CN111988407B

CN111988407B - Content pushing method and related device

Info

Publication number: CN111988407B
Application number: CN202010843952.5A
Authority: CN
Inventors: 魏望; 王业全
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-08-20
Filing date: 2020-08-20
Publication date: 2023-11-07
Anticipated expiration: 2040-08-20
Also published as: CN111988407A

Abstract

The embodiment of the application discloses a content pushing method and a related device, wherein a first pushing prediction result is obtained by utilizing a network model through obtaining content to be pushed and pushing conditions for pushing the content to be pushed. Based on the above, according to the first pushing prediction result and the difference information of the pushing target corresponding to the content to be pushed, the difference information intuitively shows the possibility of achieving the pushing target through the content to be pushed under the pushing condition. Therefore, according to the possibility represented by the difference information, the pushing mode actually adopted for the content to be pushed can be adjusted and determined in a targeted manner. Therefore, the network model based on the type determination can obtain a first pushing prediction result with high credibility, so that the influence of human experience is avoided, and the effect of efficiently and accurately determining the pushing mode of the content to be pushed is realized based on the first pushing prediction result.

Description

Content pushing method and related device

Technical Field

The present application relates to the field of data processing, and in particular, to a content pushing method and related device.

Background

The content pushing refers to a process of pushing the specified content to the specified object group according to the requirement of the content provider, and the effects of rapidly improving the heat of the specified content, increasing the audience of the specified content and the like can be achieved through the content pushing.

Content pushing is performed in the related art mainly by professionals, i.e., based on experience of professionals, and manually defined simple pushing rules. However, simple rules generally have difficulty meeting the needs of rich and varied users, and even frequent adjustment of the rules still results in a push result that often has difficulty meeting the expectations of content providers.

Disclosure of Invention

In order to solve the technical problems, the application provides a content pushing method and a related device, which can realize the effect of efficiently and accurately determining the pushing mode of the content to be pushed.

The embodiment of the application discloses the following technical scheme:

in one aspect, an embodiment of the present application provides a content pushing method, where the method includes:

acquiring content to be pushed and a pushing condition for pushing the content to be pushed;

determining a corresponding first pushing prediction result through a network model according to the content to be pushed and the pushing condition; the network model is determined according to the type corresponding to the content to be pushed;

and determining the pushing mode of the content to be pushed according to the difference information between the first pushing prediction result and the pushing target corresponding to the content to be pushed.

On the other hand, the embodiment of the application provides a content pushing device, which comprises an acquisition unit and a determination unit:

the acquisition unit is used for acquiring content to be pushed and pushing conditions for pushing the content to be pushed;

the determining unit is used for determining a corresponding first pushing prediction result through a network model according to the content to be pushed and the pushing condition; the network model is determined according to the type corresponding to the content to be pushed;

the determining unit is further configured to determine a pushing manner of the content to be pushed according to difference information between the first pushing prediction result and a pushing target corresponding to the content to be pushed.

In another aspect, an embodiment of the present application provides an apparatus for content pushing, including a processor and a memory:

the memory is used for storing program codes and transmitting the program codes to the processor;

the processor is configured to perform the method of the above aspect according to instructions in the program code.

In another aspect, embodiments of the present application provide a computer-readable storage medium storing a computer program for executing the method described in the above aspect.

In another aspect, embodiments of the present application provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the method described in the above aspect.

According to the technical scheme, the first pushing prediction result is obtained by utilizing the network model by obtaining the content to be pushed and the pushing conditions for pushing the content to be pushed. The network model is determined based on the type corresponding to the content to be pushed, has better identification sensitivity to the pushing characteristics of the content to be pushed conforming to the type, and thus the determined first pushing prediction result is more likely to objectively reflect the actual pushing result of the content to be pushed under the pushing condition and has higher credibility. Based on the above, according to the first pushing prediction result and the difference information of the pushing target corresponding to the content to be pushed, the difference information intuitively shows the possibility of achieving the pushing target through the content to be pushed under the pushing condition. Therefore, according to the possibility represented by the difference information, the pushing mode actually adopted for the content to be pushed can be adjusted and determined in a targeted manner. Therefore, the network model based on the type determination can obtain a first pushing prediction result with high credibility, and the effect of efficiently and accurately determining the pushing mode of the content to be pushed is realized based on the first pushing prediction result.

Drawings

In order to more clearly illustrate the embodiments of the application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic application scenario diagram of a content pushing method according to an embodiment of the present application;

fig. 2 is a flow chart of a content pushing method according to an embodiment of the present application;

FIG. 3 is a flowchart of a network model training method according to an embodiment of the present application;

FIG. 4 is a flowchart of another method for training a network model according to an embodiment of the present application;

FIG. 5 is a flowchart of another method for training a network model according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of a content pushing device according to an embodiment of the present application;

fig. 7 is a schematic structural diagram of a server according to an embodiment of the present application;

fig. 8 is a schematic structural diagram of a terminal device according to an embodiment of the present application.

Detailed Description

Embodiments of the present application are described below with reference to the accompanying drawings.

In the related art, content pushing may be performed based on experience of a professional and manually defined simple pushing rules. This approach relies heavily on the personal experience of professionals, resulting in inefficient, costly pushing, and the volatility of the pushing results is difficult to overcome.

In view of this, the embodiment of the application provides a content pushing method and a related device, so as to improve the satisfaction degree of a content provider for content pushing.

The content pushing method provided by the embodiment of the application is realized based on artificial intelligence, wherein the artificial intelligence (Artificial Intelligence, AI) is a theory, a method, a technology and an application system which simulate, extend and extend human intelligence by using a digital computer or a machine controlled by the digital computer, sense environment, acquire knowledge and acquire an optimal result by using the knowledge. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision.

The artificial intelligence technology is a comprehensive subject, and relates to the technology with wide fields, namely the technology with a hardware level and the technology with a software level. Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.

In the embodiment of the application, the mainly related artificial intelligence software technology comprises the machine learning/deep learning directions and the like. For example, deep Learning (ML) may be involved in Machine Learning (ML), including various types of artificial neural networks (Artificial Neural Network, ANN).

In order to facilitate understanding of the technical scheme of the present application, the content pushing method provided by the embodiment of the present application is described below in connection with an actual application scenario.

The content pushing method provided by the application can be applied to content pushing equipment with data processing capability, such as terminal equipment and servers. The terminal equipment can be a smart phone, a computer, a personal digital assistant (Personal Digital Assistant, PDA), a tablet personal computer and the like; the server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud computing service. The terminal device and the server may be directly or indirectly connected through wired or wireless communication, and the present application is not limited herein.

The cloud computing (closed computing) refers to a delivery and use mode of an IT infrastructure, and refers to obtaining required resources in an on-demand and easily-extensible manner through a network; generalized cloud computing refers to the delivery and usage patterns of services, meaning that the required services are obtained in an on-demand, easily scalable manner over a network. Such services may be IT, software, internet related, or other services. Cloud Computing is a product of fusion of traditional computer and network technology developments such as Grid Computing (Grid Computing), distributed Computing (Distributed Computing), parallel Computing (Parallel Computing), utility Computing (Utility Computing), network storage (Network Storage Technologies), virtualization (Virtualization), load balancing (Load balancing), and the like.

With the development of the internet, real-time data flow and diversification of connected devices, and the promotion of demands of search services, social networks, mobile commerce, open collaboration and the like, cloud computing is rapidly developed. Unlike the previous parallel distributed computing, the generation of cloud computing will promote the revolutionary transformation of the whole internet mode and enterprise management mode in concept.

The content processing device in the embodiment of the application has the machine learning capability. Machine learning is a multi-domain interdisciplinary, involving multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, and the like. It is specially studied how a computer simulates or implements learning behavior of a human to acquire new knowledge or skills, and reorganizes existing knowledge structures to continuously improve own performance. Machine learning is the core of artificial intelligence, a fundamental approach to letting computers have intelligence, which is applied throughout various areas of artificial intelligence. Machine learning and deep learning typically involve techniques such as artificial neural networks. The content pushing method provided by the embodiment of the application mainly relates to application to various artificial neural networks.

The following describes an embodiment of the present application with a server as an execution body.

Referring to fig. 1, fig. 1 is an application scenario schematic diagram of a content pushing method according to an embodiment of the present application. In the application scenario shown in fig. 1, a server 101 is included, and an artificial intelligence based network model is deployed in the server 101 for determining a push prediction result of content to be pushed.

During content pushing, the server 101 acquires content to be pushed and a pushing condition for pushing the content to be pushed. Wherein, the content to be pushed refers to the content that the content provider needs to push, and the expression forms include, but are not limited to: text, image, video, etc. For example, an advertiser may want to push an advertisement, i.e., the content to be pushed is an advertisement, which is shown in video form in FIG. 1. The push condition refers to a requirement that the content provider address to the content to be pushed, including but not limited to: push budget, push period, push direction, etc. In the scenario shown in fig. 1, the pushing conditions may include a placement orientation, a placement period, etc. of the advertisement. The pushing condition may be provided by the advertiser itself, or may be verified based on a cost budget provided by the advertiser, advertisement type, and the like.

Because the content pushing method in the related technology has strong dependence on professionals, a great amount of cost is consumed, and an ideal pushing result cannot be ensured, and the satisfaction degree of a content provider on the content pushing result is not high. Therefore, the embodiment of the application predicts the pushing result of the content to be pushed through the network model to determine the pushing mode of the content to be pushed, thereby improving the satisfaction degree of the content provider on the content pushing result.

In practical application, the server 101 predicts a pushing result of the content to be pushed by using the network model and takes the content to be pushed and a pushing condition for pushing the content to be pushed as inputs of the network model, so as to obtain a first pushing prediction result. The first push prediction result is used for identifying a push prediction result of the content to be pushed under the push condition.

The network model is pre-constructed based on an artificial intelligence technology, is determined according to the type corresponding to the content to be pushed, has better identification sensitivity to the pushing characteristics of the content to be pushed according to the type, and accordingly the determined first pushing prediction result is more likely to objectively show the actual pushing result of the content to be pushed under the pushing condition, and has higher credibility.

In the application scenario shown in fig. 1, the advertisement to be pushed belongs to a video type, so that the advertisement to be pushed can be predicted by using a network model corresponding to the video type, and a first pushing prediction result corresponding to the advertisement can be determined.

After determining the first pushing prediction result, the server 101 may determine difference information between the first pushing prediction result and a pushing target corresponding to the content to be pushed, and determine a pushing manner of the content to be pushed based on the difference information. The difference information can intuitively show the possibility of achieving the pushing target through the content to be pushed under the pushing condition.

And if the first pushing prediction result is determined to reach the pushing target according to the difference information. In this case, the content to be pushed may be pushed directly according to the pushing conditions as a pushing manner.

In the application scenario shown in fig. 1, if it is determined that the first pushing prediction result reaches the pushing target according to the difference information, the advertisement may be pushed directly according to the pushing condition provided by the advertiser as a pushing manner.

And if the first pushing prediction result is determined not to reach the pushing target according to the difference information. In this case, based on the difference information, the pushing conditions can be adjusted in a targeted manner, and the pushing mode actually adopted for the content to be pushed can be determined.

In the application scenario shown in fig. 1, if it is determined that the first pushing prediction result does not reach the pushing target according to the difference information, the pushing condition provided by the advertiser may be adjusted based on the difference information, for example, the manner of adjusting the delivery orientation, extending the delivery period, and the like. The advertisement can be put through a pushing mode after the difference information is determined, so that a user to be put can see the advertisement in the process of using the terminal. For example, fig. 1 illustrates a scenario in which a user may see an advertisement placed while viewing a sporting event using terminal device 102.

From the above, the difference information intuitively shows the possibility of achieving the push target through the content to be pushed under the push condition. Therefore, according to the possibility represented by the difference information, the pushing mode actually adopted for the content to be pushed can be adjusted and determined in a targeted manner. In addition, a first pushing prediction result with high credibility can be determined based on the network model with the type determination, so that the influence of human experience is avoided, and the effect of efficiently and accurately determining the pushing mode of the content to be pushed is realized based on the first pushing prediction result.

The content pushing method provided by the embodiment of the application can be applied before content pushing to determine what pushing mode is actually adopted to push the content to be pushed. The method can also be applied to the content pushing process so as to adjust the pushing mode of the content to be pushed, thereby achieving the pushing target. The content pushing method provided by the embodiment of the application is described below with reference to the accompanying drawings.

Referring to fig. 2, fig. 2 is a flow chart of a content pushing method according to an embodiment of the present application. As shown in fig. 2, the content pushing method includes the steps of:

s201: and acquiring the content to be pushed and the pushing condition for pushing the content to be pushed.

Content to be pushed refers to content that the content provider wants to push, expressed in forms including, but not limited to: text, images, video, etc.

In one possible application scenario, an automobile manufacturer may want to advertise a new car, and the automobile manufacturer may advertise by pushing an advertisement. In this scenario, the content provider is the car manufacturer and the content to be pushed is an advertisement that includes a new car. Wherein, the advertisement forms can be text, images, videos and the like. In another possible application scenario, a composer wants to advertise a new book, and the composer can advertise through a text-to-text article. In this scenario, the content provider is a composer and the content to be pushed is a promotional article that includes content related to the new book. Wherein the promotional articles may include text and images. The present application is merely provided in terms of several possible application scenarios, and in practical application, the application scenarios may be determined according to specific requirements, which are not limited in any way.

The pushing condition refers to a condition according to which the content to be pushed is pushed, and the pushing condition can be a requirement set by a content provider, a pushing mode determined based on budget, cost, requirement and the like provided by the content provider, or a pushing mode adopted when the content to be pushed is pushed last time. In an embodiment of the present application, the pushing conditions include, but are not limited to: push budget, push direction, push period, etc. In practical application, the pushing conditions may be provided by the content provider, or the corresponding pushing conditions may be determined according to the content pushing requirement of the content provider, which is not limited herein. For example, in the application scenario of pushing advertisements, the pushing conditions may include: ad listing, modifying budget limits, modifying targeting, modifying material, etc.

After the server obtains the content to be pushed and the pushing condition for pushing the content to be pushed, the content to be pushed and the pushing condition can be used as input of a network model, and the network model is used for determining what pushing mode should be adopted in practice so as to meet the pushing target of the content provider.

S202: and determining a corresponding first pushing prediction result through a network model according to the content to be pushed and the pushing condition.

In the related art, content pushing is performed based on experience of professionals and manually defined simple pushing rules, which is very dependent on personal experience of professionals, so that pushing is inefficient and costly, and fluctuation of pushing results is difficult to overcome.

Taking an application scene of advertisement pushing as an example, an advertisement advertiser can effectively conduct advertisement making, target crowd selection, price making and other works according to the requirements of advertisers, so that targets such as business growth, user increase and the like are expected to be realized. However, the drawbacks of manually placing advertisements are: in order to make the delivery better, the thrower needs to frequently make various attempts, such as modifying the bid, modifying the advertisement audience, modifying the creative scheme, etc., and the trial and error cost is too high. In addition, in many cases, an optimal delivery method is not obtained, and a large amount of system resources are consumed. The small and medium advertisers have a relatively high use cost, insufficient capability and less budget, and the advertising effect tends to be more optimistic. These drawbacks greatly reduce the aggressiveness of the advertiser.

In this embodiment, the pushing result of the content to be pushed may be predicted by constructing a network model for content pushing, and the pushing manner may be guided by using the network model. In the application process, the content to be pushed and the pushing condition can be used as the input of a network model, and the network model is utilized to determine a first pushing prediction result corresponding to the content to be pushed under the pushing condition. The network model is a network model based on machine learning, and the network structure of the network model can be determined according to an actual application scene, and is not limited in any way.

The application guides the content pushing mode by adopting a machine learning method, reduces labor cost and possibility of errors, and improves content pushing efficiency.

In the above-described related art, although content pushing is performed based on a simple pushing rule defined manually, automatic pushing of content can be achieved. However, the pushing manner is difficult to consider the difference of content types or the difference of industries related to the content, so that the pushing manner is not stable enough in different content types or different industries, and is good for a good balance.

Different types of content to be pushed are characterized by different types. In order to increase the sensitivity of the network model to different types of content to be pushed, in this embodiment, the network model is determined according to the type of content to be pushed.

In one possible implementation manner, the type corresponding to the content to be pushed includes any one or more of a content type of the content to be pushed or an industry type of an industry to which the content to be pushed relates. Among other things, the type of content to be pushed, its presentation forms include, but are not limited to: text, image, video, etc., industry types refer to industries to which content is to be pushed. Such as the automotive industry, the catering industry, the real estate industry, etc. In practical applications, the type of the content to be pushed may be determined according to the content provider, which is not limited in any way.

The content to be pushed is classified, so that the corresponding network model can be selected based on the type of the content to be pushed, and the prediction accuracy of the network model on the content to be pushed is improved.

Based on the above, a corresponding network model can be determined according to the type corresponding to the content to be pushed, and the pushing result of the content to be pushed is predicted. For example, if the type corresponding to the content to be pushed is a video type, predicting a pushing result of the content to be pushed by using a network model corresponding to the video type; and predicting the pushing result of the content to be pushed by utilizing a network model corresponding to the automobile industry when the industry related to the content to be pushed is the automobile industry.

In this embodiment, since the network model is determined according to the type corresponding to the content to be pushed, the network model has better recognition sensitivity to the push characteristics of the content to be pushed according to the type, so that the determined first push prediction result more objectively reflects the actual push result of the content to be pushed under the push condition, and has higher credibility.

It should be noted that, the first pushing prediction result determined based on the network model is a prediction result of pushing through the content to be pushed under the pushing condition. The first push prediction result may include prediction results of a plurality of evaluation dimensions. In practical applications, the evaluation dimension may be set according to the content to be pushed, which is not limited in any way.

Taking the application scenario of advertisement pushing as an example, the first pushing prediction result includes prediction results of multiple evaluation dimensions, where the multiple evaluation dimensions may include any one or more of achievement costs of delivering a single advertisement, achievement rates of delivering all advertisements, idle consumption advertisements, starting amounts, estimated deviation, idle consumption rates, and indexes related to advertisement performance such as Cost Per Action (CPA). The calculation can be specifically performed by the following formula:

the achievement cost is as follows: actual cost of advertisement-advertisement bid/advertisement bid less than or equal to a specified threshold alpha (alpha < 1);

achievement rate (in terms of consumption): ad consumption per total ad consumption per achievement cost 100%;

achievement rate (in terms of advertisement): the number of ads per total number of ads per achievement cost is 100%;

and (5) idle consumption advertisement: advertisement (N, M1 is a positive integer) with consumption of more than or equal to N conversion bids and conversion number less than M1;

the starting amount is as follows: the first day exposure is in a calculation period, the conversion reflux time is the calculation period plus the post-delay time t days, and the accumulated conversion number reaches M2 (t is a non-negative integer and M2 is a positive integer);

yield (in terms of consumption): consumption of the origin advertisement/consumption of all advertisements 100%;

yield (in terms of advertisements): number of start ads/number of all ads 100%;

Estimating deviation: predictive value-true value/true value 100%.

The first pushing prediction result may be an overall pushing prediction result of the content to be pushed under the pushing condition. In practical application, the first pushing prediction result may be a content type corresponding to the content to be pushed or a comprehensive evaluation index commonly used in the related industry. In the embodiment of the present application, the first push prediction result may be expressed in any manner, which is not limited herein.

In the related art, since pushing is performed based on experience of a professional or based on manually defined rules, only the pushing condition of the content to be pushed is known, and the state of the content to be pushed in the whole pushing environment cannot be known, so that an optimal pushing mode cannot be formulated to achieve the pushing expectations of the content provider.

In view of this, in this embodiment, historical push data related to the content to be pushed may be determined according to the push platform corresponding to the content to be pushed. The historical push data comprises any one or more of the combination of the historical conversion information, the historical targeting information or the number of the similar content of the content to be pushed.

The history conversion information refers to conversion information of the content to be pushed on the pushing platform. The historical targeting information refers to targeting pushing information of the content to be pushed on the pushing platform, such as targeting pushing objects, targeting pushing time, and the like. The number of homogeneous content refers to the total number of push content on the push platform that is of the same type as the content to be pushed.

The content to be pushed, the pushing conditions and the historical pushing data can be used as inputs of a network model, and a first pushing prediction result corresponding to the content to be pushed is determined through the network model.

According to the method, on the basis of the content to be pushed and the pushing conditions, the historical pushing data is added and used as the input of the network model, so that the network model can predict the content in the global range, the objectivity of determining the first pushing predicted result by using the network model is further improved, the first pushing predicted result is close to the actual pushing result, and the credibility of the first pushing predicted result is improved.

S203: and determining the pushing mode of the content to be pushed according to the difference information between the first pushing prediction result and the pushing target corresponding to the content to be pushed.

Based on the above S202, the first pushing prediction result is determined, and the pushing manner of the content to be pushed may be determined by determining the difference information between the first pushing prediction result and the pushing target corresponding to the content to be pushed. The pushing target refers to a target to be achieved by pushing the content to be pushed, and the pushing target can be a pushing expectation provided by a content provider, an industry average level or the like.

The measurement dimension represented by the push target may correspond to the first push prediction result described above. That is, if the pushing target is quantized by a plurality of evaluation dimensions, the first pushing prediction result includes prediction results of the plurality of evaluation dimensions; if the push target is quantized by an overall evaluation dimension, the first push prediction result includes an overall prediction result. The above S202 has already been described in detail for the first push prediction result, and will not be described herein.

And the difference information between the first pushing prediction result and the pushing target intuitively reflects the possibility of achieving the pushing target through the content to be pushed under the pushing condition. If the difference information indicates that the first pushing prediction result reaches the pushing target, the instruction can directly use the pushing condition as a pushing mode to push the content to be pushed. If the difference information indicates that the first pushing predicted result does not reach the pushing target, the pushing condition is directly used as a pushing mode to push the content to be pushed, the pushing result is not ideal enough, and the pushing target of the content provider cannot be reached. Therefore, in this case, the pushing conditions can be adjusted according to the difference information, so as to determine what pushing mode should be adopted in practice, and the pushing target can be achieved.

Because the difference information reflects the possibility of pushing the content to be pushed under the pushing condition, the pushing mode of the content to be pushed can be adaptively adjusted according to the difference information to determine what pushing mode should be adopted in practice, and the pushing target of the content provider can be achieved.

In practical applications, how to determine whether the first pushing prediction result reaches the pushing target according to the difference information can be determined by comparing the difference information with a predetermined condition. Specifically, if the difference information meets a predetermined condition, the content to be pushed may be pushed by the pushing condition. The difference information meets a preset condition, and the difference information indicates that a pushing target can be achieved through the content to be pushed under the pushing condition. In this case, the pushing conditions are not required to be adjusted any more, and the content to be pushed is pushed by directly taking the pushing conditions as a pushing mode. The predetermined conditions may be set according to the actual application scenario, and are not limited herein.

For the application scene of advertisement pushing, if the difference information meets the preset condition, an automatic advertisement putting mode can be adopted, namely, the advertisement is automatically put directly according to the putting condition provided by an advertiser. The automatic putting mode can effectively reduce the advertisement putting cost, and is suitable for miniature enterprises which cannot provide higher advertisement putting cost.

And if the difference information does not meet the preset condition, the pushing mode can be adjusted according to the difference information. The difference information does not meet the preset condition, and the difference information indicates that the pushing target is not achieved through the content to be pushed under the pushing condition. In this case, the pushing conditions may be adjusted according to the difference information, so that the pushing target may be satisfied when pushing is performed by the adjusted pushing conditions.

Similarly, for the application scenario of advertisement pushing, if the difference information does not meet the preset condition, an automatic advertisement delivery mode capable of being intervened can be adopted, that is, according to the delivery condition provided by an advertiser, the advertisement delivery mode is adjusted by combining with rich delivery experience accumulated by an advertisement advertiser by utilizing an advertisement platform, for example, dynamic budget allocation, dynamic automatic bidding, automatic media version, low-volume advertisement suspension, intelligent orientation and the like. The automatic throwing mode capable of being intervened can pertinently adjust the advertisement throwing mode when the pushing result does not reach the pushing target, so as to obtain a better pushing result, and is suitable for advertisers with throwing capability.

Because whether the difference information meets the preset condition or not is judged, whether the pushing condition can be used as a pushing mode or not can be directly determined, and the content to be pushed is pushed. And, when the difference information does not satisfy the predetermined condition, the pushing condition may be adjusted based on the difference information. Based on the method, effective guidance is provided for what pushing mode is actually adopted, and dependence on experience of professionals is reduced.

Based on the above, when the difference information does not meet the predetermined condition, the manner to be pushed may be adjusted according to the difference information. In one possible implementation manner, if the first pushing prediction result is used for identifying the prediction results of the multiple evaluation dimensions included in the pushing target, it may be determined, according to the difference information, that the prediction result does not reach the target dimension of the pushing target from the multiple evaluation dimensions, then, based on an adjustment parameter corresponding to the target dimension, any one or more of the pushing condition or the content to be pushed is adjusted to obtain an adjusted content, then, according to the adjusted content, a corresponding second pushing prediction result is determined through the network model, and if the prediction result of the second pushing prediction result for the target dimension reaches the pushing target, the pushing mode is adjusted according to the adjusted content.

The adjustment parameters corresponding to the target dimensions can be determined in advance based on experience of professionals, or can be determined based on big data analysis and historical push case analysis. In practical applications, the mapping relationship between the target dimension and the adjustment parameter may be stored in advance. For different types of content to be pushed, different respective evaluation dimensions are corresponding. In practical application, the method can be adaptively adjusted according to specific application scenes.

For example, in the application scenario shown in fig. 1, the first push prediction result includes prediction results of two indexes, i.e., the achievement rate and the conversion rate. When it is determined that the predicted result corresponding to the achievement rate does not reach the pushing target in the first pushing predicted result according to the difference information, if the adjustment parameter corresponding to the achievement rate includes the throwing orientation, the throwing orientation included in the pushing condition is adjusted, that is, the throwing orientation range of the advertisement is enlarged, for example, the user sees the thrown advertisement when watching the movie, and also sees the thrown advertisement when watching the sports event.

The difference information can clearly indicate the target dimension of which the predicted result does not reach the pushing target in a plurality of evaluation dimensions, so that the pushing condition and/or the content to be pushed are purposefully adjusted based on the adjustment parameter corresponding to the target dimension, and the problem that the pushing target cannot be reached by pushing the content to be pushed under the original pushing condition is solved.

If the predicted result of the target dimension in the second pushing predicted result does not reach the pushing target, the pushing condition and/or the pushing content are/is adjusted based on the adjustment parameter corresponding to the target dimension, and the pushing result cannot reach the pushing target. Therefore, the push method is not adjusted by using the adjusted content.

According to the content pushing method provided by the embodiment, the first pushing prediction result is obtained by utilizing the network model by acquiring the content to be pushed and the pushing conditions for pushing the content to be pushed. The network model is determined based on the type corresponding to the content to be pushed, has better identification sensitivity to the pushing characteristics of the content to be pushed conforming to the type, and thus the determined first pushing prediction result is more likely to objectively reflect the actual pushing result of the content to be pushed under the pushing condition and has higher credibility. Based on the above, according to the first pushing prediction result and the difference information of the pushing target corresponding to the content to be pushed, the difference information intuitively shows the possibility of achieving the pushing target through the content to be pushed under the pushing condition. Therefore, according to the possibility represented by the difference information, the pushing mode actually adopted for the content to be pushed can be adjusted and determined in a targeted manner. Therefore, the network model based on the type determination can obtain a first pushing prediction result with high credibility, so that the influence of human experience is avoided, and the effect of efficiently and accurately determining the pushing mode of the content to be pushed is realized based on the first pushing prediction result.

It will be appreciated that the network model performance depends on the training process of the network model. The network model training process in the above embodiment is described below with reference to the accompanying drawings.

Referring to fig. 3, fig. 3 is a flow chart of a network model training method according to an embodiment of the present application. As shown in fig. 3, the network model training method includes the following steps:

s301: and acquiring a historical content pushing task.

In practical application, a log splicing method can be adopted to obtain the historical content pushing task. The history content pushing task comprises pushed history content, a pushing mode adopted, history pushing data corresponding to the history content and a history pushing result.

The history content refers to content of which the history is pushed; the adopted pushing mode is a pushing mode actually adopted by pushing the history content, and the history pushing data is the history pushing data contained in the history content pushing process on a pushing platform for pushing the history content; the history pushing result refers to a pushing result corresponding to the pushing history content.

It should be noted that, the history content pushing task may be a history content pushing task corresponding to the content to be pushed in the above step S201, or may be a history content pushing task of the same type of content as the content to be pushed. And are not intended to be limiting in any way.

In connection with fig. 4, a network model training process is described taking an application scenario of advertisement pushing as an example. As shown in fig. 4, executing S401 obtains a history advertisement push task, the history advertisement push task including: historical advertisements, adopted advertisement putting modes, historical pushing data corresponding to the historical advertisements and historical pushing results. The expression forms of the historical advertisements can comprise texts, images and the like, the advertisement delivery mode comprises price adjustment, budget limit modification, orientation modification, material modification and the like, the historical push data comprises advertisement historical conversion number, achievement rate, comprehensive budget and the like, and the historical push results comprise a yield rate, a consumption rate, CPA and the like.

S302: and determining a training sample according to the history content pushing task.

In this implementation, the training samples include model input data and labels. The labels of the training samples are determined according to the historical pushing results. If the history pushing result shows a growing trend within the preset time, the label of the training sample can be marked as a positive example. Correspondingly, the label of the training sample is marked as a negative example.

After the training sample and the label of the training sample are determined according to the history content pushing task, the training can be performed by utilizing the undetermined network model pre-constructed by the training sample, so that the undetermined network model has the capability of predicting the pushing result of the pushing content.

In this embodiment, the tag of the training sample may compare the history pushing result with a preset threshold. If the history pushing result includes pushing results of multiple evaluation dimensions, each pushing result may be compared with a preset threshold value corresponding to each pushing result, and the comparison result is used as a label of a training sample to identify whether the history pushing result is qualified in the multiple evaluation dimensions.

For example, in the scenario shown in fig. 4, the training sample identifies whether the historical push results of the historical advertisements are qualified in three evaluation dimensions, namely, the rate of rise, the rate of consumption, and the CPA, respectively.

If the history pushing result includes the whole pushing result of the history content, the whole pushing result can be compared with a corresponding whole preset threshold value, and the comparison result is used as a label of a training sample to identify whether the whole history pushing result is qualified or not.

The label of the training sample is determined according to the history content pushing task, and the model can be trained in a supervision mode based on the training sample and the label of the training sample.

S303: and training the undetermined network model according to the training sample to obtain the network model.

In practical application, the training samples may be input into a logistic regression or other machine learning model, and according to the principle of minimizing the loss function, the optimal solution of the model loss function is calculated, and the network model in S202 may be obtained after training is completed.

In this embodiment, a model structure of the undetermined network model may be constructed based on an artificial intelligence technique according to a history content push task. In practical applications, a feature Representation (presentation) may be performed on the historical content pushing task and used as an input to the model.

As shown in fig. 5, for the application scenario shown in fig. 4, for the historical advertisement, the two aspects of text and image are divided, and a bidirectional transformer coding model (Bidirectional Encoder Representation from Transformers, BERT) and a convolutional neural network (Convolutional Neural Network, CNN) can be used for coding operation respectively, so as to obtain the characteristic representation of the historical advertisement. The pushing mode, i.e. the behavior feature, including budget quota, whether to modify orientation, and price adjustment, can use Lookup tables (Lookup tables), whether to embed (Yes/No unbedding) and linear regression (Linear Regression), respectively, to encode, and can obtain the representation of the behavior feature. In FIG. 5, table, yes/No and Linear are simplified representations, respectively. The historical push data, namely the context state characteristics, are divided into comprehensive budget, conversion number and achievement rate. The encoding may be performed using linear regression, resulting in a representation of the context state.

Based on this, after the training samples are characterized, the network model to be determined may be trained so as to obtain the network model in S202.

In the related art, since the rule-based content pushing method does not consider different content types or industry factors related to the content, when pushing different types of content, the pushing result obtained by the rule-based content pushing method is not stable enough, and it cannot be guaranteed that the different types of content can obtain better pushing results.

In view of this, in this embodiment, the type feature corresponding to the type may also be determined according to the type corresponding to the training sample. The type corresponding to the training sample comprises a content type corresponding to the historical content and/or industries related to the historical content. The presentation forms of the content types include, but are not limited to: text, images, video, etc. The type feature is used for reflecting historical content in the training sample, and the pushing feature in the pushing process of the content related to the type is used for pushing the content. For example, if fig. 5 is an automotive advertisement push scenario, determining industry characteristics of an advertisement may include: an automobile advertisement putting platform, an automobile advertisement putting time and the like. And obtaining the characteristic representation of the industry characteristic through characteristic combination and multiple full-connection operation. In practical application, the type features corresponding to the types can be determined according to the specific application scenario, and are not limited in any way.

Therefore, the network model to be determined can be trained according to the training samples and the type features, and the network model is obtained. During training, the features related to the type of features in the training sample are strengthened through an attention mechanism, and the features unrelated to the type of features are weakened.

According to the method, the type features are added in the model training process, the features related to the type features in the training samples are strengthened through the attention mechanism, the features unrelated to the type features are weakened, the features related to the type features are more focused in the network model training process, the features related to the type features are more learned, and the prediction accuracy of the network model on the pushing results of the same type of pushing content is improved.

In the training process of the network model to be determined, in one possible implementation manner, the input data of the network model to be determined can be obtained by splicing the training samples and the type features, such as the industry representation (domain representation) part shown by a dashed box in fig. 5. And then, according to the input data and the type characteristics, strengthening the characteristics related to the type characteristics in the training sample through an attention layer, weakening the characteristics unrelated to the type characteristics, and determining a push training result of the training sample according to the output data of the attention layer, so that parameter adjustment can be performed on the undetermined network model based on the push training result and the labels of the training sample, and the network model is obtained.

As shown in fig. 5, after feature enhancement is performed on the feature representation obtained after splicing by using the attention mechanism, multiple full-connection operations may be performed, and after multiplication and splicing operations are performed on the output obtained after the full-connection operation and the feature representation of the industry feature, classification is performed by using a classifier, and a first pushing prediction result is output. The classifier may include softmax, sigmoid, among others.

By splicing the type features and the training samples, the type features input by the model are enhanced, and the type features are further enhanced by combining the attention mechanism, so that the network model focuses on the type features, the recognition sensitivity of the network model on the push features conforming to the same type of content is improved, and the push prediction result determined by the network model has higher credibility.

In the application scenario shown in fig. 4 and fig. 5, after training the to-be-determined network model by using the training sample to obtain the network model for predicting the pushing result of the advertisement, the network model may be applied to an intelligent jetter for the advertisement putting scenario. Wherein, intelligent pitcher is one kind and replaces the advertisement pitcher to carry out automatic advertisement delivery, perhaps alleviates advertisement pitcher's burden, gives its more accurate professional suggestion's automated system.

According to the network model training method provided by the embodiment, the network model is trained, so that the network model can predict the pushing result of the content to be pushed under the pushing condition, the network model is utilized to predict the pushing result of the content to be pushed before pushing the content to be pushed or in the pushing process of the content to be pushed, and the pushing mode is adjusted according to the difference information between the predicting result and the pushing target, so that the pushing target of the content provider is achieved.

For better understanding, the content pushing method provided by the embodiment of the present application is described below by taking an advertiser in the automotive industry as an example to push advertisements.

The advertiser gives a push target comprising: delivery industry, delivery budget, expected cost, advertising creative material, non-negligible targeting, landing page, delivery period, media placement settings, etc. Taking the car aftermarket as an example, an advertiser selects the traffic car service industry and puts in packages (according to scenes such as new powder adding, offline to store, service purchase, and the like), uploads pictures (videos), texts, budget setting, and the like.

The intelligent thrower gives two pushing modes according to the pushing target of the advertiser and aiming at the industry related to the advertisement to be pushed and the pushing conditions provided by the advertiser, and the automatic throwing is performed or the automatic throwing can be performed in an intervening way.

And the intelligent thrower automatically predicts the pushing prediction result of the advertisement to be pushed according to the universal advertisement index of the industry and adjusts the pushing mode of the advertisement so as to quickly and effectively reach the pushing target of an advertiser. These goals may be several or all of the core metrics that include advertiser concern based on achievement cost, achievement rate, new advertising rate, outage rate, cpa, and the like.

According to the method, the pushing result of pushing the content to be pushed under the pushing condition is predicted through the intelligent pitcher (namely the network model), and the pushing mode is adjusted in a targeted manner according to the difference information between the pushing predicted result and the pushing target, so that the adjusted pushing mode can reach the pushing target, and the satisfaction degree of the content provider on content pushing is improved.

The embodiment of the application also provides a content pushing device aiming at the content pushing method provided by the embodiment.

Referring to fig. 6, fig. 6 is a content pushing device provided in an embodiment of the present application. As shown in fig. 6, the content pushing apparatus 600 includes an acquisition unit 601 and a determination unit 602:

the acquiring unit 601 is configured to acquire content to be pushed and a pushing condition for pushing the content to be pushed;

The determining unit 602 is configured to determine, according to the content to be pushed and the pushing condition, a corresponding first pushing prediction result through a network model; the network model is determined according to the type corresponding to the content to be pushed;

the determining unit 602 is further configured to determine a pushing manner of the content to be pushed according to difference information between the first pushing prediction result and a pushing target corresponding to the content to be pushed.

In a possible implementation manner, the determining unit 602 is configured to:

if the difference information meets the preset condition, pushing the content to be pushed according to the pushing condition;

and if the difference information does not meet the preset condition, adjusting the pushing mode according to the difference information.

In a possible implementation manner, the determining unit 602 is configured to:

if the first pushing prediction result is used for identifying prediction results of a plurality of evaluation dimensions included in the pushing target, determining that the prediction results do not reach the target dimension of the pushing target from the plurality of evaluation dimensions according to the difference information;

based on the adjustment parameters corresponding to the target dimension, adjusting any one or more of the pushing conditions or the contents to be pushed to obtain adjusted contents;

Determining a corresponding second pushing prediction result through the network model according to the adjusted content;

if the predicted result of the target dimension in the second pushing predicted result reaches the pushing target, adjusting the pushing mode according to the adjusted content;

and if the predicted result of the target dimension in the second pushing predicted result does not reach the pushing target, not adopting the adjusted content to adjust the pushing mode.

In one possible implementation manner, the type corresponding to the content to be pushed includes any one or more of a content type of the content to be pushed or an industry type of an industry to which the content to be pushed relates.

In a possible implementation manner, the determining unit 602 is further configured to:

according to the push platform corresponding to the content to be pushed, determining historical push data related to the content to be pushed, wherein the historical push data comprises any one or more combinations of historical conversion information, historical orientation information or the number of similar contents of the content to be pushed;

and determining a corresponding first pushing prediction result through a network model according to the content to be pushed, the pushing condition and the historical pushing data.

In a possible implementation manner, the obtaining unit 601 is further configured to:

acquiring a history content pushing task, wherein the history content pushing task comprises pushed history content, a pushing mode adopted, history pushing data corresponding to the history content and a history pushing result;

the determining unit 602 is further configured to determine a training sample according to the history content pushing task, where a label of the training sample is determined according to the history pushing result;

the device further comprises a training unit:

and the training unit is used for training the undetermined network model according to the training sample to obtain the network model.

determining type characteristics corresponding to the types according to the types corresponding to the training samples, wherein the type characteristics are used for reflecting historical contents in the training samples, and the pushing characteristics relate to the pushing process of the types of contents;

the training unit is configured to:

training the undetermined network model according to the training sample and the type characteristic to obtain the network model; during training, the features related to the type of features in the training sample are strengthened through an attention mechanism, and the features unrelated to the type of features are weakened.

In a possible implementation manner, the training unit is configured to:

the training sample and the type feature are spliced to obtain input data of the undetermined network model;

according to the input data and the type features, strengthening the features related to the type features in the training sample through an attention layer, and weakening the features unrelated to the type features;

determining a push training result of the training sample according to the output data of the attention layer;

and carrying out parameter adjustment on the undetermined network model based on the pushing training result and the label of the training sample to obtain the network model.

In one possible implementation, the label of the training sample is used to identify whether the whole historical push result is qualified, or whether the historical push result is respectively qualified in multiple evaluation dimensions.

According to the content pushing device provided by the embodiment, the first pushing prediction result is obtained by utilizing the network model by acquiring the content to be pushed and the pushing conditions for pushing the content to be pushed. The network model is determined based on the type corresponding to the content to be pushed, has better identification sensitivity to the pushing characteristics of the content to be pushed conforming to the type, and thus the determined first pushing prediction result is more likely to objectively reflect the actual pushing result of the content to be pushed under the pushing condition and has higher credibility. Based on the above, according to the first pushing prediction result and the difference information of the pushing target corresponding to the content to be pushed, the difference information intuitively shows the possibility of achieving the pushing target through the content to be pushed under the pushing condition. Therefore, according to the possibility represented by the difference information, the pushing mode actually adopted for the content to be pushed can be adjusted and determined in a targeted manner. Therefore, the network model based on the type determination can obtain a first pushing prediction result with high credibility, so that the influence of human experience is avoided, and the effect of efficiently and accurately determining the pushing mode of the content to be pushed is realized based on the first pushing prediction result.

The embodiment of the application also provides a device for pushing the content, and the device for pushing the content provided by the embodiment of the application is introduced from the perspective of hardware materialization.

Referring to fig. 7, fig. 7 is a schematic diagram of a server structure according to an embodiment of the present application, where the server 1400 may have a relatively large difference due to configuration or performance, and may include one or more central processing units (central processing units, CPU) 1422 (e.g., one or more processors) and a memory 1432, and one or more storage media 1430 (e.g., one or more mass storage devices) storing application programs 1442 or data 1444. Wherein the memory 1432 and storage medium 1430 can be transitory or persistent storage. The program stored in the storage medium 1430 may include one or more modules (not shown), each of which may include a series of instruction operations on a server. Further, the central processor 1422 may be provided in communication with a storage medium 1430 to perform a series of instruction operations in the storage medium 1430 on the server 1400.

The server 1400 may also include one or more power supplies 1426, one or more wired or wireless network interfaces 1450, one or more input/output interfaces 1458, and/or one or more operating systems 1441, such as Windows Server, mac OS XTM, unixTM, linuxTM, freeBSDTM, and the like.

The steps performed by the server in the above embodiments may be based on the server structure shown in fig. 7.

Wherein, the CPU 1422 is configured to perform the following steps:

Optionally, the CPU 1422 may further perform method steps of any specific implementation of the content pushing method in the embodiment of the present application.

For the content pushing method described above, the embodiment of the application also provides a terminal device for content pushing, so that the content pushing method is realized and applied in practice.

Referring to fig. 8, fig. 8 is a schematic structural diagram of a terminal device according to an embodiment of the present application. For convenience of explanation, only those portions of the embodiments of the present application that are relevant to the embodiments of the present application are shown, and specific technical details are not disclosed, please refer to the method portions of the embodiments of the present application. The terminal device may be any terminal device including a mobile phone, a tablet computer, a personal digital assistant (Personal Digital Assistant, PDA for short), etc., taking the terminal device as an example of the mobile phone:

Fig. 8 is a block diagram showing a part of the structure of a mobile phone related to a terminal device provided by an embodiment of the present application. Referring to fig. 8, the mobile phone includes: radio Frequency (RF) circuitry 1510, memory 1520, input unit 1530, display unit 1540, sensor 1550, audio circuitry 1560, wireless fidelity (wireless fidelity, wiFi) module 1570, processor 1580, and power supply 1590. Those skilled in the art will appreciate that the handset configuration shown in fig. 8 is not limiting of the handset and may include more or fewer components than shown, or may combine certain components, or may be arranged in a different arrangement of components.

The following describes the components of the mobile phone in detail with reference to fig. 8:

the RF circuit 1510 may be used for receiving and transmitting signals during a message or a call, and particularly, after receiving downlink information of a base station, the signal is processed by the processor 1580; in addition, the data of the design uplink is sent to the base station. Generally, RF circuitry 1510 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, LNA for short), a duplexer, and the like. In addition, the RF circuitry 1510 may also communicate with networks and other devices through wireless communication. The wireless communication may use any communication standard or protocol, including but not limited to global system for mobile communications (Global System of Mobile communication, GSM for short), general packet radio service (General Packet Radio Service, GPRS for short), code division multiple access (Code Division Multiple Access, CDMA for short), wideband code division multiple access (Wideband Code Division Multiple Access, WCDMA for short), long term evolution (Long Term Evolution, LTE for short), email, short message service (Short Messaging Service, SMS for short), and the like.

The memory 1520 may be used to store software programs and modules, and the processor 1580 implements various functional applications and data processing of the handset by running the software programs and modules stored in the memory 1520. The memory 1520 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, phonebook, etc.) created according to the use of the handset, etc. In addition, memory 1520 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device.

The input unit 1530 may be used to receive input numerical or character information and generate key signal inputs related to user settings and function control of the handset. In particular, the input unit 1530 may include a touch panel 1531 and other input devices 1532. The touch panel 1531, also referred to as a touch screen, may collect touch operations thereon or thereabout by a user (e.g., operations of the user on the touch panel 1531 or thereabout by using any suitable object or accessory such as a finger, a stylus, etc.), and drive the corresponding connection device according to a predetermined program. Alternatively, the touch panel 1531 may include two parts, a touch detection device and a touch controller. The touch detection device detects the touch azimuth of a user, detects a signal brought by touch operation and transmits the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, and sends the touch point coordinates to the processor 1580, and can receive and execute commands sent from the processor 1580. In addition, the touch panel 1531 may be implemented in various types such as resistive, capacitive, infrared, and surface acoustic wave. The input unit 1530 may include other input devices 1532 in addition to the touch panel 1531. In particular, other input devices 1532 may include, but are not limited to, one or more of a physical keyboard, function keys (e.g., volume control keys, switch keys, etc.), a trackball, mouse, joystick, etc.

The display unit 1540 may be used to display information input by a user or information provided to the user and various menus of the mobile phone. The display unit 1540 may include a display panel 1541, and optionally, the display panel 1541 may be configured in the form of a liquid crystal display (Liquid Crystal Display, LCD), an Organic Light-Emitting Diode (OLED), or the like. Further, the touch panel 1531 may cover the display panel 1541, and when the touch panel 1531 detects a touch operation thereon or thereabout, the touch operation is transferred to the processor 1580 to determine the type of touch event, and then the processor 1580 provides a corresponding visual output on the display panel 1541 according to the type of touch event. Although in fig. 8, the touch panel 1531 and the display panel 1541 are two separate components for implementing the input and input functions of the mobile phone, in some embodiments, the touch panel 1531 may be integrated with the display panel 1541 to implement the input and output functions of the mobile phone.

The handset may also include at least one sensor 1550, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor that may adjust the brightness of the display panel 1541 according to the brightness of ambient light, and a proximity sensor that may turn off the display panel 1541 and/or the backlight when the phone is moved to the ear. As one of the motion sensors, the accelerometer sensor can detect the acceleration in all directions (generally three axes), and can detect the gravity and direction when stationary, and can be used for applications of recognizing the gesture of a mobile phone (such as horizontal and vertical screen switching, related games, magnetometer gesture calibration), vibration recognition related functions (such as pedometer and knocking), and the like; other sensors such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc. that may also be configured with the handset are not described in detail herein.

Audio circuitry 1560, a speaker 1561, and a microphone 1562 may provide an audio interface between a user and a cell phone. The audio circuit 1560 may transmit the received electrical signal converted from audio data to the speaker 1561, and be converted into a sound signal by the speaker 1561 for output; on the other hand, the microphone 1562 converts the collected sound signals into electrical signals, which are received by the audio circuit 1560 for conversion into audio data, which is processed by the audio data output processor 1580 for transmission to, for example, another cellular phone via the RF circuit 1510 or for output to the memory 1520 for further processing.

WiFi belongs to a short-distance wireless transmission technology, and a mobile phone can help a user to send and receive emails, browse webpages, access streaming media and the like through a WiFi module 1570, so that wireless broadband Internet access is provided for the user. Although fig. 8 shows WiFi module 1570, it is understood that it does not belong to the necessary components of a cell phone and may be omitted entirely as desired within the scope of not changing the essence of the invention.

The processor 1580 is a control center of the mobile phone, connects various parts of the entire mobile phone using various interfaces and lines, and performs various functions of the mobile phone and processes data by running or executing software programs and/or modules stored in the memory 1520 and invoking data stored in the memory 1520. In the alternative, processor 1580 may include one or more processing units; preferably, the processor 1580 can integrate an application processor and a modem processor, wherein the application processor primarily processes operating systems, user interfaces, application programs, and the like, and the modem processor primarily processes wireless communications. It is to be appreciated that the modem processor described above may not be integrated into the processor 1580.

The handset further includes a power supply 1590 (e.g., a battery) for powering the various components, which may preferably be logically connected to the processor 1580 via a power management system so as to provide for the management of charging, discharging, and power consumption by the power management system.

Although not shown, the mobile phone may further include a camera, a bluetooth module, etc., which will not be described herein.

In an embodiment of the present application, the memory 1520 included in the mobile phone may store program codes and transmit the program codes to the processor.

The processor 1580 included in the mobile phone may execute the content pushing method provided in the foregoing embodiment according to the instruction in the program code.

The embodiment of the application also provides a computer readable storage medium for storing a computer program for executing the content pushing method provided in the above embodiment.

Embodiments of the present application also provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the content pushing method provided in the various alternative implementations of the above aspects.

Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, where the above program may be stored in a computer readable storage medium, and when the program is executed, the program performs steps including the above method embodiments; and the aforementioned storage medium may be at least one of the following media: read-only memory (ROM), RAM, magnetic disk or optical disk, etc., which can store program codes.

It should be noted that, in the present specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment is mainly described in a different point from other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, with reference to the description of the method embodiments in part. The apparatus and system embodiments described above are merely illustrative, in which elements illustrated as separate elements may or may not be physically separate, and elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

The foregoing is only one specific embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions easily contemplated by those skilled in the art within the technical scope of the present application should be included in the scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A content pushing method, the method comprising:

determining a pushing mode of the content to be pushed according to the difference information between the first pushing prediction result and the pushing target corresponding to the content to be pushed;

the training process of the network model comprises the following steps:

Determining a training sample according to the history content pushing task, wherein a label of the training sample is determined according to the history pushing result;

training a network model to be determined according to the training sample and the type characteristics to obtain the network model; during training, the features related to the type of features in the training sample are strengthened through an attention mechanism, and the features unrelated to the type of features are weakened.

2. The method of claim 1, wherein the determining the push manner of the content to be pushed comprises:

3. The method of claim 2, wherein, if the first push prediction result is used to identify a prediction result of a plurality of evaluation dimensions included in the push target, the adjusting the push manner according to the difference information includes:

Determining a target dimension of which the predicted result does not reach the pushing target from the plurality of evaluation dimensions according to the difference information;

4. The method of claim 1, wherein the type of content to be pushed comprises any one or more of a content type of the content to be pushed, or an industry type of an industry to which the content to be pushed relates.

5. The method according to claim 1, wherein the method further comprises:

the determining, according to the content to be pushed and the pushing condition, a corresponding first pushing prediction result through a network model includes:

6. The method of claim 1, wherein training the network model to be determined based on the training samples and the type features to obtain the network model comprises:

7. The method of claim 1, wherein the label of the training sample is used to identify whether the history push result is qualified as a whole or whether the history push result is respectively qualified in multiple evaluation dimensions.

8. A content pushing device, characterized in that the device comprises an acquisition unit, a determination unit and a training unit;

the determining unit is further configured to determine a pushing manner of the content to be pushed according to difference information between the first pushing prediction result and a pushing target corresponding to the content to be pushed;

the acquisition unit is further used for acquiring a history content pushing task, wherein the history content pushing task comprises pushed history content, a pushing mode adopted, history pushing data corresponding to the history content and a history pushing result;

The determining unit is further configured to determine a training sample according to the history content pushing task, where a label of the training sample is determined according to the history pushing result;

the determining unit is further configured to determine a type feature corresponding to the type according to a type corresponding to the training sample, where the type feature is used to embody a pushing characteristic of the historical content in the training sample in a content pushing process related to the type;

the training unit is used for training the network model to be determined according to the training sample and the type characteristics to obtain the network model; during training, the features related to the type of features in the training sample are strengthened through an attention mechanism, and the features unrelated to the type of features are weakened.

9. The apparatus according to claim 8, wherein the determining unit is configured to:

10. The apparatus according to claim 9, wherein the determining unit is configured to:

and if the predicted result aiming at the target dimension in the second pushing predicted result reaches the pushing target, adjusting the pushing mode according to the adjusted content.

11. The apparatus of claim 8, wherein the type of content to be pushed comprises any one or more of a content type of the content to be pushed, or an industry type of an industry to which the content to be pushed relates.

12. An apparatus for content pushing, the apparatus comprising a processor and a memory:

The processor is configured to perform the method of any of claims 1-7 according to instructions in the program code.

13. A computer readable storage medium for storing a computer program for execution by a processor to perform the method of any one of claims 1-7.