Content push method, device, computer equipment and storage medium

Abstract

The present application relates to a content push method, apparatus, computer equipment and storage medium. The method includes: obtaining a target feature combination, obtaining a content push record set, determining a first negative content push record in the content push record set in which a push object performs a negative operation on the push content, and obtaining the number of first negative records corresponding to the first negative content push record; determining a second negative content push record corresponding to the target feature combination, and obtaining the number of second negative records corresponding to the second negative content push record; determining a target feature exclusion degree corresponding to the target feature combination based on the number of first negative records and the number of second negative records, and filtering a candidate push content set for the target push object based on the target feature exclusion degree. The use of this method can improve the accuracy of content push, and the candidate push content set in the present application can be obtained using a neural network model based on artificial intelligence.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a content push method, apparatus, computer equipment and storage medium. Background Art

[0002] With the development of computer technology and Internet technology, there is a demand to push content to users in many scenarios. For example, more and more software products provide advertising slots. Advertising slots are locations in software used to place advertisements. Advertisers can place advertisements through advertising slots so that users of software products can see the placed advertisements.

[0003] In traditional content push methods, content is pushed to users based on the popularity of the content or the requirements of the content provider. For example, a content push method based on artificial intelligence can push popular content to user terminals. However, due to the different degrees of rejection of content by different users, the pushed content has low reliability, causing interference to users and low accuracy of content push. Summary of the invention

[0004] Based on this, it is necessary to provide a content push method, apparatus, computer device and storage medium that can improve push accuracy in response to the above technical problems.

[0005] A content push method, the method comprising: obtaining a target feature combination, the target feature combination being obtained by combining an object feature and a content feature; obtaining a content push record set, the content push record set comprising a plurality of content push records, the content push record comprising push content and a push object corresponding to the push content; determining a first negative content push record in the content push record set in which a push object performs a negative operation on a push content, and obtaining the number of first negative records corresponding to the first negative content push record; determining a second negative content push record corresponding to the target feature combination, and obtaining the number of second negative records corresponding to the second negative content push record; wherein the content feature of the push content in the second negative content push record is consistent with the content feature in the target feature combination, and the object feature of the push object in the second negative content push record is consistent with the object feature in the target feature combination; determining a target feature exclusion degree corresponding to the target feature combination according to the first number of negative records and the second number of negative records, the target feature exclusion degree indicating the degree of exclusion between the content feature and the object feature in the target feature combination, so as to screen candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination.

[0006] A content push device, the device comprising: a target feature combination acquisition module, used to acquire a target feature combination, the target feature combination being obtained by combining an object feature and a content feature; a content push record set acquisition module, used to acquire a content push record set, the content push record set comprising a plurality of content push records, the content push records comprising push content and a push object corresponding to the push content; a first negative record quantity acquisition module, used to determine a first negative content push record in the content push record set in which a push object performs a negative operation on a push content, and to acquire a first negative record quantity corresponding to the first negative content push record; a second negative record quantity acquisition module, used to determine a second negative record quantity corresponding to the target feature combination Push record to the content, obtain the number of second negative records corresponding to the second negative content push record; wherein the content features of the pushed content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the pushed object in the second negative content push record are consistent with the object features in the target feature combination; a target feature exclusion degree determination module is used to determine the target feature exclusion degree corresponding to the target feature combination according to the first number of negative records and the second number of negative records, the target feature exclusion degree indicates the degree of exclusion between the content features and the object features in the target feature combination, so as to screen the candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination.

[0007] In some embodiments, there are multiple target feature combinations, and the device is also used to select a target feature combination that meets the exclusion degree screening condition according to the target feature exclusion degree corresponding to the target feature combination as an exclusion feature combination, and the exclusion feature combination is used to filter the push content having the content features in the exclusion feature combination when pushing content to the push object having the object features in the exclusion feature combination; the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree being ranked before the exclusion degree ranking threshold.

[0008] In some embodiments, the device also includes: a reference push record set acquisition module, used to acquire a reference push record set, the reference push record set includes multiple reference push records, and the reference push record is a content push record corresponding to the target feature combination; a content push value acquisition module, used to acquire the content push value corresponding to the reference push record; a target push value acquisition module, used to count the content push values ​​corresponding to the reference push records in the reference push record set, and obtain the target push value corresponding to the target feature combination; and screen the candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination and the target push value.

[0009] In some embodiments, there are multiple target feature combinations, and the device is further used to select a target feature combination that meets the exclusion degree screening condition and the push value screening condition as the exclusion feature combination according to the target feature exclusion degree and the target push value corresponding to the target feature combination; the exclusion feature combination is used to filter the push content having the content features in the exclusion feature combination when pushing content to the push object having the object features in the exclusion feature combination; the push value screening condition includes at least one of the target push value being greater than the push value threshold or the target push value being ranked before the value ranking threshold; the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree being ranked before the exclusion degree ranking threshold.

[0010] In some embodiments, the target push value obtaining module includes: an overall content push value obtaining unit, which is used to sum and calculate the content push values ​​corresponding to the reference push records in the reference push record set to obtain the overall content push value; a reference content record quantity obtaining unit, which is used to obtain the number of reference push records in the reference push record set as the reference content record number; a target push value obtaining unit, which is used to calculate the ratio of the overall content push value to the number of reference content records, and obtain the target push value corresponding to the target feature combination based on the calculated ratio.

[0011] In some embodiments, the target feature exclusion degree determination module includes: a content push record quantity determination unit, used to determine the number of content push records corresponding to the content push record set; a first negative record possibility obtaining unit, used to calculate the ratio of the first number of negative records to the number of content push records, and obtain the first negative record possibility according to the calculated ratio; a second negative record possibility obtaining unit, used to calculate the ratio of the second number of negative records to the first number of negative records, and obtain the second negative record possibility according to the calculated ratio; a target feature exclusion degree obtaining unit, used to calculate the target feature exclusion degree corresponding to the target feature combination based on the first negative record possibility and the second negative record possibility.

[0012] In some embodiments, the target feature exclusion degree obtaining unit is also used to calculate the product of the first negative record possibility and the second negative record possibility to obtain the standard negative record possibility corresponding to the target feature combination; obtain the number of content push records corresponding to the target feature combination in the content push record set as the number of combined records; calculate the ratio of the number of combined records to the number of content push records, and obtain the push record possibility corresponding to the target feature combination according to the calculated ratio; and obtain the target feature exclusion degree corresponding to the target feature combination based on the standard negative record possibility and the push record possibility.

[0013] In some embodiments, the target feature combination acquisition module includes: a feature combination configuration interface sending unit, used to send a feature combination configuration interface to a configuration terminal, so that the configuration terminal displays the feature combination configuration interface, and obtains the configured target feature combination in response to a configuration operation on the feature combination configuration interface; a target feature combination receiving unit, used to receive the target feature combination sent by the configuration terminal.

[0014] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the following steps when executing the computer program: obtaining a target feature combination, wherein the target feature combination is obtained by combining an object feature and a content feature; obtaining a content push record set, wherein the content push record set includes multiple content push records, wherein the content push record includes push content and a push object corresponding to the push content; determining a first negative content push record in the content push record set in which the push object performs a negative operation on the push content, and obtaining the number of first negative records corresponding to the first negative content push record; determining a second negative content push record corresponding to the target feature combination; Push record, obtain the number of second negative records corresponding to the second negative content push record; wherein, the content features of the pushed content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the pushed object in the second negative content push record are consistent with the object features in the target feature combination; determine the target feature exclusion degree corresponding to the target feature combination according to the number of the first negative records and the number of the second negative records, the target feature exclusion degree indicates the degree of exclusion between the content features and the object features in the target feature combination, so as to screen the candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination.

[0015] A computer-readable storage medium stores a computer program, which implements the following steps when executed by a processor: obtaining a target feature combination, wherein the target feature combination is obtained by combining an object feature and a content feature; obtaining a content push record set, wherein the content push record set includes multiple content push records, wherein the content push record includes a push content and a push object corresponding to the push content; determining a first negative content push record in the content push record set in which the push object performs a negative operation on the push content, and obtaining the number of first negative records corresponding to the first negative content push record; determining a second negative content push record corresponding to the target feature combination, Obtain the number of second negative records corresponding to the second negative content push record; wherein, the content features of the pushed content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the pushed object in the second negative content push record are consistent with the object features in the target feature combination; determine the target feature exclusion degree corresponding to the target feature combination according to the number of the first negative records and the number of the second negative records, the target feature exclusion degree indicating the degree of exclusion between the content features and the object features in the target feature combination, so as to screen the candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination.

[0016] In some embodiments, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a 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 steps in the above-mentioned method embodiments.

[0017] The content push method, apparatus, computer device and storage medium described above determine a first negative content push record in which a push object performs a negative operation on a push content in a content push record set, obtain the number of first negative records corresponding to the first negative content push record, and thus the number of first negative records reflects the number of content push records that perform negative operations on the push content; determine a second negative content push record corresponding to a target feature combination, and obtain the number of second negative records corresponding to the second negative content push record; since the content features of the push content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the push object in the second negative content push record are consistent with the object features in the target feature combination, the number of second negative records reflects the number of content push records corresponding to the target feature combination in each content push record that performs a negative operation; determine the target feature repulsion degree corresponding to the target feature combination according to the number of first negative records and the number of second negative records; since the target feature repulsion degree indicates the degree of repulsion between the content features and the object features in the target feature combination, the target feature repulsion degree reflects the degree of repulsion of the object having the object features in the target feature combination to the content having the content features in the target feature combination; the greater the target feature repulsion degree, the greater the object repulsion degree. Therefore, when the candidate content to be pushed is screened based on the target feature exclusion degree corresponding to the target feature combination, the candidate content with a larger target feature exclusion degree can be filtered out from the various candidate contents, that is, the candidate content corresponding to the content features with a larger degree of exclusion between the object features of the object of the content to be pushed is filtered out, thereby reducing the content with a larger degree of exclusion between the object features and being pushed to the object, thereby reducing the possibility of pushing the excluded content to the object and improving the accuracy of content push.

[0018] A content push method, the method comprising: receiving a content push request for a target push object; in response to the content push request, obtaining a candidate push content set corresponding to the target push object; determining a target exclusion feature combination based on target object features corresponding to the target push object, the target exclusion feature combination being determined based on a target feature exclusion degree corresponding to the target feature combination; and filtering candidate push content corresponding to the target exclusion feature combination from the candidate push content set to obtain target push content corresponding to the target push object.

[0019] A content push device, the device comprising: a content push request receiving module, used to receive a content push request for a target push object; a candidate push content set acquisition module, used to respond to the content push request and acquire a candidate push content set corresponding to the target push object; a target exclusion feature combination determination module, used to determine a target exclusion feature combination based on target object features corresponding to the target push object, wherein the target exclusion feature combination is determined based on a target feature exclusion degree corresponding to the target feature combination; and a target push content acquisition module, used to filter candidate push content corresponding to the target exclusion feature combination from the candidate push content set and acquire target push content corresponding to the target push object.

[0020] In some embodiments, the target exclusion feature combination determination module includes: a first candidate exclusion feature combination obtaining unit, used to obtain multiple candidate exclusion feature combinations; the candidate exclusion feature combination is screened from each target feature combination based on the target feature exclusion degree corresponding to the target feature combination; the target exclusion feature combination obtaining unit is used to compare the target object features corresponding to the target push object with the object features corresponding to the push object in the candidate exclusion feature combination, and use the exclusion feature combination with consistent comparison as the target exclusion feature combination. In some embodiments, the device further includes a candidate exclusion feature combination obtaining module, the candidate exclusion feature combination obtaining module including: a target feature combination obtaining unit, used to obtain multiple target feature combinations; the target feature combination is obtained by combining object features and content features; a content push record set obtaining unit, used to obtain a content push record set, the content push record set including multiple content push records, the content push record including push content and a push object corresponding to the push content; a first negative record quantity obtaining unit, used to determine a first negative content push record in the content push record set in which a push object performs a negative operation on the push content, and obtain a first negative record quantity corresponding to the first negative content push record; a second negative record quantity obtaining unit, used to determine a second negative record quantity corresponding to the target feature combination; Negative content push record, obtain the number of second negative records corresponding to the second negative content push record; wherein, the content features of the pushed content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the pushed object in the second negative content push record are consistent with the object features in the target feature combination; a target feature exclusion degree determination unit, used to determine the target feature exclusion degree corresponding to the target feature combination according to the first number of negative records and the second number of negative records, the target feature exclusion degree indicating the degree of exclusion between the content features and the object features in the target feature combination; a second candidate exclusion feature combination obtaining unit, used to screen and obtain a candidate exclusion feature combination from each of the target feature combinations based on the target feature exclusion degree corresponding to the target feature combination.

[0021] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the following steps when executing the computer program: receiving a content push request for a target push object; in response to the content push request, obtaining a candidate push content set corresponding to the target push object; determining a target exclusion feature combination based on target object features corresponding to the target push object, wherein the target exclusion feature combination is determined based on a target feature exclusion degree corresponding to a target feature combination; and filtering candidate push content corresponding to the target exclusion feature combination from the candidate push content set to obtain target push content corresponding to the target push object.

[0022] A computer-readable storage medium stores a computer program, which implements the following steps when executed by a processor: receiving a content push request for a target push object; in response to the content push request, obtaining a candidate push content set corresponding to the target push object; determining a target exclusion feature combination based on target object features corresponding to the target push object, wherein the target exclusion feature combination is determined based on a target feature exclusion degree corresponding to a target feature combination; and filtering candidate push content corresponding to the target exclusion feature combination from the candidate push content set to obtain target push content corresponding to the target push object.

[0023] In some embodiments, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a 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 steps in the above-mentioned method embodiments.

[0024] The above-mentioned content push method, apparatus, computer equipment and storage medium receive a content push request for a target push object, and in response to the content push request, obtain a candidate push content set corresponding to the target push object, and determine a target exclusion feature combination based on the target object features corresponding to the target push object. Since the target exclusion feature combination is determined based on the target feature exclusion degree corresponding to the target feature combination, and the target feature exclusion degree reflects the degree of exclusion of an object having object features in the target feature combination to content having content features in the target feature combination, the larger the target feature exclusion degree, the more the object excludes the content, and thus the target feature combination with a larger target feature exclusion degree can be used as the target exclusion feature combination, thereby filtering the candidate push content corresponding to the target exclusion feature combination from the candidate push content set, and filtering out the candidate push content excluded by the target push object, thereby reducing the possibility of pushing the excluded content to the target push object and improving the accuracy of content push. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A diagram of an application environment of a content push method in some embodiments;

[0026] Figure 2 It is a flowchart of a content push method in some embodiments;

[0027] Figure 3 is a schematic diagram of a content push method in some embodiments;

[0028] Figure 4 A distribution diagram of the degree of rejection of academic education advertisements among people with different academic qualifications in some embodiments;

[0029] Figure 5 It is a flowchart of a content push method in some embodiments;

[0030] Figure 6 is a schematic diagram of a content push method in some embodiments;

[0031] Figure 7 It is a structural block diagram of a content push device in some embodiments;

[0032] Figure 8 It is a structural block diagram of a content push device in some embodiments;

[0033] Fig. 9 is an internal structure diagram of a computer device in some embodiments;

[0034] Fig.10 1 is a diagram of the internal structure of a computer device in some embodiments. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0036] The content push method provided by this application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The server 104 can be a server for content push, and the terminal 102 can be installed with a client corresponding to the server 104. The server 104 can push content to the terminal 102, and the terminal 102 can display the pushed content through the client. For example, the terminal 102 can receive an advertisement (Advertising, AD) pushed by the server 104, and display the pushed content in the client. The interface displayed by the client may include a push content display area, and the push content display area is used to display the pushed content. The push content display area can be predetermined.

[0037] Specifically, the server 104 can obtain a target feature combination, which is obtained by combining an object feature with a content feature, obtain a content push record set, the content push record set includes multiple content push records, the content push record includes push content and a push object corresponding to the push content, determine a first negative content push record in the content push record set in which the push object performs a negative operation on the push content, obtain the number of first negative records corresponding to the first negative content push record, determine a second negative content push record corresponding to the target feature combination, obtain the number of second negative records corresponding to the second negative content push record, wherein the content feature of the push content in the second negative content push record is consistent with the content feature in the target feature combination, and the object feature of the push object in the second negative content push record is consistent with the object feature in the target feature combination. The server 104 can also determine the target feature repulsion degree corresponding to the target feature combination according to the number of the first negative records and the number of the second negative records, the target feature repulsion degree indicating the degree of repulsion between the content feature and the object feature in the target feature combination, so as to screen the candidate content to be pushed based on the target feature repulsion degree corresponding to the target feature combination. For example, the server 104 can select a target feature combination that meets the exclusion degree screening condition according to the target feature exclusion degree corresponding to the target feature combination as an exclusion feature combination. When the target push object corresponding to the terminal 102 has the object feature in the exclusion feature combination, when the server 104 receives a content push request for the target push object, it can respond to the content push request and obtain the candidate push content set corresponding to the target push object. The server 104 can filter out the candidate push content with the content feature in the exclusion feature combination from the candidate push content set, and push the remaining candidate push content in the candidate push content set to the terminal 102.

[0038] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0039] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. Basic artificial intelligence technologies generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operating / interactive systems, mechatronics and other technologies. Artificial intelligence software technologies mainly include computer vision technology, speech processing technology, natural language processing technology, as well as machine learning / deep learning, autonomous driving, smart transportation and other major directions.

[0040] Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0041] With the research and advancement of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, Internet of Vehicles, automatic driving, smart transportation, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0042] The solution provided in the embodiment of the present application involves technologies such as machine learning of artificial intelligence, and is specifically explained through the following embodiments: the server 104 can use the content push model to obtain a set of candidate push content corresponding to the target push object. Among them, the content push model can be a trained neural network model. The content push model can determine the candidate object content corresponding to the push object based on the object information of the push object and the content information of the push content. For example, the server can obtain a pre-stored push object, and use the content push model to determine the candidate push content corresponding to the target push object from the pre-stored push objects to form a set of candidate push content.

[0043] The terminal 102 may be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices, and the server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0044] It can be understood that the above application scenario is only an example and does not constitute a limitation on the content push method provided in the embodiment of the present application. The method provided in the embodiment of the present application can also be applied in other application scenarios. For example, the content push method provided in the present application can be executed by the terminal 102, and the terminal 102 can upload the target feature exclusion degree corresponding to the obtained target feature combination to the server 104. The server 104 can store the target feature exclusion degree corresponding to the target feature combination, and can also forward the target feature exclusion degree corresponding to the target feature combination to other devices.

[0045] In some embodiments, Figure 2 As shown, a content push method is provided, which can be executed by a terminal or a server, or by both the terminal and the server. In the embodiment of the present application, the method is applied to Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0046] S202, obtaining a target feature combination, where the target feature combination is obtained by combining an object feature and a content feature.

[0047] Among them, the object refers to something to which content can be pushed, such as a user. The object can also be called a push object. The object feature refers to the feature of the attribute of the push object. The push object can have multiple attributes. For example, when the push object is a user, the attribute of the push object may include at least one of age, gender, marital status, occupation or education. Each object attribute may correspond to an object feature. The object attribute refers to the attribute of the push object. The object feature may be, for example, a user feature. The user feature may include at least one of age feature, gender feature or occupation feature.

[0048] Content refers to information that can be pushed, and content can be, for example, advertisements, and advertisements can be, for example, online advertisements. Online advertisements can also be called Internet advertisements, and Internet advertisements refer to advertisements placed on advertisement slots on Internet platforms. Content can also be called pushed content. Advertisers can place advertisements through advertising exchanges. Advertising exchanges (ADX, AD-Exchange) are entities that connect media owners and advertisers, and advertising exchanges place advertisers' advertisements on advertisement slots provided by media owners. Advertising exchanges can collect user information, create user portraits, and place advertisements to users based on data such as user interests or geographic locations. The content push method provided in this application can be deployed in an advertising exchange to improve the accuracy of advertising placement.

[0049] Content features refer to the features of the attributes of the content. The content can have multiple attributes. For example, when the content is an advertisement, it can include at least one of the following attributes: industry, commodity, optimization target, sensitive word, or directional POI (Point Of Interest). Each content attribute can correspond to a content feature. Content attributes refer to the attributes of the pushed content. Content features can be, for example, advertising features. Advertising features can include at least one of industry features and sensitive word features. Targeted POI refers to POI positioning, which means information point positioning. POI can be a house, a shop, a mailbox, or a bus stop. The optimization target (bid_objective) refers to the behavior that the advertiser expects the user to have in response to the advertisement. For example, when advertisers in the game industry place APP game advertisements, they can use the "download" behavior as the optimization target of the advertisement.

[0050] A feature combination is composed of an object feature and a content feature, and each feature combination includes an object feature and a content feature. At least one of the object features and content features included in different feature combinations is different. The object features included in different feature combinations can be the same or different. The content features included in different feature combinations can be the same or different. For example, target feature combination 1 is (Ua1, Ub1), and target feature combination 2 is (Ua1, Ub2), where Ua1 is an object feature, and Ub1 and Ub2 are content features. The target feature combination can be any feature combination, or it can be a feature combination screened from multiple feature combinations. There can be one or more target feature combinations, and multiple means at least two.

[0051] Specifically, the server may determine an object attribute set corresponding to the pushed object, where the object attribute set includes multiple object attributes, such as age, gender, education level, or occupation. The server may obtain an attribute value corresponding to the object attribute of the pushed object, and determine an object feature corresponding to the object attribute based on the attribute value corresponding to the object attribute. For example, the server may normalize the attribute value and use the result of the normalization as the object feature, or the server may determine a range to which the attribute value belongs, and determine the object feature based on the range to which it belongs.

[0052] In some embodiments, the server may obtain the attribute value corresponding to the object attribute of the push object, and determine the object feature based on the attribute value. For example, when the object attribute is age, if the age of the push object is 30 years old, the attribute value is 30. Among them, for the same object attribute, different push objects may obtain different attribute values. For example, when the object attribute is age, the age of push object 1 is 20, and the age of push object 2 is 50. For age, the attribute value obtained by push object 1 is 20, and the attribute value obtained by push object 2 is 50. The server can use the object features corresponding to each object attribute to represent the push object. For example, a push object can be represented by (a1, a2, ... ai ..., an), where ai represents the object feature corresponding to the i-th object attribute of the push object. i is a positive integer greater than or equal to 1 and less than or equal to n.

[0053] In some embodiments, the server may obtain a content attribute set corresponding to the pushed content, and the content attribute set includes multiple content attributes, for example, the content attribute set may include content attributes such as industry, commodity or sensitive words. The server may obtain the attribute value corresponding to the content attribute of the pushed content, and determine the content feature corresponding to the content attribute based on the attribute value of the content attribute. For example, the attribute value of the content attribute may be normalized to obtain the content feature, and of course, the content feature may be obtained in other ways, which are not limited here. For the same content attribute, the attribute values ​​obtained by different pushed contents may be different. For example, when the content attribute is the content type, the content type of the pushed content A is the marriage and love category, and the content type of the pushed content B is the non-marriage and love category. For the content type, the attribute value obtained by the pushed content A is the marriage and love category, and the attribute value obtained by the pushed content B is the non-marriage and love category. The server may use the content features corresponding to each content attribute to represent the pushed content. For example, a pushed content may be represented by (b1, b2, ... bj ..., bm), where bj represents the content attribute of the pushed content. j is a positive integer greater than or equal to 1 and less than or equal to m.

[0054] In some embodiments, the server may obtain multiple attribute combinations based on a content attribute set and an object attribute set. The attribute combination is a combination of content attributes and object attributes, and each attribute combination includes a content attribute and an object attribute. For example, the server may obtain a content attribute from the content attribute set and an object attribute from the object attribute set, and form the obtained content attribute and the obtained object attribute into an attribute combination. The server may determine the object features corresponding to the object attributes in the attribute combination, and determine the content features corresponding to the content attributes in the attribute combination, and form the determined object features and the determined content features into a candidate feature combination. Since the object features obtained by different push objects may be different for the same object attribute, there may be multiple object features corresponding to the same object attribute. Similarly, there are multiple content features corresponding to the same content attribute, so the same attribute combination may obtain multiple candidate feature combinations. For example, the attribute combination is (gender, content type). Since "gender" can include "male" and "female", and "content type" can include "marriage type" and "non-marriage type", assuming that the feature corresponding to "male" is 1, the feature corresponding to "female" is 0, the feature corresponding to "content type" is 2, and the feature corresponding to "non-marriage type" is 4, the candidate feature combinations may include at least one of (1, 2), (1, 4), (0, 2) or (0, 4).

[0055] In some embodiments, the target feature combination may be obtained by screening from a set of candidate feature combinations. The set of candidate feature combinations may include multiple candidate feature combinations. The server may obtain the reference feature exclusion degree corresponding to each candidate feature combination, and the reference feature exclusion degree may be preset, for example, it may be preset based on expert knowledge. The reference feature exclusion degree is used to predict the degree of exclusion of an object having object features in the candidate feature combination to content having content features in the candidate feature combination. The reference feature exclusion degree is a predicted value, and there may be a difference between the actual feature exclusion degree. The server may screen the target feature combination from the set of candidate feature combinations based on the reference feature exclusion degree, for example, a candidate feature combination whose reference feature exclusion degree is greater than a reference threshold value may be obtained from the candidate feature combination as the target feature combination. Alternatively, the server may obtain a candidate feature combination whose reference exclusion degree is ranked before the reference exclusion degree threshold value from the set of candidate feature combinations as the target feature combination, wherein the reference threshold value and the reference exclusion degree threshold value may be preset or may be set as required. The reference exclusion ranking refers to the ranking of the reference exclusion in the reference exclusion sequence. The reference exclusion sequence is a sequence obtained by arranging the reference feature exclusion corresponding to each candidate feature combination in order from large to small. The greater the reference feature ranking, the higher the ranking in the reference exclusion sequence. Among them, the reference feature exclusion can also be determined based on relevant field knowledge such as operations or products. For example, it can be analyzed from common sense that people with high academic qualifications have a higher degree of rejection of academic education advertisements. For people with high academic qualifications, such as PhDs, the reference feature exclusion of the candidate feature combination (features of PhDs, features corresponding to academic education) can be set to be larger.

[0056] S204, obtaining a content push record set, where the content push record set includes multiple content push records, and the content push record includes push content and a push object corresponding to the push content.

[0057] The push content refers to the pushed content, such as a pushed advertisement. The push object corresponding to the push content refers to the push object to which the push content is pushed. For example, if the push content is S1, when the push content S1 is pushed to object 1, object 1 is the push object corresponding to the push content S1. Each content push record may include a push content and a push object. For example, the content push record may be "pushed the push content S1 to push object 1".

[0058] Specifically, the content push record can be obtained based on the log of the pushed content. When the server pushes content, it can record the pushed content and the push object corresponding to the pushed content in the log, and can record the object behavior of the push object on the pushed content in the log. Object behavior refers to the behavior of the push object with respect to the pushed content. Object behavior can also be called user behavior, and user behavior can include at least one of exposure, negative feedback behavior, or positive feedback behavior, and of course can also include other behaviors, such as conversion behavior. Conversion behavior refers to the behavior of users on the advertising trading platform. For example, for APP (Application) type advertisements, conversion behavior can include at least one of downloading, activation, or payment.

[0059] In some embodiments, the content push record in the content push record set may be obtained based on the target feature combination. Specifically, the server may store multiple push objects, and the server may obtain a push object having the object features in the target feature combination from the multiple push objects as the first push object; the server may also store multiple push contents, and the server may obtain multiple push contents having the content features in the target feature combination from the multiple push contents as the first push content; the first push content is pushed to the first push object, the operation type of the first push object for the first push content is obtained, the first push content is used as the push content in the content push record, the first push object is used as the push object in the content push record, and the operation type is used as the operation type in the content push record to generate a content push record. One push may correspond to one content push record. Among them, there may be multiple target feature combinations. The content push record set may include content push records corresponding to each target feature combination.

[0060] In some embodiments, the content push record set may be a set consisting of multiple content push records obtained by performing content push within a historical time period.

[0061] S206: Determine a first negative content push record in the content push record set in which the push object performs a negative operation on the pushed content, and obtain a first negative record quantity corresponding to the first negative content push record.

[0062] Among them, there can be multiple operations for pushed content, and the operations for pushed content can be divided into multiple operation types. For example, the operations for pushed content can include at least one operation type of positive operation or negative operation. Negative operation is used to reflect that the object is repulsive to the pushed content. Negative operation can be set as needed or preset. For example, there can be a "not interested" control in the displayed advertisement, and the negative operation can be a trigger operation for the "not interested" control. Negative operation is a concept relative to positive operation. Other operations except negative operations in the operations for pushed content can be regarded as positive operations. For example, positive operations can include at least one of clicks or likes. Positive operation can also be called positive feedback behavior, and negative operation can also be called negative feedback behavior.

[0063] The content push record may also include the type of operation performed by the push object on the pushed content, and the operation type may be any one of a negative operation or a positive operation. The first negative content push record refers to a content push record whose operation type is a negative operation. The number of first negative records refers to the number of first negative content push records.

[0064] Specifically, the server can obtain the operation type corresponding to each content push record in the content push record set, take the content push record with the operation type of negative operation as the first negative content push record, count the number of each first negative content push record, and take the counted number as the first negative record number.

[0065] S208, determine the second negative content push record corresponding to the target feature combination, and obtain the number of second negative records corresponding to the second negative content push record; wherein the content features of the pushed content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the pushed object in the second negative content push record are consistent with the object features in the target feature combination.

[0066] The content features of the pushed content in the second negative content push record are the same as the content features in the target feature combination, and the object features of the pushed object in the second negative content push record are the same as the object features in the target feature combination.

[0067] Specifically, the second negative record number refers to the number of second negative content push records. The server can determine the second negative content push record corresponding to the target feature combination from the content push record set, or determine the second negative content push record corresponding to the target feature combination from each first negative content push record.

[0068] In some embodiments, the server may determine the object attribute corresponding to the object feature in the target feature combination as the target object attribute, determine the content attribute corresponding to the content feature in the target feature combination as the target content attribute, obtain the object feature of the push object in the first negative content push record on the target object attribute as the comparison object feature, obtain the content feature of the push content in the first negative content push record on the target content attribute as the comparison content feature, and filter out the second negative content push record from each first negative content push record based on the comparison object feature and the comparison content feature. For example, the server may compare the object feature in the target feature combination with the comparison object feature, and when the comparison is consistent, compare the content feature in the target feature combination with the comparison content feature, and when the comparison is consistent, use the first negative content push record as the second negative content push record. Or the server may compare the content feature in the target feature combination with the comparison content feature, and when the comparison is consistent, compare the object feature in the target feature combination with the comparison object feature, and when the comparison is consistent, use the first negative content push record as the second negative content push record.

[0069] In some embodiments, when the object features in the target feature combination are inconsistent with the comparison object features, or when the content features in the target feature combination are inconsistent with the comparison content features, it is determined that the first negative content push record is not the second negative content push record corresponding to the target feature combination.

[0070] S210, determining the target feature exclusion degree corresponding to the target feature combination according to the first number of negative records and the second number of negative records, wherein the target feature exclusion degree represents the degree of exclusion between the content features and the object features in the target feature combination, so as to screen the candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination.

[0071] Among them, the target feature exclusion degree is positively correlated with the number of second negative records. The target feature exclusion degree is used to reflect the degree of exclusion of the push object with object characteristics to the push content with content characteristics. The degree of exclusion is positively correlated with the target feature exclusion degree. The greater the target feature exclusion degree, the greater the degree of exclusion. Candidate content refers to the content to be pushed. The candidate content corresponding to different objects can be the same or different. The candidate content to be pushed refers to the push content corresponding to the push object that needs to push content. The candidate content to be pushed corresponding to different push objects can be the same or different.

[0072] Among them, positive correlation means: when other conditions remain unchanged, the two variables change in the same direction, and when one variable changes from large to small, the other variable also changes from large to small. It can be understood that the positive correlation here means that the direction of change is consistent, but it does not require that when one variable changes a little, the other variable must also change. For example, when variable a is 10 to 20, variable b can be set to 100, and when variable a is 20 to 30, variable b can be set to 120. In this way, the direction of change of a and b is that when a becomes larger, b also becomes larger. But when a is in the range of 10 to 20, b can be unchanged. Negative correlation means: when other conditions remain unchanged, the two variables change in opposite directions, and when one variable changes from large to small, the other variable changes from small to large. It can be understood that the negative correlation here means that the direction of change is opposite, but it does not require that when one variable changes a little, the other variable must also change.

[0073] Specifically, when the server obtains a content push request for a target push object, it can obtain the candidate content to be pushed corresponding to the target push object, for example, it can obtain the candidate push content set corresponding to the target push object, filter the candidate push content in the candidate push content set of the target push object based on the target feature exclusion degree corresponding to the target feature combination, and push the remaining candidate push content in the candidate push content set after filtering to the target push object. The target push object can be any push object.

[0074] In some embodiments, the server can obtain the characteristic exclusion degree corresponding to each candidate push content in the candidate push content set, and arrange the candidate push content in order of the characteristic exclusion degree from small to large to obtain a candidate push content sequence. The smaller the characteristic exclusion degree, the higher the ranking of the candidate push content in the candidate push content sequence. When pushing content to the target push object, the content with a higher ranking in the candidate push content sequence is pushed first, thereby reducing the possibility of pushing content with a higher degree of exclusion to the target push object.

[0075] In some embodiments, the server may count the number of content push records included in the content push record set, use the counted number as the number of content push records, and determine the target feature exclusion degree corresponding to the target feature combination based on the first number of negative records, the second number of negative records, and the number of content push records. For example, the server may calculate the ratio of the first number of negative records to the number of content push records, obtain the first negative record possibility according to the calculated ratio, calculate the ratio of the second number of negative records to the first number of negative records, obtain the second negative record possibility according to the calculated ratio, and calculate the target feature exclusion degree corresponding to the target feature combination based on the first negative record possibility and the second negative record possibility.

[0076] In some embodiments, there are multiple target feature combinations. The server can filter out exclusion feature combinations from various target feature combinations based on the target feature exclusion degree, and filter candidate content to be pushed based on the exclusion feature combinations.

[0077] In some embodiments, the server may obtain the target push value corresponding to the target feature combination, screen out the exclusion feature combination from each target feature combination based on the target feature exclusion degree and the target push value, and screen the candidate content to be pushed based on the exclusion feature combination. The target push value is used to reflect the benefits that can be brought to the owner of the pushed content by pushing the pushed content having the content features in the target feature combination to the push object having the object features in the target feature combination.

[0078] In the above-mentioned content push method, a first negative content push record in which a push object performs a negative operation on a push content in a content push record set is determined, and the number of first negative records corresponding to the first negative content push record is obtained, so that the number of first negative records reflects the number of content push records that perform negative operations on the push content, and a second negative content push record corresponding to a target feature combination is determined, and the number of second negative records corresponding to the second negative content push record is obtained. Since the content feature of the push content in the second negative content push record is consistent with the content feature in the target feature combination, and the object feature of the push object in the second negative content push record is consistent with the object feature in the target feature combination, the number of second negative records reflects the number of content push records corresponding to the target feature combination in each content push record that performs a negative operation, and the target feature repulsion degree corresponding to the target feature combination is determined according to the number of first negative records and the number of second negative records. Since the target feature repulsion degree indicates the degree of repulsion between the content feature and the object feature in the target feature combination, the target feature repulsion degree reflects the degree of repulsion of the object having the object feature in the target feature combination to the content having the content feature in the target feature combination, and the greater the target feature repulsion degree, the greater the object repulsion degree is. Therefore, when the candidate content to be pushed is screened based on the target feature exclusion degree corresponding to the target feature combination, the candidate content with a larger target feature exclusion degree can be filtered out from the various candidate contents, that is, the candidate content corresponding to the content features with a larger degree of exclusion between the object features of the object of the content to be pushed is filtered out, thereby reducing the content with a larger degree of exclusion between the object features and being pushed to the object, thereby reducing the possibility of pushing the excluded content to the object and improving the accuracy of content push.

[0079] In some embodiments, there are multiple target feature combinations, and the method further includes: selecting a target feature combination that meets an exclusion degree screening condition according to the target feature exclusion degree corresponding to the target feature combination, as an exclusion feature combination, the exclusion feature combination being used to filter the push content having the content features in the exclusion feature combination when pushing content to a push object having the object features in the exclusion feature combination; the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than an exclusion degree threshold or the target exclusion degree being ranked before the exclusion degree ranking threshold.

[0080] Among them, the exclusion feature combination refers to the target feature combination that meets the exclusion screening condition in each target feature combination. There can be one or more exclusion feature combinations, and multiple refers to at least two. The size of the exclusion threshold can be predetermined, and the exclusion threshold can be a fixed value such as 0.9 or 0.8. The exclusion threshold can also be determined according to the target feature exclusion corresponding to each target feature combination. For example, the server can perform a mean operation on the target feature exclusion corresponding to each target feature combination, obtain the mean exclusion, obtain a preset exclusion coefficient, multiply the mean exclusion by the exclusion coefficient, and use the result of the multiplication as the exclusion threshold. Among them, the exclusion coefficient can be a positive number greater than 1.

[0081] The target exclusion ranking refers to the ranking of the target feature exclusion in the exclusion sequence. The exclusion sequence is a sequence obtained by arranging the target feature exclusion corresponding to each target feature combination in order from large to small. The larger the target feature exclusion, the higher the ranking in the exclusion sequence.

[0082] The size of the exclusion degree sorting threshold can be preset. For example, the exclusion degree sorting threshold can be a fixed value such as 10 or 15. The exclusion degree sorting threshold can also be determined based on the number of target feature combinations. For example, the server can obtain a first sorting coefficient. The first sorting coefficient can be preset. The first sorting coefficient can be a positive number less than 1, such as 0.5 or 0.6. The first sorting coefficient is multiplied by the number of target feature combinations, and the result of the multiplication is used as the exclusion degree sorting threshold.

[0083] Specifically, the server may arrange the target feature exclusion degrees of each target feature combination in order from large to small, use the arranged sequence as the exclusion degree sequence, compare the ranking of the target feature exclusion degrees of the target feature combination in the exclusion degree sequence with the exclusion degree ranking threshold, that is, compare the target exclusion degree ranking corresponding to the target feature exclusion degree with the exclusion degree ranking threshold, and when the target exclusion degree ranking corresponding to the target feature exclusion degree is before the exclusion degree ranking threshold, the target feature combination corresponding to the target feature exclusion degree is used as the exclusion feature combination. Figure 3 As shown, the server obtains the target feature combination based on expert knowledge, obtains a content push record set based on the log of the pushed content, determines the target feature exclusion degree corresponding to the target feature combination based on the content push record set, and uses the exclusion degree screening condition to screen out the exclusion feature combination from each target feature combination.

[0084] In some embodiments, the server may compare the target feature exclusion degree with an exclusion degree threshold, and when the target feature exclusion degree is greater than the exclusion degree threshold, the target feature combination corresponding to the target feature exclusion degree is used as the exclusion feature combination.

[0085] In some embodiments, the server may take each target feature combination whose target feature exclusivity is greater than an exclusivity threshold and whose target exclusivity ranking is before the exclusivity ranking threshold as an exclusivity feature combination.

[0086] In some embodiments, when the server obtains a content push request for a target push object, the server may compare the object features of the target push object with the object features in the exclusion feature combination. When the comparison is consistent, the server may obtain the content features in the exclusion feature combination that are consistent with the comparison, obtain a candidate push content set for the target push object, compare the content features of each candidate push content in the candidate push content set with the content features in the exclusion feature combination that are consistent with the comparison, filter out the candidate push content that are consistent with the comparison from the candidate push content set, and use the remaining candidate push content in the candidate push content set as the target push content to be pushed to the target push object. The server may push each target push content to the target push object.

[0087] In this embodiment, a target feature combination that meets the exclusion degree screening condition is selected according to the target feature exclusion degree corresponding to the target feature combination as the exclusion feature combination. Since the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree ranking being before the exclusion degree ranking threshold, the target feature combination with a larger target feature exclusion degree can be used as the exclusion feature combination. Since the exclusion feature combination is used to filter the push content having the content features in the exclusion feature combination when pushing content to the push object having the object features in the exclusion feature combination, the content corresponding to the content features having a larger object feature exclusion degree can be filtered out, thereby reducing the content having a larger target feature exclusion degree being pushed to the push object, thereby reducing the situation where the content rejected by the push object is pushed to the push object, thereby improving the accuracy of content push.

[0088] In some embodiments, the method also includes: obtaining a reference push record set, the reference push record set including multiple reference push records, the reference push record being a content push record corresponding to a target feature combination; obtaining a content push value corresponding to the reference push record; performing statistics on the content push value corresponding to the reference push record in the reference push record set to obtain a target push value corresponding to the target feature combination; and screening candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination and the target push value.

[0089] Among them, the reference push record is the content push record corresponding to the target feature combination. The push object in the reference push record has the object features in the target feature combination, and the push content in the reference push record has the content features in the target feature combination. For example, the target feature combination 1 is (Ua1, Ub1), and the reference push record corresponding to the target feature combination 1 is "the push content S1 is pushed to the push object 1", then the push object 1 has the object feature of Ua1, and the push content S1 has the content feature of Ub1. The content push records in the reference push record set may be the same as or different from the content push records in the content push record set.

[0090] The content push value is used to reflect the revenue brought by the placed advertisements to the content owner. The content owner refers to the owner of the content. For example, when the content is an advertisement, the content owner can be an advertiser. The advertiser refers to the entity that displays its advertisement through the advertising space of the Internet platform. For example, when the pushed content is a product advertisement, the content push value can be the amount spent by the user to purchase the product through the product advertisement. Each reference push record can correspond to a content push value, and the content push values ​​corresponding to different reference push records can be the same or different. The target push value is the value obtained by statistically analyzing the content push values ​​corresponding to each reference push record.

[0091] Specifically, the server can obtain the content push value corresponding to each reference push record in the reference push record set, and perform statistical operations on each content push value. The statistical operation may include at least one of addition operation, mean operation, median operation or mode operation. For example, the server can perform a sum operation on each content push value, and use the result of the sum operation as the target push value corresponding to the target feature combination.

[0092] In some embodiments, the reference push record set may be content push records within a target time period. The target time period may be, for example, one month. For example, the server may push content to a push object having the object features in the target feature combination within the target time period, and form a reference push record set with content push records corresponding to the push objects having the object features in the target feature combination within the target time period.

[0093] In this embodiment, the candidate content to be pushed is screened based on the target feature exclusion degree and the target push value corresponding to the target feature combination, so that the content to be pushed can be screened based on the combined feature exclusion degree and the push value, thereby improving the accuracy of content screening and thus improving the accuracy of content push.

[0094] In some embodiments, there are multiple target feature combinations, and the method further includes: selecting a target feature combination that meets the exclusion degree screening condition and the push value screening condition as the exclusion feature combination according to the target feature exclusion degree and the target push value corresponding to the target feature combination; the exclusion feature combination is used to filter the push content having the content features in the exclusion feature combination when pushing content to the push object having the object features in the exclusion feature combination; the push value screening condition includes at least one of the target push value being greater than the push value threshold or the target push value ranking being before the value ranking threshold; the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree ranking being before the exclusion degree ranking threshold.

[0095] The exclusion feature combination may be a target feature combination that satisfies both the exclusion degree screening condition and the push value screening condition.

[0096] The size of the push value threshold can be determined in advance or can be determined based on the target push values ​​corresponding to each target feature combination. For example, the server can perform an average operation on the target push values ​​corresponding to each target feature combination to obtain the push value mean, obtain a preset push value coefficient, multiply the push value coefficient by the push value mean, and use the result of the multiplication as the push value threshold, where the push value coefficient can be a positive number less than 1.

[0097] The target push value ranking refers to the ranking of the target push value in the push value sequence. The push value sequence is a sequence obtained by arranging the target push values ​​corresponding to each target feature combination in order from small to large. The smaller the target push value, the higher the ranking in the push value sequence.

[0098] The value ranking threshold can be preset, for example, it can be a fixed value such as 5 or 10. The value ranking threshold can also be determined according to the number of target feature combinations. For example, the server can obtain the second ranking coefficient, multiply the second ranking coefficient by the number of target feature combinations, and use the result of the multiplication as the value ranking threshold. The second ranking coefficient can be the same as the first ranking coefficient, or it can be different from the first ranking coefficient. The second ranking coefficient can be a positive number less than 1, for example, 0.3 or 0.5.

[0099] Specifically, the server can obtain target feature combinations that meet the exclusion degree screening conditions from each target feature combination to form a first feature combination set, obtain target feature combinations that meet the push value screening conditions from each target feature combination to form a second feature combination set, perform an intersection operation on the first feature combination set and the second feature combination set, use the set obtained by the intersection operation as the third feature combination set, and use each target feature combination in the third feature combination set as an exclusion feature combination.

[0100] like Figure 4 As shown, the GPM effect and the distribution of rejection of academic education advertisements on people with different educational backgrounds are displayed, where academic education is the type of advertisement. GPM (Cost Per Mille) is the revenue obtained from one thousand exposures, which can also be called "charge per thousand exposures", that is, the advertisement is charged in the form of exposure, and each thousand exposures is regarded as a settlement unit. The solid rectangular box represents GPM, and the curve represents the rejection degree. It can be seen from the figure that people with a doctoral degree have a higher rejection degree for academic education advertisements, and people with a doctoral degree have a lower GPM for academic education advertisements. Therefore, when pushing advertisements to people with a doctoral degree, academic education advertisements can be filtered out from the candidate content to be pushed corresponding to the people with a doctoral degree, thereby improving the push accuracy.

[0101] In some embodiments, GPM can be calculated using formula (1). GMV(ai, bj) represents the total transaction amount of users with user characteristics ai for advertisements with advertising characteristics bj within a period of time, and expose_num represents the number of exposures. GPM(ai, bj) represents the revenue obtained by exposing an advertisement with advertising characteristics bj to a user with user characteristics ai one thousand times.

[0102]

[0103] In this embodiment, according to the target feature exclusion degree and the target push value corresponding to the target feature combination, a target feature combination that meets the exclusion degree screening condition and the push value screening condition is selected as the exclusion feature combination. Since the push value screening condition includes at least one of the target push value being greater than the push value threshold or the target push value being ranked before the value ranking threshold, and the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree being ranked before the exclusion degree ranking threshold, a target feature combination with a larger target feature exclusion degree and a smaller target push value can be used as an exclusion feature combination. When pushing content, push content with a larger degree of exclusion and a smaller value is filtered out, thereby improving the accuracy of content push.

[0104] In some embodiments, statistics are performed on the content push values ​​corresponding to the reference push records in the reference push record set to obtain the target push value corresponding to the target feature combination, including: summing up the content push values ​​corresponding to the reference push records in the reference push record set to obtain the overall content push value; obtaining the number of reference push records in the reference push record set as the number of reference content records; calculating the ratio of the overall content push value to the number of reference content records, and obtaining the target push value corresponding to the target feature combination based on the calculated ratio.

[0105] The reference content record quantity refers to the quantity of content push records included in the reference push record set.

[0106] Specifically, the server may perform a sum operation on the content push values ​​corresponding to the reference push records, and use the result of the sum operation as the overall content push value.

[0107] In some embodiments, the server may calculate the ratio of the total content push value to the number of reference content records as the unit content push value corresponding to the target feature combination, and the target push value corresponding to the target feature combination is positively correlated with the unit content push value. For example, the server may obtain the preset number of content records, perform a product operation on the preset number of content records and the unit content push value, and use the result of the operation as the target push value, wherein the preset number of content records may be a preset fixed value, such as 1000.

[0108] In this embodiment, the ratio of the overall content push value to the number of reference content records is calculated, and the target push value corresponding to the target feature combination is obtained based on the calculated ratio, thereby improving the accuracy of the target push value.

[0109] In some embodiments, determining the target feature exclusion degree corresponding to the target feature combination based on the first number of negative records and the second number of negative records includes: determining the number of content push records corresponding to the content push record set; calculating the ratio of the first number of negative records to the number of content push records, and obtaining the first negative record possibility based on the calculated ratio; calculating the ratio of the second number of negative records to the first number of negative records, and obtaining the second negative record possibility based on the calculated ratio; and calculating the target feature exclusion degree corresponding to the target feature combination based on the first negative record possibility and the second negative record possibility.

[0110] Among them, the number of content push records refers to the number of content push records included in the content push record set. The first negative record possibility is obtained based on the ratio of the first negative record number to the content push record number. For example, the ratio of the first negative record number to the content push record number can be positively correlated with the first negative record possibility. The second negative record possibility is obtained based on the ratio of the second negative record number to the first negative record number. For example, the ratio of the second negative record number to the first negative record number can be positively correlated with the second negative record possibility. The target feature repulsion is positively correlated with the first negative record possibility, and the target feature repulsion is positively correlated with the second negative record possibility.

[0111] Specifically, the server may calculate the ratio of the first number of negative records to the number of content push records, and use the calculated ratio as the first negative record possibility. The server may calculate the ratio of the second number of negative records to the first number of negative records, and use the calculated ratio as the second negative record possibility.

[0112] In some embodiments, the server may calculate the first negative record possibility and the second negative record possibility, and obtain the target feature exclusion degree corresponding to the target feature combination based on the calculated result. For example, the server may perform a product operation on the first negative record possibility and the second negative record possibility, and use the result of the product operation as the standard negative record possibility, and obtain the target feature exclusion degree corresponding to the target feature combination based on the standard negative record possibility. The target feature exclusion degree is positively correlated with the standard negative record possibility.

[0113] In this embodiment, since the first negative record possibility is obtained based on the ratio of the first negative record number to the content push record number, and the second negative record possibility is obtained based on the ratio of the second negative record number to the first negative record number, the target feature exclusion degree corresponding to the target feature combination is obtained based on the first negative record possibility and the second negative record possibility, thereby improving the accuracy of the target feature exclusion degree.

[0114] In some embodiments, calculating the target feature exclusion degree corresponding to the target feature combination based on the first negative record possibility and the second negative record possibility includes: calculating the product of the first negative record possibility and the second negative record possibility to obtain the standard negative record possibility corresponding to the target feature combination; obtaining the number of content push records corresponding to the target feature combination in the content push record set as the number of combined records; calculating the ratio of the number of combined records to the number of content push records, and obtaining the push record possibility corresponding to the target feature combination according to the calculated ratio; obtaining the target feature exclusion degree corresponding to the target feature combination based on the standard negative record possibility and the push record possibility.

[0115] The standard negative record probability is the product of the first negative record probability and the second negative record probability. The content features in the content push record corresponding to the target feature combination are consistent with the content features in the target feature combination, and the object features of the push object in the content push record corresponding to the target feature combination are consistent with the object features in the target feature combination. The number of combination records refers to the number of content push records corresponding to the target feature combination included in the content push record set.

[0116] The push record possibility is calculated based on the ratio of the number of combination records to the number of content push records. For example, the ratio of the number of combination records to the number of content push records is positively correlated with the push record possibility.

[0117] Specifically, the server may compare the object characteristics of the push object in the content push record in the content push record set with the object characteristics in the target feature combination to obtain an object characteristic comparison result, and compare the content characteristics of the push content in the content push record with the content characteristics in the target feature combination to obtain a content characteristic comparison result, and use the content push record whose object characteristic comparison result is consistent and whose content characteristic comparison result is consistent as the content push record corresponding to the target feature combination.

[0118] In some embodiments, the server may calculate the ratio of the number of combined records to the number of content push records, and use the calculated ratio as the push record possibility corresponding to the target feature combination, calculate the ratio of the standard negative record possibility to the push record possibility, and use the calculated ratio as the target feature exclusion degree corresponding to the target feature combination.

[0119] In this embodiment, the ratio of the number of combined records to the number of content push records is calculated, and the push record possibility corresponding to the target feature combination is obtained based on the calculated ratio. The target feature exclusion degree corresponding to the target feature combination is obtained based on the standard negative record possibility and the push record possibility, thereby improving the accuracy of the target feature exclusion degree.

[0120] In some embodiments, obtaining the target feature combination includes: sending a feature combination configuration interface to a configuration terminal so that the configuration terminal displays the feature combination configuration interface, and obtaining the configured target feature combination in response to a configuration operation on the feature combination configuration interface; and receiving the target feature combination sent by the configuration terminal.

[0121] The configuration terminal refers to an interface for configuring the target feature combination. The configuration terminal can be, for example, Figure 1The terminal 102 in the feature combination configuration interface. The configuration operation refers to the operation of configuring the target feature combination in the feature combination configuration interface. The feature combination configuration interface can display a feature combination editing area, and the feature combination editing area can receive the feature combination input or selected by the user. The feature combination configuration interface can also display a configuration confirmation control, and the configuration confirmation control is used to indicate the configuration of the feature combination. The configuration operation can be, for example, a trigger operation on the configuration confirmation control.

[0122] Specifically, the configuration terminal can display the entrance of the feature combination configuration interface. When the configuration terminal receives a trigger operation on the entrance of the feature combination configuration interface, it can send a feature combination configuration request to the server. The server can return the feature combination configuration interface to the configuration terminal in response to the feature combination configuration request. The configuration terminal can display the feature combination configuration interface. When the configuration terminal receives a trigger operation on the configuration confirmation control, it can obtain the feature combination entered or selected in the feature combination editing area, use the obtained feature combination as the target feature combination, and send the target feature combination to the server.

[0123] In this embodiment, a feature combination configuration interface is sent to the configuration terminal so that the configuration terminal displays the feature combination configuration interface, and in response to the configuration operation on the feature combination configuration interface, the configured target feature combination is obtained, and the target feature combination sent by the configuration terminal is received, so that feature combinations that may have a large feature exclusion degree or a small content push value can be configured as target feature combinations through the feature combination configuration interface, thereby reducing the number of target feature combinations, reducing calculation complexity, and improving calculation efficiency.

[0124] In some embodiments, Figure 5 As shown, a content push method is provided, which can be executed by a terminal or a server, or by both the terminal and the server. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, and the steps include: S502, receiving a content push request for a target push object; S504, in response to the content push request, obtaining a candidate push content set corresponding to the target push object; S506, determining a target exclusion feature combination based on target object features corresponding to the target push object; the target exclusion feature combination is determined based on a target feature exclusion degree corresponding to the target feature combination; S508, filtering candidate push content corresponding to the target exclusion feature combination from the candidate push content set to obtain a target push content corresponding to the target push object.

[0125] Among them, the target push object can be any push object. When the terminal accesses the interface corresponding to the client corresponding to the server, when the interface includes a push content display area, the server can use the user corresponding to the terminal as the target push object. The candidate push content set can be pre-set or set as needed. For example, popular content can be used as candidate push content, or candidate push content can be determined according to the needs of advertisers. The candidate push content sets corresponding to different target push objects can be the same or different.

[0126] Specifically, a plurality of preset push contents may be stored in the server, and the server may obtain push content scores corresponding to each preset push content, select candidate push contents from the plurality of preset push contents based on the push content scores, and form a candidate push content set with each candidate push content. For example, the server may obtain preset push contents with push content scores greater than a content score threshold from the plurality of preset push contents as candidate push contents, and form a candidate push content set. The content score threshold may be set as needed or may be pre-set.

[0127] In some embodiments, the push content score includes a first score, and the step of obtaining the first score may include: obtaining an object portrait corresponding to the target push object, determining the object type corresponding to the target push object based on the object portrait, as the target object type, different object types correspond to different object portraits, obtaining the predicted click probability and predicted conversion probability of the push object of the target object type for the preset push content, multiplying the predicted click probability and the predicted conversion probability, using the result of the product operation as the target conversion probability, obtaining unit conversion revenue, multiplying the target conversion probability and the unit conversion revenue, and obtaining the first score. The product of the target conversion probability and the unit conversion revenue is positively correlated with the first score, for example, the product of the target conversion probability and the unit conversion revenue can be used as the first score.

[0128] Among them, click probability refers to the probability of content being clicked. Click probability can also be called click-through rate (CTR). The ratio of the actual number of clicks on the content to the total number of exposures can be used as the click probability. Predicted click probability refers to the predicted click probability. Predicted click probability can also be called predicted click-through rate (pCTR). Conversion rate (CVR) refers to the conversion ratio of users clicking on ads to becoming an effectively activated, registered or paying user, that is, the actual number of conversions of the ad divided by the number of clicks on the ad. Predicted conversion rate (pCVR) is the estimated conversion rate. Unit conversion revenue refers to the fee that the content owner needs to pay for a successful conversion. For example, it can be a bid. The bid refers to the price offered by the advertiser for a conversion, that is, the fee paid by the advertiser to the traffic owner after a successful conversion.

[0129] In some embodiments, the server may calculate the first score using formula (2), wherein eCPM (effective cost per mille, the revenue that can be obtained for every thousand impressions) is the first score.

[0130] eCPM=bid*pCTR*pCVR*1000 (2)

[0131] In some embodiments, the push content score also includes a second score. The second score may be, for example, a quality score. The quality score in social advertising may be affected by negative feedback from users and the social activity of users. The server may calculate the sum of the first score and the second score, and use the sum as the push content score. For example, the server may calculate the push content score using formula (3). Wherein, Ranking_score represents the push content score, and quality represents the quality score, i.e., the second score.

[0132] Ranking_score=eCPM+ quality (3)

[0133] In the above-mentioned content push method, a content push request for a target push object is received, and in response to the content push request, a candidate push content set corresponding to the target push object is obtained, and a target exclusion feature combination is determined based on the target object features corresponding to the target push object. Since the target exclusion feature combination is determined based on the target feature exclusion degree corresponding to the target feature combination, and the target feature exclusion degree reflects the degree of rejection of the object having the object features in the target feature combination to the content having the content features in the target feature combination, the larger the target feature exclusion degree, the more the object rejects the content, and thus the target feature combination with a larger target feature exclusion degree can be used as the target exclusion feature combination, thereby filtering the candidate push content corresponding to the target exclusion feature combination from the candidate push content set, and filtering out the candidate push content rejected by the target push object, thereby reducing the possibility of pushing the rejected content to the target push object and improving the accuracy of content push.

[0134] In some embodiments, determining a target exclusion feature combination based on target object features corresponding to a target push object includes: obtaining a plurality of candidate exclusion feature combinations; the candidate exclusion feature combination is screened from each target feature combination based on the target feature exclusion degree corresponding to the target feature combination; comparing the target object features corresponding to the target push object with the object features corresponding to the push object in the candidate exclusion feature combination, and taking the exclusion feature combination with consistent comparison as the target exclusion feature combination.

[0135] Specifically, the server may obtain multiple target feature combinations, determine the target feature exclusion degrees corresponding to each target feature combination, and obtain target feature combinations that meet the exclusion degree screening conditions from each target feature combination based on the target feature exclusion degrees as candidate exclusion feature combinations.

[0136] In some embodiments, the server can determine the target push value corresponding to each target feature combination, and based on the target feature exclusion degree and the target push value, obtain the target feature combination that meets the exclusion degree filtering condition and the push value filtering condition from each target feature combination as a candidate exclusion feature combination.

[0137] In this embodiment, the target object features corresponding to the target pushed object are compared with the object features corresponding to the pushed objects in the candidate exclusion feature combination, and the exclusion feature combination with consistent comparison is used as the target exclusion feature combination, so as to quickly determine the target exclusion feature combination.

[0138] In some embodiments, the step of obtaining a candidate exclusion feature combination includes: obtaining multiple target feature combinations; the target feature combination is obtained by combining an object feature and a content feature; obtaining a content push record set, the content push record set including multiple content push records, the content push record including push content and a push object corresponding to the push content; determining a first negative content push record in the content push record set in which a push object performs a negative operation on the push content, and obtaining the number of first negative records corresponding to the first negative content push record; determining a second negative content push record corresponding to the target feature combination, and obtaining the number of second negative records corresponding to the second negative content push record; wherein the content features of the push content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the push object in the second negative content push record are consistent with the object features in the target feature combination; determining a target feature exclusion degree corresponding to the target feature combination according to the number of the first negative records and the number of the second negative records, the target feature exclusion degree indicating the degree of exclusion between the content features and the object features in the target feature combination; based on the target feature exclusion degree corresponding to the target feature combination, screening out a candidate exclusion feature combination from each target feature combination.

[0139] Specifically, the server may screen out target feature combinations that meet the exclusion degree screening condition and the push value screening condition from various target feature combinations as candidate exclusion feature combinations.

[0140] In this embodiment, a first negative content push record in which a push object performs a negative operation on a push content in a content push record set is determined, and the first negative record quantity corresponding to the first negative content push record is obtained, so that the first negative record quantity reflects the number of content push records that perform negative operations on the push content, and a second negative content push record corresponding to a target feature combination is determined, and the second negative record quantity corresponding to the second negative content push record is obtained. Since the content feature of the push content in the second negative content push record is consistent with the content feature in the target feature combination, and the object feature of the push object in the second negative content push record is consistent with the object feature in the target feature combination, the second negative record quantity reflects the number of content push records corresponding to the target feature combination in each content push record that performs a negative operation, and the target feature combination corresponding to the target feature combination is determined based on the first negative record quantity and the second negative record quantity. The target feature exclusion degree indicates the degree of exclusion between the content features and the object features in the target feature combination, and thus the target feature exclusion degree reflects the degree of exclusion of the object having the object features in the target feature combination to the content having the content features in the target feature combination. The greater the target feature exclusion degree, the greater the degree of exclusion. Therefore, based on the target feature exclusion degree corresponding to the target feature combination, a candidate exclusion feature combination is screened from each target feature combination, and a target feature combination with a larger target feature exclusion degree can be used as a candidate exclusion feature combination. Therefore, when the candidate push content corresponding to the target exclusion feature combination is filtered from the candidate push content set to obtain the target push content corresponding to the target push object, the candidate push content with a larger degree of exclusion from the target push object can be filtered out from the candidate push content set, thereby reducing the possibility of pushing the target push object's excluded content to the target push object and improving the accuracy of content push.

[0141] In some embodiments, a content push method is provided, comprising the following steps:

[0142] 1. The server sends a feature combination configuration interface to the configuration terminal.

[0143] The configuration terminal may display a feature combination configuration interface, and the configuration terminal may respond to the configuration operation on the feature combination configuration interface, obtain the configured feature combination as the target feature combination, and send the target feature combination to the server. The feature combination is a combination of push user features and advertising features.

[0144] like Figure 6 As shown, features can be configured in the configuration terminal according to expert knowledge to obtain a target feature combination. Specifically, the configuration terminal can obtain the feature combination obtained by the configuration user based on expert knowledge, use the configured feature combination as the target feature combination, and send the target feature combination to the server.

[0145] 2. The server receives the target feature combination sent by the configuration terminal.

[0146] There may be multiple target feature combinations.

[0147] 3. Obtain an advertisement push record set; the advertisement push record set includes multiple advertisement push records, and the advertisement push record includes the pushed advertisement and the push user to whom the pushed advertisement is pushed.

[0148] Among them, the advertisement push record in the advertisement push record set may be a record generated by pushing advertisements in a historical time period. It may also be that the push advertisement set and the push user set are determined based on the target feature combination, and the push advertisements in the push advertisement set are pushed to the push users in the push user set. The push advertisement set includes multiple push advertisements, and the push user set includes multiple push users. The push advertisement set includes push advertisements with advertisement features in the target feature combination. The push user set includes push users with push user features in the target feature combination. The server may generate an advertisement push record for each push, for example, the server may record each push advertisement and the user to which it is pushed through a log, and the server may generate an advertisement push record based on the log. The advertisement push record may also include the user's operation type for the push advertisement, and the operation type may be any one of a positive feedback behavior and a negative feedback behavior. For example, C1 may be used to represent a negative feedback behavior, and C2 may be used to represent a positive feedback behavior.

[0149] In some embodiments, the number of push advertisements included in the push advertisement set may be the same as the number of push users in the push user set, and during push, each push user pushes one push advertisement. During push, a push advertisement having advertisement features in the target feature combination may be pushed to a push user having push user features in the target feature combination.

[0150] 4. Determine the first negative advertisement push record in the advertisement push record set in which the push user provides negative feedback on the pushed advertisement, and obtain the first negative record quantity corresponding to the first negative advertisement push record.

[0151] The server may take the advertisement push record whose operation type is negative feedback behavior as the first negative advertisement push record, count the number of each first negative advertisement push record, and take the counted number as the first negative record number.

[0152] 5. Determine the second negative advertising push record corresponding to the target feature combination, and obtain the number of second negative records corresponding to the second negative advertising push record.

[0153] Among them, the advertising features of the pushed advertisement in the first negative advertisement push record are the same as the advertising features in the target feature combination, and the user features of the pushed user in the second negative advertisement push record are the same as the user features in the target feature combination.

[0154] 6. Determine the number of advertisement push records corresponding to the advertisement push record set, calculate the ratio of the first number of negative records to the number of advertisement push records, and use the calculated ratio as the first negative record possibility, calculate the ratio of the second number of negative records to the first number of negative records, and use the calculated ratio as the second negative record possibility, calculate the product of the first negative record possibility and the second negative record possibility, and use the calculated result as the standard negative record possibility corresponding to the target feature combination, obtain the number of advertisement push records corresponding to the target feature combination in the advertisement push record set as the number of combined records, calculate the ratio of the number of combined records to the number of advertisement push records, and use the calculated ratio as the push record possibility corresponding to the target feature combination, calculate the ratio of the standard negative record possibility to the push record possibility, and use the calculated ratio as the target feature repulsion corresponding to the target feature combination.

[0155] 7. Obtain a set of reference advertising push records corresponding to the target feature combination, obtain the advertising push value corresponding to the reference advertising push record, sum the advertising push values ​​corresponding to each reference advertising push record in the reference advertising push record set, and use the result of the summation as the overall advertising push value corresponding to the target feature combination, obtain the number of reference advertising push records in the reference advertising push record as the number of reference advertising records, calculate the ratio of the overall advertising push value to the number of reference advertising records, use the calculated ratio as the unit advertising push value, obtain the preset number of advertising records, multiply the unit advertising push value by the preset number of advertising records, and use the result of the product operation as the target push value corresponding to the target feature combination.

[0156] The push user in the reference advertisement push record has the user features in the target feature combination, and the push advertisement in the reference advertisement push record has the advertisement features in the target feature combination.

[0157] 8. According to the target feature exclusion degree and the target push value corresponding to the target feature combination, the target feature combination that meets the exclusion degree screening condition and the push value screening condition is selected from each target feature combination as the exclusion feature combination, and each exclusion feature combination is stored.

[0158] Among them, the push value screening condition includes at least one of the target push value being greater than the push value threshold or the target push value ranking being before the value ranking threshold; the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree ranking being before the exclusion degree ranking threshold. For example, the exclusion degree screening condition corresponding to the target feature combination (ai, bj) can be expressed as p(c1 / (ai, bj))>δ, where δ represents the exclusion degree threshold. C1 indicates that the operation type is negative feedback behavior. The push value screening condition corresponding to the target feature combination (ai, bj) can be expressed as GPM(ai, bj)<ε, where ε represents the push value threshold. The target push value can also be called the distribution value.

[0159] In some embodiments, the server may format the exclusion feature combination to obtain a formatted exclusion feature combination, and store the formatted exclusion feature combination. Figure 6 As shown, the server can store the formatted exclusion feature combination in the common sense knowledge base. The formatted exclusion feature combination can be understood by the program, that is, the formatted exclusion feature combination has a format that the program can understand. Among them, the construction of the common sense knowledge base supports two major categories of dimensions, one is the advertising dimension, which mainly supports pluggable advertising range selection, including but not limited to at least one of the advertising industry, advertisers or advertising groups, and the other is the user dimension, including but not limited to at least one of the user's age, education or marital status. The common sense knowledge base can be located in the precise ranking engine. For example, the common sense knowledge base can be configured in a database that can be directly read by the precise ranking engine, and the database stores the data result types that can be directly read and understood by the precise ranking engine.

[0160] 9. Receive an advertisement push request for a target push user, and in response to the advertisement push request, obtain a candidate push advertisement set corresponding to the target push user.

[0161] 10. Obtain the user features corresponding to the target push user, compare the user features of the target push user with the user features in the exclusion feature combination, and use the exclusion feature combination that is consistent with the comparison as the target exclusion feature combination.

[0162] There may be multiple target rejection feature combinations.

[0163] 11. Filter out advertisements with advertisement features in the target exclusion feature combination from the candidate push advertisement set, and use the remaining candidate push advertisements in the candidate push advertisement set as target push advertisements corresponding to the target push user.

[0164] Among them, a refined sorting queue corresponding to the target push user can be generated in the refined sorting engine. The server can store each candidate push advertisement in the candidate push advertisement set into the refined sorting queue, and the server can filter out the candidate push advertisements with the advertisement features in the target exclusion feature combination from the refined sorting queue. For example, assuming that the target exclusion feature combination is (a1, b2), the advertisements with the advertisement feature b2 are removed from the refined sorting queue, and the remaining advertisements in the refined sorting queue are used as the target push advertisements corresponding to the target push user.

[0165] 12. Push the target push advertisement to the target push user.

[0166] The content push method provided in the present application can predict advertisements rejected by users by obtaining a target rejection feature combination, so that when pushing advertisements to users, the advertisements rejected by users can be filtered out from the advertisements that need to be pushed, thereby reducing the situation where advertisements rejected by users are pushed to them and improving the accuracy of pushed advertisements.

[0167] It should be understood that although Figure 2-Figure 6 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2-Figure 6 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0168] In some embodiments, Figure 7 As shown, a content push device is provided, which can be a software module or a hardware module, or a combination of the two to form a part of a computer device. The device specifically includes: a target feature combination acquisition module 702, a content push record set acquisition module 704, a first negative record quantity acquisition module 706, a second negative record quantity acquisition module 708 and a target feature repulsion degree determination module 710, wherein:

[0169] A target feature combination acquisition module 702 is used to acquire a target feature combination, where the target feature combination is obtained by combining an object feature and a content feature;

[0170] The content push record set acquisition module 704 is used to acquire a content push record set, where the content push record set includes multiple content push records, and the content push record includes push content and a push object corresponding to the push content;

[0171] A first negative record quantity acquisition module 706 is used to determine a first negative content push record in which a push object performs a negative operation on a pushed content in a content push record set, and to acquire a first negative record quantity corresponding to the first negative content push record;

[0172] The second negative record quantity acquisition module 708 is used to determine the second negative content push record corresponding to the target feature combination, and obtain the second negative record quantity corresponding to the second negative content push record; wherein the content feature of the push content in the second negative content push record is consistent with the content feature in the target feature combination, and the object feature of the push object in the second negative content push record is consistent with the object feature in the target feature combination;

[0173] The target feature exclusion degree determination module 710 is used to determine the target feature exclusion degree corresponding to the target feature combination according to the first negative record number and the second negative record number. The target feature exclusion degree represents the degree of exclusion between the content features and the object features in the target feature combination, so as to screen the candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination.

[0174] In some embodiments, there are multiple target feature combinations, and the device is also used to select a target feature combination that meets the exclusion degree screening condition according to the target feature exclusion degree corresponding to the target feature combination, as an exclusion feature combination, and the exclusion feature combination is used to filter the push content having the content features in the exclusion feature combination when pushing content to the push object having the object features in the exclusion feature combination; the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree ranking being before the exclusion degree ranking threshold.

[0175] In some embodiments, the device also includes: a reference push record set acquisition module, used to acquire a reference push record set, the reference push record set includes multiple reference push records, and the reference push record is a content push record corresponding to the target feature combination; a content push value acquisition module, used to acquire the content push value corresponding to the reference push record; a target push value acquisition module, used to count the content push values ​​corresponding to the reference push records in the reference push record set, and obtain the target push value corresponding to the target feature combination; and screen the candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination and the target push value.

[0176] In some embodiments, there are multiple target feature combinations, and the device is also used to select a target feature combination that meets the exclusion degree screening condition and the push value screening condition as the exclusion feature combination according to the target feature exclusion degree and the target push value corresponding to the target feature combination; the exclusion feature combination is used to filter the push content having the content features in the exclusion feature combination when pushing content to the push object having the object features in the exclusion feature combination; the push value screening condition includes at least one of the target push value being greater than the push value threshold or the target push value being ranked before the value ranking threshold; the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree being ranked before the exclusion degree ranking threshold.

[0177] In some embodiments, the target push value obtaining module includes: an overall content push value obtaining unit, which is used to sum and calculate the content push values ​​corresponding to the reference push records in the reference push record set to obtain the overall content push value; a reference content record quantity obtaining unit, which is used to obtain the number of reference push records in the reference push record set as the reference content record number; a target push value obtaining unit, which is used to calculate the ratio of the overall content push value to the number of reference content records, and obtain the target push value corresponding to the target feature combination based on the calculated ratio.

[0178] In some embodiments, the target feature exclusion degree determination module 710 includes: a content push record quantity determination unit, which is used to determine the number of content push records corresponding to the content push record set; a first negative record possibility obtaining unit, which is used to calculate the ratio of the first negative record quantity to the content push record quantity, and obtain the first negative record possibility according to the calculated ratio; a second negative record possibility obtaining unit, which is used to calculate the ratio of the second negative record quantity to the first negative record quantity, and obtain the second negative record possibility according to the calculated ratio; a target feature exclusion degree obtaining unit, which is used to calculate the target feature exclusion degree corresponding to the target feature combination based on the first negative record possibility and the second negative record possibility.

[0179] In some embodiments, the target feature exclusion degree obtaining unit is also used to calculate the product of the first negative record possibility and the second negative record possibility to obtain the standard negative record possibility corresponding to the target feature combination; obtain the number of content push records corresponding to the target feature combination in the content push record set as the number of combined records; calculate the ratio of the number of combined records to the number of content push records, and obtain the push record possibility corresponding to the target feature combination according to the calculated ratio; obtain the target feature exclusion degree corresponding to the target feature combination based on the standard negative record possibility and the push record possibility.

[0180] In some embodiments, the target feature combination acquisition module 702 includes: a feature combination configuration interface sending unit, used to send a feature combination configuration interface to the configuration terminal, so that the configuration terminal displays the feature combination configuration interface, and obtains the configured target feature combination in response to the configuration operation on the feature combination configuration interface; a target feature combination receiving unit, used to receive the target feature combination sent by the configuration terminal.

[0181] In some embodiments, Figure 8 As shown, a content push device is provided, which can be a part of a computer device using a software module or a hardware module, or a combination of the two. The device specifically includes: a content push request receiving module 802, a candidate push content set acquisition module 804, a target exclusion feature combination determination module 806 and a target push content acquisition module 808, wherein:

[0182] The content push request receiving module 802 is used to receive a content push request for a target push object;

[0183] The candidate push content set acquisition module 804 is used to obtain the candidate push content set corresponding to the target push object in response to the content push request;

[0184] A target exclusion feature combination determination module 806 is used to determine a target exclusion feature combination based on target object features corresponding to the target push object, wherein the target exclusion feature combination is determined based on a target feature exclusion degree corresponding to the target feature combination;

[0185] The target push content obtaining module 808 is used to filter the candidate push content corresponding to the target exclusion feature combination from the candidate push content set to obtain the target push content corresponding to the target push object.

[0186] In some embodiments, the target rejection feature combination determination module 806 includes:

[0187] The first candidate exclusion feature combination obtaining unit is used to obtain multiple candidate exclusion feature combinations; the candidate exclusion feature combinations are obtained by screening from various target feature combinations based on the target feature exclusion degree corresponding to the target feature combination;

[0188] The target exclusion feature combination obtaining unit is used to compare the target object feature corresponding to the target push object with the object feature corresponding to the push object in the candidate exclusion feature combination, and use the exclusion feature combination with the consistent comparison as the target exclusion feature combination. In some embodiments, the device also includes a candidate exclusion feature combination obtaining module, and the candidate exclusion feature combination obtaining module includes:

[0189] The target feature combination obtaining unit is used to obtain multiple target feature combinations; the target feature combination is obtained by combining the object feature and the content feature;

[0190] A content push record set acquisition unit, used to acquire a content push record set, the content push record set includes multiple content push records, and the content push record includes push content and a push object corresponding to the push content;

[0191] A first negative record quantity acquisition unit, configured to determine a first negative content push record in which a push object performs a negative operation on a push content in a content push record set, and acquire a first negative record quantity corresponding to the first negative content push record;

[0192] A second negative record quantity acquisition unit is used to determine the second negative content push record corresponding to the target feature combination, and acquire the second negative record quantity corresponding to the second negative content push record; wherein the content feature of the push content in the second negative content push record is consistent with the content feature in the target feature combination, and the object feature of the push object in the second negative content push record is consistent with the object feature in the target feature combination;

[0193] A target feature repulsion degree determining unit, configured to determine a target feature repulsion degree corresponding to a target feature combination according to the first number of negative records and the second number of negative records, wherein the target feature repulsion degree indicates a degree of repulsion between content features and object features in the target feature combination;

[0194] The second candidate exclusion feature combination obtaining unit is used to screen and obtain a candidate exclusion feature combination from each target feature combination based on the target feature exclusion degree corresponding to the target feature combination.

[0195] For the specific definition of the content push device, please refer to the definition of the content push method above, which will not be repeated here. Each module in the above content push device can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0196] In some embodiments, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig. 9As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a content push method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0197] In some embodiments, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Fig.10 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data involved in the content push method. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a content push method is implemented.

[0198] Those skilled in the art will understand that Fig. 9 and Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0199] In some embodiments, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0200] In some embodiments, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0201] In some embodiments, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a 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 steps in the above-mentioned method embodiments.

[0202] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0203] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0204] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A content push method, characterized in that: The method comprises: Acquire a target feature combination, where the target feature combination is obtained by combining an object feature and a content feature; Acquire a content push record set, the content push record set including multiple content push records, the content push record including push content and a push object corresponding to the push content; Determine a first negative content push record in which a push object performs a negative operation on a pushed content in the content push record set, and obtain a first negative record quantity corresponding to the first negative content push record; Determine a second negative content push record corresponding to the target feature combination, and obtain the number of second negative records corresponding to the second negative content push record; wherein the content features of the pushed content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the pushed object in the second negative content push record are consistent with the object features in the target feature combination; The target feature exclusion degree corresponding to the target feature combination is determined according to the first number of negative records and the second number of negative records, and the target feature exclusion degree represents the degree of exclusion between content features and object features in the target feature combination, so as to screen candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination.

2. The method according to claim 1, characterized in that The target feature combination is multiple, and the method further includes: According to the target feature exclusion degree corresponding to the target feature combination, a target feature combination that meets the exclusion degree screening condition is selected as an exclusion feature combination, and the exclusion feature combination is used to filter the pushed content having the content features in the exclusion feature combination when pushing content to the push object having the object features in the exclusion feature combination; The exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than an exclusion degree threshold value or the target exclusion degree ranking being before the exclusion degree ranking threshold value.

3. The method according to claim 1, characterized in that: The method further comprises: Acquire a reference push record set, the reference push record set including a plurality of reference push records, the reference push record being a content push record corresponding to the target feature combination; Obtaining the content push value corresponding to the reference push record; The content push values ​​corresponding to the reference push records in the reference push record set are counted to obtain the target push value corresponding to the target feature combination; and the candidate content to be pushed is screened based on the target feature exclusion degree corresponding to the target feature combination and the target push value.

4. The method according to claim 3, characterized in that The target feature combination is multiple, and the method further includes: According to the target feature exclusion degree and the target push value corresponding to the target feature combination, a target feature combination that meets the exclusion degree screening condition and the push value screening condition is selected as the exclusion feature combination; the exclusion feature combination is used to filter the push content having the content features in the exclusion feature combination when pushing content to the push object having the object features in the exclusion feature combination; The push value screening condition includes at least one of the target push value being greater than the push value threshold or the target push value ranking being before the value ranking threshold; the exclusion degree screening condition includes at least one of the target feature exclusion degree being greater than the exclusion degree threshold or the target exclusion degree ranking being before the exclusion degree ranking threshold.

5. The method according to claim 3, characterized in that: The step of performing statistics on the content push values ​​corresponding to the reference push records in the reference push record set to obtain the target push value corresponding to the target feature combination includes: Summing up the content push values ​​corresponding to the reference push records in the reference push record set to obtain an overall content push value; Obtain the number of reference push records in the reference push record set as the number of reference content records; The ratio of the overall content push value to the number of reference content records is calculated, and the target push value corresponding to the target feature combination is obtained based on the calculated ratio.

6. The method according to claim 1, characterized in that The step of determining the target feature exclusion degree corresponding to the target feature combination according to the first number of negative records and the second number of negative records includes: Determining the number of content push records corresponding to the content push record set; Calculating a ratio of the first number of negative records to the number of content push records, and obtaining a first negative record possibility according to the calculated ratio; Calculating a ratio of the second number of negative records to the first number of negative records, and obtaining a second negative record possibility according to the calculated ratio; The target feature exclusion degree corresponding to the target feature combination is calculated based on the first negative record possibility and the second negative record possibility.

7. The method according to claim 6, characterized in that The target feature exclusion degree corresponding to the target feature combination calculated based on the first negative record possibility and the second negative record possibility includes: Calculating the product of the first negative record possibility and the second negative record possibility to obtain a standard negative record possibility corresponding to the target feature combination; Obtaining the number of content push records corresponding to the target feature combination in the content push record set as the number of combination records; Calculating the ratio of the number of combination records to the number of content push records, and obtaining the push record possibility corresponding to the target feature combination according to the calculated ratio; The target feature exclusion degree corresponding to the target feature combination is obtained based on the standard negative record possibility and the push record possibility.

8. The method according to claim 1, characterized in that: The acquisition target feature combination comprises: Sending a feature combination configuration interface to a configuration terminal so that the configuration terminal displays the feature combination configuration interface, and in response to a configuration operation on the feature combination configuration interface, acquiring the configured target feature combination; Receive the target feature combination sent by the configuration terminal.

9. A content push method, characterized in that: The method comprises: Receive a content push request for a target push object; In response to the content push request, obtaining a candidate push content set corresponding to the target push object; Determine a target exclusion feature combination based on the target object feature corresponding to the target push object, wherein the target exclusion feature combination is determined based on the target feature exclusion degree corresponding to the target feature combination; The candidate push content corresponding to the target exclusion feature combination is filtered from the candidate push content set to obtain the target push content corresponding to the target push object.

10. The method according to claim 9, characterized in that The determining of the target exclusion feature combination based on the target object feature corresponding to the target push object includes: Acquire multiple candidate exclusion feature combinations; the candidate exclusion feature combinations are screened from each target feature combination based on the target feature exclusion degree corresponding to the target feature combination; The target object features corresponding to the target pushed object are compared with the object features corresponding to the pushed objects in the candidate exclusion feature combination, and the exclusion feature combination with consistent comparison is used as the target exclusion feature combination.

11. The method according to claim 10, characterized in that The step of obtaining the candidate exclusion feature combination comprises: Acquire multiple target feature combinations; the target feature combinations are obtained by combining object features and content features; Acquire a content push record set, the content push record set including multiple content push records, the content push record including push content and a push object corresponding to the push content; Determine a first negative content push record in which a push object performs a negative operation on a pushed content in the content push record set, and obtain a first negative record quantity corresponding to the first negative content push record; Determine a second negative content push record corresponding to the target feature combination, and obtain the number of second negative records corresponding to the second negative content push record; wherein the content features of the pushed content in the second negative content push record are consistent with the content features in the target feature combination, and the object features of the pushed object in the second negative content push record are consistent with the object features in the target feature combination; Determine a target feature exclusion degree corresponding to the target feature combination according to the first number of negative records and the second number of negative records, wherein the target feature exclusion degree indicates a degree of exclusion between content features and object features in the target feature combination; Based on the target feature exclusion degree corresponding to the target feature combination, a candidate exclusion feature combination is screened from each of the target feature combinations.

12. A content push device, characterized in that: The device comprises: A target feature combination acquisition module is used to acquire a target feature combination, where the target feature combination is obtained by combining an object feature and a content feature; A content push record set acquisition module, used to acquire a content push record set, wherein the content push record set includes a plurality of content push records, and the content push record includes push content and a push object corresponding to the push content; A first negative record quantity acquisition module, configured to determine a first negative content push record in which a push object performs a negative operation on a pushed content in the content push record set, and acquire a first negative record quantity corresponding to the first negative content push record; A second negative record quantity acquisition module is used to determine the second negative content push record corresponding to the target feature combination, and obtain the second negative record quantity corresponding to the second negative content push record; wherein the content feature of the push content in the second negative content push record is consistent with the content feature in the target feature combination, and the object feature of the push object in the second negative content push record is consistent with the object feature in the target feature combination; A target feature exclusion degree determination module is used to determine the target feature exclusion degree corresponding to the target feature combination based on the first number of negative records and the second number of negative records, wherein the target feature exclusion degree represents the degree of exclusion between the content features and the object features in the target feature combination, so as to screen the candidate content to be pushed based on the target feature exclusion degree corresponding to the target feature combination.

13. A content push device, characterized in that: The device comprises: A content push request receiving module, used to receive a content push request for a target push object; A candidate push content set acquisition module, configured to acquire a candidate push content set corresponding to the target push object in response to the content push request; A target exclusion feature combination determination module, used to determine a target exclusion feature combination based on the target object features corresponding to the target push object, wherein the target exclusion feature combination is determined based on the target feature exclusion degree corresponding to the target feature combination; The target push content obtaining module is used to filter the candidate push content corresponding to the target exclusion feature combination from the candidate push content set to obtain the target push content corresponding to the target push object.

14. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 11 are implemented.

15. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images