Method for determining target object of promotion information and related device

Abstract

The present disclosure provides a method for determining a target object of promotion information, comprising: obtaining features of each target user respectively; determining a feature vector of each target user according to the features of each target user respectively; establishing a first graph according to a first interaction operation between each target user and a promotion information provider; determining a feature vector of the promotion information provider based on the first graph and the feature vector of each target user; establishing a second graph according to a second interaction operation between each target user and each candidate object; determining a feature vector of each candidate object based on the second graph and the feature vector of each target user respectively; and selecting a target object of promotion information from each candidate object according to the feature vector of the promotion information provider and the feature vector of each candidate object. Based on the above method, the present disclosure further provides a device for determining a target object of promotion information, an electronic device, a storage medium and a program product.

Description

Technical Field

[0001] This disclosure relates to the field of Internet technology, and in particular to a method, device, electronic device, storage medium, and program product for determining the target audience of promotional information. Background Technology

[0002] With the continuous development of internet technology, various types of information are being generated and increasing rapidly. For example, promotional information such as advertisements and product information is experiencing explosive growth. In this information explosion, to better promote advertising and product information, providers typically need to find one or more suitable figures or images to act as promoters. However, determining the right promoters to achieve better promotional results remains a significant challenge in information promotion. Summary of the Invention

[0003] In view of this, embodiments of the present disclosure provide a method for determining the target object of promotional information, which can quickly determine a suitable target object for promotional information from a large number of candidate objects.

[0004] The method for determining the target object of promotional information according to the embodiments of this disclosure may include: acquiring the characteristics of each target user; determining the feature vector of each target user based on the characteristics of each target user; establishing a first graph based on the first interaction operation between each target user and the promotional information published by the promotional information provider; determining the feature vector of the promotional information provider based on the first graph and the feature vector of each target user; establishing a second graph based on the second interaction operation between each target user and the content published by each candidate object; determining the feature vector of each candidate object based on the second graph and the feature vector of each target user; and selecting the target object of promotional information from each candidate object based on the feature vector of the promotional information provider and the feature vector of each candidate object.

[0005] Based on the above-described method for determining the target audience of promotional information, embodiments of this disclosure provide an apparatus for determining the target audience of promotional information, comprising:

[0006] The user feature acquisition module is used to acquire the features of each target user.

[0007] The user feature vector determination module is used to determine the feature vector of each target user based on the characteristics of each target user.

[0008] The first graph building module is used to build the first graph based on the first interaction between each target user and the promotional information released by the promotional information provider.

[0009] The first feature vector determination module is used to determine the feature vector of the promotion information provider based on the first graph and the feature vectors of each target user.

[0010] The second graph establishes a module for building a second graph based on the second interaction operations between the content published by each target user and each candidate object;

[0011] The second feature vector determination module is used to determine the feature vector of each candidate object based on the second graph and the feature vectors of each target user; and

[0012] The target object determination module is used to select the target object of the promotion information from the candidate objects based on the feature vector of the promotion information provider and the feature vector of each candidate object.

[0013] Furthermore, embodiments of this disclosure also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the aforementioned method for determining the target object of promotional information.

[0014] Embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the aforementioned method for determining the target object of promotional information.

[0015] Embodiments of this disclosure also provide a computer program product, including computer program instructions that, when executed on a computer, cause the computer to perform the aforementioned method for determining the target object of promotional information.

[0016] As can be seen from the above method, the method for determining the target audience of promotional information uses the target user as the intermediary between the promotional information provider and the candidate object, realizing cross-domain matching between the promotional information and the candidate object. Specifically, in the above method, on the one hand, the feature vector of the promotional information provider is established through the interaction between the target user and the promotional information provider; on the other hand, the feature vector of the candidate object is established through the interaction between the target user and the candidate object; finally, a suitable target object is selected for the promotional information based on the distance between the feature vector of the promotional information provider and the feature vector of the candidate object. It is evident that the above process can obtain accurate matching results between the promotional information provider and the content producer on the content delivery platform, achieving better information promotion effects. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this disclosure or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram illustrating the application scenario of the promotional information delivery scheme described in this embodiment is provided;

[0019] Figure 2 The implementation flow of the promotion information target object determination method described in some embodiments of this disclosure is shown;

[0020] Figure 3 The structure of the three-tower model described in some embodiments of this disclosure is shown;

[0021] Figure 4 The training process of the three-tower model described in some embodiments of this disclosure is shown;

[0022] Figure 5 Examples of the first spectra described in some embodiments of this disclosure are shown;

[0023] Figure 6 Examples of the second spectra described in some embodiments of this disclosure are shown;

[0024] Figure 7 This disclosure illustrates the specific implementation process of selecting target objects for promotional information as described in some embodiments;

[0025] Figure 8 The internal structure of the promotional information target object determination device described in some embodiments of this disclosure is shown; and

[0026] Figure 9 A schematic diagram of a more specific electronic device hardware structure described in some embodiments of this disclosure is shown. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0028] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0029] As mentioned earlier, in the current information explosion phase, in order to better deliver advertising and product information, the providers of promotional information usually need to find one or more suitable target audiences as promoters for the promotional information. However, how to determine the right promoters for the promotional information to achieve better promotional results is also one of the problems that information promotion needs to solve.

[0030] Currently, with the development of the internet and mobile internet, many content delivery platforms have emerged, such as news information platforms, short video application platforms, and audio / video application platforms. These platforms typically have a large number of active users. Some users act as content producers, publishing their own content on these platforms, such as text, images, audio, and / or video. Other users act as content consumers, browsing various information or multimedia files to obtain information or for entertainment. Furthermore, content consumers can interact with content producers through these platforms. It should be noted that on these platforms, content producers and consumers are distinguished by user roles. In practice, a user can be both a content producer and a content consumer.

[0031] Because such content delivery platforms typically have a large number of active users, especially a large number of active content consumers, many providers of promotional information currently choose to place their promotional information on one or a few such platforms. Furthermore, many providers of promotional information also hope to select one or more popular content creators on these platforms as promoters of their promotional information, thereby achieving better promotional results. For ease of description, in this application, the promoters of the aforementioned promotional information will be referred to as the target audience of the promotional information.

[0032] It is understood that before using the technical solutions of the various embodiments in this disclosure, users will be informed of the type, scope of use, and usage scenarios of the personal information involved in an appropriate manner, and user authorization will be obtained.

[0033] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose, based on the prompt message, whether to provide personal information to the software or hardware such as electronic devices, applications, servers, or storage media performing the operations of this disclosed technical solution.

[0034] As an optional but not limited implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0035] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0036] Figure 1 This illustration shows an application scenario diagram of promotional information delivery on a content provisioning platform, as described in an embodiment of this disclosure. For example... Figure 1As shown, in the above application scenario, the provider of promotional information 110 needs to distribute the promotional information on the content delivery platform 100. The content delivery platform 100 has a large number of active content producers 120 and content consumers 130. The content producers 120 serve as candidates for the promotional information provider 110; the content consumers 130 serve as the target users of the promotional information provider 110. Furthermore, to complete the promotion of the information, the content delivery platform 100 will set up a promotional information management server 102 to manage the promotional information provider 110 and the promotional information, select suitable users from the content consumers 130 as the audience for the promotional information, and select suitable target audiences from the content producers 120, etc.

[0037] The following will describe in detail, with reference to the accompanying drawings and specific examples, the method for determining the target audience of promotional information according to the embodiments of this disclosure. This method can be... Figure 1 The promotion information management server 102 on the content providing platform 100 is implemented.

[0038] Figure 2 The implementation flow of the method for determining the target audience of promotional information according to an embodiment of this disclosure is shown. For example... Figure 2 As shown, the method may include the following steps.

[0039] In step 202, the characteristics of each target user are obtained.

[0040] In some embodiments of this disclosure, as previously described, the target users can refer to all or some of the content consumers 130 on the content providing platform 100. The promotional information management server 102 will select a suitable audience for the promotional information provider from among the target users.

[0041] Furthermore, in the embodiments of this disclosure, the characteristics of the target user may include: user attribute information and historical behavior information of the target user, etc. The aforementioned user attribute information generally includes: interest and hobby information, such as the user's actively selected subscription topics / content. It should be noted that the methods for obtaining and using user information are described in the preceding section of the specific embodiments of this disclosure, and comply with relevant laws and regulations.

[0042] In step 204, the feature vector of each target user is determined according to the characteristics of each target user.

[0043] In the embodiments of this disclosure, various feature vector extraction methods can be used to determine the feature vectors of each target user. This disclosure does not limit the specific user feature vector extraction method.

[0044] In some embodiments of this disclosure, the feature vectors of the target users can be obtained using a three-tower model. Specifically, the features of each target user can be input into a pre-trained three-tower model to obtain the feature vectors of the target users.

[0045] In the embodiments of this disclosure, the structure of the above-mentioned three-tower model can be as follows: Figure 3 As shown, the system specifically includes: a user feature tower 302, a promotional information feature tower 304, a content feature tower 306, a first merging layer 308, a second merging layer 310, a first output layer 312, a second output layer 314, and a training module 316. Specifically, the user feature tower 302 can be used to map the features of the target user into a fixed-length feature vector of the target user. The promotional information feature tower 304 can be used to map the features of the promotional information into a fixed-length feature vector of the promotional information. The content feature tower 306 can be used to map the features of the content into a fixed-length feature vector of the content. Specifically, the user feature tower 302, the promotional information feature tower 304, and the content feature tower 306 can be implemented by feature extraction networks such as neural networks or deep neural networks; the specific structure is not limited in this application.

[0046] In addition, the first merging layer 308 is used to merge the feature vector of the target user and the feature vector of the promotion information to obtain the merged first feature vector.

[0047] The first output layer 312 is used to output a prediction result that predicts the probability of a first interactive operation between the target user and the promotional information based on the first feature vector. The first interactive operation may include: the target user clicking on the promotional information or the target user converting their view on the promotional information, etc. For example, the first output layer 312 may be used to output a prediction result that predicts the probability of the target user clicking and / or converting their view on the promotional information based on the first feature vector.

[0048] The second merging layer 310 is used to merge the feature vector of the target user and the feature vector of the content to obtain the merged second feature vector.

[0049] The second output layer 314 is used to output a prediction result based on the second feature vector, forecasting the probability of a second interactive operation between the target user and the content. This second interactive operation may include: the target user clicking the content, the target user liking the content, the target user completing the playback of the content, and / or the target user commenting on the content. Furthermore, the second operation may also include: the target user following the content creator corresponding to the content. For example, the first output layer 312 may be used to output a prediction result based on the first feature vector, forecasting the probability of the content being clicked, liked, completed, and / or commented on by the target user; or, the first output layer 312 may be used to output the probability based on the first feature vector that the content creator of the content is followed by the target user.

[0050] The training module 316 is used to calculate the loss of the current three-tower model based on the prediction results of the first output layer 312 and the second output layer 314, and to adjust the parameters of the three-tower model according to the loss, thereby completing the training of the three-tower model.

[0051] Specifically, the training process of the above three-tower model can be as follows: Figure 4 As shown, the main steps include the following:

[0052] In step 402, sample data is obtained.

[0053] In embodiments of this disclosure, the sample data may include: characteristics of multiple target users, characteristics of multiple promotional information, characteristics of multiple content, and the first interactive operation between each target user and each promotional information, and the second interactive operation between each target user and each content.

[0054] The features of the aforementioned promotional information may include information about the promotional information itself (e.g., the name, category, conversion goal, and click-through rate of the promotional information) and information about the provider of the promotional information (e.g., the name or logo of the provider of the promotional information and its industry and other attribute information); the features of the aforementioned content may include information about the content itself (e.g., the name, category or theme, keywords, and duration of the content) and information about the content producer (e.g., the name or logo of the content producer and its attribute information).

[0055] In step 404, the characteristics of the target user, the characteristics of the promotional information, and the characteristics of the content in the sample data are input into the user feature tower 302, the promotional information feature tower 304, and the content feature tower 306 of the three-tower model, respectively.

[0056] In step 406, the probability prediction results of the first interactive operation between the target user and the promotional information and the probability prediction results of the second interactive operation between the target user and the content are obtained from the first output layer 312 and the second output layer 314 of the above three-tower model, respectively.

[0057] In step 408, the loss of the three-tower model is determined based on the first interaction between each target user and each promotional information and the second interaction between each target user and each content in the sample data. Then, the parameters in the three-tower model are adjusted by gradient descent based on the determined loss. Then, the process returns to step 404 until the model converges or the number of training iterations reaches a preset training iteration threshold.

[0058] It should be noted that the loss of the aforementioned three-tower model may include a first loss (e.g., cross-entropy) determined based on the probability prediction of a first interaction between each target user and each promotional message, and the circumstances of such first interaction between each target user and each promotional message; and a second loss (e.g., cross-entropy) determined based on the probability prediction of a second interaction between each target user and each content, and the circumstances of such second interaction between each target user and each content. Specifically, the loss of the aforementioned three-tower model may be a weighted average or a weighted sum of the aforementioned first and second losses, etc.

[0059] Thus, it can be seen that the feature vectors of each target user determined by the above method can fully consider the relationship between the target user and the promotional information and content published on the content providing platform. That is, when determining the user's feature vector, not only the user's features themselves are considered, but also the features of the promotional information that the user is interested in and the features of the content that the user is interested in. By associating the features of the promotional information with the features of the content, a more accurate user feature vector that simultaneously reflects the features of the promotional information and the features of the content can be obtained in the current information promotion scenario.

[0060] In step 206, a first graph is established based on the first interaction between each target user and the promotional information released by the promotional information provider.

[0061] In some embodiments of this disclosure, the first atlas described above can be established by the following method:

[0062] First, the information providers and target users are defined as points in the first graph;

[0063] Secondly, in response to determining that a first interaction exists between a target user and the promotional information published by the promotional information provider, an edge is established between the point corresponding to the target user and the point corresponding to the promotional information provider; and

[0064] Finally, the first weight value of the edge between the point corresponding to the target user and the point corresponding to the promotional information provider is determined based on the number of first interaction operations between the target user and the promotional information provider.

[0065] As mentioned above, in some embodiments of this disclosure, the first interactive operation may include: the operation of clicking on promotional information published by the promotional information provider (hereinafter referred to as the click operation), or the conversion operation through the promotional information (hereinafter referred to as the conversion operation).

[0066] In the above steps, if there is a first interaction between the target user and the promotional information published by the promotional information provider, such as the target user clicking on a promotional information published by the promotional information provider, or the target user achieving a conversion through the promotional information, such as the target user purchasing related products or completing the task specified in the promotional information, then an edge can be established between the point corresponding to the target user and the point corresponding to the promotional information provider.

[0067] Furthermore, as mentioned earlier, the first weight of this edge can also be determined based on the number of first interactive operations between the target user and the promotional information published by the promotional information provider. Specifically, the number of first interactive operations between the target user and the promotional information published by the promotional information provider can be directly counted and used as the first weight of this edge. Alternatively, different weights can be set for different types of first interactive operations, and the number of first interactive operations between the target user and the promotional information published by the promotional information provider can be weighted and summed to obtain the first weight of this edge. For example, assuming that the above first interactive operations only include click operations and conversion operations, the weight of click operations can be set to W1, and the weight of conversion operations can be set to W2. Assuming that the number of times the promotional information published by the promotional information provider is clicked by the target user is N1, and the number of times conversion operations are achieved is N2, the first weight value of this edge can be obtained as W1*N1+W2*N2. By setting different weights for different first interactive operations, the degree of preference of the target user for the promotional information represented by different interactive operations can be truly and effectively reflected, so that the calculated first weight value can more realistically and effectively reflect the degree of association between the target user and the promotional information provider.

[0068] Figure 5 An example of a first-level map is shown. Figure 5 In this example, triangles are used to represent promotional information providers, and circles are used to represent target users. For example... Figure 5As shown, the promotion information provider in the first graph is associated with the target users T1, T3, T4, T6 and T8, and the first weights of the edges established between the promotion information provider and the target users T1, T3, T4, T6 and T8 are w1, w3, w4, w6 and w8, respectively.

[0069] Therefore, it can be seen that the first graph determined by the above method not only represents the relationship between each target user and the promotional information provider, but also the degree of relationship between each target user and the promotional information provider, that is, it reflects to a certain extent the degree of preference of each target user for the promotional information provider. Moreover, in the above scheme, different weights are assigned to different interactive operations, which can more accurately reflect the degree of preference of each target user for the promotional information provider.

[0070] In step 208, the feature vector of the promotion information provider is determined based on the first graph and the feature vectors of each target user.

[0071] In some embodiments of this disclosure, the feature vector of the aforementioned promotional information provider can be determined by the following method:

[0072] First, based on the first graph, identify at least one target user corresponding to a point that has an edge with the point corresponding to the promotion information provider, and use this user as the associated user of the promotion information provider.

[0073] Then, the first weight values ​​of the edges between the points corresponding to the promotion information provider and the points corresponding to the associated users are obtained respectively; and

[0074] Finally, the user feature vectors of the associated users are weighted and summed according to the first weight value to obtain the feature vector of the promotion information provider.

[0075] Still with Figure 5 Taking the first spectrum shown as an example, from Figure 5 It can be seen that target users T1, T3, T4, T6, and T8 are associated users of the promotional information provider. Assuming the feature vectors of target users T1, T3, T4, T6, and T8 are U1, U3, U4, U6, and U8 respectively, the feature vector of the promotional information provider can be determined by the following expression: AD = w1*U1 + w3*U3 + w4*U4 + w6*U6 + w8*U8.

[0076] In step 210, a second graph is established based on the second interaction operations between each target user and the content published by each candidate object.

[0077] In some embodiments of this disclosure, the aforementioned candidate objects may specifically be all or a portion of the content producers on the content providing platform. In this case, the aforementioned second graph can be established using the following method:

[0078] First, each candidate object and each target user are treated as points in the second graph;

[0079] Secondly, in response to determining that a second interaction exists between the content posted by a target user and a candidate object, an edge is established between the point corresponding to the target user and the point corresponding to the candidate object; and

[0080] Finally, the second weight value of the edge between the point corresponding to the target user and the point corresponding to the candidate object is determined based on the number of second interaction operations between the content published by the target user and the candidate object.

[0081] In the embodiments of this disclosure, the aforementioned second interactive operation includes: the operation of the content published by the candidate being clicked by the target user (hereinafter referred to as the click operation), the operation of being liked (hereinafter referred to as the like operation), the operation of being broadcast in its entirety (hereinafter referred to as the broadcast operation), or the operation of being commented on (hereinafter referred to as the comment operation); or the operation of the candidate being followed by the target user (hereinafter referred to as the follow operation), etc.

[0082] In the above steps, if there is a second interaction between the target user and the content published by the candidate, such as the target user forwarding, liking or commenting on a piece of content published by the candidate, or the target user following a candidate, then an edge can be established between the point corresponding to the target user and the point corresponding to the candidate.

[0083] Furthermore, as mentioned earlier, the second weight of this edge can also be determined based on the number of second interaction operations between the target user and the content published by the candidate. Specifically, the number of second interaction operations between the target user and the content published by the candidate can be directly counted and used as the second weight of this edge. Alternatively, a weight can be set for different types of second interaction operations, and the number of second interaction operations between the target user and the content published by the candidate can be weighted and summed to obtain the second weight of this edge. For example, assuming that the above-mentioned second interaction operations only include click operations, like operations, and comment operations, the weight of click operations can be set to W3, the weight of like operations to W4, and the weight of comment operations to W5. Assuming that the number of click operations between the target user and the content published by the candidate is N3, the number of like operations is N4, and the number of comment operations is N5, the second weight of this edge can be obtained as W3*N3 + W4*N4 + W5*N5. By assigning different weights to different second interactive operations, the degree of preference of the target user for the content represented by different interactive operations can be reflected in a true and effective manner. Thus, the calculated second weight value can more realistically and effectively reflect the degree of association between the target user and the candidate object.

[0084] Figure 6 An example of a second map is shown. Figure 6 In this example, square dots represent candidate objects (i.e., content producers); circular dots represent target users. For example... Figure 6 As shown, candidate object C1 in the second graph is associated with target users T1, T3, T4, T6, and T8, and the second weight values ​​of the edges between candidate object C1 and target users T1, T3, T4, T6, and T8 are v1, v3, v4, v6, and v8, respectively. Furthermore, candidate object C2 is associated with target users T1, T2, T3, T9, and T10. Figure 6 The second weight value of the edge between candidate C2 and target users T1, T2, T3, T9 and T10 is not shown.

[0085] Therefore, it can be seen that the second graph determined by the above method not only represents the relationship between each target user and each candidate object (i.e., content producer), but also the degree of relationship between each target user and each candidate object, which to a certain extent reflects the degree of preference of each target user for each candidate object. Moreover, in the above scheme, different weights are assigned to different interactive operations, which can more accurately reflect the degree of preference of each target user for each candidate object.

[0086] In step 212, the feature vectors of each candidate object are determined based on the second graph and the feature vectors of each target user.

[0087] In some embodiments of this disclosure, the following operations can be performed for each candidate object to determine the feature vector of each candidate object.

[0088] First, based on the second graph, at least one target user corresponding to a point that has an edge with the point corresponding to the candidate object is identified, and this user is taken as the associated user of the candidate object.

[0089] Secondly, obtain the second weight value of the edge between the point corresponding to the candidate object and the point corresponding to the associated user; and

[0090] Finally, the user feature vectors of the associated users are weighted and summed according to the second weight value to obtain the feature vectors of the candidate objects.

[0091] Still with Figure 6 Taking the second spectrum shown as an example, from Figure 6 It can be seen that target users T1, T3, T4, T6, and T8 are associated users of candidate object C1. Assuming the feature vectors of target users T1, T3, T4, T6, and T8 are U1, U3, U4, U6, and U8 respectively, then the feature vector of candidate object C1 will be expressed as C1 = v1*U1 + v3*U3 + v4*U4 + v6*U6 + v8*U8.

[0092] In step 214, the target object of the promotion information is selected from each candidate object based on the feature vector of the promotion information provider and the feature vector of each candidate object.

[0093] In some embodiments of this disclosure, Figure 7 This illustrates the specific implementation process of selecting target objects for promotional information as described in embodiments of this disclosure. For example... Figure 7 As shown, the method includes:

[0094] In step 702, the feature vector of the promotion information provider and the distance between the feature vectors of each candidate object are determined.

[0095] In some embodiments of this disclosure, the aforementioned distance may specifically refer to the inner product between vectors. That is, in the above steps, the distance between the feature vector of the promotion information provider and the feature vectors of each candidate object can be determined by determining the inner product between the feature vector of the promotion information provider and the feature vectors of each candidate object.

[0096] In other embodiments of this disclosure, the aforementioned distance may specifically refer to the cosine distance between vectors. That is, in the above steps, the distance between the feature vector of the promotion information provider and the feature vectors of each candidate object can be determined by determining the cosine distance between them.

[0097] In step 704, the candidate objects are sorted from near to far according to the distance mentioned above.

[0098] In step 706, one or more candidate objects corresponding to one or more feature vectors that are closest to the feature vector of the promotion information provider are selected as the target objects of the promotion information.

[0099] In embodiments of this disclosure, if the aforementioned distance specifically refers to the inner product between vectors, then in this step, one or more candidate objects with the largest inner product can be selected from among the candidate objects as the target object of the aforementioned promotion information.

[0100] In embodiments of this disclosure, if the aforementioned distance specifically refers to the cosine distance between vectors, then in this step, one or more candidate objects with the smallest cosine distance can be selected from among the candidate objects as the target object of the aforementioned promotion information.

[0101] It can be understood that the cosine distance between two vectors primarily characterizes the degree of similarity between the two vectors in terms of angle. The smaller the cosine distance between two vectors, the more similar the two vectors are in terms of angle, meaning the two vectors are more similar in direction. Besides the similarity of the vectors, the inner product between vectors can further reflect the number of users associated with the candidate object and the number of users associated with the promotional information provider. Therefore, the above two different schemes for determining the feature vector of the promotional information provider and the distance between the feature vectors of each candidate object provide two different strategies for matching target objects for the promotional information provider.

[0102] As can be seen from the above method, it uses the target user as a medium between the promotional information provider and the candidate, achieving cross-domain matching between the promotional information and the candidate. Specifically, in this method, on the one hand, a feature vector of the promotional information provider is established through the interaction between the target user and the promotional information provider; on the other hand, a feature vector of the candidate is established through the interaction between the target user and the candidate; finally, a suitable target is selected for the promotional information based on the distance between the feature vectors of the promotional information provider and the candidate. Therefore, this process can achieve accurate matching results between the promotional information provider and the content producer on the content delivery platform, resulting in better information promotion effects.

[0103] Furthermore, the feature vector of the target user determined by the three-tower model takes into account both the target user's preference for the provider of promotional information and the target user's preference for the content. In other words, the feature vector of the target user determined by the three-tower model contains both the features of promotional information and the features of the content, thus enabling more accurate cross-domain matching of promotional information and content.

[0104] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0105] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0106] Corresponding to the above method, embodiments of this disclosure also disclose a device for determining the target object of promotional information. Figure 8 The internal structure of the device described in the embodiments of this disclosure is shown. For example... Figure 8 As shown, the device may include:

[0107] User feature acquisition module 802 is used to acquire the features of each target user respectively;

[0108] User feature vector determination module 804 is used to determine the feature vector of each target user according to the features of each target user;

[0109] The first graph establishment module 806 is used to establish a first graph based on the first interactive operation between each target user and the promotional information released by the promotional information provider.

[0110] The first feature vector determination module 808 is used to determine the feature vector of the promotion information provider based on the first map and the feature vectors of each target user.

[0111] The second graph establishment module 810 is used to establish the second graph based on the second interaction operation between the content published by each target user and each candidate object;

[0112] The second feature vector determination module 812 is used to determine the feature vector of each candidate object based on the second map and the feature vectors of each target user; and

[0113] The target object determination module 814 is used to select the target object of the promotion information from the candidate objects based on the feature vector of the promotion information provider and the feature vector of each candidate object.

[0114] Specifically, the user feature vector determination module 804 can use a pre-trained three-tower model to determine the feature vector of the target user. The structure and training method of the three-tower model have been described above and will not be repeated here.

[0115] The specific implementation of each of the above modules can be found in the aforementioned methods and accompanying drawings, and will not be repeated here. For ease of description, the above apparatus is described in terms of function, divided into various modules. Of course, in implementing this disclosure, the functions of each module can be implemented in one or more software and / or hardware.

[0116] The apparatus described above is used to implement the method for determining the target object of the promotional information in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0117] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for determining the target object of promotional information as described in any of the above embodiments.

[0118] Figure 9 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 2010, a memory 2020, an input / output interface 2030, a communication interface 2040, and a bus 2050. The processor 2010, memory 2020, input / output interface 2030, and communication interface 2040 are interconnected internally via the bus 2050.

[0119] The processor 2010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0120] The memory 2020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 2020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 2020 and is called and executed by the processor 2010.

[0121] The input / output interface 2030 is used to connect input / output modules to enable information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0122] The communication interface 2040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (e.g., USB, Ethernet cable) or wireless means (e.g., mobile network, Wi-Fi, Bluetooth).

[0123] Bus 2050 includes a pathway for transmitting information between various components of the device, such as processor 2010, memory 2020, input / output interface 2030, and communication interface 2040.

[0124] It should be noted that although the above-described device only shows the processor 2010, memory 2020, input / output interface 2030, communication interface 2040, and bus 2050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0125] The electronic device described above is used to implement the method for determining the target object of the promotional information in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0126] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the method for determining the target object of promotional information as described in any of the above embodiments.

[0127] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0128] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the task processing method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0129] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.

[0130] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuitry) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that the embodiments of this disclosure may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0131] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0132] This disclosure is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for determining the target audience of promotional information, comprising: Obtain the characteristics of each target user separately; Determine the feature vector of each target user based on the characteristics of each target user; A first graph is established based on the first interaction between each target user and the promotional information released by the promotional information provider. The feature vector of the promotion information provider is determined based on the first graph and the feature vectors of each target user. A second graph is established based on the second interaction operations between the content published by each target user and each candidate object; Based on the second map and the feature vectors of each target user, the feature vectors of each candidate object are determined respectively. as well as The target object for the promotional information is selected from the candidate objects based on the feature vector of the promotional information provider and the feature vector of each candidate object.

2. The method according to claim 1, wherein, The step of determining the feature vector of each target user based on the features of each target user includes: inputting the features of each target user into a pre-trained three-tower model to obtain the feature vector of each target user.

3. The method according to claim 2, wherein, The three-tower model includes: a user feature tower, a promotional information feature tower, a content feature tower, a first merging layer, a first output layer, a second merging layer, a second output layer, and a training module; wherein, The user feature tower is used to map the features of the target user into a feature vector of the target user; The promotion information feature tower is used to map the features of promotion information into a feature vector of promotion information; The content feature tower is used to map the features of the content into a feature vector of the content. The first merging layer is used to merge the feature vector of the target user and the feature vector of the promotion information to obtain a merged first feature vector; The first output layer is used to output a prediction result that predicts the probability of a first interactive operation between the target user and the promotional information based on the first feature vector; The second merging layer is used to merge the feature vector of the target user and the feature vector of the content to obtain a merged second feature vector; The second output layer is used to output a prediction result that predicts the probability of a second interactive operation between the target user and the content based on the second feature vector; and The training module is used to train the three-tower model based on the prediction results.

4. The method according to claim 3, wherein, The step of training the three-tower model based on the prediction results includes: Acquire sample data; wherein, the sample data includes: characteristics of multiple target users, characteristics of multiple promotional information, characteristics of multiple content, and the first interactive operation between each target user and each promotional information and the second interactive operation between each target user and each content; The characteristics of target users, promotional information, and content in the sample data are respectively input into the user feature tower, promotional information feature tower, and content feature tower; The probability prediction results of a first interactive operation between the target user and the promotional information and the probability prediction results of a second interactive operation between the target user and the content are obtained from the first output layer and the second output layer, respectively. The loss of the three-tower model is determined based on the first interaction between each target user and each promotional message and the second interaction between each target user and each content in the sample data. Based on the determined loss, the parameters in the three-tower model are adjusted using gradient descent; and Return to the step of inputting the features of the target user, the features of the promotional information, and the features of the content from the sample data into the user feature tower, the promotional information feature tower, and the content feature tower, respectively, until the model converges.

5. The method according to claim 1, wherein, The step of establishing the first graph based on the first interaction between each target user and the promotional information published by the promotional information provider includes: The promotional information provider and each target user are defined as points in the first graph. In response to determining that a first interaction exists between a target user and promotional information published by the promotional information provider, an edge is established between the point corresponding to the target user and the point corresponding to the promotional information provider; and The first weight value of the edge between the point corresponding to the target user and the point corresponding to the promotional information provider is determined based on the number of first interaction operations between the target user and the promotional information provider.

6. The method of claim 5, further comprising: Different weights are pre-assigned to different types of first-interaction actions; among them, The step of determining the first weight value of the edge between the point corresponding to the target user and the point corresponding to the promotional information provider based on the number of first interaction operations between the target user and the promotional information provider includes: Based on the weights of different types of first interactive operations, the number of first interactive operations between the target user and the promotional information published by the promotional information provider is weighted and summed to obtain the first weight value of the edge between the point corresponding to the target user and the point corresponding to the promotional information provider.

7. The method according to claim 1, wherein, The step of establishing the second graph based on the second interaction operation between the content published by each target user and each candidate includes: Each candidate object and each target user are used as points in the second graph; In response to determining that a second interaction exists between content posted by a target user and a candidate object, an edge is established between the point corresponding to the target user and the point corresponding to the candidate object; and The second weight value of the edge between the point corresponding to the target user and the point corresponding to the candidate object is determined based on the number of second interaction operations between the content published by the target user and the candidate object.

8. The method of claim 7, further comprising: Different weights are pre-assigned to different types of second interactive operations; among them, The step of determining the second weight value of the edge between the point corresponding to the target user and the point corresponding to the candidate object based on the number of second interaction operations between the content published by the target user and the candidate object includes: Based on the weights of different types of second interactive operations, the number of second interactive operations between the target user and the candidate object's published content is weighted and summed to obtain the second weight value of the edge between the point corresponding to the target user and the point corresponding to the candidate object.

9. The method according to claim 1, wherein, The step of determining the feature vector of the promotion information provider based on the first graph and the feature vectors of each target user includes: Based on the first graph, at least one target user is identified as a point that has an edge with the point corresponding to the promotion information provider, and this user is identified as an associated user of the promotion information provider. Obtain the first weight value of the edge between the point corresponding to the promotion information provider and the point corresponding to the associated user; and The user feature vectors of the associated users are weighted and summed according to the first weight value to obtain the feature vector of the promotion information provider.

10. The method according to claim 1, wherein, The step of determining the feature vector of each candidate object based on the second map and the feature vectors of each target user includes: Perform the following operations for each candidate object: Based on the second graph, at least one target user is identified as a point that has an edge with the point corresponding to the candidate object, and this target user is identified as an associated user of the candidate object. Obtain the second weight values ​​of the edges between the points corresponding to the candidate objects and the points corresponding to the associated users; and The user feature vectors of the associated users are weighted and summed according to the second weight value to obtain the feature vector of the candidate object.

11. The method according to claim 1, wherein, The step of selecting the target object for promotional information from the candidate objects based on the feature vector of the promotional information provider and the feature vector of each candidate object includes: Determine the feature vector of the promotion information provider and the distance between the feature vectors of each candidate object; and From the candidate objects, select one or more candidate objects corresponding to one or more feature vectors that are closest to the feature vector of the promotion information provider as the target objects of the promotion information.

12. The method according to claim 11, wherein, Determining the distance between the feature vector of the promotional information provider and the feature vectors of each candidate object includes: calculating the inner product of the feature vector of the promotional information provider and the feature vectors of each candidate object; and The step of selecting one or more candidate objects corresponding to one or more feature vectors that are closest to the feature vector of the promotion information provider from among the candidate objects includes: selecting one or more candidate objects with the largest inner product from among the candidate objects as the promotion information target objects.

13. The method according to claim 11, wherein, Determining the distance between the feature vector of the promotional information provider and the feature vectors of each candidate object includes: calculating the cosine distance between the feature vector of the promotional information provider and the feature vectors of each candidate object; and The step of selecting one or more candidate objects corresponding to one or more feature vectors that are closest to the feature vector of the promotion information provider from among the candidate objects includes: selecting one or more candidate objects with the smallest cosine distance from among the candidate objects as the promotion information target objects.

14. A device for determining the target audience of promotional information, comprising: The user feature acquisition module is used to acquire the features of each target user. The user feature vector determination module is used to determine the feature vector of each target user based on the features of each target user. The first graph establishment module is used to establish a first graph based on the first interaction operation between each target user and the promotional information released by the promotional information provider. The first feature vector determination module is used to determine the feature vector of the promotion information provider based on the first map and the feature vectors of each target user. The second graph is established by a module for creating a second graph based on the second interaction operation between the content published by each target user and each candidate object. The second feature vector determination module is used to determine the feature vector of each candidate object based on the second map and the feature vectors of each target user. as well as The target object determination module is used to select the target object of the promotion information from the candidate objects based on the feature vector of the promotion information provider and the feature vector of each candidate object.

15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method for determining the target of promotional information as described in any one of claims 1-13.

16. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method for determining the target object of promotional information as described in any one of claims 1-13.

17. A computer program product comprising computer program instructions that, when executed on a computer, cause the computer to perform the method for determining the target audience of promotional information as described in any one of claims 1-13.

Images