Content push method and device
By obtaining and processing content, device and user feature vectors and selecting appropriate push methods, the problem of mismatch between content and user needs in the application is solved, and the push accuracy and user experience are improved.
Patent Information
- Application Number
- CN202010195145.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-03-19
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2040-03-19
AI Technical Summary
In the prior art, application push content does not match user needs, resulting in user resistance and reduce user experience and application activity.
By obtaining content feature vectors, device feature vectors and user feature vectors, processing using a full-connected network, generating message push tags, and selecting appropriate push methods based on the tags to improve the accuracy of content push.
It improves the accuracy of content push, reduces inappropriate push methods, avoids user resistance, and enhances user stickiness.
Smart Images

Figure CN112749327B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a content push method and device. Background Art
[0002] At present, in some applications used by users, these applications will push various contents for users to read, in order to improve the product activity and user stickiness of the application. The content pushed by the application to the user is generally displayed by directly pushing it in the user's terminal interface. This message push method does not take into account the matching degree between the pushed content and the user's needs. When the pushed content does not meet the user's needs, it will easily cause users to resist, reduce the user experience, and thus reduce the frequency of users opening the application. Summary of the invention
[0003] The embodiments of the present application provide a content push method and device, which can improve the accuracy of the message push mode selected for pushing content.
[0004] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present application.
[0005] According to one aspect of an embodiment of the present application, a content push method is provided, including: obtaining a content feature vector corresponding to a target push content, a device feature vector corresponding to a client of the target push content, and a user feature vector of a user currently logged in to the client; inputting the content feature vector, the device feature vector, and the user feature vector into a push mode prediction model; utilizing multiple fully connected networks within the push mode prediction model to process the content feature vector, the device feature vector, and the user feature vector to obtain a message push mode label; and pushing the target push content to the client based on the message push mode corresponding to the message push mode label.
[0006] According to one aspect of an embodiment of the present application, a content push device is provided, including: a first acquisition unit, used to acquire a content feature vector corresponding to a target push content, a device feature vector corresponding to a client of the target push content, and a user feature vector of a user currently logged in to the client; a first input unit, used to input the content feature vector, the device feature vector, and the user feature vector into a push mode prediction model; an execution unit, used to process the content feature vector, the device feature vector, and the user feature vector using multiple fully connected networks within the push mode prediction model to obtain a message push mode label; a push unit, used to push the target push content to the client based on the message push mode corresponding to the message push mode label.
[0007] In some embodiments of the present application, based on the aforementioned scheme, the execution unit is configured as: a first mapping subunit, used to map the content feature vector using the first fully connected network within the push mode prediction model to generate a first feature vector; a second mapping subunit, used to map the user feature vector using the second fully connected network within the push mode prediction model to generate a second feature vector; a first execution subunit, used to obtain a message push mode label through the third fully connected network within the push mode prediction model and the first feature vector, the second feature vector, and the device feature vector.
[0008] In some embodiments of the present application, based on the aforementioned scheme, the first execution sub-unit is configured to: perform semantic matching processing based on the first feature vector and the second feature vector to generate a third feature vector; classify the first feature vector, the second feature vector, the third feature vector and the device feature vector through the third fully connected network within the push mode prediction model to obtain a message push mode label.
[0009] In some embodiments of the present application, based on the aforementioned scheme, the first execution subunit is configured to: concatenate the first feature vector, the second feature vector, the third feature vector and the device feature vector to generate a concatenated vector; classify the concatenated vector through the third fully connected network within the push mode prediction model to obtain a message push mode label.
[0010] In some embodiments of the present application, based on the aforementioned scheme, the first acquisition unit is configured to: generate a content feature vector according to the target pushed content; generate the device feature vector according to the device information corresponding to the client; and generate a user feature vector according to the user information of the user.
[0011] In some embodiments of the present application, based on the aforementioned scheme, the first acquisition unit is configured to: obtain historical push records for all clients of each push content with the same target classification label as the target push content; determine, based on the historical push records, content statistical parameters for pushing the push content under the target classification label using the message push method corresponding to each message push method label; and generate a content feature vector based on the content statistical parameters.
[0012] In some embodiments of the present application, based on the aforementioned scheme, the first acquisition unit is configured to: obtain, according to the user identifier, the historical viewing records of the currently logged-in user for each pushed content with the same target classification label as the target pushed content; determine, according to the historical viewing records, the content viewing parameters for pushing using the message push method corresponding to each message push method label; and generate a user feature vector according to the content viewing parameters.
[0013] In some embodiments of the present application, based on the aforementioned scheme, the content push device also includes: a second acquisition unit, used to obtain training set sample data for training the machine learning model to be trained, each sample data in the training set sample data includes a content feature vector, a device feature vector, a user feature vector and a message push method label; a training unit, used to train the machine learning model to be trained through the training set sample data to obtain a trained machine learning model.
[0014] In some embodiments of the present application, based on the aforementioned scheme, the content push device also includes: a third acquisition unit, used to acquire test set sample data for verifying the trained machine learning model, each sample data in the test set sample data includes a content feature vector, a device feature vector, a user feature vector and a message push method label; a second input unit, used to input the content feature vector, the device feature vector and the user feature vector of each sample data of the test set sample data into the trained machine learning model, and output a predicted message push method label; a verification unit, used to identify the trained machine learning model as the push method prediction model if the proportion of sample data in the test set sample data whose message push method label is consistent with the predicted message push method label accounts for more than a predetermined proportion threshold in the total number of sample data in the test set sample data.
[0015] According to one aspect of an embodiment of the present application, a computer-readable medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the content push method described in the above embodiment is implemented.
[0016] According to one aspect of an embodiment of the present application, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the content push method as described in the above embodiment.
[0017] In the technical solutions provided in some embodiments of the present application, by selecting a suitable message push method based on the content feature vector corresponding to the target pushed content, the device feature vector corresponding to the client, and the user feature vector of the user currently logged in to the client, taking into account the degree of matching between the pushed content and the needs of the user who reads the content, so as to improve the accuracy of the selected message push method for pushing the pushed content, thereby reducing the situation where the message is pushed using an inappropriate message push method, thereby avoiding user resistance and improving the user stickiness of the product.
[0018] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present application, and together with the specification, are used to explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:
[0020] Figure 1 A schematic diagram of an exemplary system architecture to which the technical solution of the embodiments of the present application can be applied is shown.
[0021] Figure 2 A flow chart of a content push method according to an embodiment of the present application is shown.
[0022] Figure 3 A specific flow chart of step S210 of the content push method according to an embodiment of the present application is shown.
[0023] Figures 4A-4C A schematic diagram showing a content pushing method used for pushing content according to an embodiment of the present application is shown.
[0024] Figure 5 A specific flow chart of step S310 of the content push method according to an embodiment of the present application is shown.
[0025] Figure 6 A specific flow chart of step S330 of the content push method according to an embodiment of the present application is shown.
[0026] Figure 7 A flow chart of a content push method according to an embodiment of the present application is shown.
[0027] Figure 8 A flow chart of a content push method according to an embodiment of the present application is shown.
[0028] Fig. 9 A specific flow chart of step S230 of the content push method according to an embodiment of the present application is shown.
[0029] Fig.10 A schematic diagram of the internal structure of a push mode prediction model according to an embodiment of the present application is shown.
[0030] Fig.11 A specific flow chart of step S930 of the content push method according to an embodiment of the present application is shown.
[0031] Fig.12 A specific flow chart of step S1120 of the content push method according to an embodiment of the present application is shown.
[0032] Fig.13 A block diagram of a content pushing device according to an embodiment of the present application is shown.
[0033] Fig.14 A schematic diagram of the structure of a computer system suitable for implementing an electronic device of an embodiment of the present application is shown. DETAILED DESCRIPTION
[0034] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be more comprehensive and complete and fully convey the concept of the example embodiments to those skilled in the art.
[0035] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, known methods, devices, realizations or operations are not shown or described in detail to avoid blurring the various aspects of the application.
[0036] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.
[0037] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.
[0038] Figure 1 A schematic diagram of an exemplary system architecture to which the technical solution of the embodiments of the present application can be applied is shown.
[0039] like Figure 1 As shown, the system architecture may include a client 101 (eg Figure 1 The embodiment of the present invention is a method for manufacturing a network 102 and a server 103. The network 102 is a medium for providing a communication link between the client 101 and the server 103. The network 102 may include various connection types, such as a wired communication link, a wireless communication link, etc.
[0040] It should be understood that Figure 1 The number of clients 101, networks 102, and servers 103 in the embodiment is only for illustration. According to the implementation requirements, there may be any number of clients 101, networks 102, and servers 103. For example, the server 103 may be a server cluster composed of multiple servers. The client 101 interacts with the server 103 through the network 102 to receive or send messages, etc. The server 103 may be a server that provides various services.
[0041] The server 103 obtains the content feature vector corresponding to the target push content to be pushed, the device feature vector corresponding to the client 101 that receives the target push content, and the user feature vector of the user currently logged in to the client 101; inputs the content feature vector, the device feature vector, and the user feature vector into the push mode prediction model; uses the multiple fully connected networks inside the push mode prediction model to process the content feature vector, the device feature vector, and the user feature vector to obtain a message push mode label; pushes the target push content to the client 101 based on the message push mode corresponding to the message push mode label, so that the user can view the target push content through the client 101, and according to the content feature vector corresponding to the target push content, the device feature vector corresponding to the client 101, and the user feature vector of the user currently logged in to the client 101, it is possible to select a suitable message push mode under the premise of considering the matching degree between the pushed content and the user demand for reading the content, so as to improve the accuracy of the selected message push mode for pushing the push content, thereby reducing the situation of pushing due to the use of inappropriate message push mode, thereby avoiding user resistance and improving the user stickiness of the product.
[0042] It should be noted that the content push method provided in the embodiment of the present application is generally executed by the server 103, and accordingly, the content push device is generally arranged in the server 103. However, in other embodiments of the present application, the client 101 may also have similar functions to the server 103, so as to execute the scheme of the content push method provided in the embodiment of the present application. For example, the client 101 receives the target push content issued by the server 103, and the client 101 obtains the content feature vector corresponding to the target push content, the device feature vector corresponding to the client 101 receiving the target push content, and the user feature vector of the user currently logged in by the client 101; the content feature vector, the device feature vector and the user feature vector are input into the push mode prediction model; the content feature vector, the device feature vector and the user feature vector are processed by the multiple fully connected networks inside the push mode prediction model to obtain the message push mode label; the client 101 pushes the target push content on the client 101 based on the message push mode corresponding to the message push mode label, so that the user can view the target push content through the client 101.
[0043] The implementation details of the technical solution of the embodiment of the present application are elaborated in detail below.
[0044] Figure 2 A flow chart of a content push method according to an embodiment of the present application is shown. The content push method can be executed by a server, which can be Figure 1 The server side 103 shown in FIG. Figure 2 As shown, the content push method at least includes steps S210 to S240, which are described in detail as follows:
[0045] In step S210, a content feature vector corresponding to the target pushed content, a device feature vector corresponding to the client of the target pushed content, and a user feature vector of the user currently logged in to the client are obtained.
[0046] In one embodiment, the target pushed content is some kind of pushed content pushed from the server to the client for providing to the user for reading, such as an article or a video.
[0047] The content feature vector is used as the characteristic information of the target pushed content, the device feature vector is used as the characteristic information of the device corresponding to the client receiving the target pushed content, and the user feature vector is used as the characteristic information of the user currently logged in by the client. The content feature vector, the device feature vector and the user feature vector can be pre-generated when the target pushed content is received.
[0048] refer to Figure 3 , Figure 3 The specific flow chart of step S210 of the content push method according to an embodiment of the present application is shown, and the step S210 may include steps S310 to S330. It should be noted that there is no definite order of steps S310 to S330, such as steps S310 to S330 can be executed simultaneously in different threads, and the detailed description of steps S310 to S330 is as follows.
[0049] In step S310, a content feature vector is generated according to the target pushed content.
[0050] The content feature vector is feature vector information generated according to the attribute information contained in the target push content, and is used to characterize the characteristic information of the target push content. The attribute information contained in the target push content may include classification label information, content record information, and operation record information of each push content.
[0051] The classification label information may be a type label set for each pushed content, such as finance, education, games, etc. The content record information may be information recorded for the content contained in each pushed content, such as the number of words, punctuation marks, emoticons, and embedded images in the article, etc. The operation record information may be operation information recorded after the user of each client performs a corresponding reading operation on a certain pushed content, such as the number of clicks, click-through rate, number of comments, number of favorites, number of likes, and number of exposures of the pushed content, etc.
[0052] The content feature vector generated based on the target push content can accurately reflect the characteristics contained in the target push content, thereby facilitating the matching between the characteristics of the target push content and the characteristics of the user currently logged in on the client, thereby improving the accuracy of the selected message push method for pushing content to a certain extent.
[0053] refer to Figures 4A-4C , Figures 4A-4C A schematic diagram showing a content pushing method used for pushing content according to an embodiment of the present application is shown.
[0054] In one embodiment, the content can be pushed on the client application interface in a content push manner as shown in the figure. Figure 4A The content push method shown is the lock screen bar push method, which is used to continuously display the push message corresponding to the push content in the middle area of the client interface until the user enters an instruction to close the push message or clicks on the push message to view the corresponding push content, and then closes the push message; Figure 4BThe content push method shown is the status bar push method, which is used to display the push message corresponding to the push content in the upper area of the client interface, and only display it for a predetermined time, for example, only display it for 3 seconds. The push message will be placed in the client's message background. When the user reads the push message, the user can click on the push message on the client's message background to view the push content. Figure 4C The content push method shown is the notification bar push method, which is used to display in the message push interface of the application. The user needs to enter the message push interface of the application to view the push message corresponding to the displayed push content, and then view the push content. It should be pointed out that when the target push content to be pushed has a high degree of match with user needs, the lock screen bar push method can be used to push the target push content; when the target push content to be pushed has a low degree of match with user needs, the notification bar push method or the notification bar push method can be used to push the target push content.
[0055] refer to Figure 5 , Figure 5 A specific flow chart of step S310 of the content push method according to an embodiment of the present application is shown. Step S310 may include steps S510 to S530, which are described in detail as follows.
[0056] In step S510, historical push records of each push content having the same target classification tag as the target push content to all clients are obtained.
[0057] The attribute information included in the target pushed content may also include content statistical parameters, which are statistical parameters for pushing the push content of each classification label using the message push methods corresponding to each corresponding message push method label.
[0058] In order to generate a content feature vector based on content statistical parameters, it is necessary to first obtain the historical push records of all clients for each push content with the same target classification label as the target push content. The historical push record can be the log information generated when the client operates on each push content. The log information at least includes the classification label of the push content, the message push method label for pushing the push content, and the click record information of whether the client clicks on the push content. It should be pointed out that the historical push record can be the historical push record within a certain period of time before the current time, such as all historical push records within one month from the current time.
[0059] In step S520, according to the historical push records, content statistical parameters of the push content under the target classification label that is pushed using the message push methods corresponding to the message push method labels are determined.
[0060] In order to determine the content statistical parameters of the message push methods corresponding to each message push method label, the content statistical parameters of the message push methods corresponding to different message push method labels will be determined based on the message push method label used to push each pushed content and the click record information of whether the client clicks on the pushed content information.
[0061] The types of statistical parameters may include the number of pushes, the number of clicks, and the click-through rate. The number of pushes is the total number of pushes using different message push methods. The number of clicks is the total number of clicks on the push content when the push method is used to push the push content. The click-through rate is the ratio of the total number of clicks for the push method to the total number of pushes.
[0062] Taking the target push content as an example, if the target push content is financial push content, if the A push method is used to push the financial push content 6,000 times, the number of clicks is 3,000 times; the B push method is used to push the financial push content 3,000 times, the number of clicks is 1,800 times; the C push method is used to push the financial push content 1,000 times, the number of clicks is 900 times; then for the financial push content, the content statistical parameters of using the A push method to push the financial push content are 6,000 pushes, 3,000 clicks, and a click-through rate of 50%; the content statistical parameters of using the B push method to push the financial push content are 3,000 pushes, 1,800 clicks, and a click-through rate of 60%; the content statistical parameters of using the C push method to push the financial push content are 1,000 pushes, 900 clicks, and a click-through rate of 90%.
[0063] In step S530, a content feature vector is generated according to the content statistical parameters.
[0064] After the content statistical parameters are obtained, the corresponding content feature vector is generated according to the content statistical parameters and the corresponding relationship between the content statistical parameters and the vector value.
[0065] Figure 5 In the technical solution of the embodiment shown, by using various message push methods to push content according to the category label to which the target push content belongs, the message push method suitable for the target push content can be determined more accurately, thereby improving the accuracy of the selected message push method for pushing the content to a certain extent.
[0066] Please continue to refer to Figure 3 In step S320, the device feature vector is generated according to the device information corresponding to the client.
[0067] In one embodiment, the device feature vector is feature vector information generated according to device information corresponding to the client receiving the target pushed content, and is used to characterize the characteristics of the device corresponding to the client receiving the target pushed content.
[0068] The device information may include the current time period, the current device manufacturer, the current network type, the current device screen brightness, the current device power, the current motion state, etc. Specifically, the current time period may be morning, afternoon, evening, or early morning; the current device manufacturer may be the current device manufacturer label of the device corresponding to the client; the current network type may be WIFI, 2G, 3G, 4G, etc.; the current device screen brightness may be the brightness level of the device display, such as very dark, dark, bright, very bright, etc.; the current device power may be the current power level of the device, such as full, sufficient, low, very low, etc.; the current motion state may be moving or stationary.
[0069] The current time period, current device manufacturer, current network type, current device brightness, and current device power of the device corresponding to the client can be obtained from the device attribute information corresponding to the device, and the current motion state can be determined by detecting the built-in gyroscope sensor component of the device.
[0070] After obtaining the device information corresponding to the client, a device feature vector may be generated according to the device information and the corresponding relationship between the device information and the vector value.
[0071] Different device information corresponding to the client will affect the different message push methods. For example, when the user is in an outdoor sports scene, the user is in motion and not using the device. In the device information corresponding to the client, the current motion state information is in motion, the current device screen brightness level is very dim or dim, and the current network type is 2G, 3G or 4G. Since the user is in an outdoor sports scene, it is likely that the user does not need to preview the push content at the moment. Therefore, you can use Figure 4A The lock screen bar push method shown is used to push the push content or to push the push content, so that the user can directly click on the push message in the middle area of the client interface after exercising, so as to conveniently view the push content and avoid the user missing the push content of interest, thereby increasing the user's activity on the target application and improving the product stickiness of the application.
[0072] For example, when a user is using a mobile phone before going to bed, the user is using the device corresponding to the client. In the device information corresponding to the client, the current motion status information is still. Since the device is currently in use, the current device brightness is bright or very bright. Since the user is indoors, the current network type is WIFI network, and the current time period is night. Figure 4B The status bar push method shown in the figure should be avoided to push the push content. Figure 4A The lock screen bar push method shown is used to push the content, so that the user can not only be informed of the push content, but also avoid the push message for the push content being directly displayed in the middle area of the client, thereby avoiding harassment to the user and further improving the product stickiness of the application.
[0073] Based on the device feature vector generated based on the device information corresponding to the client, the appropriate message push method can be selected according to the current usage status of the device corresponding to the client, which improves the accuracy of the selected message push method for pushing content to a certain extent, thereby increasing the user's activity on the target application and improving the product stickiness of the application.
[0074] In step S330, a user feature vector is generated according to the user information.
[0075] In one embodiment, the user feature vector is user feature vector information generated according to the user information of the user currently logged in by the client, and is used to characterize the characteristics of the user currently logged in by the client. The user information may include gender, age, tags, permanent geographic location, and common device manufacturers.
[0076] The content feature vector generated based on the user information of the user currently logged in on the client can accurately mine the user characteristics of the user currently logged in on the client, thereby facilitating the matching between the characteristics of the target pushed content and the characteristics of the user currently logged in on the client, thereby improving the accuracy of the selected message push method for pushing content to a certain extent.
[0077] refer to Figure 6 , Figure 6 A specific flow chart of step S330 of the content push method according to an embodiment of the present application is shown. Step S330 may include steps S610 to S630, which are described in detail as follows.
[0078] In step S610, historical reading records of the currently logged-in user for each pushed content having the same target classification tag as the target pushed content are obtained according to the user identifier.
[0079] The user information may also include content reading parameters, which are parameter information for pushing messages using message push methods corresponding to each message push method label. The types of parameter information may include the number of pushes, the number of clicks, and the click rate.
[0080] In order to generate a user feature vector based on content viewing parameters, it is necessary to first obtain the historical viewing records of the currently logged-in user of the client for each pushed content with the same target classification label as the target pushed content. The historical viewing records can be log information generated when the currently logged-in user of the client operates each pushed content. The log information at least includes the classification label of the pushed content, the message push method label for pushing the pushed content, and the click record information of whether the client clicks on the pushed content. It should be pointed out that the historical viewing records can be historical viewing records within a certain period of time before the current time, such as all historical viewing records within three months from the current time.
[0081] In step S620, based on the historical reading records, content reading parameters pushed using the message push methods corresponding to the message push method labels are determined.
[0082] When obtaining the historical reading records, for the user currently logged in on the client, the content reading parameters pushed using the message push methods corresponding to each message push method label will be determined based on the message push method label used to push each pushed content on the client and whether the client clicks on the pushed content information. The types of content reading parameters may include the number of pushes, the number of clicks, and the click-through rate. The number of pushes is the total number of pushes using different message push methods, the number of clicks is the total number of clicks on the pushed content when the push method is used to push the pushed content, and the click-through rate is the ratio of the total number of clicks for the push method to the total number of pushes.
[0083] Taking financial push content as an example, for financial push content, if the A push method is used to push the financial push content 120 times, the number of clicks is 90 times; the B push method is used to push the financial push content 52 times, the number of clicks is 26 times; the C push method is used to push the financial push content 10 times, the number of clicks is 3 times; then for the financial push content, the content statistical parameters of using the A push method to push the financial push content are 120 pushes, 90 clicks, and a click-through rate of 75%; the content statistical parameters of using the B push method to push the financial push content are 52 pushes, 26 clicks, and a click-through rate of 50%; the content statistical parameters of using the C push method to push the financial push content are 10 pushes, 3 clicks, and a click-through rate of 30%.
[0084] In step S630, a user feature vector is generated according to the content browsing parameters.
[0085] After the content browsing parameters are obtained, the corresponding user feature vector is generated according to the content browsing parameters and the corresponding relationship between the content browsing parameters and the vector value.
[0086] Figure 6 In the technical solution of the embodiment shown, the message push method suitable for the user is determined based on the click situation of the client user on the pushed content of the category label to which the target push content belongs when pushed using various message push methods, thereby improving the accuracy of the selected message push method for pushing content to a certain extent.
[0087] Please continue to refer to Figure 2 In step S220, the content feature vector, the device feature vector and the user feature vector are input into a push mode prediction model.
[0088] In one embodiment, the push mode prediction model is obtained by training the machine learning model with training sample data. The push mode prediction model is used to process the input content feature vector, device feature vector and user feature vector to determine the message push mode label for pushing the push content.
[0089] refer to Figure 7 , Figure 7 A flowchart of a content push method according to an embodiment of the present application is shown. The content push method may include steps S710 to S720, which are described in detail as follows.
[0090] In step S710, training set sample data for training the machine learning model to be trained is obtained, and each sample data in the training set sample data includes a content feature vector, a device feature vector, a user feature vector and a message push method label.
[0091] The training set sample data is generated by recording the relevant data when all pushed content in the historical segment is pushed. Each sample data in the training set sample data includes a content feature vector, a device feature vector, a user feature vector and a message push method label, wherein the content feature vector is generated based on the pushed content, the device feature vector is generated based on the device information corresponding to the client, and the user feature vector is generated based on the user information.
[0092] In step S720, the machine learning model to be trained is trained using the training set sample data to obtain a trained machine learning model.
[0093] The training set sample data is input into the machine learning model, and the machine learning model to be trained is trained with the training set sample data to obtain the trained machine learning model. The process of training the machine learning model is to adjust the coefficients in the network structure corresponding to the machine learning model, so that for the input content feature vector, device feature vector and user feature vector, after the coefficients in the network structure corresponding to the machine learning model are calculated, the output result is the determined message push method label.
[0094] In one embodiment, during the training process of the machine learning model, the various coefficients in the network structure corresponding to the machine learning model can also be optimized based on a preset optimization algorithm. For example, the various coefficients in the network structure corresponding to the machine learning model can be optimized based on the Adam optimization algorithm, thereby improving the efficiency of training the machine learning model.
[0095] refer to Figure 8 , Figure 8 A flow chart of a content push method according to an embodiment of the present application is shown. After step S720, the content push method may include steps S810 to S830, which are described in detail as follows.
[0096] In step S810, test set sample data is obtained for verifying the trained machine learning model, and each sample data in the test set sample data includes a content feature vector, a device feature vector, a user feature vector, and a message push method label.
[0097] The trained machine learning model also needs to be verified to ensure that the machine learning model meets the expected effect. When verifying the trained machine learning model, a test set sample data for verifying the trained machine learning model can be obtained, wherein each sample data in the test set sample data includes a content feature vector, a device feature vector, a user feature vector, and a message push method label.
[0098] In step S820, the content feature vector, device feature vector and user feature vector of each sample data of the test set sample data are input into the trained machine learning model, and the predicted message push mode label is output.
[0099] The content feature vector, device feature vector and user feature vector of each sample data in the test set are input into the trained machine learning model, and each sample data including the content feature vector, device feature vector and user feature vector is processed by the coefficients in the network structure corresponding to the machine learning model to obtain the predicted message push method label for each sample data.
[0100] In step S830, if the proportion of the number of sample data items in the test set sample data whose message push method labels are consistent with the predicted message push method labels exceeds a predetermined proportion threshold value, the trained machine learning model is identified as the push method prediction model.
[0101] For each piece of sample data in the test set sample data, if the proportion of the number of sample data pieces whose message push method labels in the test set sample data are consistent with the predicted message push method labels exceeds the predetermined proportion threshold, it means that the machine learning model meets the expected functional effect, and the trained machine learning model can be identified as a push method prediction model; otherwise, it needs to be trained using the training set sample data until the machine learning model meets the expected functional effect.
[0102] Please continue to refer to Figure 2 In step S230, the content feature vector, the device feature vector and the user feature vector are processed using multiple fully connected networks within the push mode prediction model to obtain a message push mode label.
[0103] The push mode prediction model processes the content feature vector, device feature vector and user feature vector through multiple fully connected networks within it, and outputs a prediction result, which is a message push mode label determined based on the content feature vector, device feature vector and user feature vector.
[0104] refer to Fig. 9 , Fig. 9 A specific flow chart of step S230 of the content push method according to an embodiment of the present application is shown, and step S230 may include steps S910 to S930.
[0105] refer to Fig.10 , Fig.10 A schematic diagram of the internal structure of a push mode prediction model according to an embodiment of the present application is shown. The push mode prediction model includes a first fully connected network 1010, a second fully connected network 1020, and a third fully connected network 1030.
[0106] The following references Fig. 9 and Fig.10 Steps S910 to S930 are described in detail.
[0107] In step S910, the content feature vector is mapped using a first fully connected network within the push mode prediction model to generate a first feature vector.
[0108] In one embodiment, for the content feature vector input into the push mode prediction model, the content feature vector is mapped by the first fully connected network 1010 inside the push mode prediction model to generate a first feature vector. Through the processing of the first fully connected network 1010, the information contained in the content feature vector can be fully mined, thereby improving the accuracy of the predicted message push mode label.
[0109] In step S920, the user feature vector is mapped using a second fully connected network within the push mode prediction model to generate a second feature vector.
[0110] In one embodiment, for the user feature vector input into the push mode prediction model, the user feature vector is mapped to generate a second feature vector through the second fully connected network 1020 within the push mode prediction model. Through the processing of the second fully connected network 1020, the information contained in the user feature vector can be fully mined, thereby improving the accuracy of the predicted message push mode label.
[0111] In step S930, a message push mode label is obtained through the third fully connected network within the push mode prediction model and the first feature vector, the second feature vector, and the device feature vector.
[0112] After obtaining the first feature vector and the second feature vector, for the target push content to be pushed, in order to determine the appropriate message push method label for the user logged in on the client, the third fully connected network 1030 within the push method prediction model classifies the first feature vector, the second feature vector and the device feature vector to obtain the message push method label. Through the processing of the third fully connected network 1030, the correlation between the first feature vector, the second feature vector and the device feature vector can be fully explored, so as to fully consider the matching degree between the pushed content and the user's needs, and then facilitate the determination of the accuracy of the selected message push method.
[0113] refer to Fig.11 , Fig.11 A specific flow chart of step S930 of the content push method according to an embodiment of the present application is shown. Step S930 may include steps S1110 to S1130, which are described in detail as follows.
[0114] In step S1110, semantic matching processing is performed based on the first feature vector and the second feature vector to generate a third feature vector.
[0115] Please continue to refer to Fig.10, the push mode prediction model also includes a matching operator network 1040. When the message push mode label is obtained through the third fully connected network and the first feature vector, the second feature vector, and the device feature vector inside the push mode prediction model, the matching operator network 1040 will perform semantic matching processing based on the first feature vector and the second feature vector to generate a corresponding fourth feature vector. The process of performing semantic matching processing on the first feature vector and the second feature vector is actually performing at least one of an addition operation and a multiplication operation on the elements at corresponding positions contained in the first feature vector and the second feature vector. It should be noted that the corresponding elements contained in the first feature vector and the second feature vector can fully mine the information contained in the content feature vector and the user feature vector through semantic matching processing, and then fully consider the matching degree between the pushed content and the user, thereby improving the accuracy of the selected message push mode.
[0116] In step S1120, the first feature vector, the second feature vector, the third feature vector and the device feature vector are classified and processed by a third fully connected network within the push mode prediction model to obtain a message push mode label.
[0117] In one embodiment, after determining to obtain the third feature vector, the first feature vector, the second feature vector, the third feature vector and the device feature vector are classified and processed by the third fully connected network 1030 within the push mode prediction model to obtain a message push mode label.
[0118] refer to Fig.12 , Fig.12 A specific flow chart of step S1120 of the content push method according to an embodiment of the present application is shown. Step S1120 may include steps S1210 to S1220, which are described in detail as follows.
[0119] Step S1210: concatenate the first feature vector, the second feature vector, the third feature vector and the device feature vector to generate a concatenated vector.
[0120] Step S1220: Classify the concatenated vector through the third fully connected network within the push mode prediction model to obtain a message push mode label.
[0121] After determining that the third eigenvector is obtained, in order to facilitate classification processing by the third fully connected network 1030 within the push mode prediction model, the third eigenvector can also be spliced with the first eigenvector, the second eigenvector and the device eigenvector. When splicing, while ensuring the independence of the third eigenvector and the first eigenvector, the second eigenvector and the device feature vector, the splicing vector generated by the splicing process is also convenient for the third fully connected network 1030 within the push mode prediction model to perform classification processing, which can improve the processing efficiency of the push mode prediction model.
[0122] Please continue to refer to Figure 2 In step S240, the target push content is pushed to the client based on the message push mode corresponding to the message push mode label.
[0123] After determining the message push method tag, the target push content will be pushed to the client according to the message push method corresponding to the message push method tag, so that the user can read the push content on the client.
[0124] By selecting a suitable message push method based on the content feature vector corresponding to the target pushed content, the device feature vector corresponding to the client, and the user feature vector of the user currently logged in on the client, the method can be configured to match the pushed content with the needs of users who want to read the content. This allows the selected message push method to match the pushed content with user needs, thereby improving the accuracy of the selected message push method for pushing content. This avoids user resistance when the pushed content does not meet user needs, thereby improving user stickiness of the product.
[0125] The following describes an embodiment of the device of the present application, which can be used to execute the content push method in the above embodiment of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the content push method in the above embodiment of the present application.
[0126] Fig.13 A block diagram of a content pushing device according to an embodiment of the present application is shown.
[0127] Reference Fig.13As shown, according to an embodiment of the present application, a content push device 1300 includes: a first acquisition unit 1310, a first input unit 1320, an execution unit 1330 and a push unit 1340, wherein the first acquisition unit 1310 is used to acquire a content feature vector corresponding to a target push content, a device feature vector corresponding to a client of the target push content, and a user feature vector of a user currently logged in to the client; the first input unit 1320 is used to input the content feature vector, the device feature vector and the user feature vector into a push mode prediction model; the execution unit 1330 is used to process the content feature vector, the device feature vector and the user feature vector using multiple fully connected networks within the push mode prediction model to obtain a message push mode label; the push unit 1340 is used to push the target push content to the client based on the message push mode corresponding to the message push mode label.
[0128] In some embodiments of the present application, based on the aforementioned scheme, the execution unit 1330 is configured as: a first mapping subunit, used to map the content feature vector using the first fully connected network inside the push mode prediction model to generate a first feature vector; a second mapping subunit, used to map the user feature vector using the second fully connected network inside the push mode prediction model to generate a second feature vector; a first execution subunit, used to obtain a message push mode label through the third fully connected network inside the push mode prediction model and the first feature vector, the second feature vector, and the device feature vector.
[0129] In some embodiments of the present application, based on the aforementioned scheme, the first execution sub-unit is configured to: perform semantic matching processing based on the first feature vector and the second feature vector to generate a third feature vector; classify the first feature vector, the second feature vector, the third feature vector and the device feature vector through the third fully connected network within the push mode prediction model to obtain a message push mode label.
[0130] In some embodiments of the present application, based on the aforementioned scheme, the first execution subunit is configured to: concatenate the first feature vector, the second feature vector, the third feature vector and the device feature vector to generate a concatenated vector; classify the concatenated vector through the third fully connected network within the push mode prediction model to obtain a message push mode label.
[0131] In some embodiments of the present application, based on the aforementioned scheme, the first acquisition unit 1310 is configured to: generate a content feature vector according to the target pushed content; generate the device feature vector according to the device information corresponding to the client; and generate a user feature vector according to the user information of the user.
[0132] In some embodiments of the present application, based on the aforementioned scheme, the first acquisition unit 1310 is configured to: obtain historical push records for all clients of each push content with the same target classification label as the target push content; determine, based on the historical push records, content statistical parameters for pushing the push content under the target classification label using the message push method corresponding to each message push method label; and generate a content feature vector based on the content statistical parameters.
[0133] In some embodiments of the present application, based on the aforementioned scheme, the first acquisition unit 1310 is configured to: obtain, according to the user identifier, the historical viewing records of the currently logged-in user for each pushed content with the same target classification label as the target pushed content; determine, according to the historical viewing records, the content viewing parameters for pushing using the message push methods corresponding to each message push method label; and generate a user feature vector according to the content viewing parameters.
[0134] In some embodiments of the present application, based on the aforementioned scheme, the content push device also includes: a second acquisition unit, used to obtain training set sample data for training the machine learning model to be trained, each sample data in the training set sample data includes a content feature vector, a device feature vector, a user feature vector and a message push method label; a training unit, used to train the machine learning model to be trained through the training set sample data to obtain a trained machine learning model.
[0135] In some embodiments of the present application, based on the aforementioned scheme, the content push device also includes: a third acquisition unit, used to acquire test set sample data for verifying the trained machine learning model, each sample data in the test set sample data includes a content feature vector, a device feature vector, a user feature vector and a message push method label; a second input unit, used to input the content feature vector, the device feature vector and the user feature vector of each sample data of the test set sample data into the trained machine learning model, and output a predicted message push method label; a verification unit, used to identify the trained machine learning model as the push method prediction model if the proportion of sample data in the test set sample data whose message push method label is consistent with the predicted message push method label accounts for more than a predetermined proportion threshold in the total number of sample data in the test set sample data.
[0136] In some embodiments of the present application, based on the aforementioned scheme, the content push device also includes at least one of the following units: a second sorting unit, used to reorder the content push list based on a preset recommended audio blacklist; and a second or third sorting unit, used to reorder the content push list based on the similarity of the audio tags of each audio in the content push list.
[0137] Fig.14 A schematic diagram of the structure of a computer system suitable for implementing an electronic device of an embodiment of the present application is shown.
[0138] It should be noted that Fig.14 The computer system 1100 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0139] like Fig.14 As shown, the computer system 1400 includes a central processing unit (CPU) 1401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1402 or the program loaded from the storage part 1408 to the random access memory (RAM) 1403, such as executing the method described in the above embodiment. In the RAM 1403, various programs and data required for system operation are also stored. The CPU 1401, the ROM 1402 and the RAM 1403 are connected to each other through the bus 1404. The input / output (I / O) interface 1405 is also connected to the bus 1404.
[0140] The following components are connected to the I / O interface 1405: an input section 1406 including a keyboard, a mouse, etc.; an output section 1407 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1408 including a hard disk, etc.; and a communication section 1409 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 1409 performs communication processing via a network such as the Internet. A drive 1410 is also connected to the I / O interface 1405 as needed. A removable medium 1411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1410 as needed so that a computer program read therefrom is installed into the storage section 1408 as needed.
[0141] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication section 1409, and / or installed from a removable medium 1411. When the computer program is executed by a central processing unit (CPU) 1401, various functions defined in the system of the present application are executed.
[0142] It should be noted that the computer-readable medium shown in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by an instruction execution system, device or device or used in combination with it. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable computer program is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.
[0143] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. Wherein, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0144] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.
[0145] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the method described in the above embodiment.
[0146] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.
[0147] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.
[0148] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary technical means in the art that are not disclosed in the present application.
[0149] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.
Claims
1. A content push method, It is characterized in that include: Obtaining a content feature vector corresponding to the target pushed content, a device feature vector corresponding to the client of the target pushed content, and a user feature vector of a user currently logged in to the client; Inputting the content feature vector, the device feature vector and the user feature vector into a push mode prediction model; Mapping the content feature vector using a first fully connected network within the push mode prediction model to generate a first feature vector; Mapping the user feature vector using a second fully connected network within the push mode prediction model to generate a second feature vector; Performing semantic matching processing based on the first feature vector and the second feature vector to generate a third feature vector; Splicing the first feature vector, the second feature vector, the third feature vector and the device feature vector to generate a splicing vector; Classify the concatenated vectors using a third fully connected network within the push mode prediction model to obtain a message push mode label; The target push content is pushed to the client based on the message push method corresponding to the message push method label, and the message push method is one of the following: lock screen bar push method, status bar push method, and notification bar push method.
2. The content push method according to claim 1, It is characterized in that The step of acquiring a content feature vector corresponding to the target pushed content, a device feature vector corresponding to the client of the target pushed content, and a user feature vector of a user currently logged in by the client includes: Generate a content feature vector according to the target pushed content; Generate the device feature vector according to the device information corresponding to the client; A user feature vector is generated according to the user information of the user.
3. The content push method according to claim 2, It is characterized in that The generating of the content feature vector according to the target pushed content also includes: Acquire historical push records of each push content having the same target classification label as the target push content for all clients; Determine, based on the historical push records, content statistical parameters for pushing the push content under the target classification label using the message push methods corresponding to each message push method label; A content feature vector is generated according to the content statistical parameters.
4. The content push method according to claim 2, It is characterized in that The generating of the user feature vector according to the user information of the user further includes: Acquire, according to the user identifier, historical reading records of the currently logged-in user for each pushed content having the same target classification label as the target pushed content; Determine content reading parameters for pushing messages using message push methods corresponding to each message push method tag according to the historical reading records; A user feature vector is generated according to the content browsing parameters.
5. The content push method according to claim 1, It is characterized in that The content push method further includes: Acquire training set sample data for training the machine learning model to be trained, wherein each piece of sample data in the training set sample data includes a content feature vector, a device feature vector, a user feature vector, and a message push mode label; The machine learning model to be trained is trained using the training set sample data to obtain a trained machine learning model.
6. The content push method according to claim 5, It is characterized in that After the machine learning model to be trained is trained using the training set sample data to obtain the trained machine learning model, the content push method further includes: Acquire test set sample data for verifying the trained machine learning model, wherein each piece of sample data in the test set sample data includes a content feature vector, a device feature vector, a user feature vector, and a message push mode label; Inputting the content feature vector, device feature vector and user feature vector of each piece of sample data in the test set into the trained machine learning model, and outputting the predicted message push mode label; If the proportion of the number of sample data items in the test set sample data whose message push method labels are consistent with the predicted message push method labels exceeds a predetermined proportion threshold, the trained machine learning model is identified as the push method prediction model.
7. A content push device, It is characterized in that include: A first acquisition unit is used to acquire a content feature vector corresponding to a target pushed content, a device feature vector corresponding to a client of the target pushed content, and a user feature vector of a user currently logged in to the client; A first input unit, configured to input the content feature vector, the device feature vector, and the user feature vector into a push mode prediction model; An execution unit, configured to process the content feature vector, the device feature vector, and the user feature vector by using a plurality of fully connected networks within the push mode prediction model to obtain a message push mode label; A push unit, configured to push the target push content to the client based on a message push mode corresponding to the message push mode label, wherein the message push mode is one of the following: a lock screen bar push mode, a status bar push mode, and a notification bar push mode; Wherein, the execution unit includes: A first mapping subunit, configured to map the content feature vector using a first fully connected network within the push mode prediction model to generate a first feature vector; A second mapping subunit, configured to map the user feature vector using a second fully connected network within the push mode prediction model to generate a second feature vector; A first execution subunit is used to obtain a message push mode label through a third fully connected network in the push mode prediction model and the first feature vector, the second feature vector, and the device feature vector; The first execution subunit is configured as follows: Based on the first feature vector and the second feature vector, semantic matching processing is performed to generate a third feature vector; the first feature vector, the second feature vector, the third feature vector and the device feature vector are spliced to generate a spliced vector; the spliced vector is classified by the third fully connected network within the push mode prediction model to obtain a message push mode label.
8. A computer readable medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the content pushing method according to any one of claims 1 to 6 is implemented.
9. An electronic device, It is characterized in that include: one or more processors; A storage device, used to store one or more programs, when the one or more programs are executed by the one or more processors, enables the one or more processors to implement the content push method according to any one of claims 1 to 6.
Citation Information
Cited By
Content pushing method based on artificial intelligence
CN119415764A
