Resource recommendation method and device, electronic equipment and computer readable storage medium

Abstract

The invention provides a resource recommendation method and device, electronic equipment and a computer readable storage medium, relates to the technical field of artificial intelligence, in particular to the technical fields of intelligent recommendation, deep learning and the like, and can be used for application scenes such as generative search, document intelligent editing, intelligent assistants, virtual assistants, intelligent e-commerce and the like. The resource recommendation method comprises the steps of obtaining respective first interaction data of a plurality of candidate resources in response to a received resource browsing request from a to-be-recommended object; according to the first interaction data of each candidate resource, an evaluation index of each candidate resource is determined, the evaluation index comprises an interaction item, and the interaction item represents the probability that other objects become the target object type after the other objects browse the candidate resources; and according to the respective evaluation index of each candidate resource, recommending at least one target resource in the plurality of candidate resources to the to-be-recommended object.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to technical fields such as intelligent recommendation and deep learning, and can be used in application scenarios such as generative search, intelligent document editing, intelligent assistants, virtual assistants, and intelligent e-commerce. More specifically, it relates to a resource recommendation method and device, an electronic device, and a computer-readable storage medium. Background Art

[0002] With the development of the Internet and mobile communication technology, there are more and more scenarios for recommending Internet information to people, and e-commerce has become an indispensable part of people's daily lives. For example, live broadcast rooms, videos, articles, news, and pictures can be recommended to people.

[0003] Invention Resources

[0004] The present disclosure provides a resource recommendation method and device, an electronic device, and a computer-readable storage medium.

[0005] According to one aspect of the present disclosure, a resource recommendation method is provided, comprising: in response to receiving a resource browsing request from an object to be recommended, obtaining first interaction data of each of a plurality of candidate resources; determining an evaluation index of each of the candidate resources based on the first interaction data of each of the candidate resources, wherein the evaluation index includes an interaction item, and the interaction item represents a probability that the other object becomes a target object type after the other object browses the candidate resource; and recommending at least one target resource among the plurality of candidate resources to the object to be recommended based on the evaluation index of each of the candidate resources.

[0006] According to another aspect of the present disclosure, a resource recommendation device is provided, including: an acquisition module, used to obtain first interaction data of each of a plurality of candidate resources in response to receiving a resource browsing request from an object to be recommended; a first determination module, used to determine an evaluation index of each of the candidate resources according to the first interaction data of each of the candidate resources, wherein the evaluation index includes an interaction item, and the interaction item represents a probability that the other object becomes a target object type after the other object browses the candidate resource; and a recommendation module, used to recommend at least one target resource among the plurality of candidate resources to the object to be recommended according to the evaluation index of each of the candidate resources.

[0007] According to another aspect of the present disclosure, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0008] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the steps of the above method are implemented.

[0009] According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program or instructions, wherein the computer program or instructions implement the steps of the above method when executed by a processor.

[0010] It should be understood that the resources 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 easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0012] Figure 1 The system architecture to which the resource recommendation method according to the embodiment of the present disclosure can be applied is schematically shown;

[0013] Figure 2 A flowchart of a resource recommendation method according to an embodiment of the present disclosure is schematically shown;

[0014] Figure 3A An example schematic diagram schematically shows a process of obtaining a trained model according to an embodiment of the present disclosure;

[0015] Figure 3B An example schematic diagram of a process of determining an evaluation index for each candidate resource according to first interaction data for each candidate resource according to an embodiment of the present disclosure is schematically shown;

[0016] Figure 4 An example schematic diagram of a process of determining an evaluation index for each candidate resource according to first interaction data for each candidate resource according to another embodiment of the present disclosure is schematically shown;

[0017] Figure 5A An exemplary schematic diagram schematically illustrates a process of recommending at least one target resource from among a plurality of candidate resources to a to-be-recommended object according to an evaluation index of each candidate resource according to an embodiment of the present disclosure;

[0018] Figure 5B An example schematic diagram schematically shows a process of obtaining an initial evaluation value according to an embodiment of the present disclosure;

[0019] Figure 6 An example schematic diagram schematically shows a resource recommendation process according to an embodiment of the present disclosure;

[0020] Figure 7 A block diagram schematically shows a resource recommendation device according to an embodiment of the present disclosure; and

[0021] 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

[0022] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0023] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0024] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0025] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0026] E-commerce live streaming combines live streaming technology with the convenience of online shopping, providing consumers with a brand new shopping experience. Live streaming e-commerce platforms display products in the form of live videos, allowing consumers to interact with the anchor in real time, understand product details, and make purchasing decisions.

[0027] In one example, the recommendation algorithm of the live broadcast e-commerce platform may include at least one of the following: a rule-based recommendation method, a collaborative filtering-based recommendation method, a content-based recommendation method, and a deep learning model-based recommendation method. The above methods all rely on two evaluation dimensions: content quality and e-commerce performance. Content quality is usually measured by indicators such as live broadcast duration and viewing duration, while e-commerce performance is evaluated by indicators such as click-through rate, conversion rate, and transaction amount. However, the above two evaluation dimensions do not take into account the interactivity between the anchor and the audience, which limits the effectiveness of the live broadcast recommendation system, resulting in the potential of user experience and platform commercial benefits not being fully tapped. Therefore, how to accurately and efficiently recommend high-quality live broadcast content to users is an urgent problem to be solved.

[0028] To this end, the embodiment of the present disclosure proposes a resource recommendation scheme. For example, in response to receiving a resource browsing request from an object to be recommended, first interaction data of each of multiple candidate resources is obtained; based on the first interaction data of each candidate resource, an evaluation index of each candidate resource is determined, wherein the evaluation index includes an interaction item, which represents the probability that the other object becomes a target object type after the other object browses the candidate resource; and, based on the evaluation index of each candidate resource, at least one target resource among the multiple candidate resources is recommended to the object to be recommended.

[0029] According to an embodiment of the present disclosure, by collecting and analyzing the first interaction data of candidate resources, it is possible to conduct a real-time personalized evaluation of each candidate resource, and obtain the respective evaluation index of each candidate resource, which helps to improve the accuracy of subsequent resource recommendations. Since the evaluation index includes an interaction item, the interaction item takes into account the probability that other objects become the target object type after other objects browse the candidate resources. By introducing an interaction item that can characterize the fan conversion rate, the attractiveness of candidate resources and user retention can be measured, reflecting the importance of interaction between candidate resources and fans for resource recommendations. On this basis, by combining other evaluation items included in the evaluation index, it is helpful to improve user engagement and retention rate, and better capture the user's diverse needs and real-time interests, so as to achieve a more comprehensive and accurate evaluation of candidate resources. This multi-dimensional evaluation method helps to achieve more accurate and personalized resource recommendations.

[0030] In the technical solution of the present invention, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0031] In the technical solution of the present invention, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0032] Figure 1The system architecture to which the resource recommendation method according to the embodiment of the present disclosure can be applied is schematically shown. It should be noted that: Figure 1 What is shown is merely an example of a system architecture to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical resources of the present disclosure, but it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0033] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0034] The user may use at least one of the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0035] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0036] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0037] It should be noted that the resource recommendation method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the resource recommendation device provided in the embodiment of the present disclosure can generally be set in the server 105. The resource recommendation method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the resource recommendation device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0038] Alternatively, the resource recommendation method provided in the embodiment of the present disclosure may also be executed by the first terminal device 101, the second terminal device 102 or the third terminal device 103, or may also be executed by other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103. Accordingly, the resource recommendation device provided in the embodiment of the present disclosure may also be arranged in the first terminal device 101, the second terminal device 102 or the third terminal device 103, or may be arranged in other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103.

[0039] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0040] It should be noted that the sequence numbers of the operations in the following method are only used as representations of the operations for the purpose of description, and should not be regarded as representing the execution order of the operations. Unless explicitly stated, the method does not need to be executed completely in the order shown.

[0041] The above describes the system architecture to which the resource recommendation method provided by the present disclosure can be applied. Figure 2 As an example, the resource recommendation process of the present disclosure is further explained.

[0042] Figure 2 The flowchart of the resource recommendation method according to the embodiment of the present disclosure is schematically shown.

[0043] like Figure 2 As shown, the resource recommendation method 200 includes operations S210 to S230.

[0044] In operation S210, in response to receiving a resource browsing request from an object to be recommended, first interaction data of each of a plurality of candidate resources is obtained.

[0045] In operation S220, an evaluation index for each candidate resource is determined according to the first interaction data for each candidate resource, where the evaluation index includes an interaction item, which represents a probability that the other object becomes a target object type after the other object browses the candidate resource.

[0046] In operation S230, at least one target resource among the plurality of candidate resources is recommended to the object to be recommended according to the evaluation index of each candidate resource.

[0047] The recommended object refers to the user who is browsing resources. The resource browsing request refers to the user's request to browse and view specific resources in the network or system. After the user sends a resource browsing request, the server will return the corresponding target resource to the user from multiple candidate resources so that the user can view or obtain the required information.

[0048] Candidate resources can refer to resources with real-time interactivity, high user engagement, content diversity, social attributes and commercialization potential. Real-time interactivity means that users can participate in real time during the live broadcast and interact with the anchor or other users. High user engagement means that users not only watch the live broadcast content, but also express their preferences and attitudes through likes, comments, sharing, purchases, etc. Content diversity refers to content that can cover multiple fields such as life, entertainment and education. Social attributes mean that users can make new friends, join fan groups or communities, and form close social relationships during the live broadcast. Commercialization potential means that profits can be achieved through product sales, advertising, rewards, e-commerce sales, course sales, membership services, etc.

[0049] The type of resources can be configured according to actual business needs and is not limited here. For example, the resources may include at least one of the following: e-commerce platform live broadcast, social media, short video platform, music player, and online education platform live broadcast, etc.

[0050] After receiving a resource browsing request, the first interaction data of each of the multiple candidate resources can be obtained. The first interaction data refers to the statistical data of the interaction between the audience and the anchor in each candidate resource, and may include the real-time interaction data and real-time user behavior data of the candidate resource. Real-time interaction data refers to the data generated by the communication and interaction between users, for example, it may include the number of comments, the number of likes, and the number of gifts. Real-time user behavior data refers to the data generated by various operations and behaviors of users on the platform, for example, it may include the live broadcast time, viewing time, click behavior, search history, number of logins, and live broadcast information followed.

[0051] After obtaining the first interaction data of each candidate resource, each first interaction data may be processed separately to obtain the evaluation index of each candidate resource. The method for obtaining the evaluation index may be configured according to actual business needs and is not limited here. For example, the method for obtaining the evaluation index may include a model-based method and a rule-based method.

[0052] The evaluation index may include multiple evaluation items, which may be configured according to actual business needs and are not limited here. For example, if the candidate resource is news report information, the multiple evaluation items may include indicators indicating the timeliness of the candidate resource, click-through rate, indicators indicating interactive data, etc. For another example, if the candidate resource is a game, the multiple evaluation items may include online time, number of games started, total transaction amount, etc.

[0053] In one example, multiple evaluation items may include an interaction item. The interaction item may represent the probability that other objects become the target object type after other objects browse the candidate resource. The target object type may be, for example, a fan. The interaction item may represent the probability that other objects become fans of the candidate resource after the candidate resource is browsed by other objects. The interaction item may emphasize the importance of the interaction between the anchor and the fans for the recommendation system. It can be used not only to measure the attractiveness of the candidate resource, but also to reflect the closeness of the relationship between the anchor and the fans.

[0054] After obtaining the evaluation index of each candidate resource, the predicted evaluation values ​​of the multiple evaluation items for the candidate resource can be used as variables to obtain the total evaluation value of each candidate resource. For example, the average daily online time of other objects can be determined based on the first interaction data, and the average value can be percentileized to serve as the predicted evaluation value of the online time. For another example, the click-through rate of other objects for multiple real-time news can be determined based on the first interaction data, and the predicted evaluation value of the click-through rate can be obtained by percentileizing the click-through rate.

[0055] After obtaining the respective predicted evaluation values ​​of each candidate resource, at least one target resource can be determined from the multiple candidate resources based on the predicted evaluation values, and the at least one target resource can be recommended to the recommended object. The method of determining the target resource can be configured according to actual business needs and is not limited here. For example, multiple candidate resources can be sorted from high to low according to the predicted evaluation values, and at least some or all of the candidate resources with higher predicted evaluation values ​​can be used as target resources.

[0056] In one example, the resource recommendation method can be applied to scenarios such as e-commerce platforms, content platforms, social platforms, and advertising marketing. For example, for e-commerce platforms, the resource recommendation method provided in the present disclosure can be used to improve the effect of live broadcasts of product sales, and to increase sales conversion rates by optimizing live content recommendations. Alternatively, for content platforms that attract users with content, the resource recommendation method provided in the present disclosure can be used to consider user interaction to enhance the attractiveness of the content and viewing time. Alternatively, for social platforms with strong social attributes, the resource recommendation method provided in the present disclosure can be used to enhance interaction between users and community stickiness. Alternatively, for advertising marketing scenarios, the resource recommendation method provided in the present disclosure can be used to improve the effect and return on advertising.

[0057] According to an embodiment of the present disclosure, by collecting and analyzing the first interaction data of candidate resources, it is possible to conduct a real-time personalized evaluation of each candidate resource, and obtain the respective evaluation index of each candidate resource, which helps to improve the accuracy of subsequent resource recommendations. Since the evaluation index includes an interaction item, the interaction item takes into account the probability that other objects become the target object type after other objects browse the candidate resources. By introducing an interaction item that can characterize the fan conversion rate, the attractiveness of candidate resources and user retention can be measured, reflecting the importance of interaction between candidate resources and fans for resource recommendations. On this basis, by combining other evaluation items included in the evaluation index, it is helpful to improve user engagement and retention rate, and better capture the user's diverse needs and real-time interests, so as to achieve a more comprehensive and accurate evaluation of candidate resources. This multi-dimensional evaluation method helps to achieve more accurate and personalized resource recommendations.

[0058] In the embodiment of the present disclosure, the method for obtaining the interaction item may include a model-based method, that is, the interaction item may be obtained by processing the first interaction data using a trained model. Figure 3A A schematic description of how to obtain a trained model is given.

[0059] Figure 3A An example schematic diagram of a process for obtaining a trained model according to an embodiment of the present disclosure is schematically shown.

[0060] like Figure 3AAs shown, in 300A, historical interaction data 301 can be obtained. Historical interaction data 301 refers to the statistical data of the interaction between the audience and the anchor in each candidate resource, and may include the historical interaction data and historical user behavior data of the candidate resource. Historical interaction data refers to the data generated by the communication and interaction between users, for example, it may include the number of comments, the number of likes, and the number of gifts. Historical user behavior data refers to the data generated by various operations and behaviors of users on the platform, for example, it may include live broadcast duration, viewing duration, click behavior, search history, number of logins, and live broadcast information followed. The historical interaction data 301 may also include an interaction item label 302, which indicates whether the historical object becomes a target object type after browsing the candidate resources.

[0061] After obtaining the historical interaction data 301, feature extraction processing can be performed on the historical interaction data 301 to obtain historical interaction features 302. On this basis, a training sample set can be generated according to the historical interaction features 302 and the interaction item labels 302, and the model to be trained 304 can be trained with this. The model to be trained 304 can be configured according to actual business needs, which is not limited here.

[0062] For example, the model 304 to be trained may include at least one of the following: Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Logistic Regression (LR), Decision Tree (Decision Tree), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and a model based on the attention mechanism (Attention), etc.

[0063] In one example, the historical interaction feature 302 can be input into the model to be trained 304 to obtain the predicted interaction item 305. The predicted interaction item 305 can represent the predicted probability that the historical object becomes the target object type after the historical object browses the candidate resources. On this basis, the predicted interaction item 305 and the interaction item label 302 can be input into the loss function 306 to obtain the loss function value 307. Based on the loss function value 307, the model parameters of the model to be trained 304 are adjusted to obtain a trained model. In this process, a cross-validation method such as K-Fold Cross-Validation can also be used to more accurately estimate the generalization ability of the model through multiple training and testing.

[0064] According to the embodiments of the present disclosure, by utilizing historical interaction features and interaction item labels to train the model, based on a closed-loop mechanism of feature extraction, model prediction and feedback optimization, and by utilizing the comparison between predicted interaction items and interaction item labels, the model can automatically adjust parameters so that the model to be trained can continuously optimize its prediction ability for interaction items, thereby gradually improving the prediction accuracy, so as to provide more accurate recommendation services in the future.

[0065] The above describes how to obtain a trained model. Figure 3B The process of using the trained model to process the first interaction data to obtain the interaction term is schematically described.

[0066] Figure 3B An example schematic diagram of a process of determining an evaluation index for each candidate resource according to first interaction data for each candidate resource according to an embodiment of the present disclosure is schematically shown.

[0067] like Figure 3B As shown, in 300B, in the real-time recommendation stage, the first interaction data 308 may be feature extracted to obtain the first interaction feature 309. On this basis, the first interaction feature 309 may be input into the trained model 310, so as to use the trained model 310 to predict the interaction item 311 according to the real-time data.

[0068] According to an embodiment of the present disclosure, by performing feature extraction on the first interaction data to obtain the first interaction feature, and using the trained model to process the first interaction feature, automatic interaction item prediction can be achieved, thereby improving the efficiency and accuracy of obtaining interaction items, thereby being able to accurately characterize the importance of the interaction between candidate resources and fans for resource recommendation, and helping to improve the overall resource recommendation effect.

[0069] In the embodiment of the present disclosure, the method for obtaining the interaction item may include a method based on a preset rule, that is, the interaction item may be obtained by processing the first interaction data using the preset rule. Figure 4 The process of processing the first interaction data by using preset rules to obtain interaction items is schematically described.

[0070] Figure 4 An example schematic diagram of a process of determining an evaluation index for each candidate resource according to first interaction data for each candidate resource according to another embodiment of the present disclosure is schematically shown.

[0071] like Figure 4As shown, in 400, at the real-time recommendation stage, at least one of a method based on domain knowledge and a method based on data analysis can also be used to obtain a preset rule 403. The preset rule 403 is a set of rules for predicting interaction items that are manually designed and defined based on the relationship between the historical interaction data 401 and the historical interaction items 402. On this basis, the first interaction data 404 can be processed based on the preset rule 403 to obtain the interaction item 405.

[0072] The domain knowledge-based approach is to develop rules based on expert knowledge and experience, using the understanding of the problem domain. In this approach, domain experts create a set of rules based on relevant knowledge and assumptions, which can help predict interaction items. The data analysis-based approach is to analyze historical interaction data, explore potential patterns and regularities in the data, and convert them into rules for predicting interaction items.

[0073] According to the embodiments of the present disclosure, by using the relationship between historical interaction data and historical interaction items to determine preset rules, the computational complexity can be simplified and the efficiency of data processing can be improved.

[0074] On this basis, by processing the first interaction data based on the preset rules, automatic interaction item prediction can be achieved, which improves the efficiency and accuracy of obtaining interaction items, thereby accurately characterizing the importance of interaction between candidate resources and fans for resource recommendation, and helping to improve the overall resource recommendation effect.

[0075] In the embodiment of the present disclosure, after obtaining the evaluation index of each candidate resource, the resource evaluation value can be determined, and at least one target resource can be determined from multiple candidate resources based on the resource evaluation value. Figure 5A The process of obtaining resource assessment values ​​and determining target resources is schematically described.

[0076] Figure 5A An example schematic diagram of a process of recommending at least one target resource from a plurality of candidate resources to a to-be-recommended object according to an evaluation index of each candidate resource according to an embodiment of the present disclosure is schematically shown.

[0077] like Figure 5A As shown, in 500A, for each candidate resource, a resource evaluation value 507 may be determined according to preset weights and evaluation indicators. Resource evaluation value 507 represents a comprehensive evaluation result of the candidate resource under specific conditions according to the set weights and evaluation indicators.

[0078] In one example, the evaluation indicators also include multiple content items 504 and multiple entity items 505. For the interaction item 501, the interaction evaluation value 503 can be determined based on the interaction item 501 and the first weight 502. The first weight 502 can be determined according to the degree of influence of the interaction item 501 on the resource recommendation, and is used to characterize the importance of the interaction item 501 to the resource evaluation value 507. For example, indicators with high importance should have higher weights. In addition, the initial evaluation value 506 can also be determined based on the multiple content items 504 and the multiple entity items 505. On this basis, the resource evaluation value 507 can be determined based on the interaction evaluation value 503 and the initial evaluation value 506.

[0079] In a specific example, the resource evaluation value 507 is determined as shown in the following formula (1).

[0080] score new =score old *(1+w0*follow_q) (1)

[0081] Among them, score new Characterizes resource evaluation value 507, score old represents the initial evaluation value 506 , w0 represents the first weight, and follow_q represents the interaction term 501 .

[0082] According to the embodiments of the present disclosure, since the first weight represents the importance of the interaction item to the resource evaluation value, by introducing the interaction item and the first weight, the user behavior can be accurately quantified. In addition, in the process of obtaining the resource evaluation value, not only the interaction evaluation value is considered, but also the initial evaluation value determined based on the content item and the entity item is combined. The interaction evaluation value can reflect the interaction between the user and the candidate resource, and the initial evaluation value can reflect the objective attributes of the candidate resource. This multi-dimensional resource evaluation method can make the resource evaluation value more accurate and comprehensive, which helps to further improve the accuracy of subsequent resource recommendations.

[0083] After obtaining the resource evaluation value 507 of each candidate resource, multiple candidate resources (e.g., candidate resource 508_1, candidate resource 508_2, ..., candidate resource 508M) can be sorted according to the resource evaluation value 507 of each candidate resource to obtain multiple sorted candidate resources. On this basis, at least one target resource 509 can be recommended to the recommended object in the order of the multiple sorted candidate resources. The recommendation method can be configured according to actual business needs and is not limited here. For example, the recommendation method can include at least one of the following: list recommendation, flow layout recommendation, card recommendation, carousel / slide recommendation, pop-up / floating window recommendation, etc.

[0084] According to the embodiments of the present disclosure, since the resource evaluation value is determined based on preset weights and evaluation indicators, it is possible to ensure that the resource evaluation is more comprehensive and accurate during the resource recommendation process, and the flexible setting of evaluation indicators and preset weights can improve the accuracy of the resource evaluation value, which helps to further improve the accuracy of subsequent resource recommendations. On this basis, by sorting the candidate resources according to the resource evaluation value of each candidate resource, high-quality resources can be quickly identified and recommended with priority, thereby improving the speed of resource screening and optimizing the responsiveness of resource recommendations. By determining at least one target resource recommended as a to-be-recommended object among multiple candidate resources in the order of the sorted candidate resources, it is possible to ensure that the target resource is the optimal resource generated after multi-dimensional evaluation, thereby helping to improve the user experience and better meet user needs.

[0085] The above describes the process of obtaining resource evaluation values ​​and determining target resources. Figure 5B A schematic description is given of a process of obtaining an initial resource value based on a content item and an entity item.

[0086] Figure 5B The following is a schematic diagram schematically showing an example of a process for obtaining an initial evaluation value according to an embodiment of the present disclosure.

[0087] like Figure 5B As shown, in 500B, for each content item, a content evaluation value 512 can be determined based on the content data 510 of each of the multiple content items and the second weight 511 of the content item. For example, the content data 510 of each content item can be multiplied by the corresponding second weight 511 to obtain a first product value; the first product values ​​of each content item can be added to obtain a first cumulative value; and the first cumulative value can be determined as the content evaluation value 512. The second weight 511 can be determined based on the influence of the content item on the resource recommendation, and is used to characterize the importance of the content item to the resource evaluation value. For example, a content item with high importance should have a higher second weight 511.

[0088] The content item refers to an evaluation item related to the user's access duration to the candidate resource. The content item may include at least one of the following: a first access duration and a second access duration. The second access duration may be greater than the first access duration. For example, the first access duration may be 3 seconds, and the second access duration may be 20 seconds. Alternatively, the first access duration may be 30 seconds, and the second access duration may be 60 seconds, 200 seconds, etc.

[0089] For each entity item, an entity evaluation value 515 can be determined based on the entity data 513 of each of the multiple entity items and the third weight 514 of the entity item. For example, the entity data 513 of each entity item can be multiplied by the corresponding third weight 514 to obtain a second product value; the second product values ​​of each entity item can be added to obtain a second cumulative value; and the second cumulative value can be determined as the entity evaluation value 515. The third weight 514 can be determined based on the degree of influence of the entity item on the resource recommendation, and is used to characterize the importance of multiple entity items to the resource evaluation value. For example, an entity item with high importance should have a higher third weight 514.

[0090] Entity items refer to evaluation items used to measure the effectiveness of resource recommendations in different aspects. Entity items may include at least one of the following: click-through rate (CTR), click-through conversion rate (CVR), conversion rate (CTCVR) and gross merchandise volume (GMV). Click-through rate refers to the proportion of users who actually click to enter the live broadcast room after seeing the live broadcast recommendation. Click-to-conversion rate refers to the proportion of users who enter the live broadcast room after clicking on the live broadcast recommendation and further complete the preset conversion behavior (such as purchasing goods, following the anchor, etc.). The conversion rate refers to the overall efficiency of users from seeing the recommendation, clicking on the recommendation to finally completing the purchase or conversion behavior. The gross merchandise volume refers to the total amount of commodity transactions achieved through live broadcast recommendations within a certain period of time.

[0091] On this basis, the initial evaluation value 516 can be determined according to the content evaluation value 512 and the entity evaluation value 515. In a specific example, the initial evaluation value 516 can be determined as shown in the following formula (2) and formula (3).

[0092] score old =score base *(1+click dura_20 )·(1+2·click dura_60 )·(1+2·click dura_200 )·(1+5·dura_30) (2)

[0093]

[0094] Among them, click dura_20 Represents the probability that the viewing time of the candidate resource is greater than 20 seconds, click dura_60 Represents the probability that the viewing time of the candidate resource is greater than 60 seconds, click dura_200represents the probability that the viewing time of a candidate resource is greater than 200 seconds, dura_30 represents the probability that the stay time of a candidate resource is greater than 30 seconds, w1~w8 represent the weights, dura_3s q Indicates the probability that the click duration of a candidate resource is greater than 3 seconds. dura_30s_q Indicates the probability that the click duration of a candidate resource is greater than 30 seconds, ctr q Characterize click-through rate, ctcvr q represents the conversion rate, price_factor represents the price factor, gmv q Indicators that represent conversion benefits.

[0095] In one example, different calculation methods can be used according to different object types. Object types can be distinguished by preset type classification rules, and the preset type classification rules consider users who have purchased products in a preset time period as core users. For example, for non-core users, methods w1 to w4 can be used for calculation. Alternatively, for core users, methods w5 to w8 can be used for calculation.

[0096] According to the embodiments of the present disclosure, different weights are used to evaluate content items and entity items, respectively. Since different weights characterize the importance of each content item and entity item to the resource evaluation value, accurate evaluation of the content evaluation value and the entity evaluation value is achieved. In addition, since the weights can be adjusted according to the actual scenario, this weight adjustment mechanism can also ensure the flexibility and dynamism of the evaluation process. On this basis, a more comprehensive initial evaluation value is generated by integrating the content evaluation value and the entity evaluation value, which improves the accuracy and comprehensiveness of the subsequent resource evaluation value, thereby contributing to the resource recommendation effect.

[0097] In the embodiment of the present disclosure, after at least one target resource is recommended to the object to be recommended, the weights of various evaluation items can also be dynamically adjusted based on the resource recommendation effect. Figure 6 The dynamic adjustment process of obtaining weights is schematically described.

[0098] Figure 6 An example schematic diagram of a resource recommendation process according to an embodiment of the present disclosure is schematically shown.

[0099] like Figure 6As shown, in 600, after obtaining the first interaction data 601, the second interaction data 602 can also be obtained at a predetermined interval. According to the changes between the first interaction data 601 and the second interaction data 602 of each of the multiple candidate resources, the resource recommendation effect 603 is determined. The resource recommendation effect 604 may refer to evaluating whether these recommended target resources meet the needs of the object to be recommended and whether they improve the experience of the object to be recommended after recommending at least one target resource to the object to be recommended. The resource recommendation effect 604 may include at least one of the following: accuracy, viewing experience, engagement, retention rate, satisfaction, etc.

[0100] After obtaining the resource recommendation effect 603, the target weight 603 can be adjusted according to the resource recommendation effect 603, thereby being able to continue to adjust the weights of various indicators according to the actual application effect, thereby further optimizing the recommendation process. The target weight 603 may include at least one of a first weight representing the importance of the interaction item to the resource evaluation value, a second weight representing the importance of the content item to the resource evaluation value, and a third weight representing the importance of the entity item to the resource evaluation value.

[0101] According to the embodiments of the present disclosure, by evaluating the resource recommendation effect based on the second interaction data and dynamically adjusting the target weight according to the resource recommendation effect, the resource recommendation process is continuously optimized according to real-time data, thereby quickly adapting to market changes and changes in user needs, improving the flexibility, accuracy and response speed of resource recommendation, and realizing adaptive optimization of the resource recommendation process. By combining real-time feedback from users, the system can dynamically adjust the weights of different dimensions to continuously optimize the recommendation strategy and ensure that the recommendation results are more in line with user needs.

[0102] In one example, taking the personalized recommendation scenario of an e-commerce live broadcast platform as an example, the historical interaction data of each live broadcast room in the e-commerce live broadcast platform can be used to train the model to be trained to obtain a trained model. When resource recommendation is required, the first interaction data is processed using the trained model to predict the fan conversion rate of each live broadcast room, and the resource evaluation value of each live broadcast room is recalculated. Based on the resource evaluation value of each live broadcast room, priority is given to recommending live broadcast rooms that are not only rich in content and have good commercial conversion, but also in line with user interests, highly interactive with users, and have high fan conversion potential.

[0103] On this basis, the resource recommendation effect can also be determined based on the second interaction data after the recommendation, and the model parameters and weights can be continuously adjusted based on the resource recommendation effect to achieve the best recommendation effect. As a result, the e-commerce live broadcast platform can more effectively improve user experience and e-commerce conversion, increase user stickiness and platform e-commerce conversion rate, and at the same time motivate anchors to improve the quality of live broadcast interaction, creating a win-win result.

[0104] In one example, taking the growth scenario of live content creators as an example, the e-commerce live broadcast platform can provide interactive feedback to creators to help creators understand audience needs, optimize live broadcast strategies, analyze user preferences and optimize live broadcast content and style to increase user engagement. For example, the conversion rate can be increased by enhancing interaction with users (for example, interactive games, question-and-answer sessions, gift giving), and personalized recommendations can be made based on the above methods to gain more exposure opportunities, thereby incentivizing the production of more high-quality content. In addition, the analysis tools provided by the platform can be used to track the conversion rate and other evaluation indicators, so as to continuously optimize the live broadcast strategy, improve the live broadcast performance, and thus increase the commercial value of the live broadcast room.

[0105] In one example, taking marketing activities and advertising delivery scenarios as an example, the potential effect of marketing activities can be predicted by analyzing the user portraits and interaction items of different live broadcast content, and the best advertising delivery or cooperative live broadcast room can be determined. In addition, marketing strategies and delivery content can be dynamically adjusted based on the real-time feedback and conversion rate of the activities, effectively driving higher e-commerce conversion rates, that is, more effectively converting traffic into sales to achieve maximum marketing effects, thereby improving return on investment, brand awareness and sales.

[0106] The above are merely exemplary embodiments, but are not limited thereto, and may also include other resource recommendation methods known in the art, as long as more accurate and personalized resource recommendations can be achieved.

[0107] Figure 7 The block diagram of the resource recommendation device according to an embodiment of the present disclosure is schematically shown.

[0108] like Figure 7 As shown, the resource recommendation device 700 may include an acquisition module 710 , a first determination module 720 and a recommendation module 730 .

[0109] The acquisition module 710 is configured to acquire first interaction data of each of a plurality of candidate resources in response to receiving a resource browsing request from the object to be recommended.

[0110] The first determination module 720 is used to determine the evaluation index of each candidate resource according to the first interaction data of each candidate resource, wherein the evaluation index includes an interaction item, and the interaction item represents the probability that the other object becomes the target object type after the other object browses the candidate resource.

[0111] The recommendation module 730 is used to recommend at least one target resource from among multiple candidate resources to the object to be recommended according to the evaluation index of each candidate resource.

[0112] According to an embodiment of the present disclosure, the first determining module 720 may include a feature extraction unit and a first processing unit.

[0113] The feature extraction unit is used to extract features from the first interaction data for each candidate resource to obtain a first interaction feature.

[0114] The first processing unit is used to process the first interaction feature by using the trained model to obtain the interaction item.

[0115] According to an embodiment of the present disclosure, a trained model is obtained in the following manner: obtaining a training sample set, wherein the training sample set includes historical interaction features and interaction item labels, the historical interaction features are obtained by performing feature extraction processing on historical interaction data, and the interaction item labels represent whether the historical object becomes the target object type after browsing candidate resources; inputting the historical interaction features into the model to be trained to obtain predicted interaction items, wherein the predicted interaction items represent the predicted probability that the historical object becomes the target object type after the historical object browses candidate resources; and, using the predicted interaction items and interaction item labels, adjusting the model parameters of the model to be trained to obtain a trained model.

[0116] According to an embodiment of the present disclosure, the first determining module 720 may include a second processing unit.

[0117] The second processing unit is used to process the first interaction data for each candidate resource based on a preset rule to obtain an interaction item, wherein the preset rule is determined according to the relationship between the historical interaction data and the historical interaction item.

[0118] According to an embodiment of the present disclosure, the recommendation module 730 may include a determination unit, a sorting unit, and a recommendation unit.

[0119] The determination unit is used to determine the resource evaluation value for each candidate resource according to the preset weight and evaluation index, so as to obtain the resource evaluation value of each candidate resource.

[0120] The sorting unit is used to sort multiple candidate resources according to the resource evaluation value of each candidate resource to obtain multiple sorted candidate resources.

[0121] The recommendation unit is used to recommend at least one target resource to the object to be recommended according to the order of multiple sorted candidate resources.

[0122] According to an embodiment of the present disclosure, the evaluation index further includes a plurality of content items and a plurality of entity items.

[0123] According to an embodiment of the present disclosure, the determining unit may include a first determining subunit and a second determining subunit.

[0124] The first determining subunit is used to determine the interaction evaluation value based on the interaction item and the first weight, wherein the first weight represents the importance of the interaction item to the resource evaluation value.

[0125] The second determining subunit is used to determine the resource evaluation value according to the interaction evaluation value and the initial evaluation value obtained based on the multiple content items and the multiple entity items.

[0126] According to an embodiment of the present disclosure, an initial evaluation value is obtained in the following manner: for each content item, a content evaluation value is determined based on content data of each of the multiple content items and a second weight of the content item, wherein each second weight represents the importance of the multiple content items to the resource evaluation value; for each entity item, an entity evaluation value is determined based on entity data of each of the multiple entity items and a third weight of the entity item, wherein each third weight represents the importance of the multiple entity items to the resource evaluation value; and, based on the content evaluation values ​​and the entity evaluation values, an initial evaluation value is determined.

[0127] According to an embodiment of the present disclosure, the resource recommendation device 700 may further include a second determination module and an adjustment module.

[0128] The second determination module is used to determine the resource recommendation effect according to the second interaction data of each of the plurality of candidate resources, wherein the second interaction data is obtained after a predetermined time interval with the first interaction data.

[0129] The adjustment module is used to adjust the target weight according to the resource recommendation effect, wherein the target weight is at least one of the first weight, the second weight and the third weight.

[0130] 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, workbenches, 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.

[0131] like Figure 8As shown, the 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. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0132] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0133] The computing unit 801 may be a variety of general and / or special processing components 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 dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, 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, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on 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 appropriate manner (e.g., by means of firmware).

[0134] Various implementations of the systems and techniques described above herein 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), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including 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.

[0135] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0136] 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, device, or equipment. 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, device, or equipment, or any suitable combination of the above resources. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above resources.

[0137] 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).

[0138] The systems and techniques described herein may 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 may 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.

[0139] A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on 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 combined with a blockchain.

[0140] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0141] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A resource recommendation method, comprising: In response to receiving a resource browsing request from an object to be recommended, obtaining first interaction data of each of a plurality of candidate resources; Determine, according to the respective first interaction data of each candidate resource, an evaluation index of each candidate resource, wherein the evaluation index includes an interaction item, and the interaction item represents the probability that the other object becomes a target object type after the other object browses the candidate resource; as well as At least one target resource among the multiple candidate resources is recommended to the object to be recommended according to the evaluation index of each candidate resource.

2. The method according to claim 1, wherein: Determining the evaluation index of each candidate resource according to the first interaction data of each candidate resource includes: For each of the candidate resources, performing feature extraction on the first interaction data to obtain a first interaction feature; and The first interaction feature is processed using the trained model to obtain the interaction term.

3. The method according to claim 2, wherein: The trained model is obtained as follows: Acquire a training sample set, wherein the training sample set includes historical interaction features and interaction item labels, wherein the historical interaction features are obtained by performing feature extraction processing on historical interaction data, and the interaction item labels represent whether the historical objects become the target object type after browsing the candidate resources; Inputting the historical interaction features into a model to be trained to obtain a predicted interaction item, wherein the predicted interaction item represents a predicted probability that the historical object becomes a target object type after the historical object browses the candidate resource; and The model parameters of the model to be trained are adjusted using the predicted interaction term and the interaction term label to obtain the trained model.

4. The method according to claim 1, wherein: Determining the evaluation index of each candidate resource according to the first interaction data of each candidate resource includes: For each of the candidate resources, The first interaction data is processed based on a preset rule to obtain the interaction item, wherein the preset rule is determined according to the relationship between the historical interaction data and the historical interaction item.

5. The method according to any one of claims 1 to 4, wherein: The recommending at least one target resource from among the plurality of candidate resources to the object to be recommended according to the evaluation index of each candidate resource comprises: For each of the candidate resources, determine a resource evaluation value according to a preset weight and the evaluation index, and obtain a resource evaluation value of each of the candidate resources; sorting the plurality of candidate resources according to the resource evaluation value of each of the candidate resources to obtain a plurality of sorted candidate resources; and The at least one target resource is recommended to the object to be recommended according to the order of the plurality of sorted candidate resources.

6. The method according to claim 5, wherein: The evaluation index also includes a plurality of content items and a plurality of entity items; Determining the resource evaluation value according to the preset weight and the evaluation index to obtain the resource evaluation value of each candidate resource includes: Determining an interaction evaluation value based on the interaction item and the first weight, wherein the first weight represents the importance of the interaction item to the resource evaluation value; and The resource evaluation value is determined according to the interaction evaluation value and initial evaluation values ​​obtained based on the multiple content items and the multiple entity items.

7. The method according to claim 6, wherein: The initial evaluation value is obtained by: For each of the content items, determining a content evaluation value based on the content data of each of the plurality of content items and a second weight of the content item, wherein each of the second weights represents the importance of the plurality of content items to the resource evaluation value; For each of the entity items, determining an entity evaluation value based on the entity data of each of the plurality of entity items and a third weight of the entity item, wherein each of the third weights represents the importance of the plurality of entity items to the resource evaluation value; and The initial evaluation value is determined according to the content evaluation value and the entity evaluation value.

8. The method according to claim 6 or 7, further comprising, after recommending at least one target resource from among the plurality of candidate resources to the object to be recommended according to the respective evaluation indicators of each of the candidate resources: Determine the resource recommendation effect according to the second interaction data of each of the plurality of candidate resources, wherein: The second interaction data is obtained after a predetermined time interval from the first interaction data; as well as According to the resource recommendation effect, a target weight is adjusted, wherein the target weight is at least one of the first weight, the second weight, and the third weight.

9. A resource recommendation device, comprising: An acquisition module, configured to acquire first interaction data of each of a plurality of candidate resources in response to receiving a resource browsing request from a to-be-recommended object; A first determination module is used to determine an evaluation index of each candidate resource according to the first interaction data of each candidate resource, wherein the evaluation index includes an interaction item, and the interaction item represents the probability that the other object becomes a target object type after the other object browses the candidate resource; as well as The recommendation module is used to recommend at least one target resource among the multiple candidate resources to the object to be recommended according to the evaluation index of each candidate resource.

10. The device according to claim 9, wherein: The first determining module comprises: a feature extraction unit, configured to extract features from the first interaction data for each of the candidate resources to obtain a first interaction feature; and The first processing unit is used to process the first interaction feature by using the trained model to obtain the interaction item.

11. The device according to claim 10, wherein: The trained model is obtained as follows: Acquire a training sample set, wherein the training sample set includes historical interaction features and interaction item labels, wherein the historical interaction features are obtained by performing feature extraction processing on historical interaction data, and the interaction item labels represent whether the historical objects become the target object type after browsing the candidate resources; Inputting the historical interaction features into a model to be trained to obtain a predicted interaction item, wherein the predicted interaction item represents a predicted probability that the historical object becomes a target object type after the historical object browses the candidate resource; and The model parameters of the model to be trained are adjusted using the predicted interaction term and the interaction term label to obtain the trained model.

12. The device according to claim 9, wherein: The first determining module comprises: The second processing unit is used to process the first interaction data based on a preset rule for each candidate resource to obtain the interaction item, wherein the preset rule is determined according to the relationship between the historical interaction data and the historical interaction item.

13. The device according to any one of claims 9 to 12, wherein: The recommendation module includes: A determination unit, configured to determine a resource evaluation value for each candidate resource according to a preset weight and the evaluation index, and obtain a resource evaluation value for each candidate resource; a sorting unit, configured to sort the plurality of candidate resources according to the resource evaluation value of each of the candidate resources to obtain a plurality of sorted candidate resources; and A recommendation unit is used to recommend the at least one target resource to the object to be recommended according to the order of the plurality of sorted candidate resources.

14. The device according to claim 13, wherein: The evaluation index also includes a plurality of content items and a plurality of entity items; The determining unit comprises: A first determining subunit is configured to determine an interaction evaluation value based on the interaction item and the first weight, wherein the first weight represents the importance of the interaction item to the resource evaluation value; and The second determining subunit is configured to determine the resource evaluation value according to the interaction evaluation value and an initial evaluation value obtained based on the plurality of content items and the plurality of entity items.

15. The device according to claim 14, wherein: The initial evaluation value is obtained by: For each of the content items, determining a content evaluation value based on the content data of each of the plurality of content items and a second weight of the content item, wherein each of the second weights represents the importance of the plurality of content items to the resource evaluation value; For each of the entity items, determining an entity evaluation value based on the entity data of each of the plurality of entity items and a third weight of the entity item, wherein each of the third weights represents the importance of the plurality of entity items to the resource evaluation value; and The initial evaluation value is determined according to the content evaluation value and the entity evaluation value.

16. The device according to claim 14 or 15, further comprising: A second determination module, configured to determine a resource recommendation effect according to second interaction data of each of the plurality of candidate resources, wherein the second interaction data is obtained after a predetermined time interval with the first interaction data; and An adjustment module is used to adjust a target weight according to the resource recommendation effect, wherein the target weight is at least one of the first weight, the second weight and the third weight.

17. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, The method is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 8.

18. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

19. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images