A method and system for sending push messages

Abstract

The present invention discloses a method and system for sending push messages. In the process of configuring push messages, an embodiment of the present invention obtains the sending-related numerical features of push messages, the product identification features in the push messages, and the recommendation features in the push messages; the sending-related numerical features of the push messages, the product identification features in the push messages, and the recommendation features of the push messages are input into a trained first neural network for matching processing, and a push message including matching copy content is output, wherein the copy content includes product information and the recommendation features of the corresponding product; after the push message is configured, the push message including the matching copy content is sent to a user terminal device. Therefore, the embodiment of the present invention accurately sets a push message with matching copy content according to the needs of the user and sends it to the user terminal device, thereby improving the timeliness of the push message and improving the operational efficiency of the application that provides the push message.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method and system for sending a push message. Background Art

[0002] With the development of communication technology, more and more users like to shop online. In order to attract more users within a set range, such as users who buy specific products (such as low-priced products), it is necessary to recommend products to users within the set range. At this time, push messages are introduced. By sending push messages including copy content to users within the set range, the copy content contains product information and recommendation information, so that users within the set range can directly know which products they can buy online.

[0003] Applications that provide products and services to users have the function of setting and sending such push messages, determine the user range and content of the push messages from users who use or potentially use the application, and send the set push messages within a set time. In particular, for applications that provide specific products, sending push messages with content that meets their needs to users within a set range can increase the number and frequency of users using the products and services provided by the application.

[0004] However, currently when sending push messages, the copy content in the push messages is pre-set, and there is no push message setting with different copy content based on user needs, which results in low timeliness in sending push messages, making it difficult to increase the number of visits to the application where the product involved in the push message is located and the number of purchases of the product involved, thereby reducing the efficiency of operating the application. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a method for sending push messages, which can accurately set push messages with matching text content according to user needs and send them to user-end devices, thereby improving the timeliness of push messages and improving the operational efficiency of applications that provide push messages.

[0006] The embodiment of the present invention also provides a push message sending system, which can accurately set matching push messages according to user needs and send them to user terminal devices, thereby improving the timeliness of push messages and improving the operating efficiency of applications that provide push messages.

[0007] The embodiment of the present invention is implemented as follows:

[0008] A method for sending a push message, the method comprising:

[0009] Obtaining the sending-related numerical features of push messages, product identification features in push messages, and recommendation features in push messages;

[0010] Inputting the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message into the trained first neural network for matching processing, and outputting a push message including matching copy content, wherein the copy content includes product information and the recommendation features of the corresponding product;

[0011] A push message including matching copy content is sent to the user's terminal device.

[0012] Preferably, the first neural network obtained through training is a neural network based on an attention mechanism, and the inputting into the first neural network obtained through training for matching processing comprises:

[0013] Processing the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message by a pairwise cross-attention mechanism in a first neural network;

[0014] The obtained vector results are classified and processed to obtain push messages including matching copy content.

[0015] Preferably, the outputting of different push messages corresponding to different user identifiers and including different copy contents further includes:

[0016] Processing the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message by a pairwise cross-attention mechanism in a first neural network;

[0017] A second neural network is obtained by training, and the second neural network performs convolution processing on the input sending-related numerical features of the push message and the product identification features in the push message;

[0018] The vector result obtained by the convolution processing and the vector result obtained by the pairwise cross attention mechanism processing are fully connected and then classified to obtain a push message including matching copy content.

[0019] Preferably, the sending-related numerical features of the push message include:

[0020] One or more combinations of the number of times push messages are sent, the number of product categories in the push messages, the number of product brands in the push messages, the number of times push messages are received, the number of product categories in the received push messages, and the number of product brands in the received push messages within different units of time;

[0021] The product identification features in the push message include:

[0022] The product identifiers of the most products in the push messages and / or the product identifiers of the most products in the received push messages in different units of time;

[0023] The recommended features in the push message include:

[0024] The frequency with which user-end devices access or / and obtain product-related information in different time units.

[0025] Preferably, the method for sending the push message is applied in the process of recommending products in an application providing products to a user terminal;

[0026] The recommendation features in the push message are set according to the recommendation features set by the application side that provides the product to the user terminal, and / or the recommendation features obtained when the user terminal uses the application that provides the product.

[0027] A system for sending push messages, the system comprising: an acquisition unit, a matching unit and a sending unit, wherein:

[0028] An acquisition unit, used to acquire the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features in the push message;

[0029] a matching unit, configured to input the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message into the trained first neural network for matching processing, and output a push message including matching copy content, wherein the copy content includes product information and the recommendation features of the corresponding product;

[0030] The sending unit is used to send a push message including the matching copy content to the user terminal device.

[0031] Preferably, the matching unit is further used to process the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message in a pairwise cross-attention mechanism in the trained first neural network;

[0032] The obtained vector results are classified and processed in the trained first neural network to obtain a push message including matching copy content.

[0033] Preferably, the sending-related numerical features of the push message include:

[0034] One or more combinations of the number of times push messages are sent, the number of product categories in the push messages, the number of products in the push messages, the number of times push messages are received, the number of product categories in the received push messages, and the number of products in the received push messages within different units of time;

[0035] The product identification features in the push message include:

[0036] The product identifiers of the most products in the push messages and / or the product identifiers of the most products in the received push messages in different units of time;

[0037] The content features of the push message include:

[0038] The frequency with which user-end devices access or / and obtain product-related information in different time units.

[0039] A device for sending a push message, comprising:

[0040] A memory; and a processor coupled to the memory, the processor being configured to execute the above-mentioned method for sending a push message based on instructions stored in the memory.

[0041] A computer-readable storage medium, characterized in that a computer program is stored thereon, and when the program is executed by a processor, the above-mentioned method for sending a push message is implemented.

[0042] As can be seen above, in the process of configuring the push message, the embodiment of the present invention obtains the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features in the push message; the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features in the push message are input into the trained first neural network for matching processing, and a push message including matching copy content is output, and the copy content includes product information and the recommendation features of the corresponding product; after the push message is configured, the push message including the matching copy content is sent to the user terminal device. Since in the process of configuring the push message, the embodiment of the present invention uses the features that can reflect user needs and product recommendation information as the input of the trained first neural network, and uses the trained first neural network for matching processing, the copy content in the obtained push message can reflect user needs and product recommendation information. Therefore, the embodiment of the present invention accurately sets the push message with matching copy content according to the user's needs and sends it to the user terminal device, thereby improving the timeliness of the push message and improving the operational efficiency of the application that provides the push message. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of the framework structure of push message configuration and sending provided in an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of software modules for configuring and sending push messages provided by an embodiment of the present invention;

[0045] Figure 3 A flow chart of a method for sending a push message is provided for an embodiment of the present invention;

[0046] Figure 4 A schematic diagram of the structure of the combination of the first neural network and the second neural network provided in an embodiment of the present invention;

[0047] Figure 5 A schematic diagram of the structure of a push message sending system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0049] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0050] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0051] In order to solve the problem that the timeliness of sending push messages is not high, it is difficult to increase the number of visits to the application where the product involved in the push message is located and the number of purchases of the involved products, thereby reducing the efficiency of operating applications, the embodiment of the present invention obtains the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features in the push message during the configuration of the push message; the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message are input into the trained first neural network for matching processing, and a push message including matching copy content is output, wherein the copy content includes product information and the recommendation features of the corresponding product; after the push message is configured, the push message including the matching copy content is sent to the user terminal device.

[0052] Since the embodiment of the present invention uses the features that can reflect user needs and product recommendation information as the input of the trained first neural network during the process of configuring the push message, and uses the trained first neural network for matching processing, the copy content in the obtained push message can reflect user needs and product recommendation information. Therefore, the embodiment of the present invention accurately sets the push message with matching copy content according to the user's needs and sends it to the user-end device, thereby improving the timeliness of the push message and improving the operational efficiency of the application that provides the push message.

[0053] Since, in the process of configuring push messages, the embodiment of the present invention uses features that can reflect user needs and product recommendation information as the input of the trained first neural network, and uses the trained first neural network for matching processing, the copy content in the obtained push message can reflect user needs and product recommendation information. Therefore, the embodiment of the present invention accurately sets push messages with different copy content according to user needs and sends them to users, thereby improving the timeliness of push messages and improving the operational efficiency of applications that provide push messages.

[0054] When configuring and sending push messages, it usually includes the operating side network entity of the application involved in the push message, the push message matching entity and the push message center, wherein the operating side network entity is used to determine the recommended features of the product and the range of users to send the push message, the push message matching entity is used to fill in the recommended features according to the set copy template to obtain the push message; the push message center is used to send the obtained push message within the determined range of users to send the push message. In the scheme of the embodiment of the present invention, the features to be obtained are reset in the operating side network entity so that the user needs and the recommended features corresponding to different products can be accurately reflected. In the matching process on the push message matching entity, a first neural network is used, which is a neural network based on the attention mechanism, to configure the push message.

[0055] In the implementation of the present invention, the network entity on the operating side of the application involved in the push message can control the configuration and sending of the push message according to different scenarios and services.

[0056] Figure 1 The schematic diagram of the framework structure for configuring and sending push messages provided for an embodiment of the present invention includes an application operation-side network entity, a push message matching entity, a push message center, and a user-end device, wherein the application operation-side network entity has multiple functional units, including a scene determination unit, a recommendation feature extraction unit, a sending user range determination unit, and a recommendation feature recall unit; the user-end device includes a receiving unit and a display unit for receiving push messages.

[0057] The scenario determination unit in the network entity on the operating side of the application determines the scenario information of the application's current activity operation purpose, such as attracting new users, promoting products and supporting products, etc.; the recommendation feature extraction unit reports the recommendation feature based on the scenario information of the application's current activity operation purpose; the sending user range determination unit determines the sending user range involved in the application's current activity, and the determination basis can be based on the set rules or the set model, etc.; the recommendation feature recall unit is used to determine the recalled recommendation features based on the user-related behavior information fed back by the user-end device. In this way, the network entity on the operating side of the application determines the complete recommendation features and the sending user range, and provides them to the push message matching entity.

[0058] Here, the recommended features can also be called interest points, including reported recommended features and recalled recommended features. The process of recalling recommended features is: when the user-end device clicks on the received push message, the relevant information of the product involved in the copy content in the push message will be displayed in the display unit of the user-end device in the form of a landing page, and the user-related behavior information of the user operating the relevant information of the product in the landing page can be obtained. Of course, when the relevant information of the product involved in the copy content in the push message is displayed in the display unit in the form of a landing page, the user-end device is also matched and screened based on the user portrait and product similarity of the user-end device, and a set number of products with the highest matching degree and the highest product similarity are displayed in the form of a landing page.

[0059] Push message matching entity, after filling in the recommended features according to the set copy template, a push message is obtained; multiple copy templates can be set, after selecting one of the copy templates, the recommended features are filled in, and a push message is obtained; when matching, the copy template of the push message can also be selected according to the user click information fed back by the receiving unit of the user terminal device;

[0060] The push message center sends the received push message within the determined push message sending user range;

[0061] The receiving unit of the user terminal device receives the push message and obtains the user's click information on the push message;

[0062] The display unit of the user terminal device is used to obtain the user's click information on the push message, and then display the relevant information of the product involved in the copy content in the push message in the form of a landing page.

[0063] The solution of the embodiment of the present invention resets the features to be obtained in the operating side network entity so as to accurately reflect user needs and recommended features for different products. In the matching process on the push message matching entity, a first neural network is adopted. The first neural network is a neural network based on the attention mechanism to configure the push messages, thereby improving the timeliness of the push messages and improving the operational efficiency of the applications that provide the push messages.

[0064] Figure 2 The software module diagram of the push message configuration and sending process provided by the embodiment of the present invention includes the business demand module of push messages, the strategy opening module of push messages, the configuration upgrade module of push messages, the generation model module of the copy content in push messages, the business launch module of push messages and the platform configuration module of push messages. After the implementation of these software modules, Figure 1 The described architecture loads the coordination and sending process of push messages.

[0065] Here, the business demand module for push messages, the strategy implementation module for push messages, the business launch module for push messages and the platform configuration module for push messages are all existing modules. The sending user range determination submodule, the recommended feature selection submodule, and the recommended feature matching submodule in the configuration upgrade module for push messages are existing modules. The embodiment of the present invention improves the copy content generation submodule in the configuration upgrade module for push messages, and proposes a generation model module for the copy content in push messages, which are described in detail below.

[0066] The business requirement modules for push messaging include:

[0067] The product SKU recommendation submodule proposes a business requirement for push messages: after sending a push message including recommended products to the user-end device, determine the products obtained by the user-end device;

[0068] The landing page recommendation submodule displays the relevant information of the products involved in the copy content of the push message in the form of a landing page. If the number of related products obtained based on the similarity or relevance of the products involved is insufficient, the products that the user terminal device is interested in or the hot-selling products are displayed on the landing page; that is, a set number of related products involved in the copy content of the push message are displayed on the landing page, and the set number can be 100;

[0069] Operation strategy submodule, based on Figure 1 The architecture shown establishes the configuration and sending process of push messages;

[0070] The business goal submodule sets the business goals for push messages, such as recommending new products or promoting a certain type of product, increasing the number of daily active users of the app, etc.

[0071] The push message strategy module specifically includes:

[0072] The recall strategy submodule sets the product sales volume and product click volume within a set time, sorts the products, and determines the set number of recalled recommended products; sets the recall of similar or related recommended products based on user behavior information;

[0073] The sorting model submodule uses the first neural network provided in the embodiment of the present invention to push the text content in the message;

[0074] The diversity strategy submodule filters the products in the recommendation features, such as: 1) Filter out-of-stock products or discontinued products. According to the user's delivery address, out-of-stock products or discontinued products are filtered out; 2) Purchased filtering: filter out products that the user has purchased within a certain period; 3) Title keyword filtering: filter products with sensitive words in the title; 4) Non-compliant price filtering: filter products that do not meet the price standards, low-priced and high-priced products; 4) Season filtering: filter products that are not suitable for the season according to the season of the user's region; 5) User purchasing power filtering: filter out products that do not meet the user's purchasing power; 6) Blacklist filtering: filter products that are consistent with artificially formulated products, first-level categories, second-level categories, and third-level categories, mainly some sensitive products; 7) Same picture filtering: filter products with the same main picture; 8) Same standardized product unit (SPU) filtering: filter products with the same SPU; 9) Product filtering during the review period.

[0075] The push message integration subunit integrates the configured push messages, synchronizes them to the push message center, and sends them within the set user range.

[0076] The configuration upgrade module for push messages specifically includes:

[0077] The sending user range determination submodule performs push message sensitivity modeling and delivery value modeling based on user behavior information, and selects the sending user range of the push message based on the model;

[0078] Recommended feature selection submodule to select the pushed information;

[0079] The recommendation feature matching submodule matches the products in the push information, including recall, sorting and user-specific product matching (different users' sensitivity to products and selection of appropriate product types);

[0080] The copywriting content generation submodule uses the first neural network to generate copywriting content, including the construction of copywriting library, phrase library, and creation rule construction. Among them, the copywriting library construction: the materials of the copywriting library are mainly divided into two sources, internal and external. The internal source is accumulated through long-term operations, and the external source determines keywords based on the business. The algorithm side uses the classification algorithm to mark the copywriting and divide the copywriting according to the following dimensions, including copywriting source, industry business, style, copywriting scene, function, user group, etc.; phrase library construction: using a large amount of copywriting materials, phrases are extracted according to the rules. The phrase library needs to be constantly refreshed and optimized, and the output includes: hot word copywriting, crowd description, marketing vocabulary, advertising slogans, keyword copywriting, business vocabulary, industry copywriting, style copywriting, scene copywriting, etc.; creation rule construction: although the copywriting library can provide resources and inspiration for copywriting creation, it often cannot fully meet the push message requirements for the activity at the time, and needs to be manually reorganized or modified. Therefore, some creation rules need to be established. The creation rules need to consider the fluency, relevance and click effect of the copywriting.

[0081] The generation model module of the copy content in the push message is the focus of the embodiment of the present invention, and specifically includes:

[0082] The feature engineering submodule is mainly used to obtain feature definitions that reflect user needs and product recommendation information, including the sending-related numerical features of push messages, product identification features in push messages, and recommendation features in push messages.

[0083] The numerical features related to the sending of push messages are called dense features. They are numerical features, and their values ​​reflect their numerical meanings, including: the number of push message sending times, the number of product categories in the push messages, the number of product brands in the push messages, the number of push message receiving times, the number of product categories in the received push messages, and the number of product brands in the received push messages, or one or more combinations thereof, in different unit times;

[0084] For example: the total number of push messages sent to users in 7 days; the total number of push messages sent to users in 15 days; the total number of push messages sent to users in 30 days; the total number of third-level product categories in push messages sent to users in 7 days; the total number of third-level product categories in push messages sent to users in 15 days; the total number of third-level product categories in push messages sent to users in 30 days; the total number of product brands in push messages sent to users in 7 days; the total number of product brands in push messages sent to users in 15 days; the total number of product brands in push messages sent to users in 30 days; the total number of users clicking on push messages in 7 days; the total number of users clicking on push messages in 15 days; the total number of users clicking on push messages in 30 days; the total number of third-level product categories in push messages that users clicked on in 7 days; the total number of third-level product categories in push messages that users clicked on in 15 days; the total number of third-level product categories in push messages that users clicked on in 30 days; the total number of product brands in push messages that users clicked on in 7 days; the total number of product brands in push messages that users clicked on in 15 days; the total number of product brands in push messages that users clicked on in 30 days.

[0085] The product identification feature in the push message is called an ID feature, which reflects the product attributes, including:

[0086] The product identifiers of the most products in the push messages and / or the product identifiers of the most products in the received push messages in different units of time;

[0087] For example, the product brand logos with the most push messages sent to users in 30 days; the product brand logos with the most push messages sent to users in 15 days; the product brand logos with the most push messages sent to users in 7 days; the product brand logos with the most push messages clicked by users in 30 days; the product brand logos with the most push messages clicked by users in 15 days; the product brand logos with the most push messages clicked by users in 7 days.

[0088] The recommendation features in the push message are actually the interest point features, including: the frequency and frequency of the user terminal device accessing or / and obtaining the relevant information of the product in different unit time; the frequency and frequency of the user terminal device accessing or / and obtaining the relevant information of the product in different unit time;

[0089] For example: the content included in the user's browsing history in the past day; the frequency of the user's browsing history in the past day; the content included in the user's browsing history in the past 7 days; the frequency of the user's browsing history in the past 7 days; the content included in the user's browsing history in the past 30 days; the frequency of the user's browsing history in the past 30 days; the content included in the user's purchase history in the past 7 days; the frequency of the user's purchase history in the past 7 days; the content included in the user's purchase history in the past 30 days; the frequency of the user's purchase history in the past 30 days; the content included in the user's purchase history in the past 30 days; the frequency of the user's purchase history in the past 30 days; the content included in the user's purchase history in the past 60 days; the frequency of the user's purchase history in the past 60 days.

[0090] The label submodule is used to mark whether the push message is clicked on the user's device. If it is clicked, it is set to 1, and if it is not opened, it is set to 0;

[0091] An attention mechanism module, used to train a first neural network based on the attention mechanism and apply the already set first neural network, the training and application of the first neural network are described in detail below;

[0092] The network structure submodule is used to train the first neural network based on the second neural network, and combine the two network models as a network structure that matches the text content in the push message.

[0093] The business launch module for push messages specifically includes:

[0094] Multi-party joint debugging of sub-modules to ensure that the content launched online is logically correct and has good performance;

[0095] Whitelist test submodule, for setting and testing the user's whitelist;

[0096] The push timing modeling submodule is used to model the sending timing of push messages, such as modeling based on user activity, and setting the sending timing of push messages according to the established model;

[0097] The front-end user display submodule is used for landing pages. When the user clicks on the push message to enter the landing page, if they cannot find the desired product information, they will easily lose interest in the push message. Therefore, the relevant information related to the product in the push message displayed on the landing page is very important, and its sorting and display content are related to the push message.

[0098] The platform configuration modules for push messages include:

[0099] The message configuration center submodule is responsible for maintaining the core data model of the platform, providing the operation side with scenarios and materials for page operation configuration push messages, reducing the access cost of the operation side; and synchronizing the configuration data to the operation side in offline full and real-time incremental ways;

[0100] The message matching center submodule filters and pushes messages for each user based on the scenario and material configuration, sorts the materials based on the historical click-through rate, and recalls the materials and personalized content that the user is most likely to be interested in based on the user's recent behavior, and fills them into the push messages;

[0101] The scheduling task center submodule supports timed triggering, real-time triggering, and time-customized triggering, which closes the reach and facilitates the platform to arrange the reach plan, including triggering time and sending volume, to ensure that business goals are achieved without causing excessive pressure on downstream business systems;

[0102] Push frequency setting submodule is used to set the frequency of push messages.

[0103] The embodiment of the present invention actually resets the features to be acquired in the network entity on the operation side so as to accurately reflect the user needs and the recommended features for different products. The first neural network is used in the matching process on the push message matching entity. The first neural network is a neural network based on the attention mechanism to configure the push message. The following is a detailed description.

[0104] Figure 3 A flow chart of a method for sending a push message is provided for an embodiment of the present invention, and the specific steps thereof include:

[0105] Step 301: Obtaining the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features in the push message;

[0106] Step 302: input the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message into the trained first neural network for matching processing, and output a push message including matching copy content, wherein the copy content includes product information and the recommendation features of the corresponding product;

[0107] Step 303: Send a push message including the matching copy content to the user terminal device.

[0108] In the method, the first neural network obtained by training is a neural network based on an attention mechanism, and the inputting into the first neural network obtained by training for matching processing includes:

[0109] Processing the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message by a pairwise cross-attention mechanism in a first neural network;

[0110] The obtained vector results are classified and processed to obtain push messages corresponding to different user identifiers including matching copy content.

[0111] That is to say, only the first neural network set in the embodiment of the present invention can accurately set a push message with matching copy content according to the user's needs and the recommended features of the product.

[0112] In the method, the outputting of a push message of the matched copy content also includes:

[0113] Processing the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message by a pairwise cross-attention mechanism in a first neural network;

[0114] A second neural network is obtained by training, and the second neural network performs convolution processing on the input sending-related numerical features of the push message and the product identification features in the push message;

[0115] The vector result obtained by the convolution processing and the vector result obtained by the pairwise cross attention mechanism processing are fully connected and then classified to obtain a push message with matching copy content.

[0116] In this case, the trained second neural network has been set in the push message matching entity. In this case, the two neural network models of the first neural network and the second neural network are combined to process the input information so that the different copy contents in the obtained push messages can better reflect the user's needs and the information to be recommended, such as Figure 4 As shown, Figure 4 A schematic diagram of the structure of a combination of a first neural network and a second neural network provided in an embodiment of the present invention, wherein the second neural network adopts a convolutional neural network (CNN).

[0117] In this method, the sending-related numerical features of the push message include:

[0118] One or more combinations of the number of times push messages are sent, the number of product categories in the push messages, the number of product brands in the push messages, the number of times push messages are received, the number of product categories in the received push messages, and the number of product brands in the received push messages within different units of time;

[0119] The product identification features in the push message include:

[0120] The product identifiers of the most products in the push messages and / or the product identifiers of the most products in the received push messages in different units of time;

[0121] The recommended features in the push message include:

[0122] The frequency with which user-end devices access or / and obtain product-related information in different time units.

[0123] The method further includes:

[0124] After receiving the information of whether the push information is clicked, a corresponding label is set based on the clicked information, and is used as an input feature when subsequently determining the content of the push message, and is input into the first neural network for processing. In other words, the label of whether the push information is obtained is used as an input parameter for the training and application of the first neural network, so that the copy content matched by it is more accurate.

[0125] The method for sending a push message provided by an embodiment of the present invention is applied in a product recommendation process of an application providing products to a user terminal device;

[0126] The recommendation features in the push message are set according to the recommendation features set by the application side that provides the product for the user terminal device, and / or the recommendation features obtained when the user terminal device uses the application that provides the product. In other words, the recommendation features input into the first neural network include two parts, one part is set by the operation side when promoting the product activities, and the other part is the recommendation features recalled based on the user behavior information, so that it better reflects the user needs and the needs of the promotion activities of the operation side.

[0127] Figure 5 A schematic diagram of a push message sending system provided by an embodiment of the present invention includes: an acquisition unit, a matching unit and a sending unit, wherein:

[0128] An acquisition unit, used to acquire the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features in the push message;

[0129] a matching unit, configured to input the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message into the trained first neural network for matching processing, and output a push message with matching copy content, wherein the copy content includes product information and the recommendation features of the corresponding product;

[0130] The sending unit is used to send a push message of the matching copy content to the user terminal device.

[0131] In the system, the matching unit is further used to process the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message in a pairwise cross-attention mechanism in the trained first neural network;

[0132] The obtained vector results are classified and processed in the trained first neural network to obtain a push message with matching copy content.

[0133] In the system, the matching unit is further used to process the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message in a pairwise cross-attention mechanism in the first neural network;

[0134] A second neural network is obtained by training, and the second neural network performs convolution processing on the input sending-related numerical features of the push message and the product identification features in the push message;

[0135] The vector result obtained by the convolution processing and the vector result obtained by the pairwise cross attention mechanism processing are fully connected and then classified to obtain a push message with matching copy content.

[0136] In this system, the sending-related numerical features of the push message include:

[0137] One or more combinations of the number of times push messages are sent, the number of product categories in the push messages, the number of products in the push messages, the number of times push messages are received, the number of product categories in the received push messages, and the number of products in the received push messages within different units of time;

[0138] The product identification features in the push message include:

[0139] The product identifiers of the most products in the push messages and / or the product identifiers of the most products in the received push messages in different units of time;

[0140] The content features of the push message include:

[0141] The frequency with which user-end devices access or / and obtain product-related information in different time units.

[0142] The push message sending system provided in an embodiment of the present invention is applied in the process of recommending products by an application that provides products to a user terminal; the recommendation features in the push message are set based on the recommendation features set on the application side that provides products to the user terminal, and / or the recommendation features obtained when the user terminal uses the application that provides products.

[0143] The method and system provided in the embodiment of the present invention can be specifically applied in an application that provides low-priced products to a user terminal, and the user who uses the user terminal device is a sinking user.

[0144] When an application that provides products on the user side wants to send a push message for product recommendation, the embodiment of the present invention redefines and obtains the features input into the trained first neural network in combination with the sending scenario and sending location, so that it can reflect user needs and recommended features of the product, and trains the first neural network based on the attention mechanism, so that the final push message including the copy content can also fully reflect the user needs and recommended features of the product, thereby realizing intelligent matching of user-product-copy content, improving the DAU indicator of the application and improving the user experience.

[0145] The embodiment of the present application also provides a computer-readable medium, the computer-readable storage medium stores instructions, and the instructions can execute the steps in the method for sending a push message as described above when executed by a processor. In practical applications, the computer-readable medium may be included in the device / device / system described in the above embodiment, or it may exist alone without being assembled into the device / device / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the implementation method of the sharing display described in the above embodiments can be implemented. According to the embodiment disclosed in the present application, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above, but it is not used to limit the scope of protection of the present application. In the embodiment disclosed in the present application, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by an instruction execution system, a device or a device or used in combination with it.

[0146] The embodiment of the present invention further provides an electronic device, in which the device for implementing the method of the embodiment of the present application can be integrated. Specifically:

[0147] The electronic device may include one or more processors with processing cores, one or more computer-readable storage media memories, and a computer program stored in the memories and executable on the processors. When executing the program in the memory, the above-mentioned method for implementing sales by the automatic vending device may be implemented.

[0148] Specifically, in practical applications, the electronic device may further include components such as a power supply, an input unit, and an output unit. Those skilled in the art will appreciate that the structure of the electronic device in the embodiment of the present invention does not limit the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. Among them:

[0149] The processor is the control center of the electronic device. It uses various interfaces and lines to connect the various parts of the entire electronic device. It performs various functions of the server and processes data by running or executing software programs and / or modules stored in the memory, and calling data stored in the memory, thereby monitoring the electronic device as a whole.

[0150] The memory can be used to store software programs and modules, that is, the above-mentioned computer-readable storage medium. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the server, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory may also include a memory controller to provide the processor with access to the memory.

[0151] The electronic device also includes a power supply for supplying power to various components, which can be logically connected to the processor through a power management system, so as to manage charging, discharging, and power consumption through the power management system. The power supply can also include any components such as one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators.

[0152] The electronic device may further include an input unit, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.

[0153] The electronic device may further include an output unit, which may be used to display information input by a user or information provided to a user and various graphical user interfaces, which may be composed of graphics, texts, icons, videos and any combination thereof.

[0154] The flow chart and block diagram in the accompanying drawings of the present application show the possible architecture, function and operation of the system, method and computer program product according to the various embodiments disclosed in the present application. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the box can also occur in the order of the standards in different figures. For example, the boxes represented by two connections 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 flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0155] Those skilled in the art will appreciate that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present application. In particular, without departing from the spirit and teachings of the present application, the features described in the various embodiments and / or claims of the present application may be combined and / or combined in a variety of ways, and all of these combinations and / or combinations fall within the scope disclosed in the present application.

[0156] Specific embodiments are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas, and is not used to limit the present application. For those skilled in the art, changes can be made in the specific implementation methods and application scopes according to the ideas, spirits and principles of the present invention, and any modifications, equivalent substitutions, improvements, etc. made therein should be included in the scope of protection of this application.

Claims

1. A method for sending a push message, characterized in that: The method comprises: Obtaining the sending-related numerical features of push messages, product identification features in push messages, and recommendation features in push messages; Inputting the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message into the trained first neural network for matching processing, and outputting a push message including matching copy content, wherein the copy content includes product information and the recommendation features of the corresponding product; Sending a push message including matching copy content to a user terminal device; The sending-related numerical characteristics of the push message include: In different unit times, one or more combinations of the number of times push messages are sent, the number of product categories in push messages, the number of product brands in push messages, the number of times push messages are received, the number of product categories in received push messages, and the number of product brands in received push messages.

2. The method according to claim 1, characterized in that The first neural network obtained by training is a neural network based on an attention mechanism, and the inputting into the first neural network obtained by training for matching processing includes: Processing the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message by a pairwise cross-attention mechanism in a first neural network; The obtained vector results are classified and processed to obtain push messages including matching copy content.

3. The method according to claim 1, characterized in that The output is different push messages corresponding to different user identifiers, including different copy contents, and also includes: Processing the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message by a pairwise cross-attention mechanism in a first neural network; A second neural network is obtained by training, and the second neural network performs convolution processing on the input sending-related numerical features of the push message and the product identification features in the push message; The vector result obtained by the convolution processing and the vector result obtained by the pairwise cross attention mechanism processing are fully connected and then classified to obtain a push message including matching copy content.

4. The method according to claim 1, 2 or 3, characterized in that: The product identification features in the push message include: The product identifiers of the most products in the push messages and / or the product identifiers of the most products in the received push messages in different units of time; The recommended features in the push message include: The frequency with which user-end devices access or / and obtain product-related information in different time units.

5. The method according to claim 1, characterized in that The method for sending the push message is applied in the process of recommending products to the application providing products to the user end; The recommendation features in the push message are set according to the recommendation features set by the application side that provides the product to the user terminal, and / or the recommendation features obtained when the user terminal uses the application that provides the product.

6. A push message sending system, characterized in that: The system comprises: an acquisition unit, a matching unit and a sending unit, wherein: An acquisition unit, used to acquire the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features in the push message; a matching unit, configured to input the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message into the trained first neural network for matching processing, and output a push message including matching copy content, wherein the copy content includes product information and the recommendation features of the corresponding product; A sending unit, used to send a push message including the matching copy content to a user terminal device; The sending-related numerical characteristics of the push message include: In different unit times, one or more combinations of the number of times push messages are sent, the number of product categories in push messages, the number of product brands in push messages, the number of times push messages are received, the number of product categories in received push messages, and the number of product brands in received push messages.

7. The system according to claim 6, characterized in that The matching unit is further used to process the sending-related numerical features of the push message, the product identification features in the push message, and the recommendation features of the push message in a pairwise cross-attention mechanism in the trained first neural network; The obtained vector results are classified and processed in the trained first neural network to obtain a push message including matching copy content.

8. The system according to claim 6, characterized in that The sending-related numerical characteristics of the push message include: One or more combinations of the number of times push messages are sent, the number of product categories in the push messages, the number of products in the push messages, the number of times push messages are received, the number of product categories in the received push messages, and the number of products in the received push messages within different units of time; The product identification features in the push message include: The product identifiers of the most products in the push messages and / or the product identifiers of the most products in the received push messages in different units of time; The content features of the push message include: The frequency with which user-end devices access or / and obtain product-related information in different time units.

9. A device for implementing the sending of a push message, characterized in that: include: Memory; and a processor coupled to the memory, wherein the processor is configured to execute the push message sending method according to any one of claims 1 to 5 based on instructions stored in the memory.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the method for sending a push message as claimed in any one of claims 1 to 5 is implemented.

Images