatively, the similarity between the coding result representing the target object and the coding result of the presentation information and the inner product of the coding results corresponding to the target object and the presentation information may be determined. In this case, after the inner product of the encoding result of the target object and the encoding result of the presentation information is calculated, the data index value of the presentation information may be calculated based on the calculated inner product. And representing the similarity between the coding result of the target object and the coding result of the information for representation by using the inner product of the coding result of the target object and the coding result of the information for representation so as to simplify the calculation process of the similarity, further reduce the complexity of the prediction model and improve the prediction efficiency of the prediction model.
In some optional implementations of this embodiment, the predictive model may be trained based on rank learning. Among them, Rank learning (Learn to Rank) may be various existing techniques for ranking applied in machine learning. For example, ranking learning includes, but is not limited to, Pointwise, Pairwise, Listwise, and the like. The training process can be simplified by obtaining the prediction model through sequencing learning training, so that the training speed of the prediction model is improved.
It should be noted that the aforementioned ordering learning such as Pointwise, Pairwise, Listwise, etc. is a well-known technology that is widely researched and applied at present, and is not described herein again.
Taking Pairwise as an example, the prediction model can be obtained by training as follows:
step one, training data is obtained.
In this step, the training data may include a preset object information set, a plurality of expression information pairs each including any two pieces of expression information in an expression information set that is historically displayed on an object indicated by each piece of object information in the preset object information set, and indication information indicating a magnitude relationship between data index values respectively corresponding to the two pieces of expression information in each of the expression information pairs. Wherein the object information may be used to characterize the object. The training data may be obtained from a designated data platform or database, or the like.
And step two, training the initial model by using the acquired training data based on a machine learning method, and determining the trained initial model as a prediction model.
In this step, the initial model may be various double tower models. For example, the initial model may be an existing open source, some two-tower model that has been trained or untrained. Such as DSSM (Deep Structured semantic models), etc. For another example, the initial model may be a two-tower model built by a technician using a deep learning framework such as Keras, tensrflow, or the like.
The initial model may then be trained using the training data. Specifically, in the training process, the target information in the target information set and one piece of performance information in the pair of performance information corresponding to the target information may be input as an initial model to obtain a first predicted value corresponding to the performance information, the other piece of performance information in the pair of performance information corresponding to the target information may be input as an initial model to obtain a second predicted value corresponding to the other piece of performance information, and then parameters of each network layer of the initial model may be adjusted according to a preset loss function so that a magnitude relationship between the first predicted value and the second predicted value is consistent with a magnitude relationship between the two pieces of performance information in the pair of performance information corresponding to the target information. The loss function can be designed by a technician in advance according to the actual application requirements.
In some optional implementations of this embodiment, the loss function of the prediction model may be hindeloss. By using the Hinge Loss, the dependency on the number of training data can be reduced, and the training efficiency of the prediction model is improved.
Step 203, selecting the presentation information and the presentation from the presentation information set according to the corresponding data index values.
In this embodiment, the presentation information may be flexibly selected from the presentation information set for presentation according to the data index values corresponding to the presentation information, respectively, based on different application requirements and application scenarios.
For example, the presentation information may be selected from the presentation information set in descending order of the data index value. For another example, the presentation information having the corresponding data index value in the predetermined section may be selected from the presentation information set.
The execution main body may push the selected presentation information to the user terminal after selecting the presentation information from the presentation information set, so that the user terminal may present the received presentation information. The user terminal may be a terminal used by the owner of the target object, or may be each user terminal corresponding to a group of people designated in advance.
Optionally, the other related information of the target object and the selected information for performance may be pushed to the user terminal together for presentation. For example, when the target object is an advertisement video and the presentation information is a document of the advertisement video, the advertisement video and the selected document may be displayed on the user terminal in combination.
With continued reference to fig. 4, fig. 4 is an illustrative application scenario 400 of the method for presenting information in accordance with the present embodiment. In the application scenario of fig. 4, an advertisement 401 to be presented and an advertisement copy set 402 composed of a plurality of advertisement copies matching the advertisement 401 may be obtained first. Then, the advertisement 401 and each advertisement pattern 4021 in the advertisement pattern set 402 may be input to the pre-trained prediction model 403 belonging to the double tower model, so as to obtain a click rate 4041 corresponding to the advertisement pattern 4021. The click-through rates corresponding to the various advertising copy in the set of advertising copies 402 comprise a set of click-through rates 404. Then, the advertisement copy with the highest click rate may be selected from the advertisement copy set, and the selected advertisement 401 with the highest click rate may be pushed to the user terminal 405 for display.
The method provided by the above embodiment of the present disclosure predicts the data index values of the representation information of the target object respectively by using the pre-trained prediction model belonging to the double-tower model, and further selects the representation information from the representation information set for display according to the data index values corresponding to the representation information respectively according to the requirement of the data index values of the representation information in practical application. Therefore, targeted information display can be achieved, the condition that useless information display appears due to the fact that all information for performance is displayed blindly is avoided, cost can be saved for an information display end and an information sending end, and information display efficiency and quality are improved.
With further reference to fig. 5, a flow 500 of yet another embodiment of a method for presenting information is illustrated. The process 500 of the method for presenting information includes the following steps:
in step 501, an information set for expression of a target object is acquired.
Step 502 is to predict the data index value of the expression information from the target object and the expression information by using a prediction model trained in advance for the expression information in the expression information set.
The specific implementation process of steps 501 and 502 may refer to the related description of steps 201 and 202 in the corresponding embodiment of fig. 2, and will not be described herein again.
And 503, selecting the information for expression from the information set for expression and displaying the information at the user terminal according to the data index values corresponding to the data index values and the user behavior information corresponding to the user terminal.
In this embodiment, the user behavior information corresponding to the user terminal may be user behavior information of a user using the user terminal. The user behavior information can be acquired from a database or a data platform according to actual application requirements. For example, the user behavior information may record categories of information that the user prefers to view, times when the information is viewed frequently, and so forth.
After the user behavior information is acquired, the presentation information may be selected from the presentation information set by using various selection methods according to the data index values corresponding to the presentation information in the presentation information set and the user behavior information of the user terminal, and the selected presentation information may be displayed at the user terminal.
For example, the matching degree between each piece of performance information in the set of performance information and the user behavior information may be determined, and then the performance information having the corresponding matching degree greater than a preset threshold may be selected as the candidate performance information. Then, the candidate presentation information having the corresponding data index value belonging to the preset section may be selected from the candidate presentation information and presented in the user terminal.
For another example, after the matching degree between each piece of performance information in the set of performance information and the user behavior information is determined, a weighted sum of a data index value corresponding to each piece of performance information in the set of performance information and a corresponding matching value may be calculated, and then the performance information belonging to a predetermined section and the corresponding weighted sum may be selected from the set of performance information and displayed on the user terminal. The weight corresponding to the data index and the weight corresponding to the user behavior information may be preset by a technician.
The method provided by the above embodiment of the present disclosure predicts the data index value of each presentation information in advance, so as to select the presentation information with pertinence according to the data index value, and further selects the presentation information to be displayed at the user terminal by combining the user behavior information corresponding to the user terminal on the basis, so that personalized presentation information display for different user terminals can be realized on the basis of ensuring the data index value of the displayed presentation information. .
With further reference to fig. 6, as an implementation of the methods shown in the above-mentioned figures, the present disclosure provides an embodiment of an apparatus for presenting information, which corresponds to the method embodiment shown in fig. 2, and which is particularly applicable in various electronic devices.
As shown in fig. 6, the apparatus 600 for presenting information provided in this embodiment includes an obtaining unit 601, a predicting unit 602, and a presenting unit 603. Wherein the acquisition unit is configured to acquire an information set for expression of the target object, wherein information for expression in the information set for expression is used for expressing the target object; a prediction unit configured to predict a data index value of expression information from a target object and the expression information using a pre-trained prediction model for the expression information in the expression information set, wherein the prediction model is a two-tower model; and a presentation unit configured to select the presentation information from the presentation information set and present the presentation information according to the respective corresponding data index values.
In the present embodiment, in the apparatus 600 for presenting information: the specific processing of the obtaining unit 601, the predicting unit 602, and the displaying unit 603 and the technical effects thereof can refer to the related descriptions of step 201, step 202, and step 203 in the corresponding embodiment of fig. 2, which are not repeated herein.
In some optional implementations of the embodiment, the prediction unit 602 is further configured to encode the target object and the representation information by using a prediction model, and calculate the data index value of the representation information according to the encoding result corresponding to the target object and the representation information.
In some optional implementations of the embodiment, the prediction unit 602 is further configured to determine similarities of the encoding results corresponding to the target object and the information for representation, respectively, and calculate the data index value of the information for representation using the similarities.
In some optional implementations of the present embodiment, the prediction model is obtained based on a rank learning training.
In some optional implementations of this embodiment, the prediction model includes a first coding network for coding the target object and a second coding network for coding the active information, where the first coding network includes an input layer, a feature embedding representation layer, a self-attention layer, and an output layer, and the second coding network includes an input layer, a feature embedding representation layer, a self-attention layer, and an output layer.
In some optional implementations of the present embodiment, the prediction unit 602 is further configured to encode the target object according to attribute information of the target object by using a prediction model, where the attribute information of the target object includes, but is not limited to: the display content information of the target object, the user group information of the target object, the type information of the target object and the owner information of the target object.
In some optional implementation manners of this embodiment, the presenting unit 603 is further configured to select the presentation information from the presentation information set and present the presentation information at the user terminal according to the corresponding data index value and the user behavior information corresponding to the user terminal.
The apparatus provided by the above embodiment of the present disclosure acquires, by the acquisition unit, an information set for expression of a target object, wherein information for expression in the information set for expression is used for expressing the target object; a prediction unit that predicts a data index value of the information for expression in the information for expression set from the target object and the information for expression using a pre-trained prediction model, wherein the prediction model is a two-tower model; the presentation unit selects presentation information from the presentation information set and presents the presentation information according to the data index values respectively corresponding to the presentation information set. In this way, by predicting the data index value of each piece of presentation information in the presentation information set of the target object, the presentation information set is screened in advance based on the data index value, and the targeted presentation of the presentation information of the target object is realized.
Referring now to FIG. 7, a block diagram of an electronic device (e.g., the server of FIG. 1) 700 suitable for use in implementing embodiments of the present disclosure is shown. The server shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 7, electronic device 700 may include a processing means (e.g., central processing unit, graphics processor, etc.) 701 that may perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)702 or a program loaded from storage 708 into a Random Access Memory (RAM) 703. In the RAM703, various programs and data necessary for the operation of the electronic apparatus 700 are also stored. The processing device 701, the ROM 702, and the RAM703 are connected to each other by a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.
Generally, the following devices may be connected to the I/O interface 705: input devices 706 including, for example, a touch screen, touch pad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; an output device 707 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage 708 including, for example, magnetic tape, hard disk, etc.; and a communication device 709. The communication means 709 may allow the electronic device 700 to communicate wirelessly or by wire with other devices to exchange data. While fig. 7 illustrates an electronic device 700 having various means, it is to be understood that not all illustrated means are required to be implemented or provided. More or fewer devices may alternatively be implemented or provided. Each block shown in fig. 7 may represent one device or may represent multiple devices as desired.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such embodiments, the computer program may be downloaded and installed from a network via the communication means 709, or may be installed from the storage means 708, or may be installed from the ROM 702. The computer program, when executed by the processing device 701, performs the above-described functions defined in the methods of embodiments of the present disclosure.
It should be noted that the computer readable medium described in the embodiments of the present disclosure may be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In embodiments of the disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In embodiments of the present disclosure, however, a computer readable signal medium may comprise a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.
The computer readable medium may be embodied in the server; or may exist separately and not be assembled into the server. The computer readable medium carries one or more programs which, when executed by the server, cause the server to: acquiring an information set for expression of a target object, wherein the information for expression in the information set for expression is used for expressing the target object; predicting data index values of the information for expression in the information set for expression according to the target object and the information for expression by using a pre-trained prediction model, wherein the prediction model is a double-tower model; selecting the presentation information and the presentation from the presentation information set based on the corresponding data index values.
Computer program code for carrying out operations for embodiments of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units described in the embodiments of the present disclosure may be implemented by software or hardware. The described units may also be provided in a processor, and may be described as: a processor includes an acquisition unit, a prediction unit, and a presentation unit. Here, the names of these units do not constitute a limitation to the unit itself in some cases, and for example, the acquisition unit may also be described as a "unit that acquires an information set for representation of a target object".
The foregoing description is only exemplary of the preferred embodiments of the disclosure and is illustrative of the principles of the technology employed. It will be appreciated by those skilled in the art that the scope of the invention in the embodiments of the present disclosure is not limited to the specific combination of the above-mentioned features, but also encompasses other embodiments in which any combination of the above-mentioned features or their equivalents is made without departing from the inventive concept as defined above. For example, the above features and (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure are mutually replaced to form the technical solution.

Claims (10)

1. A method for presenting information, comprising:
acquiring a presentation information set of a target object, wherein presentation information in the presentation information set is used for presenting the target object;
predicting data index values of the representation information in the representation information set according to the target object and the representation information by using a pre-trained prediction model, wherein the prediction model is a double-tower model;
and selecting the active representation information from the active representation information set and displaying the active representation information according to the corresponding data index values.
2. The method according to claim 1, wherein the predicting the data index value of the information for expression from the target object and the information for expression using a pre-trained prediction model comprises:
the target object and the information for representation are encoded by the prediction model, and a data index value of the information for representation is calculated from the encoding result corresponding to each of the target object and the information for representation.
3. The method according to claim 2, wherein said calculating a data index value of the presentation information based on the coding results corresponding to the target object and the presentation information, respectively, comprises:
and determining the similarity of the coding results corresponding to the target object and the representation information respectively, and calculating the data index value of the representation information according to the similarity.
4. The method of claim 1, wherein the predictive model is trained based on rank learning.
5. The method of claim 2, wherein the prediction model comprises a first coding network for coding the target object, wherein the first coding network comprises an input layer, a feature embedded representation layer, a self-attention layer, an output layer, and a second coding network for coding active information, wherein the second coding network comprises an input layer, a feature embedded representation layer, a self-attention layer, an output layer.
6. The method according to claim 2, wherein said encoding the target object and the representation information using the prediction model, respectively, comprises:
encoding the target object according to the attribute information of the target object by using the prediction model, wherein the attribute information of the target object includes but is not limited to: the display content information of the target object, the user group information of the target object, the type information of the target object and the owner information of the target object.
7. A method as in claim 1, wherein said selecting presentation information from the presentation information set based on the respective corresponding data indicator values comprises:
and selecting the information for expression from the information set for expression and displaying the information at the user terminal according to the data index values respectively corresponding to the data index values and the user behavior information corresponding to the user terminal.
8. An apparatus for presenting information, wherein the apparatus comprises:
an acquisition unit configured to acquire a set of information for expression of a target object, wherein information for expression in the set of information for expression is used to express the target object;
a prediction unit configured to predict a data index value of the representation information from the target object and the representation information by using a pre-trained prediction model for the representation information in the representation information set, wherein the prediction model is a two-tower model;
and the display unit is configured to select the active representation information from the active representation information set and display the active representation information according to the corresponding data index values.
9. A server, comprising:
one or more processors;
a storage device having one or more programs stored thereon;
when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.
10. A computer-readable medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1-7.
CN202010721543.8A 2020-07-24 2020-07-24 Method and device for displaying information Pending CN111859220A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010721543.8A CN111859220A (en) 2020-07-24 2020-07-24 Method and device for displaying information

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010721543.8A CN111859220A (en) 2020-07-24 2020-07-24 Method and device for displaying information

Publications (1)

Publication Number Publication Date
CN111859220A true CN111859220A (en) 2020-10-30

Family

ID=72949388

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010721543.8A Pending CN111859220A (en) 2020-07-24 2020-07-24 Method and device for displaying information

Country Status (1)

Country Link
CN (1) CN111859220A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114579869A (en) * 2022-05-05 2022-06-03 腾讯科技(深圳)有限公司 Model training method and related product

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109522483A (en) * 2018-11-14 2019-03-26 北京百度网讯科技有限公司 Method and apparatus for pushed information
CN109981787A (en) * 2019-04-03 2019-07-05 北京字节跳动网络技术有限公司 Method and apparatus for showing information
US20200012953A1 (en) * 2018-07-03 2020-01-09 Baidu Online Network Technology (Beijing) Co., Ltd. Method and apparatus for generating model
CN111008335A (en) * 2019-12-20 2020-04-14 腾讯科技(深圳)有限公司 Information processing method, device, equipment and storage medium
CN111125502A (en) * 2018-10-31 2020-05-08 北京字节跳动网络技术有限公司 Method and apparatus for generating information
CN111143610A (en) * 2019-12-30 2020-05-12 腾讯科技(深圳)有限公司 Content recommendation method and device, electronic equipment and storage medium
CN111414561A (en) * 2019-01-08 2020-07-14 百度在线网络技术(北京)有限公司 Method and apparatus for presenting information

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200012953A1 (en) * 2018-07-03 2020-01-09 Baidu Online Network Technology (Beijing) Co., Ltd. Method and apparatus for generating model
CN111125502A (en) * 2018-10-31 2020-05-08 北京字节跳动网络技术有限公司 Method and apparatus for generating information
CN109522483A (en) * 2018-11-14 2019-03-26 北京百度网讯科技有限公司 Method and apparatus for pushed information
CN111414561A (en) * 2019-01-08 2020-07-14 百度在线网络技术(北京)有限公司 Method and apparatus for presenting information
CN109981787A (en) * 2019-04-03 2019-07-05 北京字节跳动网络技术有限公司 Method and apparatus for showing information
CN111008335A (en) * 2019-12-20 2020-04-14 腾讯科技(深圳)有限公司 Information processing method, device, equipment and storage medium
CN111143610A (en) * 2019-12-30 2020-05-12 腾讯科技(深圳)有限公司 Content recommendation method and device, electronic equipment and storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114579869A (en) * 2022-05-05 2022-06-03 腾讯科技(深圳)有限公司 Model training method and related product
CN114579869B (en) * 2022-05-05 2022-07-22 腾讯科技(深圳)有限公司 Model training method and related product

Similar Documents

Publication Publication Date Title
CN109460514B (en) Method and device for pushing information
CN109460513B (en) Method and apparatus for generating click rate prediction model
US10650311B2 (en) Suggesting resources using context hashing
CN109981785B (en) Method and device for pushing information
CN110489582B (en) Method and device for generating personalized display image and electronic equipment
CN107577807B (en) Method and device for pushing information
CN110688528B (en) Method, apparatus, electronic device, and medium for generating classification information of video
CN113688310B (en) Content recommendation method, device, equipment and storage medium
CN112380331A (en) Information pushing method and device
CN111738010B (en) Method and device for generating semantic matching model
CN111897950A (en) Method and apparatus for generating information
CN112149699A (en) Method and device for generating model and method and device for recognizing image
CN110866040A (en) User portrait generation method, device and system
WO2024099171A1 (en) Video generation method and apparatus
JP2023550211A (en) Method and apparatus for generating text
CN112182281B (en) Audio recommendation method, device and storage medium
CN112328889A (en) Method and device for determining recommended search terms, readable medium and electronic equipment
CN110059172B (en) Method and device for recommending answers based on natural language understanding
CN112182255A (en) Method and apparatus for storing media files and for retrieving media files
JP2024508502A (en) Methods and devices for pushing information
CN111859220A (en) Method and device for displaying information
CN111026849A (en) Data processing method and device
CN109857838B (en) Method and apparatus for generating information
CN111949860B (en) Method and apparatus for generating a relevance determination model
CN113742593A (en) Method and device for pushing information

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