Resource recommendation method, device, electronic device, and storage medium

Abstract

The present disclosure provides a resource recommendation method, device, electronic device, and storage medium, which relate to the fields of computer technology, and in particular to the fields of big data technology, intelligent search technology, and artificial intelligence technology. A specific implementation scheme is as follows: in response to a resource recommendation request, obtaining candidate resource features related to each of a plurality of candidate resources; processing the plurality of candidate resource features according to a pre-trained deep learning model to obtain at least one first recommended resource from the plurality of candidate resources; determining at least one second recommended resource from the plurality of candidate resources based on first recommended resource features related to at least one of the first recommended resources; and displaying a recommended resource set on an interactive page, wherein the recommended resource set includes the first recommended resource and the second recommended resource.

Description

Technical Field

[0001] The present disclosure relates to the fields of computer technology, and in particular to the fields of big data technology, intelligent search technology, and artificial intelligence technology. Background Art

[0002] With the rapid development of Internet technology, users can conveniently browse resources such as videos and images through mobile phones, tablets, and other terminal devices. Relevant Internet service platforms will also recommend resources to users based on their needs. For example, they may recommend marketing resources such as advertisements on the browsing page of the terminal device, or they may also recommend other types of resources such as videos and text on the browsing page to meet user needs. Summary of the Invention

[0003] The present disclosure provides a resource recommendation method, device, electronic device, and storage medium.

[0004] According to one aspect of the present disclosure, a resource recommendation method is provided, comprising: in response to a resource recommendation request, obtaining candidate resource features related to each of a plurality of candidate resources; processing the plurality of candidate resource features according to a pre-trained deep learning model to obtain at least one first recommended resource from the plurality of candidate resources; determining at least one second recommended resource from the plurality of candidate resources according to the first recommended resource features related to the at least one first recommended resource; and displaying a recommended resource set in an interactive page, wherein the recommended resource set includes the first recommended resource and the second recommended resource.

[0005] According to another aspect of the present disclosure, a resource recommendation device is provided, including: an acquisition module for acquiring candidate resource features related to each of multiple candidate resources in response to a resource recommendation request; a first recommended resource acquisition module for processing the multiple candidate resource features according to a pre-trained deep learning model to obtain at least one first recommended resource from the multiple candidate resources; a second recommended resource acquisition module for determining at least one second recommended resource from the multiple candidate resources based on the first recommended resource features related to the at least one first recommended resource; and a first display module for displaying a recommended resource set in an interactive page, wherein the recommended resource set includes the first recommended resource and the second recommended resource.

[0006] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method provided according to an embodiment of the present disclosure.

[0007] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method provided according to an embodiment of the present disclosure.

[0008] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements the method provided according to the embodiment of the present disclosure when executed by a processor.

[0009] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings are used to better understand the present invention and do not constitute a limitation of the present invention.

[0011] Figure 1 Schematically illustrates an exemplary system architecture to which the resource recommendation method and apparatus according to an embodiment of the present disclosure can be applied;

[0012] Figure 2 The following schematically shows a flow chart of a resource recommendation method according to an embodiment of the present disclosure;

[0013] Figure 3A The following schematically shows a principle diagram of a resource recommendation method according to an embodiment of the present disclosure;

[0014] Figure 3B Schematically shows a principle diagram of a resource recommendation method according to another embodiment of the present disclosure;

[0015] Figure 4 The following schematically shows a principle diagram of a resource recommendation method according to another embodiment of the present disclosure;

[0016] Figure 5 A schematic diagram of an interactive page according to an embodiment of the present disclosure is schematically shown;

[0017] Figure 6 Schematically shows a principle diagram of updating a deep learning model according to an embodiment of the present disclosure;

[0018] Figure 7 A block diagram schematically shows a resource recommendation device according to an embodiment of the present disclosure; and

[0019] Figure 8 A block diagram of an electronic device suitable for implementing a resource recommendation method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0020] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0021] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

[0022] With the rapid development of internet technology, relevant internet platforms can build advertising retrieval systems to select suitable advertising resources from vast advertising resource libraries for user recommendations. By recommending appropriate advertising resources to users, users can easily access information about goods or services, and operators of related goods or services can be provided with efficient and accurate marketing strategies, thereby meeting users' actual needs while improving transaction efficiency for operators. However, conventional advertising resource recommendation methods struggle to accurately meet user needs, especially when recommending multiple advertising resources. The accuracy of recommendations is low, the computational overhead is high, and they are unable to meet the actual needs of users or operators.

[0023] Embodiments of the present disclosure provide a resource recommendation method, apparatus, electronic device, and storage medium. The resource recommendation method includes: in response to a resource recommendation request, obtaining candidate resource features associated with each of a plurality of candidate resources; processing the plurality of candidate resource features according to a pre-trained deep learning model to obtain at least one first recommended resource from the plurality of candidate resources; determining at least one second recommended resource from the plurality of candidate resources based on the first recommended resource features associated with the at least one first recommended resource; and presenting a recommended resource set on an interactive page, wherein the recommended resource set includes the first recommended resource and the second recommended resource.

[0024] According to an embodiment of the present disclosure, a first recommended resource is generated by processing multiple candidate resource features according to a pre-trained deep learning model, and a second recommended resource is iteratively determined according to the first recommended resource features of the generated first recommended resource. It is possible to further generate new recommended resources based on the already generated recommended resources, so that the characteristic attributes of the multiple recommended resources in the obtained recommended resource set are adapted to each other, thereby reducing the characteristic attribute conflicts between the multiple recommended resources, improving the adaptability between the recommended resources, and making the recommended resources adaptable to the recommended targets. By displaying the recommended resource set in the interactive page, user satisfaction can be improved.

[0025] Figure 1An exemplary system architecture to which the resource recommendation method and apparatus according to an embodiment of the present disclosure can be applied is schematically shown.

[0026] It should be noted that Figure 1 The examples shown are merely examples of system architectures to which the embodiments of the present disclosure may be applied, to help those skilled in the art understand the technical content of the present disclosure. This does not mean that the embodiments of the present disclosure cannot be applied to other devices, systems, environments, or scenarios. For example, in another embodiment, an exemplary system architecture to which the resource recommendation method and apparatus may be applied may include a terminal device, but the terminal device may implement the resource recommendation method and apparatus provided by the embodiments of the present disclosure without interacting with a server.

[0027] like Figure 1 As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0028] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software (for example only).

[0029] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0030] Server 105 may be a server that provides various services, such as a background management server (for example only) that supports content browsed by users using terminal devices 101, 102, and 103. The background management server may analyze and process received data such as user requests, and feed back processing results (e.g., web pages, information, or data obtained or generated based on user requests) to the terminal device.

[0031] It should be noted that the resource recommendation method provided in the embodiment of the present disclosure can generally be executed by the terminal device 101, 102, or 103. Accordingly, the resource recommendation apparatus provided in the embodiment of the present disclosure can also be set in the terminal device 101, 102, or 103.

[0032] Alternatively, the resource recommendation method provided in the embodiment of the present disclosure may also be generally executed by the server 105. Accordingly, the resource recommendation apparatus provided in the embodiment of the present disclosure may generally be provided in the server 105. The resource recommendation method provided in the embodiment of the present disclosure may also be performed by a server or server cluster that is different from the server 105 and that is capable of communicating with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the resource recommendation apparatus provided in the embodiment of the present disclosure may also be provided in a server or server cluster that is different from the server 105 and that is capable of communicating with the terminal devices 101, 102, 103 and / or the server 105.

[0033] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0034] Figure 2 The flowchart of the resource recommendation method according to an embodiment of the present disclosure is schematically shown.

[0035] like Figure 2 As shown, the resource recommendation method includes operations S210 to S240.

[0036] In operation S210 , in response to a resource recommendation request, candidate resource features associated with each of a plurality of candidate resources are obtained.

[0037] In operation S220 , multiple candidate resource features are processed according to a pre-trained deep learning model to obtain at least one first recommended resource from the multiple candidate resources.

[0038] In operation S230 , at least one second recommended resource is determined from a plurality of candidate resources based on a first recommended resource feature related to the at least one first recommended resource.

[0039] In operation S240 , a recommended resource set is displayed on the interactive page, where the recommended resource set includes a first recommended resource and a second recommended resource.

[0040] According to embodiments of the present disclosure, resource recommendation requests can be generated based on a user's resource demand operation. For example, resource recommendation requests can be generated based on any type of user operation, such as a search operation or login operation. However, this is not limited to this method. Resource recommendation requests can also be determined based on other methods, such as requests sent by other devices such as terminals or servers. The embodiments of the present disclosure do not limit the method for determining resource recommendation requests, and those skilled in the art can make their own selection based on actual needs.

[0041] According to the embodiments of the present disclosure, the pre-trained deep learning model can be constructed based on any type of deep learning algorithm, for example, a deep learning model can be constructed based on a convolutional neural network algorithm, or a deep learning model can be constructed based on an attention network algorithm. The embodiments of the present disclosure do not limit the type of algorithm for constructing the deep learning model. The pre-trained deep learning model can be trained based on any training method, for example, a deep learning model can be trained based on a supervised method, but is not limited thereto. A pre-trained deep learning model can also be trained based on an unsupervised method, a semi-supervised method, or a reinforcement learning strategy.

[0042] According to the embodiments of the present disclosure, the candidate resources in the embodiments of the present disclosure may include any type of resource information, such as video resources, text resources, image resources, audio resources, page resources, advertising component resources, etc. The embodiments of the present disclosure do not limit the specific resource types of the candidate resources, as long as they can be browsed on terminal devices such as mobile phones, or can be transmitted through Internet links.

[0043] According to embodiments of the present disclosure, candidate resource features may include any type of feature data representing the resource attributes of the candidate resource. For example, this may include text information, image information, etc. The embodiments of the present disclosure do not limit the specific attribute types of the resource attributes represented by the candidate resource features, as long as they can meet actual needs.

[0044] According to an embodiment of the present disclosure, determining at least one second recommended resource from a plurality of candidate resources based on first recommended resource features associated with at least one first recommended resource may include processing the first recommended resource features using a pre-trained deep learning model to obtain the second recommended resource. However, this is not limited to this, and may also include processing candidate resource features and first recommended resource features using a pre-trained deep learning model to obtain the second recommended resource.

[0045] It should be noted that the generation time of the first recommended resource may be earlier than the generation time of the second recommended resource. When the second recommended resource is generated, the second recommended resource can be determined as the first recommended resource, and then the second recommended resource can be iteratively determined based on the first recommended resource characteristics of at least one first recommended resource that has been generated.

[0046] According to an embodiment of the present disclosure, displaying the recommended resource set in the interactive page may include displaying all or part of the recommended resources in the recommended resource set in the interactive page.

[0047] According to an embodiment of the present disclosure, displaying a recommended resource set in an interactive page may further include displaying the recommended resources in the recommended resource set in order of their respective attributes, so as to meet personalized recommendation needs.

[0048] According to an embodiment of the present disclosure, the candidate resource feature includes at least one of the following: search information corresponding to the candidate resource, resource title information of the candidate resource, resource description information of the candidate resource, and value information of the candidate resource.

[0049] According to an embodiment of the present disclosure, the search information corresponding to the candidate resource may include a search keyword (or Query) that matches the resource description text representing the candidate resource, or may also include a word or field that has a mapping relationship with the candidate resource.

[0050] According to an embodiment of the present disclosure, the resource title information may include subject information for describing the candidate resource, such as an advertisement slogan title, a file title of a file resource, and the like.

[0051] According to an embodiment of the present disclosure, the resource description information may include text information representing attributes of the candidate resource, or may also include text contained in the candidate resource, such as an advertisement sentence in an advertisement resource.

[0052] According to an embodiment of the present disclosure, the value information of the candidate resources may include the transaction value of the candidate resources, such as the revenue value of pushing advertising resources, the value that users need to pay for browsing the candidate resources, and the like.

[0053] According to the embodiments of the present disclosure, the candidate resource features may be represented by any type of data, for example, text, image, code, or other types of data. The embodiments of the present disclosure do not limit the data types representing the candidate resource features.

[0054] According to an embodiment of the present disclosure, the deep learning model includes a feature fusion network and a recommended resource generation network.

[0055] According to an embodiment of the present disclosure, the feature fusion network and the recommended resource generation network can be constructed based on the attention network model. For example, the feature fusion network can be constructed based on the encoder of the Transformer model, and the recommended resource generation network can be constructed based on the decoder of the Transformer model.

[0056] According to an embodiment of the present disclosure, processing multiple candidate resource features according to a pre-trained deep learning model, and obtaining at least one first recommended resource from the multiple candidate resources may include: inputting the multiple candidate resource features into a feature fusion network and outputting the fused features; and inputting the fused features into a recommended resource generation network and outputting at least one first recommended resource.

[0057] According to the embodiments of the present disclosure, the feature fusion network can fuse multiple candidate resource features based on the attention mechanism, so that the fused features can fully learn the resource attributes of each of the multiple candidate resources, as well as the attribute relationships such as dependency and exclusion relationships between the resource attributes of the multiple candidate resources. Then, based on the recommended resource generation network, the fused features can be processed, and the first recommended resource can be obtained after learning the global resource attributes of multiple candidate resources, thereby improving the prediction accuracy of the first recommended resource.

[0058] Figure 3A The schematic diagram shows a principle diagram of a resource recommendation method according to an embodiment of the present disclosure.

[0059] like Figure 3A As shown, the resource recommendation method provided by the embodiment of the present disclosure can be implemented based on a pre-trained deep learning model 300. The deep learning model 300 may include a feature fusion network 310 and a recommended resource generation network 320. The feature fusion network 310 and the recommended resource generation network 320 may be constructed based on the encoder and decoder of the Transformer model, respectively. The multiple candidate resource features may include a first candidate resource feature T3011, a second candidate resource feature T3012, a third candidate resource feature T3013, and a fourth candidate resource feature T3014.

[0060] For the prediction task of determining the first recommended resource, the first candidate resource feature T3011, the second candidate resource feature T3012, the third candidate resource feature T3013, and the fourth candidate resource feature T3014 can be input into the feature fusion network 310, which can fuse multiple candidate resource features based on the attention mechanism and output the fused features. The fused features can be input into the recommended resource generation network 320, which outputs the predicted first recommended resource 3011. The first recommended resource 3011 can, for example, be the first candidate resource among the first to fourth candidate resources. Obtaining the first recommended resource through the end-to-end network structure of a pre-trained deep learning model can reduce the computational complexity of predicting recommended resources and save computational overhead.

[0061] According to an embodiment of the present disclosure, determining at least one second recommended resource from multiple candidate resources based on a first recommended resource feature related to at least one first recommended resource may include: processing the first recommended resource feature and the fusion feature according to a recommended resource generation network to obtain at least one second recommended resource.

[0062] According to an embodiment of the present disclosure, the recommendation resource generation network can fuse the first recommendation resource features and fusion features of the generated first recommendation resource based on the attention mechanism, so that the recommendation resource generation network can predict the second recommendation resource under the condition of fully learning the local resource attributes of multiple candidate resources, the global resource attributes of multiple candidate resources, and the dependency, mutual exclusion and other related attributes among the resource attributes of the generated recommendation resources, thereby improving the matching degree between the second recommendation resource and the generated first recommendation resource, reducing the degree of contradiction between the recommended resources, and thus improving the adaptability of multiple recommendation resources to the recommendation needs.

[0063] Figure 3B The diagram schematically shows a principle diagram of a resource recommendation method according to another embodiment of the present disclosure.

[0064] like Figure 3B As shown, when first recommended resource 3011 has been generated, for the prediction task of predicting a second recommended resource, recommended resource generation network 320 of deep learning model 300 can process the fused features output by feature fusion network 310 and first candidate resource features T3011, and output second recommended resource 3012. Second recommended resource 3012 can be, for example, a second candidate resource corresponding to the second candidate resource features.

[0065] It should be understood that the first candidate resource feature T3011 may be the first recommended resource feature corresponding to the generated first recommended resource 3011 .

[0066] It should be noted that after the second recommended resource 3012 is generated, the second recommended resource 3012 may be determined as the first recommended resource, and a new second recommended resource may be determined based on the two generated first recommended resources.

[0067] According to an embodiment of the present disclosure, the recommended resource generation network may include a recommended resource evaluation layer.

[0068] According to an embodiment of the present disclosure, the recommended resource evaluation layer can be constructed based on an attention network algorithm, for example, the recommended resource evaluation layer can be constructed based on a self-attention head.

[0069] According to an embodiment of the present disclosure, processing the first recommended resource feature and the fusion feature according to the recommended resource generation network to obtain at least one second recommended resource includes: processing the first recommended resource feature and the fusion feature according to the recommended resource evaluation layer to obtain intermediate resource fusion features corresponding to each of the multiple candidate resources; masking the mutually exclusive resource fusion features corresponding to the mutually exclusive resources in the multiple intermediate resource fusion features according to the mutually exclusive resource attributes corresponding to the mutually exclusive resources to obtain the target resource fusion feature; and generating at least one second recommended resource according to the target resource fusion feature.

[0070] According to an embodiment of the present disclosure, mutually exclusive resources may be determined from a plurality of candidate resources based on a preset resource mutual exclusion rule. Mutually exclusive resource attributes may include a resource location or a resource identifier of the mutually exclusive resource.

[0071] According to an embodiment of the present disclosure, the intermediate resource fusion feature may include intermediate resource fusion feature elements corresponding to multiple candidate resources respectively, and the intermediate resource fusion feature elements corresponding to mutually exclusive resources may be masked by mutually exclusive resource attributes to obtain the target resource fusion feature.

[0072] In one embodiment of the present disclosure, the intermediate resource fusion features may include initial evaluation information representing the recommendation probability. For example, multiple intermediate resource fusion features may be represented as [1, 2, -5, 9], where the intermediate resource fusion features "1", "2", "-5", and "9" may correspond to the first to fourth candidate resources, respectively. In the case where the first candidate resource is a mutually exclusive resource, the mutually exclusive resource fusion feature "1" in [1, 2, -5, 9] may be masked, and the resulting target resource fusion feature may be represented as [mask, 2, -5, 9], where "2", "-5", and "9" may be target resource fusion features that are not masked.

[0073] According to an embodiment of the present disclosure, generating at least one second recommended resource based on the target resource fusion feature may include processing the target resource fusion feature based on the activation function layer, thereby obtaining the normalized recommendation probability corresponding to each of the multiple candidate resources, thereby determining at least one second recommended resource from the multiple candidate resources based on the recommendation probability.

[0074] According to an embodiment of the present disclosure, the preset resource mutual exclusion rules can be determined based on actual recommendation needs. For example, a mutual exclusion relationship can be set between multiple candidate resources for advertising and marketing the same model of items, so that resource mutual exclusion rules for multiple candidate resources can be obtained.

[0075] Figure 4 The following schematically shows a principle diagram of a resource recommendation method according to another embodiment of the present disclosure.

[0076] like Figure 4As shown, the recommended resource generation network 420 may include a recommended resource evaluation layer 421 and a recommendation evaluation information output layer 422. Once a first recommended resource 4011 has been generated, for the prediction task of a second recommended resource, the fused feature T410, the first recommended resource feature T4011, and the location information w4011 corresponding to the first recommended resource 4011 may be input into the recommended resource evaluation layer 421. The intermediate resource fused feature is then output. The location information w4011 may indicate that the first recommended resource 4011 is ranked first within the recommended resource set.

[0077] like Figure 4 As shown, the mutually exclusive relationship matrix 430 can represent the mutually exclusive relationship between the first candidate resource to the fourth candidate resource. For example, the black elements in the mutually exclusive relationship matrix 430 can represent the existence of a mutually exclusive relationship between the first candidate resource and the third candidate resource. Therefore, when the first recommended resource 4011 that has been generated is the first candidate resource, the third candidate resource can be in a mutually exclusive relationship. The mutually exclusive resource attributes can be determined by the mutually exclusive relationship matrix 430, and then the mutually exclusive relationship between the third candidate resource and the first recommended resource 4011 that has been generated can be represented based on the mutually exclusive resource attributes. In addition, the mutually exclusive resource attributes determined based on the location information w4011 can also represent that the intermediate resource fusion feature corresponding to the first candidate resource in the intermediate resource fusion feature is a mutually exclusive resource fusion feature.

[0078] By inputting mutually exclusive resource attributes and intermediate resource fusion features into the recommendation evaluation information output layer 422, masking of the mutually exclusive resource fusion features within the intermediate resource fusion features can be achieved, allowing the recommendation evaluation information output layer 422 to output recommendation evaluation information P402 for the second recommended resource. Recommendation evaluation information P402 can be represented as [0.0, 0.8, 0.0, 0.2], where the recommendation probabilities of "0.0," "0.8," "0.0," and "0.2" correspond to the first through fourth candidate resources, respectively. The recommendation probabilities of "0.0" corresponding to the first and third candidate resources in the recommendation evaluation information P402 are obtained after masking. Based on the recommendation evaluation information P402, the second candidate resource can be determined as the second recommended resource from the first through fourth candidate resources.

[0079] For the first recommended resource 4011 that has been generated, mask processing can be used to shield the mutually exclusive resources in the candidate resources, thereby avoiding setting up multi-objective optimization solution problems based on mutually exclusive rules, reducing the excessive computational overhead and gradient explosion problems caused by solving multi-objective optimization problems, thereby saving computational overhead and improving computational efficiency.

[0080] According to an embodiment of the present disclosure, a pre-trained deep learning model can be obtained by training based on a reinforcement learning method.

[0081] According to an embodiment of the present disclosure, the recommended resources include multiple resources, each of the multiple recommended resources has a recommendation order attribute, and the interactive page has multiple resource locations to be updated.

[0082] According to an embodiment of the present disclosure, displaying a recommended resource set in an interactive page includes: determining a target resource location related to the recommended resource from multiple resource locations to be updated based on multiple recommendation order attributes; and displaying the recommended resource corresponding to the target resource location at the target resource location in the interactive page.

[0083] According to an embodiment of the present disclosure, the recommendation order attribute may represent the ranking position of each of the multiple recommended resources, or the recommendation order attribute may further represent the recommendation level of each of the multiple recommended resources. The recommendation order attribute may be determined based on the generation time of each of the recommended resources, or may further be determined based on a preset recommendation level rule.

[0084] According to an embodiment of the present disclosure, displaying recommended resources corresponding to the target resource location in the interactive page may include rendering a recommended resource component corresponding to the recommended resource at the target resource location, thereby displaying the recommended resource at the target resource location.

[0085] Figure 5 A schematic diagram of an interactive page according to an embodiment of the present disclosure is schematically shown.

[0086] like Figure 5 As shown, the interactive page 500 can display recommended resources 510 , 520 , and 530 at each target resource location. The interactive page 500 can also include page display content 540 .

[0087] According to an embodiment of the present disclosure, the resource recommendation method may further include: updating the recommended resources according to the page operation on the interactive page to obtain new recommended resources; and displaying the new recommended resources on the interactive page.

[0088] According to an embodiment of the present disclosure, a page operation may include any type of operation performed by a user on an interactive page, such as a click operation, a page sliding operation, and the like.

[0089] According to an embodiment of the present disclosure, when a page operation is detected, multiple recommended resources displayed on the interactive page can be updated, so that new recommended resources are displayed on the interactive page, thereby meeting the user's recommendation needs in a timely manner.

[0090] According to an embodiment of the present disclosure, updating recommended resources based on page operations for interactive pages includes: identifying operation intentions based on page operation attributes corresponding to the page operations to obtain intention recognition results; and updating recommended resources when the intention recognition results indicate that update conditions are met.

[0091] According to embodiments of the present disclosure, page operation attributes can be processed based on a pre-trained intent recognition model to obtain intent recognition results. The intent recognition model can be constructed based on a neural network algorithm, such as a convolutional neural network algorithm. However, this is not limited to this, and intent recognition models can also be constructed based on other types of neural network algorithms.

[0092] According to an embodiment of the present disclosure, the page operation attribute includes at least one of the following: page operation duration, page operation type, and operation input information corresponding to the page operation.

[0093] According to an embodiment of the present disclosure, the page operation duration may include the duration of the page operation, such as the duration of a sliding operation or a browsing operation, or may also include the operation interval duration between multiple page operations.

[0094] According to an embodiment of the present disclosure, the operation input information corresponding to the page operation may include any type of information such as text information, audio information, etc.

[0095] According to an embodiment of the present disclosure, operation intention identification is performed based on page operation attributes corresponding to the page operation, and obtaining the intention identification result may also include: comparing the page operation attributes with preset update conditions to obtain a comparison result; and determining the intention identification result based on the comparison result.

[0096] According to an embodiment of the present disclosure, the preset update condition can be determined based on page operation attributes, for example, a duration threshold can be set for the page operation duration, and the comparison result between the page operation duration and the duration threshold is used to determine whether the comparison result meets the update condition. However, this is not limited to this. The preset update condition can also be set based on a preset operation combination, so that when the comparison result between the page operation attributes and the preset update condition indicates that the update condition is met, the intent recognition result is determined to meet the update condition.

[0097] According to an embodiment of the present disclosure, updating recommended resources according to page operations on interactive pages to obtain new recommended resources includes: updating the current deep learning model to obtain an updated deep learning model when the page operation meets the update conditions; and processing at least one candidate resource feature according to the updated deep learning model to obtain new recommended resources.

[0098] According to an embodiment of the present disclosure, by updating the current deep learning model and determining new recommended resources from candidate resources based on the updated deep learning model, it is possible to timely optimize recommended resources based on the user's page operations, and to determine new recommended resources for other users based on the updated deep learning model, thereby conveniently improving the accuracy and efficiency of resource recommendations.

[0099] According to an embodiment of the present disclosure, updating the current deep learning model may include: determining resource interaction information corresponding to the current recommended resource set; determining reward information based on the resource interaction information; and updating model parameters of the current deep learning model based on the reward information.

[0100] According to an embodiment of the present disclosure, resource interaction information may be determined based on page operations on an interactive page displaying a recommended resource set, such as a browse operation on the interactive page, a click operation on a recommended resource, a purchase operation, and the like.

[0101] According to an embodiment of the present disclosure, the resource interaction information includes at least one of the following indicators corresponding to the recommended resource set: a resource click-through rate indicator, a resource conversion rate indicator, a resource transaction value indicator, and a resource browsing time indicator.

[0102] According to an embodiment of the present disclosure, the resource conversion rate indicator may include a commodity purchase rate, a repurchase rate, etc. corresponding to the resource.

[0103] According to an embodiment of the present disclosure, the resource transaction value indicator may include the revenue value, profit value, etc. of the transaction completed based on the recommended resources.

[0104] According to an embodiment of the present disclosure, the resource browsing time indicator may include the browsing time of an interactive page displaying recommended resources, or may also include the browsing time of jumping to a recommended resource page after clicking on a recommended resource.

[0105] According to embodiments of the present disclosure, determining reward information based on resource interaction information may include processing the resource interaction information based on a pre-trained reward information prediction model to generate corresponding reward information. Alternatively, the method may include processing the resource interaction information based on a reward information determination rule to obtain reward information. The embodiments of the present disclosure do not limit the specific method for determining reward information.

[0106] Figure 6 The figure schematically shows a principle diagram of updating a deep learning model according to an embodiment of the present disclosure.

[0107] like Figure 6As shown, the current deep learning model 610 can obtain a recommended resource set 601 by processing multiple candidate resource features. By pushing the recommended resource set 601 to the environment model 620, the recommended resource set 601 can be displayed on the interactive page, and resource interaction information such as resource click-through rate indicators can be collected based on the target object's page operations on the recommended resource set. The environment model 620 can send the resource interaction information to the reward information generation module 630, and the reward information generation module 630 can determine N reward influencing factors based on the resource interaction information. …to . And the reward information is determined according to the following formula (1).

[0108] (1);

[0109] In formula (1), Indicates the generated reward information, …to Represents the hyperparameters of each of the N reward influencing factors, which are used to balance the weights between different reward influencing factors. Reward influencing factors can include multiple optimization objectives to be achieved by the recommended resource, such as factors such as advertising reach, resource transaction value, and advertising search expectations. Updating the model parameters of the current deep learning model 610 using the calculated reward information allows the updated deep learning model to more accurately match the optimization objectives to be achieved by the recommended resource, thereby improving the prediction accuracy of the recommended resource, accommodating the complex nonlinear relationships between multiple recommended resources, and improving the adaptability of the recommended resource.

[0110] It should be noted that the information collected in the embodiments of this disclosure, including but not limited to resource interaction information and page operation information, is conducted with user authorization. The user is informed of the purpose of the information collection before it is collected. The goal is to provide users with resources that meet their actual needs. Necessary encryption and desensitization measures are also employed during information processing to prevent the leakage of personal information.

[0111] According to embodiments of the present disclosure, a pre-trained deep learning model can be trained using a reinforcement learning approach. For example, historical resource features of historical resources generated during a historical time period can be obtained, processed using a pre-trained resource interaction information prediction model to obtain predicted resource interaction information, and processed using a reward information generation module to obtain predicted reward information. The model parameters of the deep learning model can then be iteratively adjusted using the predicted reward information until the predicted reward information converges.

[0112] Figure 7 The block diagram of the resource recommendation device according to an embodiment of the present disclosure is schematically shown.

[0113] like Figure 7 As shown, the resource recommendation device 700 includes: an acquisition module 710 , a first recommended resource acquisition module 720 , a second recommended resource acquisition module 730 and a first display module 740 .

[0114] The acquisition module 710 is configured to acquire candidate resource features related to each of a plurality of candidate resources in response to a resource recommendation request.

[0115] The first recommended resource obtaining module 720 is configured to process features of multiple candidate resources according to a pre-trained deep learning model and obtain at least one first recommended resource from the multiple candidate resources.

[0116] The second recommended resource obtaining module 730 is configured to determine at least one second recommended resource from a plurality of candidate resources according to a first recommended resource feature associated with at least one first recommended resource.

[0117] The first display module 740 is configured to display a recommended resource set on an interactive page, wherein the recommended resource set includes a first recommended resource and a second recommended resource.

[0118] According to an embodiment of the present disclosure, the resource recommendation device further includes: an updating module and a second display module.

[0119] The update module is used to update the recommended resources according to the page operation on the interactive page to obtain new recommended resources.

[0120] The second display module is used to display new recommended resources on the interactive page.

[0121] According to an embodiment of the present disclosure, the update module includes: an intention recognition submodule and a first update submodule.

[0122] The intention recognition submodule is used to identify the operation intention according to the page operation attributes corresponding to the page operation and obtain the intention recognition result.

[0123] The first updating submodule is used to update the recommended resources when the intention recognition result representation meets the updating conditions.

[0124] According to an embodiment of the present disclosure, the intention recognition submodule includes: a comparison unit and an intention recognition result determination unit.

[0125] The comparison unit is used to compare the page operation attribute with the preset update condition to obtain a comparison result.

[0126] The intention recognition result determination unit is used to determine the intention recognition result based on the comparison result.

[0127] According to an embodiment of the present disclosure, the page operation attribute includes at least one of the following: page operation duration, page operation type, and operation input information corresponding to the page operation.

[0128] According to an embodiment of the present disclosure, the updating module includes: a second updating submodule and a recommended resource obtaining submodule.

[0129] The second updating submodule is used to update the current deep learning model when the page operation meets the update conditions to obtain an updated deep learning model.

[0130] The recommended resource acquisition submodule is used to process at least one candidate resource feature according to the updated deep learning model to obtain a new recommended resource.

[0131] According to an embodiment of the present disclosure, the second updating submodule includes: a resource interaction information determining unit, a reward information determining unit, and a model parameter updating unit.

[0132] The resource interaction information determining unit is configured to determine the resource interaction information corresponding to the current recommended resource set.

[0133] The reward information determining unit is used to determine the reward information according to the resource interaction information.

[0134] The model parameter updating unit is used to update the model parameters of the current deep learning model according to the reward information.

[0135] According to an embodiment of the present disclosure, the resource interaction information includes at least one of the following indicators corresponding to the recommended resource set: a resource click-through rate indicator, a resource conversion rate indicator, a resource transaction value indicator, and a resource browsing time indicator.

[0136] According to an embodiment of the present disclosure, the deep learning model includes a feature fusion network and a recommended resource generation network.

[0137] According to an embodiment of the present disclosure, the first recommended resource obtaining module includes: a fusion submodule and a first recommended resource obtaining submodule.

[0138] The fusion submodule is used to input multiple candidate resource features into the feature fusion network and output fusion features.

[0139] The first recommended resource obtaining submodule is used to input the fusion features into the recommended resource generation network and output at least one first recommended resource.

[0140] According to an embodiment of the present disclosure, the second recommended resource obtaining module includes a second recommended resource obtaining sub-module.

[0141] The second recommended resource obtaining submodule is configured to process the first recommended resource features and the fusion features according to the recommended resource generation network to obtain at least one second recommended resource.

[0142] According to an embodiment of the present disclosure, a recommended resource generation network includes a recommended resource evaluation layer.

[0143] According to an embodiment of the present disclosure, the second recommended resource obtaining submodule includes: an intermediate resource fusion feature obtaining unit, a target resource fusion feature obtaining unit, and a second recommended resource obtaining unit.

[0144] The intermediate resource fusion feature obtaining unit is used to process the first recommended resource feature and the fusion feature according to the recommended resource evaluation layer to obtain the intermediate resource fusion features corresponding to each of the multiple candidate resources.

[0145] The target resource fusion feature acquisition unit is used to mask the mutually exclusive resource fusion features corresponding to the mutually exclusive resources among multiple intermediate resource fusion features according to the mutually exclusive resource attributes corresponding to the mutually exclusive resources, so as to obtain the target resource fusion feature, wherein the mutually exclusive resources are determined from multiple candidate resources based on preset resource mutual exclusion rules.

[0146] The second recommended resource obtaining unit is configured to generate at least one second recommended resource according to the target resource fusion feature.

[0147] According to an embodiment of the present disclosure, the candidate resource feature includes at least one of the following: search information corresponding to the candidate resource, resource title information of the candidate resource, resource description information of the candidate resource, and value information of the candidate resource.

[0148] According to an embodiment of the present disclosure, the recommended resources include multiple resources, each of the multiple recommended resources has a recommendation order attribute, and the interactive page has multiple resource locations to be updated.

[0149] According to an embodiment of the present disclosure, the first presentation module includes: a target resource location determination submodule and a presentation submodule.

[0150] The target resource location determination submodule is used to determine a target resource location related to the recommended resource from a plurality of resource locations to be updated according to a plurality of recommendation order attributes.

[0151] The display submodule is used to display the recommended resources corresponding to the target resource location at the target resource location in the interactive page.

[0152] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0153] According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described above.

[0154] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the method described above.

[0155] According to an embodiment of the present disclosure, a computer program product includes a computer program, and when the computer program is executed by a processor, the computer program implements the method described above.

[0156] Figure 8 A block diagram of an electronic device suitable for implementing a resource recommendation method according to an embodiment of the present disclosure is schematically shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0157] like Figure 8 As shown, device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. RAM 803 may also store various programs and data required for the operation of device 800. Computing unit 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to bus 804.

[0158] Various components in device 800 are connected to I / O interface 805, including an input unit 806, such as a keyboard, mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, optical disk, etc.; and a communication unit 809, such as a network card, modem, wireless communication transceiver, etc. The communication unit 809 allows device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0159] The computing unit 801 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the resource recommendation method. For example, in some embodiments, the resource recommendation method may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed onto the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the resource recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the resource recommendation method in any other suitable manner (e.g., via firmware).

[0160] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0161] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0162] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0163] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0164] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0165] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0166] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0167] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A resource recommendation method, comprising: In response to the resource recommendation request, obtaining candidate resource features related to each of the plurality of candidate resources; Processing a plurality of candidate resource features according to a pre-trained deep learning model to obtain at least one first recommended resource from the plurality of candidate resources, wherein the deep learning model includes a recommended resource generation network; Processing the fused feature and the first recommended resource feature related to the first recommended resource using the recommended resource evaluation layer of the recommended resource generation network to obtain intermediate resource fused features corresponding to each of the plurality of candidate resources, wherein the fused feature is determined by fusing the plurality of candidate resource features based on an attention mechanism; performing masking processing on mutually exclusive resource fusion features corresponding to mutually exclusive resources among the plurality of intermediate resource fusion features according to mutually exclusive resource attributes corresponding to the mutually exclusive resources, to obtain a target resource fusion feature, wherein the mutually exclusive resources are determined from the plurality of candidate resources based on a resource mutual exclusion rule; Obtaining a second recommended resource from the plurality of candidate resources according to the target resource fusion feature; A recommended resource set is displayed on an interactive page, wherein the recommended resource set includes the first recommended resource and the second recommended resource.

2. The method according to claim 1, further comprising: updating the recommended resources according to the page operation on the interactive page to obtain new recommended resources; as well as The new recommended resource is displayed on the interactive page.

3. The method according to claim 2, wherein: The updating of the recommended resources according to the page operation on the interactive page includes: Performing operation intention recognition based on the page operation attribute corresponding to the page operation to obtain an intention recognition result; and When the intent recognition result indicates that an update condition is satisfied, the recommended resource is updated.

4. The method according to claim 3, wherein: The operation intention is identified according to the page operation attribute corresponding to the page operation, and the intention identification result is obtained, which includes: Comparing the page operation attribute with a preset update condition to obtain a comparison result; and The intention recognition result is determined based on the comparison result.

5. The method according to claim 3, wherein: The page operation attributes include at least one of the following: Page operation duration, page operation type, and operation input information corresponding to the page operation.

6. The method according to claim 2, wherein: The updating of the recommended resources according to the page operation on the interactive page to obtain new recommended resources includes: When the page operation satisfies an update condition, updating the current deep learning model to obtain an updated deep learning model; and At least one of the candidate resource features is processed according to the updated deep learning model to obtain the new recommended resource.

7. The method according to claim 6, wherein: Updating the current deep learning model includes: Determining resource interaction information corresponding to the current recommended resource set; Determining reward information based on the resource interaction information; and Update the model parameters of the current deep learning model according to the reward information.

8. The method according to claim 7, wherein: The resource interaction information includes at least one of the following indicators corresponding to the recommended resource set: Resource click-through rate indicator, resource conversion rate indicator, resource transaction value indicator, and resource browsing time indicator.

9. The method according to claim 1, wherein: The deep learning model includes a feature fusion network; The processing of the plurality of candidate resource features according to the pre-trained deep learning model to obtain at least one first recommended resource from the plurality of candidate resources includes: Inputting a plurality of the candidate resource features into the feature fusion network and outputting a fusion feature; and The fused features are input into the recommended resource generation network, and the at least one first recommended resource is output.

10. The method according to any one of claims 1 to 9, wherein The candidate resource characteristics include at least one of the following: Search information corresponding to the candidate resource, resource title information of the candidate resource, resource description information of the candidate resource, and value information of the candidate resource.

11. The method according to any one of claims 1 to 9, wherein The recommended resources include a plurality of resources, each of the plurality of recommended resources has a recommendation order attribute, and the interactive page includes a plurality of resource locations to be updated; The display of the recommended resource set on the interactive page includes: Determining a target resource location related to the recommended resource from the plurality of resource locations to be updated according to a plurality of recommendation sequence attributes; and The target resource location in the interactive page displays recommended resources corresponding to the target resource location.

12. A resource recommendation device, comprising: An acquisition module, configured to obtain, in response to a resource recommendation request, candidate resource features associated with each of a plurality of candidate resources; a first recommended resource obtaining module, configured to process features of a plurality of candidate resources according to a pre-trained deep learning model to obtain at least one first recommended resource from the plurality of candidate resources, wherein the deep learning model includes a feature fusion network and a recommended resource generation network; a second recommended resource obtaining module, configured to determine at least one second recommended resource from a plurality of the candidate resources based on a first recommended resource feature associated with at least one of the first recommended resources; as well as A first display module, configured to display a recommended resource set on an interactive page, wherein the recommended resource set includes the first recommended resource and the second recommended resource; The deep learning model includes a recommended resource generation network; the second recommended resource acquisition module is configured as follows: Processing the fused feature and the first recommended resource feature related to the first recommended resource using the recommended resource evaluation layer of the recommended resource generation network to obtain intermediate resource fused features corresponding to each of the plurality of candidate resources, wherein the fused feature is determined by fusing the plurality of candidate resource features based on an attention mechanism; performing masking processing on mutually exclusive resource fusion features corresponding to mutually exclusive resources among the plurality of intermediate resource fusion features according to mutually exclusive resource attributes corresponding to the mutually exclusive resources, to obtain a target resource fusion feature, wherein the mutually exclusive resources are determined from the plurality of candidate resources based on a resource mutual exclusion rule; According to the target resource fusion feature, a second recommended resource among the plurality of candidate resources is obtained.

13. The apparatus according to claim 12, further comprising: An updating module, configured to update the recommended resources according to a page operation on the interactive page to obtain new recommended resources; as well as The second display module is used to display the new recommended resource on the interactive page.

14. The device according to claim 13, wherein The update module includes: an intention recognition submodule, configured to recognize the operation intention according to the page operation attribute corresponding to the page operation, and obtain an intention recognition result; and The first updating submodule is configured to update the recommended resources when the intention recognition result representation satisfies an updating condition.

15. The device according to claim 14, wherein The intention recognition submodule includes: a comparing unit, configured to compare the page operation attribute with a preset update condition to obtain a comparison result; and The intention recognition result determination unit is used to determine the intention recognition result according to the comparison result.

16. The device according to claim 14, wherein The page operation attributes include at least one of the following: Page operation duration, page operation type, and operation input information corresponding to the page operation.

17. The device according to claim 13, wherein The update module includes: A second updating submodule is configured to update the current deep learning model to obtain an updated deep learning model when the page operation satisfies an update condition; and The recommended resource acquisition submodule is used to process at least one of the candidate resource features according to the updated deep learning model to obtain the new recommended resource.

18. The device according to claim 17, wherein The second updating submodule includes: A resource interaction information determining unit, configured to determine resource interaction information corresponding to the current recommended resource set; a reward information determining unit, configured to determine reward information based on the resource interaction information; and A model parameter updating unit is used to update the model parameters of the current deep learning model according to the reward information.

19. The device according to claim 18, wherein The resource interaction information includes at least one of the following indicators corresponding to the recommended resource set: Resource click-through rate indicator, resource conversion rate indicator, resource transaction value indicator, and resource browsing time indicator.

20. The device according to claim 12, wherein The deep learning model includes a feature fusion network; The first recommended resource obtaining module includes: a fusion submodule, configured to input the plurality of candidate resource features into the feature fusion network and output a fusion feature; and The first recommended resource obtaining submodule is configured to input the fusion feature into the recommended resource generation network and output the at least one first recommended resource.

21. The device according to any one of claims 13 to 20, wherein The candidate resource characteristics include at least one of the following: Search information corresponding to the candidate resource, resource title information of the candidate resource, resource description information of the candidate resource, and value information of the candidate resource.

22. The device according to any one of claims 13 to 20, wherein The recommended resources include a plurality of resources, each of the plurality of recommended resources has a recommendation order attribute, and the interactive page includes a plurality of resource locations to be updated; Wherein, the first display module includes: a target resource location determination submodule, configured to determine a target resource location associated with the recommended resource from the plurality of resource locations to be updated according to a plurality of recommendation order attributes; and The display submodule is used to display the recommended resources corresponding to the target resource location at the target resource location in the interactive page.

23. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 11.

24. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 11.

25. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 11.

Images