Information pushing method and device, electronic equipment and storage medium
By calling the pre-trained target information push model during the information push process to recall information and cache this information in the storage device, the problem of low correlation caused by simple recall methods in the prior art is solved, and the response speed of information push is improved and user experience.
Patent Information
- Application Number
- CN202311028402.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-08-15
- Publication Date
- 2025-02-21
AI Technical Summary
The existing information push method is simple to recall, resulting in low correlation between recall results and user interest.
By receiving the client's request, the pre-trained target information push model is called for information recall processing, the target push information is obtained, and this information is cached in the storage device.
It improves the response speed and user experience of information push, enhances the correlation between target push information and user preferences, and provides push information more in line with user preferences.
Smart Images
Figure CN119493898A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to an information push method, device, electronic device and storage medium. Background Art
[0002] With the development of the Internet, users will be exposed to a large amount of information that they are interested in or not interested in. When users browse the Internet, it is necessary to display the information of interest to them in real time. However, the existing information push method only calculates and recalls the information when a recall request is received, and the recall model is simple, resulting in a low correlation between the recall results and the user's interests. Summary of the invention
[0003] The present application provides an information push method, device, electronic device, computer-readable storage medium and computer program product to solve the problem that the recall method is simple and leads to non-optimal recall results during the information push process. The technical solution of the present application is as follows:
[0004] According to the first aspect of an embodiment of the present application, an information push method is provided, including: receiving a first request sent by a client; calling a pre-trained target information push model to perform information recall processing according to the first request, obtaining the target push information of the client, and caching the target push information in a storage device; receiving a second request from the client, reading the target push information from the storage device according to the second request, and sending one or more of the target push information to the client.
[0005] In one embodiment of the present application, the pre-trained target information push model is called according to the first request to perform information recall processing to obtain the target push information of the client, including: obtaining the user information corresponding to the client, the current user behavior information of the client and the candidate push information; inputting the user information, user behavior information and the candidate push information into the target push information model, and the target push information model performs feature extraction to obtain target feature information; based on the target feature information, preference prediction is performed on the candidate push information to recall the target push information.
[0006] In one embodiment of the present application, the user information, user behavior information and the candidate push information are input into the target push information model, and the target push information model performs feature extraction to obtain target feature information, including: performing feature extraction on the user information to obtain user feature information; performing feature extraction on the candidate push information to obtain feature information; based on the user behavior information, obtaining cross-feature information between the client and the candidate push information; and splicing the user feature information, the feature information and the cross-feature information to obtain the target feature information.
[0007] In one embodiment of the present application, caching the target push information in a storage device includes: associating the identification information of the client with the target push information, and storing the target push information in the storage device.
[0008] In one embodiment of the present application, the second request includes identification information of the client, and reading the target push information from the storage device according to the second request includes: based on the identification information in the second request, obtaining push information associated with the identification information from the storage device as the target push information of the client.
[0009] In one embodiment of the present application, caching the target push information in a storage device includes: allocating a storage area in the storage device to the client, associating address information of the storage area with identification information of the client, and storing the target push information in the storage area.
[0010] In one embodiment of the present application, the second request includes identification information of the client, and reading the target push information from the storage device according to the second request includes: based on the identification information in the second request, obtaining address information associated with the identification information from the storage device, and accessing the storage area corresponding to the client based on the address information to obtain the target push information.
[0011] In one embodiment of the present application, after reading the target push information from the storage device according to the second request, the method further includes: releasing the storage area occupied by the target push information.
[0012] In one embodiment of the present application, the pre-training process of the target information push model includes: obtaining training data, the training data including a sample client, sample push information corresponding to the sample client, user feedback information of the sample push information, and label information of the sample push information; performing offline pre-training on the initial information push model based on the training data, and the information push model performs sample user feature information of the sample client and sample feature information of the sample push information, as well as sample cross-feature information of the sample client and the sample push information; based on the sample user feature information and the sample feature information, as well as the sample cross-feature information, obtaining a prediction result of the information push model; based on the label information and the prediction result, adjusting the information push model and returning to continue training until the training is completed to obtain the target information push model.
[0013] According to the first aspect of an embodiment of the present application, another information push method is provided, including: in response to satisfying an asynchronous recall condition, sending a first request to a server, the first request being used to request the server to call a pre-trained target information push model to perform information recall processing, so as to obtain the target push information corresponding to the client and cache it in a storage device; in response to detecting a recall result access operation, sending a second request to the server, the second request being used to request the server to read the target push information from the storage device; and receiving one or more of the target push information returned by the server.
[0014] In one embodiment of the present application, in response to satisfying the asynchronous recall condition, sending a first request to the server includes: monitoring user behavior, and determining that the asynchronous recall condition is satisfied in response to a change in the user behavior.
[0015] According to the second aspect of an embodiment of the present application, an information push device is provided, including: a receiving module, used to receive a first request sent by a client; a recall module, used to call a pre-trained target information push model to perform information recall processing according to the first request, obtain the target push information of the client, and cache the target push information in a storage device; a transceiver module, used to receive a second request from the client, read the target push information from the storage device according to the second request, and send one or more of the target push information to the client.
[0016] In one embodiment of the present application, the recall module is also used to: obtain user information corresponding to the client, current user behavior information of the client and candidate push information; input the user information, user behavior information and candidate push information into the target push information model, and perform feature extraction by the target push information model to obtain target feature information; based on the target feature information, perform preference prediction on the candidate push information to recall the target push information.
[0017] In one embodiment of the present application, the recall module is also used to: perform feature extraction on the user information to obtain user feature information; perform feature extraction on the candidate push information to obtain feature information; based on the user behavior information, obtain cross-feature information between the client and the candidate push information; and splice the user feature information, the feature information and the cross-feature information to obtain the target feature information.
[0018] In one embodiment of the present application, the recall module is further used to: associate the identification information of the client with the target push information, and store the target push information in the storage device.
[0019] In one embodiment of the present application, the transceiver module is further used to: based on the identification information in the second request, obtain push information associated with the identification information from the storage device as target push information of the client.
[0020] In one embodiment of the present application, the recall module is further used to: allocate a storage area in the storage device to the client, associate the address information of the storage area with the identification information of the client, and store the target push information in the storage area.
[0021] In one embodiment of the present application, the transceiver module is also used to: based on the identification information in the second request, obtain address information associated with the identification information from the storage device, and access the storage area corresponding to the client based on the address information to obtain the target push information.
[0022] In one embodiment of the present application, the transceiver module is further used to: release the storage area occupied by the target push information.
[0023] In one embodiment of the present application, the recall module is also used to: obtain training data, the training data including a sample client, sample push information corresponding to the sample client, user feedback information of the sample push information, and label information of the sample push information; perform offline pre-training on an initial information push model based on the training data, and the information push model performs pre-training on sample user feature information of the sample client and sample feature information of the sample push information, as well as sample cross-feature information of the sample client and the sample push information; obtain a prediction result of the information push model based on the sample user feature information and the sample feature information, as well as the sample cross-feature information; adjust the information push model based on the label information and the prediction result and return to continue training until the training is completed to obtain the target information push model.
[0024] According to the second aspect of an embodiment of the present application, another information push device is provided, including: a first sending module, used to send a first request to a server in response to satisfying an asynchronous recall condition, the first request being used to request the server to call a pre-trained target information push model to perform information recall processing, so as to obtain the target push information corresponding to the client and cache it in a storage device; a second sending module, used to send a second request to the server in response to detecting a recall result access operation, the second request being used to request the server to read the target push information from the storage device; and a receiving module, used to receive one or more of the target push information returned by the server.
[0025] In one embodiment of the present application, the first sending module is further used to: monitor user behavior, and in response to changes in the user behavior, determine whether the asynchronous recall condition is satisfied.
[0026] According to a third aspect of an embodiment of the present application, there is provided an electronic device, comprising a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to implement the steps of the method of the first aspect of the embodiment of the present application.
[0027] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which computer program instructions are stored. When the program instructions are executed by a processor, the steps of the method of the first aspect of the embodiment of the present application are implemented.
[0028] According to a fifth aspect of an embodiment of the present application, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor of an electronic device, the steps of the method of the first aspect of the embodiment of the present application are implemented.
[0029] The technical solution provided by the embodiment of the present application brings at least the following beneficial effects: the server receives the first request sent by the client, calls the target information push model to predict the target push information according to the first request, and stores the target push information in the storage device, based on the first request sent by the client when it perceives the change in user behavior information, improves the model's exploration of user preferences, improves the correlation between the target push information and user preferences, and helps to provide push information that is more in line with user preferences. When the server receives the second request, it reads the target push information from the storage device and sends one or more of the target push information to the client, thereby improving the response speed of the information push and the user's experience. The present application can predict the target push information based on the model in advance, and through the mechanism of storing the target push information in the storage device, the purpose of asynchronously obtaining the target push information is achieved.
[0030] 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
[0031] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the application.
[0032] Figure 1 The figure is a flow chart of an information push method according to an exemplary embodiment.
[0033] Figure 2 The figure is a flowchart of an information pushing method according to another exemplary embodiment.
[0034] Figure 3 The present invention is a flowchart of a pre-training process of a target information push model in an information push method according to an exemplary embodiment.
[0035] Figure 4 It is a structural diagram of a target information push model according to an exemplary embodiment.
[0036] Figure 5 The figure is a flow chart of an information pushing method according to another exemplary embodiment.
[0037] Figure 6 The figure is a structural block diagram of an information push system according to an exemplary embodiment.
[0038] Figure 7 The figure is a flow chart of an information pushing method according to another exemplary embodiment.
[0039] Figure 8 The figure is a structural block diagram of an information pushing device according to an exemplary embodiment.
[0040] Fig. 9 is a structural block diagram of an information pushing device according to another exemplary embodiment.
[0041] Fig.10 The present invention is a block diagram of an electronic device according to an exemplary embodiment.
[0042] Fig.11 The present invention is a structural block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION
[0043] In order to enable ordinary persons in the art to better understand the technical solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.
[0044] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application 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 application described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims.
[0045] The acquisition, storage, use, and processing of user information in this application's technical solution must comply with relevant laws and regulations, respect user privacy and legal rights, and must not infringe on user privacy. At the same time, adequate protection measures must be provided to protect the security of users' personal information.
[0046] Figure 1 is a flow chart of an information push method according to an exemplary embodiment. Figure 1 As shown, the information push method of the embodiment of the present application includes but is not limited to the following steps.
[0047] S101, receiving a first request sent by a client.
[0048] It should be noted that the execution subject of the information push method provided in the embodiment of the present application is an electronic device, and the electronic device includes a server and a client. Among them, the server includes the near-line recall part and the offline training part of the information push method; the client includes the online access part of the information push method. The electronic device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook or a personal digital assistant (PDA), etc., and the non-mobile electronic device can be a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present application.
[0049] In some implementations, the electronic device can push information to the user based on the user's preferences. Since the user's preferences are dynamically changing, the electronic device needs to perceive the user's behavior information to determine the user's preferences and ensure that the target push information is highly correlated with the user's preferences. When the client of the electronic device perceives that the user's behavior has changed, such as generating new browsing, purchasing, and other behavior information, a first request is generated, and the server receives the first request sent by the client. Among them, the first request can call a pre-trained target information push model to obtain the target push information.
[0050] S102: Call a pre-trained target information push model to perform information recall processing according to the first request, obtain the target push information of the client, and cache the target push information in a storage device.
[0051] In some implementations, the first request includes a calling instruction for calling a pre-trained target push information model. The pre-trained target information push model can recall candidate push information and determine the target push information of the client from the candidate push information.
[0052] Optionally, the user's behavior can be analyzed to determine the user's preference, and the preference prediction of the candidate push information can be performed to predict the push information that the user is most interested in as the target push information of the client. Optionally, the user's preference score for the candidate push information can be predicted, and the candidate push information with the highest score can be used as the target push information. The target push information includes one or more candidate push information.
[0053] In some implementations, a large amount of target push information is obtained, and in order to achieve the purpose of asynchronous push, a storage device can be configured, through which the target push information obtained by the storage device cache model is waited for asynchronous push. That is to say, after the target push information is obtained, the target push information can be stored in the storage device. Optionally, the target push information can be stored in the memory of the electronic device. It should be noted that due to the limited memory space of the electronic device, the storage device can optionally include but is not limited to memory, hard disk, cloud storage device, storage server, etc., and the type of storage device is not limited in the embodiments of the present application.
[0054] Optionally, an association relationship between the client and the target push information may be established during storage, so that related target push information may be pushed to the client later.
[0055] Optionally, the first request includes identification information of the client. The identification information of the client may be login information of the user. When storing, the storage area of the target push information may be associated with the identification information of the client, so as to obtain the target push information corresponding to the client from the storage area based on the identification information of the client.
[0056] S103, receiving a second request from the client, reading target push information from the storage device according to the second request, and sending one or more of the target push information to the client.
[0057] In some implementations, the client generates a second request based on the user's access operation, such as a click or a pull-down operation, and the second request is used to push the target push information to the user. The server receives the second request sent by the client, and the second request includes a read instruction. The server reads the target push information associated with the client from the storage device and sends one or more of the target push information to the client.
[0058] Optionally, the second request includes identification information of the client, and based on the identification information of the client, the target push information associated with the client is read from the storage device, and one or more of the target push information is sent to the client.
[0059] In the information push method provided in the embodiment of the present application, the server receives a first request sent by the client, calls the target information push model to predict the target push information according to the first request, and stores the target push information in a storage device. Based on the first request sent when the client perceives a change in user behavior information, the model's exploration of user preferences is improved, and the relevance of the target push information to the user's preferences is improved, which helps to provide push information that is more in line with user preferences. When the server receives the second request, it reads the target push information from the storage device and sends one or more of the target push information to the client, thereby improving the response speed of the information push and the user's experience. The present application can predict the target push information based on the model in advance, and through the mechanism of storing the target push information in the storage device, the purpose of asynchronously obtaining the target push information is achieved.
[0060] Figure 2 is a flow chart of an information push method according to an exemplary embodiment. Figure 2 As shown, the information push method of the embodiment of the present application includes but is not limited to the following steps.
[0061] S201, receiving a first request sent by a client, and calling a pre-trained target information push model according to the first request.
[0062] In the embodiment of the present application, the implementation method of step S201 can be implemented by any method in the embodiments of the present application, which is not limited here and will not be repeated.
[0063] S202, obtaining user information corresponding to the client, current user behavior information of the client, and candidate push information.
[0064] In an embodiment of the present application, user information includes information such as the user's age, gender, region and historical behavior; user behavior information includes new browsing, purchasing and other information; candidate push information includes information such as the identification, category, brand, title, model, color and other information of the push information.
[0065] In some implementations, the user information is stored in a user information storage, and the candidate push information is stored in a material information storage. When the server receives the first request sent by the client, it sends the user information and candidate push information corresponding to the client to the user information storage and the material information storage.
[0066] S203: input the user information, user behavior information and candidate push information into a target push information model, and extract features from the target push information model to obtain target feature information.
[0067] S204: Based on the target feature information, a preference prediction is performed on the candidate push information to recall the target push information.
[0068] In some implementations, in order to push information to the client, the server needs to obtain the user feature information corresponding to the client and the feature information of the candidate push information. Furthermore, in order to improve the accuracy of the push, the cross feature information between the client and the candidate push information can be obtained. Optionally, the server can perform information recall based on the user feature information, the feature information of the candidate push information, and the cross feature information to obtain the target push information corresponding to the client.
[0069] Optionally, the user information, user behavior information and candidate push information are input into a target push information model, and the target push information model extracts features of the user information, user behavior information and candidate push information through a feature extractor to obtain target feature information.
[0070] Optionally, feature extraction is performed on user information to obtain user feature information; feature extraction is performed on candidate push information to obtain feature information; and based on user behavior information, cross feature information between the client and the candidate push information is obtained. It is understandable that the cross feature information can be obtained by combining the user feature information with the candidate push information to indicate the association between the user and the candidate push information. For example, the cross feature information can indicate whether the user has browsed the candidate push information.
[0071] Furthermore, the user characteristic information, the characteristic information and the cross characteristic information are spliced to obtain the target characteristic information. Optionally, the user characteristic information, the characteristic information and the cross characteristic information can be spliced in a weighted manner to obtain the target characteristic information.
[0072] In some implementations, the target push information model can predict the user's preference for candidate push information based on target feature information to determine the target push information.
[0073] Optionally, the user's preference score for the candidate push information can be predicted based on the target feature information, and one or more candidate push information with the highest score can be used as the target push information. Optionally, the correlation between the user's preference and the candidate push information can be predicted based on the target feature information, and the candidate push information can be sorted based on the prediction result, and one or more candidate push information with high correlation can be used as the target push information.
[0074] S205: Cache the target push information in the storage device.
[0075] In the embodiment of the present application, the target push information of the corresponding client can be obtained according to the identification information of the client, and the target push information can be stored in the storage device. The identification information of the client can determine the user information of the client, and optionally, the identification information of the client can be determined based on the user identifier.
[0076] Optionally, the server may associate the identification information of the client with the target push information, and store the target push information in a storage device.
[0077] Optionally, the server may allocate a storage area in a storage device to the client, associate address information of the storage area with identification information of the client, and store the target push information in the storage area.
[0078] S206, receiving a second request from the client, reading target push information from the storage device according to the second request, and sending one or more of the target push information to the client.
[0079] In some implementations, the client may generate a second request based on the user access operation, the second request including the identification information of the client, the server receives the second request from the client, reads the target push information from the storage device based on the second request, and sends one or more candidate push information in the target push information to the client. Directly reading the target push information from the storage device according to the user's access operation improves the response speed of information push and helps to provide recommendation information that better meets the user's preferences.
[0080] It can be understood that the identification information of the client is associated with the target push information; the identification information of the client is associated with the address information of the storage device. Optionally, based on the identification information in the second request, the push information associated with the identification information can be obtained from the storage device as the target push information of the client. Optionally, based on the identification information in the second request, the address information associated with the identification information can be obtained from the storage device, and the storage area corresponding to the client can be accessed based on the address information to obtain the target push information.
[0081] In some implementations, after reading the target push information from the storage device, the server releases the storage area occupied by the target push information to provide more storage space. Releasing storage space also helps to keep data organized and tidy, improve data management efficiency, and facilitate faster reading of the target push information from the storage device.
[0082] In the information push method provided in the embodiment of the present application, the server receives the first request sent by the client, and obtains the feature information of the user and the candidate push information, as well as the cross-feature information according to the first request, and performs preference prediction on the candidate push information through the target information push model based on the cross-feature information to obtain the target push information, thereby improving the model's exploration of user preferences, improving the relevance of the target push information with the user preferences, and helping to provide push information that is more in line with the user's preferences. The target push information is stored in a storage device, reducing the complex reasoning process. When the server receives the second request, it can directly read the target push information from the storage device and send one or more of the target push information to the client, thereby improving the response speed of the information push and the user's experience.
[0083] Based on the above embodiments, the present application embodiment explains the pre-training process of the target information push model. Figure 3 As shown, the pre-training process of the target information push model includes the following steps.
[0084] S301, obtaining training data, where the training data includes a sample client, sample push information corresponding to the sample client, user feedback information of the sample push information, and label information of the sample push information.
[0085] In some implementations, the information push model is pre-trained by an offline training part of the server, and the information push model training includes modules such as a user feedback information collector, a training data generator, and a model training module. The information push model may use a deep model.
[0086] Optionally, the training data generator may obtain training data such as sample clients, sample push information corresponding to the sample clients, and label information of the sample push information from the user feature storage and the material information storage. Optionally, the user feedback information collector may record feedback information of the user after receiving the candidate push information as user feedback information of the sample push information.
[0087] S302, performing offline pre-training on the initial information push model based on the training data, and the information push model performs pre-training on sample user feature information of the sample client and sample feature information of the sample push information, as well as sample cross-feature information of the sample client and the sample push information.
[0088] In some implementations, the pre-trained target information push model extracts features from the training data through a feature extractor to obtain sample user feature information of the sample client and sample feature information of the sample push information, as well as sample cross-feature information of the sample client and the sample push information.
[0089] By way of example, sample user characteristic information may include user account, age, gender, historical behavior (such as browsing, purchasing, and collecting), and region; sample characteristic information of sample push information may include information identification, category (such as electronic products, clothing products), title information, and other information; sample cross-characteristic information may include whether the user has browsed the push information.
[0090] S303, based on the sample user feature information and the sample feature information, as well as the sample cross feature information, obtain the prediction result of the information push model.
[0091] In some implementations, the sample user feature information and the sample feature information, as well as the sample cross feature information, may be concatenated to obtain the target sample feature information, and input into the model training module to train the information push model to obtain the prediction result of the information push model, wherein the prediction result includes one or more sample push information.
[0092] S304: Based on the label information and the prediction results, the information recommendation model is adjusted and returned to continue training until the training is completed to obtain the target information push model.
[0093] During the training process, the difference between the label information and the prediction result can be obtained, the difference can be calculated to determine the correction gradient, and then the model parameters of the information push model can be adjusted based on the correction gradient, and the adjusted information push model can be trained based on the subsequent training data until the training end condition is met to obtain the target information push model. Optionally, the loss function of the model can be calculated based on the error between the label information of the sample push information and the prediction result, the information push model can be adjusted based on the loss function, and the adjusted information push model can be trained based on the subsequent training data until the training end condition is met to obtain the target information push model.
[0094] It is understandable that the training end condition may be that the training time of the information push model reaches a set value. The training end condition may also be that the correlation of the prediction result output by the information push model reaches a set value.
[0095] In the information push method provided in the embodiment of the present application, the server trains the pre-trained information push model based on the sample user feature information of the sample client and the sample feature information of the sample push information, as well as the sample cross-feature information of the sample client and the sample push information, so as to improve the model's exploration of user preferences. The model is corrected based on the label information of the sample push information and the model prediction results, so as to improve the accuracy and reliability of the model prediction results.
[0096] like Figure 4The structural diagram of the information push model shown, wherein the information push model includes: an activation unit, a cross-feature extraction unit, a feature extraction unit, a weighting module, a convolution module, a normalization module and a scoring module. Among them, the cross-feature extraction unit connects the activation unit and the feature extraction unit. The cross-feature extraction unit also includes a cross-feature extraction stage 1, a cross-feature extraction stage 2, ..., a cross-feature extraction stage n for cross-feature extraction. The cross-feature extraction unit is connected to the weighting module, the weighting module is connected to the convolution module, the convolution module is connected to the normalization module, and the normalization module is connected to the scoring module.
[0097] The behavior information of the sample user is input into the activation unit. In the activation unit, the feature vectors of the extracted multiple sample user behaviors are outer-producted, and the results of the outer-products are connected to obtain a one-dimensional feature vector as the feature information of the sample user behavior. The sample push information is input into the feature extraction unit to obtain the feature information of the sample push information. The feature information of the sample user behavior and the feature information of the sample push information are input into the cross-feature extraction unit together to extract the sample cross-feature information. Among them, the cross-feature extraction unit performs multiple cross-feature extractions, and weights the extracted multiple sample cross-feature information to obtain the weighted average of the sample cross-feature information. At the same time, the sample user information (age, gender) is convolved with the weighted average of the sample cross-feature information to obtain the fusion expression of the sample cross-feature information. And based on the set activation function, normalization processing is performed to obtain a value in (0,1). The closer the value is to 1, the higher the user's preference score for the push information is. Based on the score, the prediction result (that is, the prediction label) of the sample push information can be output. Furthermore, based on the prediction results and label information of the sample push information, the model parameters are adjusted, and the adjusted information push model is continuously trained based on subsequent training data until the training end condition is met to obtain the target information push model.
[0098] Figure 5 is a flow chart of an information push method according to an exemplary embodiment. Figure 5 As shown, the information push method of the embodiment of the present application includes but is not limited to the following steps.
[0099] S501: In response to satisfying an asynchronous recall condition, sending a first request to a server.
[0100] Among them, the first request is used to request the server to call the pre-trained target information push model to perform information recall processing, so as to obtain the target push information corresponding to the client and cache it in the storage device.
[0101] It should be noted that the target information push model can be adopted Figure 4The training method of the information push model shown is obtained and will not be repeated here.
[0102] In some implementations, the client monitors user behavior, determines that the asynchronous recall condition is met in response to changes in user behavior, and sends a first request to the server. The changes in user behavior may be new browsing, collection, purchase, and other behaviors.
[0103] Optionally, the first request includes a call instruction for requesting the server to call a pre-trained target information push model. The pre-trained target information push model can recall candidate push information, determine the client's target push information from the candidate push information, and then cache the recalled target push information in a storage device.
[0104] Optionally, the first request may include identification information of the client. The identification information of the client may be login information of the user. When storing, the storage area of the target push information may be associated with the identification information of the client, so as to obtain the target push information corresponding to the client from the storage area based on the identification information of the client.
[0105] S502: In response to detecting a recall result access operation, a second request is sent to the server, where the second request is used to request the server to read target push information from a storage device.
[0106] In some implementations, the client monitors the user operation and sends a second request to the server in response to the user performing an access operation. The second request may request the server to read the target push information corresponding to the client from the storage device. The target push information includes one or more push information.
[0107] Optionally, the second request includes identification information of the client, and based on the identification information of the client, the server may be requested to read target push information associated with the client from the storage device.
[0108] S503: Receive one or more target push information returned by the server.
[0109] In some implementations, the client receives one or more push information in the target push information read by the server from the storage device and displays it to the user.
[0110] In the information push method provided in the embodiment of the present application, the client sends a first request to the server based on the user's behavior changes. The first request is used to request the server to call a pre-trained target information push model to perform information recall processing. The first request sent when the client perceives the change in user behavior information improves the model's exploration of user preferences, improves the relevance of target push information and user preferences, and helps to provide push information that is more in line with user preferences. Storing the target push information in a storage device reduces the complex reasoning process. When the user performs an access operation, a second request is sent to the server, and the target push information can be directly read from the storage device, and one or more of the target push information is sent to the client, which improves the response speed of the information push and the user's experience. The present application can predict the target push information based on the model in advance, and through the mechanism of storing the target push information in the storage device, the purpose of asynchronously obtaining the target push information is achieved.
[0111] like Figure 6 As shown in the structural block diagram of the information push system, the offline part and the near-line part are executed by the server, and the online part is executed by the client.
[0112] In the near-line recall part, the user behavior change is obtained through the user behavior sensor, and the user behavior is sent to the user information storage and material information storage. The user information storage stores all user information, including the user's age, gender, region and historical behavior. At the same time, the material information storage stores all candidate push information, including information identification, category, brand, title, model, color, etc. The user information, user behavior information and candidate push information are input into the pre-trained target push information model, feature extraction is performed, the target feature information is obtained, and the user's preference score for the candidate push information is estimated. The candidate push information with the highest score is selected as the target push information and stored in the memory.
[0113] In the offline training part, first access the material information storage and user feature storage to obtain material information and user information respectively, and perform feature extraction. Then extract cross features (such as whether the user has browsed the material), concatenate the feature information, user features and cross features of the candidate push information to obtain the target feature information, and input it into the training data generator. At the same time, collect user feedback information and input it into the training data generator. The training data generator generates training data with the received target feature information and user feedback information to train the pre-trained information push model.
[0114] In the online access part, an access operation of the user, such as a click or a pull-down operation, is received, a second request is generated, one or more push information in the target push information is read from the storage device, and pushed to the user.
[0115] Figure 7 is a flow chart of an information push method according to an exemplary embodiment. Figure 7 As shown, the information push method of the embodiment of the present application includes but is not limited to the following steps.
[0116] S701, in response to satisfying an asynchronous recall condition, the client sends a first request to the server.
[0117] S702: The server calls a pre-trained target information push model according to the first request.
[0118] S703, the server obtains the user information corresponding to the client, the current user behavior information of the client, and the candidate push information.
[0119] S704, the server inputs the user information, user behavior information and candidate push information into the target push information model, and the target push information model performs feature extraction to obtain target feature information, and based on the target feature information, performs preference prediction on the candidate push information to recall the target push information.
[0120] S705, the server pushes the cache target information to the storage device.
[0121] S706: In response to detecting the recall result access operation, the client sends a second request to the server.
[0122] S707: The server reads the target push information from the storage device according to the second request.
[0123] S708: Receive one or more target push information returned by the server.
[0124] In the information push method provided in the embodiment of the present application, the client sends a first request to the server based on the user's behavior changes. The server calls the pre-trained target information push model to perform information recall processing based on the first request. Based on the first request sent when the client perceives the change in user behavior information, the model's exploration of user preferences is improved. The target push information is stored in a storage device, which reduces the complex reasoning process. When the user performs an access operation, a second request is sent to the server. The server can directly read the target push information from the storage device and send one or more of the target push information to the client, thereby improving the response speed of the information push. Predicting the target push information based on the user's behavior helps to provide push information that is more in line with the user's preferences and improves the user's experience. The present application can predict the target push information based on the model in advance, and through the mechanism of storing the target push information in the storage device, the purpose of asynchronously obtaining the target push information is achieved.
[0125] This application Figure 8is a structural block diagram of an information push device according to an exemplary embodiment. Figure 8 The information push device 800 of the embodiment of the present application includes: a receiving module 801, a recall module 802 and a transceiver module 803.
[0126] The receiving module 801 is used to receive a first request sent by a client.
[0127] The recall module 802 is used to call the pre-trained target information push model to perform information recall processing according to the first request, obtain the target push information of the client, and cache the target push information in the storage device.
[0128] The transceiver module 803 is configured to receive a second request from the client, read target push information from the storage device according to the second request, and send one or more of the target push information to the client.
[0129] In one embodiment of the present application, the recall module 802 is also used to: obtain user feature information corresponding to the client and feature information of candidate push information; obtain user information corresponding to the client, current user behavior information of the client and candidate push information; input user information, user behavior information and candidate push information into a target push information model, and perform feature extraction by the target push information model to obtain target feature information; perform preference prediction on candidate push information based on the target feature information to recall the target push information.
[0130] In one embodiment of the present application, the recall module 802 is also used to: extract features from user information to obtain user feature information; extract features from candidate push information to obtain feature information; obtain cross-feature information between the client and the candidate push information based on user behavior information; and splice user feature information, feature information and cross-feature information to obtain target feature information.
[0131] In one embodiment of the present application, the recall module 802 is further used to: associate the identification information of the client with the target push information, and store the target push information in a storage device.
[0132] In one embodiment of the present application, the transceiver module 803 is further used to: based on the identification information in the second request, obtain push information associated with the identification information from the storage device as the target push information of the client.
[0133] In one embodiment of the present application, the recall module 802 is further used to: allocate a storage area in a storage device to the client, associate address information of the storage area with identification information of the client, and store the target push information in the storage area.
[0134] In one embodiment of the present application, the transceiver module 803 is also used to: based on the identification information in the second request, obtain address information associated with the identification information from the storage device, and access the storage area corresponding to the client based on the address information to obtain the target push information.
[0135] In one embodiment of the present application, the transceiver module 803 is further used to release the storage area occupied by the target push information.
[0136] In one embodiment of the present application, the recall module 802 is also used to: obtain training data, the training data includes a sample client, sample push information corresponding to the sample client, user feedback information of the sample push information, and label information of the sample push information; perform offline pre-training in the initial information push model based on the training data, and the information push model performs sample user feature information of the sample client and sample feature information of the sample push information, as well as sample cross-feature information of the sample client and the sample push information; obtain the prediction result of the information push model based on the sample user feature information and the sample feature information, as well as the sample cross-feature information; adjust the information push model based on the label information and the prediction result and return to continue training until the training is completed to obtain the target information push model. The present application can predict the target push information based on the model in advance, and achieve the purpose of asynchronously obtaining the target push information through a mechanism of storing the target push information in a storage device.
[0137] In the information push method provided in the embodiment of the present application, the server receives a first request sent by the client, calls the target information push model to predict the target push information according to the first request, and stores the target push information in a storage device. Based on the first request sent when the client perceives a change in user behavior information, the model's exploration of user preferences is improved, and the relevance of the target push information to the user's preferences is improved, which helps to provide push information that is more in line with user preferences. When the server receives the second request, it reads the target push information from the storage device and sends one or more of the target push information to the client, thereby improving the response speed of the information push and the user's experience. The present application can predict the target push information based on the model in advance, and through the mechanism of storing the target push information in the storage device, the purpose of asynchronously obtaining the target push information is achieved.
[0138] This application Fig. 9 is a structural block diagram of another information push device according to an exemplary embodiment. Fig. 9 The information push device 900 of the embodiment of the present application includes: a first sending module 901 , a second sending module 902 and a receiving module 903 .
[0139] The first sending module 901 is used to send a first request to the server in response to meeting the asynchronous recall condition, and the first request is used to request the server to call the pre-trained target information push model to perform information recall processing, so as to obtain the target push information corresponding to the client and cache it in the storage device;
[0140] A second sending module 902 is used to send a second request to the server in response to detecting the recall result access operation, where the second request is used to request the server to read the target push information from the storage device;
[0141] The receiving module 903 is used to receive one or more target push information returned by the server.
[0142] In one embodiment of the present application, the first sending module 901 is further used to: monitor user behavior, and in response to changes in user behavior, determine whether an asynchronous recall condition is met.
[0143] In the information push method provided in the embodiment of the present application, the client sends a first request to the server based on the user's behavior changes. The first request is used to request the server to call the pre-trained target information push model for information recall processing, improve the model's exploration of user preferences, and improve the relevance of target push information and user preferences, which helps to provide push information that is more in line with user preferences. The target push information is stored in a storage device, which reduces the complex reasoning process. When the user performs an access operation, a second request is sent to the server, which can directly read the target push information from the storage device and send one or more of the target push information to the client, thereby improving the response speed of information push and the user's experience.
[0144] Fig.10 The present invention is a block diagram of an electronic device according to an exemplary embodiment. Fig.10 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0145] like Fig.10 As shown, the electronic device 1000 includes a processor 1001, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a memory 1006 to a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for the operation of the electronic device 1000 are also stored. The processor 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.
[0146] The following components are connected to the I / O interface 1005: a memory 1006 including a hard disk, etc.; and a communication part 1007 including a network interface card such as a local area network (LAN) card, a modem, etc., and the communication part 1007 performs communication processing via a network such as the Internet; a drive 1008 is also connected to the I / O interface 1005 as needed.
[0147] 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 carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1007. When the computer program is executed by the processor 1001, the above-mentioned functions defined in the method of the present application are executed.
[0148] In an exemplary embodiment, a storage medium including instructions is also provided, such as a memory including instructions, and the instructions can be executed by the processor 1001 of the electronic device 1000 to complete the above method. Optionally, the storage medium can be a non-transitory computer-readable storage medium, for example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.
[0149] In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device. 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, which carries a computer-readable program code. 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. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.
[0150] Fig.11 The present invention is a structural block diagram of an electronic device according to an exemplary embodiment. Fig.11 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present application. Fig.11As shown, the electronic device 1100 includes a processor 1101 and a memory 1102. The memory 1102 is used to store program codes, and the processor 1101 is connected to the memory 1102 and is used to read program codes from the memory 1102 to implement the information push method in the above embodiment.
[0151] Optionally, the number of processors 1101 may be one or more.
[0152] Optionally, the electronic device may further include an interface 1103, and the number of the interface 1103 may be multiple. The interface 1103 may be connected to an application program and may receive data from an external device such as a sensor.
[0153] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed in this application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.
[0154] 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. An information push method, characterized in that: Executed by a server, the method includes: Receive a first request sent by a client; Calling a pre-trained target information push model to perform information recall processing according to the first request, obtaining the target push information of the client, and caching the target push information in a storage device; A second request from the client is received, the target push information is read from the storage device according to the second request, and one or more of the target push information is sent to the client.
2. The method according to claim 1, characterized in that The calling of a pre-trained target information push model to perform information recall processing according to the first request to obtain the target push information of the client includes: Obtaining user information corresponding to the client, current user behavior information of the client, and candidate push information; Inputting the user information, user behavior information and the candidate push information into the target push information model, and extracting features from the target push information model to obtain target feature information; Based on the target feature information, preference prediction is performed on the candidate push information to recall the target push information.
3. The method according to claim 2, characterized in that The step of inputting the user information, the user behavior information and the candidate push information into the target push information model, and extracting features from the target push information model to obtain target feature information includes: Extracting features from the user information to obtain user feature information; Extracting features from the candidate push information to obtain feature information; Based on the user behavior information, obtaining cross-feature information between the client and the candidate push information; The user characteristic information, the characteristic information and the cross characteristic information are concatenated to obtain the target characteristic information.
4. The method according to claim 1, characterized in that: The caching the target push information in a storage device includes: The identification information of the client is associated with the target push information, and the target push information is stored in the storage device.
5. The method according to claim 4, characterized in that The second request includes identification information of the client, and reading the target push information from the storage device according to the second request includes: Based on the identification information in the second request, push information associated with the identification information is acquired from the storage device as target push information of the client.
6. The method according to claim 1, characterized in that The caching the target push information in a storage device includes: A storage area in the storage device is allocated to the client, address information of the storage area is associated with identification information of the client, and the target push information is stored in the storage area.
7. The method according to claim 6, characterized in that The second request includes identification information of the client, and reading the target push information from the storage device according to the second request includes: Based on the identification information in the second request, address information associated with the identification information is acquired from the storage device, and based on the address information, a storage area corresponding to the client is accessed to acquire the target push information.
8. The method according to any one of claims 1 to 7, characterized in that After reading the target push information from the storage device according to the second request, the method further includes: The storage area occupied by the target push information is released.
9. The method according to any one of claims 1 to 7, characterized in that: The pre-training process of the target information push model includes: Acquire training data, the training data including sample clients, sample push information corresponding to the sample clients, user feedback information of the sample push information, and label information of the sample push information; Based on the training data, an initial information push model is pre-trained offline, and the information push model performs sample user feature information of a sample client and sample feature information of the sample push information, as well as sample cross-feature information of the sample client and the sample push information; Based on the sample user feature information, the sample feature information, and the sample cross feature information, obtaining a prediction result of the information push model; Based on the label information and the prediction result, the information push model is adjusted and returned to continue training until the training is completed to obtain the target information push model.
10. An information push method, characterized in that: Executed by a client, the method includes: In response to satisfying the asynchronous recall condition, sending a first request to the server, wherein the first request is used to request the server to call a pre-trained target information push model to perform information recall processing, so as to obtain the target push information corresponding to the client and cache it in a storage device; In response to detecting the recall result access operation, sending a second request to the server, wherein the second request is used to request the server to read the target push information from the storage device; Receive one or more of the target push information returned by the server.
11. The method according to claim 10, characterized in that The step of sending a first request to a server in response to satisfying an asynchronous recall condition comprises: The user behavior is monitored, and in response to a change in the user behavior, it is determined that the asynchronous recall condition is satisfied.
12. An information push device, characterized in that: Executed by a server, the device includes: A receiving module, used for receiving a first request sent by a client; A recall module, configured to call a pre-trained target information push model to perform information recall processing according to the first request, obtain the target push information of the client, and cache the target push information in a storage device; The transceiver module is used to receive the second request from the client, read the target push information from the storage device according to the second request, and send one or more of the target push information to the client.
13. An information push device, characterized in that: Executed by a client, the device includes: A first sending module, configured to send a first request to a server in response to satisfying an asynchronous recall condition, wherein the first request is configured to request the server to call a pre-trained target information push model to perform information recall processing, so as to obtain target push information corresponding to the client and cache the information in a storage device; A second sending module, configured to send a second request to the server in response to detecting a recall result access operation, wherein the second request is used to request the server to read the target push information from the storage device; The receiving module is used to receive one or more of the target push information returned by the server.
14. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to implement the steps of any one of the methods of claims 1-11.
15. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by a processor, the steps of any method in claims 1-11 are implemented.