Multi-target reordering method and device, electronic equipment and storage medium

Abstract

The invention provides a multi-target reordering method and device, electronic equipment and a storage medium, and can be applied to the technical field of artificial intelligence and digital marketing. The method comprises: in response to an access request for a target page, obtaining a pre-recommendation result about the target page, the pre-recommendation result comprising a plurality of material features of a plurality of to-be-displayed objects, the plurality of material features being respectively related to a plurality of indexes; determining respective current weight coefficients of the plurality of indexes at the current moment; processing the respective current weight coefficients of the plurality of indexes by using a super network to obtain dynamic network parameters; performing parameter adjustment on the initial model based on the dynamic network parameters to obtain a reordering model; and inputting the plurality of material features of the plurality of to-be-displayed objects into the reordering model to obtain a reordering result.

Description

Technical Field

[0001] The present disclosure relates to the fields of artificial intelligence and digital marketing technology, and more specifically, to a multi-target reordering method, device, electronic device and storage medium. Background Art

[0002] In online shopping scenarios, when users enter the homepage or search for products, e-commerce platforms can select specific products to display to users based on user interests and merchant information to provide users with a better shopping experience. Since the platform's reserve materials come from thousands of merchants, it is necessary to use appropriate sorting algorithms to select the most suitable products to recommend to users.

[0003] For e-commerce platforms, each item from advertisers has multiple attributes as an estimate of its contribution to multiple goals. For example, if a user searches for a category of electrical appliances, the platform will estimate the probability of clicks, conversions, and billing for each brand of electrical appliances for the user. These multiple attributes are very important in determining their sorting and display process. Summary of the invention

[0004] In view of this, the present disclosure provides a multi-target reordering method, device, electronic device and storage medium.

[0005] One aspect of the present disclosure provides a multi-objective reordering method, comprising: in response to an access request for a target page, obtaining a pre-recommendation result about the target page, the pre-recommendation result including a plurality of material features of each of a plurality of objects to be displayed, the plurality of material features being respectively related to a plurality of indicators; determining a current weight coefficient of each of the plurality of indicators at a current moment; processing the current weight coefficients of each of the plurality of indicators using a hypernetwork to obtain dynamic network parameters; adjusting parameters of an initial model based on the dynamic network parameters to obtain a reordering model; and inputting the plurality of material features of each of the plurality of objects to be displayed into the reordering model to obtain a reordering result.

[0006] Another aspect of the present disclosure provides a multi-objective reordering device, including: a first acquisition module, used to respond to an access request for a target page, obtain a pre-recommendation result about the target page, the pre-recommendation result includes a plurality of material features of each of a plurality of objects to be displayed, and the plurality of material features are respectively related to a plurality of indicators; a first determination module, used to determine a current weight coefficient of each of the plurality of indicators at a current moment; a processing module, used to process the current weight coefficients of each of the plurality of indicators using a hypernetwork to obtain dynamic network parameters; a parameter adjustment module, used to adjust the parameters of an initial model based on the dynamic network parameters to obtain a reordering model; and a reordering module, used to input the plurality of material features of each of the plurality of objects to be displayed into the reordering model to obtain a reordering result.

[0007] Another aspect of the present disclosure provides an electronic device, including: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described above.

[0008] Another aspect of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the above method when executed.

[0009] Another aspect of the present disclosure provides a computer program product, which includes computer executable instructions, and when the instructions are executed, are used to implement the method as described above.

[0010] According to the embodiments of the present disclosure, the current weight coefficients of multiple indicators at the current moment can be determined in real time based on the traffic feature information at the current moment. In combination with the super network, the network parameters of the reordering model can be adjusted based on the current weight coefficients of the multiple indicators, and the scores of multiple objects to be displayed at the current moment can be performed based on the adjusted reordering model to achieve dynamic multi-target reordering. The optimality of the platform sorting strategy can be guaranteed in real time when the time periods are different and the user characteristics change, so that the displayed objects are more in line with the objects required by the users, thereby improving the click-through rate of the objects to be displayed and enhancing the user experience. 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 An exemplary system architecture to which a multi-objective reordering method or apparatus according to an embodiment of the present disclosure can be applied is schematically shown.

[0013] Figure 2 The flowchart of the multi-objective reordering method according to an embodiment of the present disclosure is schematically shown.

[0014] Figure 3 The following schematically shows a schematic diagram of a recommendation system according to an embodiment of the present disclosure.

[0015] Figure 4 A schematic diagram of a constraint integrated controller according to an embodiment of the present disclosure is schematically shown.

[0016] Figure 5 The figure schematically shows a hypernetwork model according to an embodiment of the present disclosure.

[0017] Figure 6 A schematic diagram of a reordering system according to an embodiment of the present disclosure is schematically shown.

[0018] Figure 7 A block diagram of a multi-target reordering device according to an embodiment of the present disclosure is schematically shown.

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

[0020] 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.

[0021] 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.

[0022] 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.

[0023] 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.).

[0024] Reordering is an important component of the recommendation system of e-commerce platforms. It receives a part of the items that have been screened from the upstream system and reorders them so that the item sequence finally presented to the user can meet the assessment requirements of multiple goals. The multiple goals may include total clicks, total charges, conversion rate, etc.

[0025] In the related art, the multi-objective re-ranking methods used by e-commerce platforms are roughly divided into two categories. One is the fixed slot allocation method, which usually pre-allocates different ad slots to items with different attributes. For example, there are 6 ad slots in total. The 1st, 3rd, and 5th ad slots can be allocated to items with higher click-through rates, and the 2nd, 4th, and 6th ad slots can be allocated to items with higher billings, so as to ensure that multiple goals are taken into account. The other is a sorting method based on static multi-objective weighting. In order to uniformly sort all materials, the e-commerce platform can score all materials uniformly according to some indicators, such as click-through rate, conversion rate, billing, etc. Due to the different dimensions of different indicators, a weighted average method can usually be used to obtain a unified score. Goods with higher scores can be allocated to the front ad slots. For example, all items are uniformly sorted according to a weight of 50% for clicks, 30% for billing, and 20% for conversions.

[0026] However, the related art usually adopts a heuristic approach to design multi-objective re-ranking rules. For example, based on experience, it is judged that different advertising positions should be allocated to materials that are more click-oriented or more consumption-oriented. This approach makes it difficult to ensure that the ranking rules are optimal. As the user and traffic characteristics of different time periods change, it can be theoretically proved that the optimal ranking rules will also change, that is, the importance of various indicators such as clicks, billing, and conversions will change. The methods proposed in the related art are based on a preset goal. When traffic characteristics change, such as when user groups and consumption habits are different, these technologies cannot guarantee real-time optimality. For example, when a user contributes more to clicks during this time period, the weight of the click should theoretically be increased, but the static multi-objective re-ranking in the related art cannot achieve adaptive adjustment of the weights.

[0027] In view of this, the embodiments of the present disclosure provide a multi-objective reordering method, device, electronic device and storage medium, which can ensure the optimality of the platform sorting strategy in real time when different time periods and user characteristics change. The multi-objective reordering method includes: in response to an access request for a target page, obtaining a pre-recommendation result about the target page, the pre-recommendation result includes multiple material features of multiple objects to be displayed, and the multiple material features are respectively related to multiple indicators; determining the current weight coefficients of the multiple indicators at the current moment; using a hypernetwork to process the current weight coefficients of the multiple indicators to obtain dynamic network parameters; adjusting the parameters of the initial model based on the dynamic network parameters to obtain a reordering model; and inputting the multiple material features of the multiple objects to be displayed into the reordering model to obtain a reordering result.

[0028] In the embodiments of the present disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage, etc. of the data involved (for example, including but not limited to user personal information) are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security, and national security.

[0029] In the embodiments of the present disclosure, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0030] Figure 1 The exemplary system architecture to which the multi-objective reordering method or apparatus 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 content 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.

[0031] 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 and / or wireless communication links, etc.

[0032] The user may use 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, and / or social platform software, etc. (only for example).

[0033] 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.

[0034] 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.

[0035] It should be noted that the multi-target reordering method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the multi-target reordering device provided in the embodiment of the present disclosure can generally be set in the server 105. The multi-target reordering 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. Correspondingly, the multi-target reordering 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. Alternatively, the multi-target reordering method provided in the embodiment of the present disclosure can also be executed by the first terminal device 101, the second terminal device 102 or the third terminal device 103, or by other terminal devices that are different from the first terminal device 101, the second terminal device 102 or the third terminal device 103. Accordingly, the multi-target reordering 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 in other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103.

[0036] It should be understood that Figure 1The 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.

[0037] Figure 2 The flowchart of the multi-objective reordering method according to an embodiment of the present disclosure is schematically shown.

[0038] like Figure 2 As shown, the method includes operations S201 to S205.

[0039] In operation S201 , in response to an access request for a target page, a pre-recommendation result about the target page is obtained.

[0040] In operation S202, current weight coefficients of each of the plurality of indicators at the current moment are determined.

[0041] In operation S203, the current weight coefficients of the respective multiple indicators are processed using the hypernetwork to obtain dynamic network parameters.

[0042] In operation S204, the initial model is adjusted based on the dynamic network parameters to obtain a reordering model.

[0043] In operation S205, a plurality of material features of the plurality of objects to be displayed are input into a reordering model to obtain a reordering result.

[0044] When a user visits a website homepage, searches for a specific object, etc., the device may generate an access request. Accordingly, the target page for the access request may be the website homepage, a search result page for a specific object, etc.

[0045] The pre-recommendation results for the target page can be obtained from the upstream recommendation system. The recommendation system can use recommendation models, such as GLRM (Generalized Low Rank Models), to filter multiple objects to be displayed from multiple objects and output pre-recommendation results. For example, in an online shopping scenario, the recommendation system can be represented as a subsystem of an e-commerce platform. The subsystem can use the recommendation model to filter multiple commodities to be displayed from various commodities based on the material information of various commodities provided by many merchants and the user information of the visiting users, that is, the multiple objects to be displayed.

[0046] The pre-recommendation result may include multiple material features of each of the multiple objects to be displayed. Each material feature may be related to an indicator used in reordering, and accordingly, the multiple material features may correspond to multiple indicators. The material features related to the indicator may be obtained by extracting features of the material information related to the indicator, and the specific feature extraction method is not limited here. The multiple indicators may include click-through rate, conversion rate, bid, etc. For example, for the object to be displayed, the multiple material features of the object to be displayed may include features corresponding to the click-through rate, features corresponding to the conversion rate, features corresponding to the bid, etc. The multiple indicators can be represented as multiple targets for reordering.

[0047] The current moment may be the triggering moment of the access request. The weight coefficients of the multiple indicators may be adjusted in real time based on the user characteristic information, traffic characteristic information, etc. at the current moment. Optionally, the sum of the multiple weight coefficients corresponding to the multiple indicators may be 1. The current weight coefficients of the multiple indicators may be the weight coefficients obtained after real-time adjustment at the current moment.

[0048] The reordering model can score the object to be displayed based on the material characteristics of each of the multiple indicators of the object to be displayed. The model parameters of the reordering model, or at least part of the model parameters, can change with the change of the weight coefficients of each of the multiple indicators, that is, based on the current weight coefficients of each of the multiple indicators, the current values ​​of the model parameters of the reordering model, or at least part of the model parameters, can be determined to complete the reconstruction of the reordering model.

[0049] The updating method of the model parameters of the re-ranking model is not limited here. For example, the current weight coefficients of the multiple indicators can be mapped to the model parameters of the ranking model through a set mapping relationship, or the current weight coefficients of the multiple indicators can be mapped to the change in the model parameters of the ranking model. For another example, a hypernetwork can be combined and used to update the model parameters of the re-ranking model.

[0050] Hypernetworks can be a neural network that generates weights for another neural network. The current weight coefficients of multiple indicators can be input into the hypernetwork, and the hypernetwork can output dynamic network parameters, which can be used to adjust the parameters of the reordering model. The adjustment method can include overall replacement, partial replacement, weighted adjustment, etc., which are not limited here.

[0051] The re-ranking model in this scheme can be implemented based on the SlateQ recommendation system re-ranking model, the PRM re-ranking model, etc., and the re-ranking result can be obtained by implementing advertising auctions through the DNA (Deep Neural Auction, an auction mechanism based on deep learning) model and the Deep-GSP (Deep Generalized Second Price, a generalized second price auction based on deep learning) model.

[0052] The multiple material features of the object to be displayed can form a feature matrix, which can be input into the re-ranking model to obtain the ranking score of the object to be displayed. Based on the ranking scores of the multiple objects to be displayed, a re-ranking result can be obtained.

[0053] According to the embodiments of the present disclosure, the current weight coefficients of multiple indicators at the current moment can be determined in real time based on the traffic feature information at the current moment. In combination with the super network, the network parameters of the reordering model can be adjusted based on the current weight coefficients of the multiple indicators, and the scores of multiple objects to be displayed at the current moment can be performed based on the adjusted reordering model to achieve dynamic multi-target reordering. The optimality of the platform sorting strategy can be guaranteed in real time when the time periods are different and the user characteristics change, so that the displayed objects are more in line with the objects required by the users, thereby improving the click-through rate of the objects to be displayed and enhancing the user experience.

[0054] Reference below Figure 3~Figure 6 , combined with specific embodiments Figure 2 The method shown is further explained.

[0055] When a user requests traffic for a target page, an access request for the target page can be generated, and the upstream recommendation system can respond to the access request and determine multiple objects to be displayed from the objects based on the materials of the objects. The objects to be displayed can be objects that meet the user's traffic characteristics, that is, objects that are likely to be liked by the user. The user's traffic characteristics can be determined based on the user's behavior information, Internet environment information, etc., which are not limited here.

[0056] Figure 3 The following is a schematic diagram of a recommendation system according to an embodiment of the present disclosure.

[0057] like Figure 3As shown, the recommendation system may be configured with a recommendation model 301, and the advertising data 3021, material data 3022, etc. of each of the n objects 302 may be respectively input into the recommendation model 301 to obtain the recommendation score 3023 of each of the n objects. Based on the recommendation score 3023 of each of the n objects and the number of objects to be selected, m objects to be displayed 303 with higher scores may be selected from the n objects 302. The number of objects to be selected may be determined, for example, according to the number of display positions on the target page, which is not limited here. The advertising data, material data, etc. of each of the m objects to be displayed 303 may be input into the feature extraction model 304 to obtain k material features 305 of each of the m objects to be displayed 303. The k material features 305 may be respectively related to k indicators, and the k indicators may be, for example, click-through rate, bid, conversion rate, etc. n≥m≥1, k≥1.

[0058] Optionally, in actual application scenarios, the recommendation model 301 can also score the object 302 based on other data besides the advertisement data 3021 and material data 3022 of the object 302. For example, in an advertisement recommendation scenario, the recommendation model can also score the object based on user feature data, historical session data, context of the scenario, etc.

[0059] While the recommendation system is making pre-recommendations, the current weight coefficients of each of the multiple indicators can be determined in real time.

[0060] For example, the current weight coefficients of the multiple indicators may be determined at the same time when the access request is received. At this time, determining the current weight coefficients of the multiple indicators at the current time may include the following operations:

[0061] Acquire flow characteristic data collected within a preset time period; and determine current weight coefficients of multiple indicators based on the flow characteristic data.

[0062] The preset time period may be represented as a time period with the current time as the end point and a set duration. For example, if the current time is 10:00:00 and the set duration is 2 hours, the preset time period may be represented as a time period between 08:00:00 and 10:00:00.

[0063] Traffic characteristic data may include traffic characteristic data on the user side, such as data generated from the user's clicks and access behaviors, etc., and may also include traffic characteristic data on the merchant side, such as data related to activities launched by the merchant, data generated during the merchant's sales process, etc. The traffic characteristic data may be collected when the user or merchant uses the platform after obtaining the user's or merchant's consent.

[0064] According to an embodiment of the present disclosure, the weight coefficients of the previously determined multiple indicators can be feedback-adjusted based on the traffic characteristic data in a feedback-adjusted manner to obtain the current weight coefficients of the multiple indicators. Specifically, determining the current weight coefficients of the multiple indicators based on the traffic characteristic data may include the following operations:

[0065] Based on the preset constraint conditions and the flow characteristic data, an adjustment error is obtained; and based on the adjustment error, feedback adjustment is performed on the initial weight coefficients of each of the multiple indicators to obtain the current weight coefficients of each of the multiple indicators.

[0066] The preset constraint condition can be set based on the specific application scenario. For example, in an advertising marketing scenario, the preset constraint condition can be expressed as the proportion of items with higher billing in all displayed items is close to the target value. Based on the traffic characteristics data, it can be determined that the current proportion is , the current adjustment error can be obtained as Optionally, a constraint condition may be set for each indicator respectively, and accordingly, the obtained adjustment error may also be the adjustment error of each indicator, which will not be described in detail here.

[0067] The initial weight coefficients of the multiple indicators are the weight coefficients of the multiple indicators determined previously.

[0068] Optionally, feedback regulation can be implemented using various controllers, such as a PID (proportional integral derivative) controller, a first-order differential controller, etc., which are not limited here.

[0069] According to an embodiment of the present disclosure, based on the adjustment error, feedback adjustment is performed on the initial weight coefficients of the multiple indicators to obtain the current weight coefficients of the multiple indicators, which may include the following operations:

[0070] Based on the preset control coefficient, the adjustment error and the historical adjustment error determined in the historical period, a feedback adjustment coefficient is obtained; and based on the feedback adjustment coefficient, feedback adjustment is performed on the initial weight coefficients of each of the multiple indicators to obtain the current weight coefficients of each of the multiple indicators.

[0071] The preset control coefficient may be related to the type of controller specifically selected. For example, when the controller used for feedback regulation is a PID controller, the preset control coefficient may include a proportional control coefficient, an integral control coefficient, and a differential control coefficient.

[0072] Figure 4 A schematic diagram of a constraint integrated controller according to an embodiment of the present disclosure is schematically shown.

[0073] like Figure 4As shown, the constraint integrated controller can use the PID controller to perform feedback adjustment on the weight coefficient of each indicator separately.

[0074] For example, the initial weight coefficient of an indicator can be , the PID controller can obtain the feedback adjustment coefficient of the indicator based on the adjustment error of the indicator, the adjustment errors of the indicator in multiple historical time periods, the proportional control coefficient, the integral control coefficient and the differential control coefficient, as shown in formula (1):

[0075]

[0076] In formula (1), u(t) can represent the feedback adjustment coefficient of the indicator; k p Can represent the proportional control coefficient; k i Can represent the differential control coefficient; k d can represent the integral control coefficient; i can represent the number of historical time periods. p , k i , k d Both π and m can be hyperparameters of the PID controller, and their values ​​can be determined based on the specific application scenario and are not limited here.

[0077] After determining the feedback adjustment coefficient of the indicator, the PID controller can use the feedback adjustment coefficient to adjust the initial weight coefficient of the indicator to obtain the current weight coefficient of the indicator, as shown in formula (2):

[0078]

[0079] In formula (2), It can be expressed as the current weight coefficient of the indicator.

[0080] For another example, in order to reduce the consumption of computing resources, a day can be divided into 240 intervals, each interval is 6 minutes long, the weight coefficient of the indicator can be updated at the beginning of each interval, and the weight coefficient of the indicator updated at the beginning is used as the weight coefficient of the indicator in the interval. After receiving an access request, the weight coefficient of the indicator in the corresponding interval can be selected as the current weight coefficient of the indicator based on the triggering time of the access request.

[0081] According to an embodiment of the present disclosure, a periodic task may be set at the start time of each interval. In response to triggering the periodic task, flow characteristic data collected within a first target period may be acquired; and, based on the flow characteristic data, weight coefficients of multiple indicators within the first target period may be determined.

[0082] The first target time period is determined based on a preset duration and a triggering time of a periodic task. The preset duration is the duration of the interval, and the triggering time of the periodic task is the starting time of the interval.

[0083] According to an embodiment of the present disclosure, determining the weight coefficients of the respective multiple indicators in the first target period based on the traffic characteristic data may include the following operations:

[0084] Based on preset constraints and flow characteristic data, an adjustment error is obtained; and based on the adjustment error, feedback adjustment is performed on the initial weight coefficients of each of the multiple indicators to obtain the weight coefficients of each of the multiple indicators in the first target time period, wherein the initial weight coefficients of each of the multiple indicators are determined based on the flow characteristic data collected in a time period before the first target time period.

[0085] For example, the first target time period may refer to the time period of the t-th interval, and the initial weight coefficients of the multiple indicators may be determined based on the traffic characteristic data collected in the time period of the t-1-th interval.

[0086] According to an embodiment of the present disclosure, based on the adjustment error, feedback adjustment is performed on the initial weight coefficients of the multiple indicators to obtain the weight coefficients of the multiple indicators in the first target period, which may include the following operations:

[0087] Based on the preset control coefficient, the adjustment error and the historical adjustment error determined in the historical period, a feedback adjustment coefficient is obtained; and based on the feedback adjustment coefficient, the initial weight coefficients of each of the multiple indicators are feedback-adjusted to obtain the weight coefficients of each of the multiple indicators in the first target period.

[0088] Based on the traffic characteristic data collected during the first target period, the weight coefficients of the multiple indicators during the first target period are determined. The same or similar method as that for real-time updating of the weight coefficients may be used, which will not be repeated here.

[0089] Optionally, the amount of traffic characteristic data collected during the first target period can be small, for example, less than the data volume threshold. In this case, it can be considered that the traffic characteristic data collected during the first target period does not meet the update conditions. At this time, the weight coefficients of the multiple indicators determined in the previous period of the first target period can be directly used as the current weight coefficients of the multiple indicators. Alternatively, the traffic characteristic data collected during the first period and the traffic characteristic data collected during one or more periods before the first period can be combined to perform feedback adjustment on the weight coefficients of the multiple indicators, which is not limited here.

[0090] Furthermore, through the method described above, the weight coefficients of multiple indicators in the current interval period can be determined. Therefore, when responding to an access request, the current interval period can be determined according to the current moment, and the weight coefficients of multiple indicators in the current interval period can be used as the current weight coefficients of the multiple indicators.

[0091] For example, if the current moment is within the second target period, the second target period corresponding to the current moment can be determined; and based on the weight coefficients of the multiple indicators within the second target period, the current weight coefficients of the multiple indicators can be obtained.

[0092] According to the embodiments of the present disclosure, by collecting information of the current online environment in real time and performing feedback adjustment on the weight coefficients of multiple indicators based on the collected data, the real-time optimality of the weight coefficients can be guaranteed to improve the user experience and display effect.

[0093] The reordering model may include a static part and a dynamic part, and the dynamic network parameters output by the hypernetwork may only adjust the model parameters of the dynamic part of the reordering model. Specifically, the initial model may include a static network module and a dynamic network module. Adjusting the parameters of the initial model based on the dynamic network parameters to obtain the reordering model may include the following operations:

[0094] The network parameters of the dynamic network module in the initial model are replaced by the dynamic network parameters to obtain an adjusted dynamic network module; and the static network module is spliced ​​with the adjusted dynamic network module to obtain a reordering model.

[0095] That is, when adjusting the parameters of the reordering model, only the dynamic network parameters can be used to adjust the parameters of the dynamic network module of the initial model, and the parameters of the static network module can be frozen to obtain the reordering model.

[0096] Figure 5 The figure schematically shows a hypernetwork model according to an embodiment of the present disclosure.

[0097] like Figure 5 As shown, the super network model can be composed of a super network 501 and a reordering model 502 , and the reordering model 502 can be divided into a static network module 5021 and a dynamic network module 5022 .

[0098] The current weight coefficients 503 of each of the multiple indicators can be input into the hypernetwork 501 to obtain the dynamic network parameters 504. The dynamic network parameters 504 can be used to adjust the model parameters of the dynamic network module 5022, and the adjustment method includes but is not limited to overall replacement, adjustment based on a set functional relationship, etc. The dynamic network model after the parameter replacement can constitute a reordering model after parameter adjustment together with the static network module 5021. Multiple material features 505 of each of the multiple objects to be displayed can be input into the adjusted reordering model to output the ranking scores 506 of each of the multiple objects to be displayed. The model parameters of the hypernetwork 501 and the model parameters of the static network module 5021 can be obtained by training through unsupervised training based on the training method of the hypernetwork, and the specific training method is not limited here.

[0099] For example, the current weight coefficients of multiple indicators can be expressed as , the hypernetwork can be a two-layer fully connected neural network. The hypernetwork can be the current weight coefficients of multiple indicators Perform mapping processing to obtain dynamic network parameters ,Should It can be expressed as the model parameters of the hypernetwork. The model parameters of the dynamic network module can be replaced by the dynamic network parameters, so that the model parameters of the reordering model can be expressed as ,Should It can represent the model parameters of the static network module. Multiple material characteristics of the object to be displayed can be represented as , the material characteristics By inputting the re-ranking model, the ranking score of the object to be displayed can be obtained, as shown in formula (3):

[0100]

[0101] In formula (3), It can represent the ranking score of the object to be displayed.

[0102] For multiple objects to be displayed, their respective ranking scores are sorted from large to small to obtain a re-ranking result.

[0103] After the reordering result is obtained, the multiple objects to be displayed can be displayed on the target page based on the reordering result. Specifically, the target page can include multiple display areas, and the pre-recommendation result can also include display materials of the multiple objects to be displayed. Based on the reordering result, the display materials of the multiple objects to be displayed can be respectively set in the multiple display areas to display the multiple objects to be displayed on the target page.

[0104] For example, the multiple display areas may be multiple advertising slots, and the multiple advertising slots may be sorted according to their importance. Advertising slots with higher importance, such as page pop-ups and advertising slots at the top of the page, may be allocated to objects to be displayed with higher sorting scores, and used to display the display materials of the objects to be displayed with higher sorting scores.

[0105] According to the embodiments of the present disclosure, by implementing dynamic multi-target reordering, the optimality of the platform sorting strategy can be guaranteed in real time in different time periods and when user characteristics change, so that the displayed objects are more in line with the objects required by users, thereby increasing the click-through rate of the objects to be displayed and improving the user experience.

[0106] Figure 6 A schematic diagram of a reordering system according to an embodiment of the present disclosure is schematically shown.

[0107] like Figure 6 As shown, after receiving the user traffic request, first, the multiple material features of the multiple objects to be displayed generated by the upstream recommendation system 601 can be obtained, and these material features can be added to the processing queue. Afterwards, the constraint integrated controller 602 can be used for feedback adjustment according to all the feature information of the current traffic, including the user traffic feature information and the environmental traffic feature information, to change the weight coefficient to obtain the current weight coefficient of each of the multiple indicators. The current weight coefficients of each of the multiple indicators can be output as the model parameters of the dynamic network module in the reordering model 604 through the super network 603, which is combined with the model parameters of the static network module to form the reordering network 604. The reordering network 604 can score all the material features in the queue to obtain the ranking scores of the multiple objects to be displayed, that is, the reordering results. According to the ranking scores, the display materials of the multiple objects to be displayed can be allocated to the advertising positions of the target page in turn, and finally the target page can be displayed to the user.

[0108] Figure 7 A block diagram of a multi-target reordering device according to an embodiment of the present disclosure is schematically shown.

[0109] like Figure 7 As shown, the multi-target reordering device 700 includes a first acquisition module 710 , a first determination module 720 , a processing module 730 , a parameter adjustment module 740 and a reordering module 750 .

[0110] The first acquisition module 710 is used to obtain a pre-recommendation result about the target page in response to an access request for the target page, where the pre-recommendation result includes a plurality of material features of each of a plurality of objects to be displayed, and the plurality of material features are respectively related to a plurality of indicators.

[0111] The first determination module 720 is used to determine the current weight coefficient of each of the multiple indicators at the current moment.

[0112] The processing module 730 is used to process the current weight coefficients of the multiple indicators by using the hypernetwork to obtain dynamic network parameters.

[0113] The parameter adjustment module 740 is used to adjust the parameters of the initial model based on the dynamic network parameters to obtain a reordering model.

[0114] The reordering module 750 is used to input a plurality of material features of a plurality of objects to be displayed into a reordering model to obtain a reordering result.

[0115] According to an embodiment of the present disclosure, the initial model includes a static network module and a dynamic network module.

[0116] According to an embodiment of the present disclosure, the parameter adjustment module 740 includes a first parameter adjustment unit and a second parameter adjustment unit.

[0117] The first parameter adjustment unit is used to replace the network parameters of the dynamic network module in the initial model with the dynamic network parameters to obtain the adjusted dynamic network module.

[0118] The second parameter adjustment unit is used to splice the static network module with the adjusted dynamic network module to obtain a reordering model.

[0119] According to an embodiment of the present disclosure, the multi-target reordering device 700 further includes a second acquisition module and a second determination module.

[0120] The second acquisition module is used to acquire the traffic characteristic data collected within a first target time period in response to triggering the periodic task, wherein the first target time period is determined based on a preset duration and a triggering time of the periodic task.

[0121] The second determination module is used to determine the weight coefficients of the multiple indicators in the first target time period based on the traffic characteristic data.

[0122] According to an embodiment of the present disclosure, the second determining module includes a first determining unit and a second determining unit.

[0123] The first determination unit is used to obtain the adjustment error based on the preset constraint condition and the flow characteristic data.

[0124] The second determination unit is used to perform feedback adjustment on the initial weight coefficients of each of the multiple indicators based on the adjustment error, so as to obtain the weight coefficients of each of the multiple indicators in the first target time period, wherein the initial weight coefficients of each of the multiple indicators are determined based on the flow characteristic data collected in the previous time period of the first target time period.

[0125] According to an embodiment of the present disclosure, the second determining unit includes a first determining subunit and a second determining subunit.

[0126] The first determination subunit is used to obtain a feedback adjustment coefficient based on a preset control coefficient, an adjustment error, and a historical adjustment error determined in a historical period.

[0127] The second determination subunit is used to perform feedback adjustment on the initial weight coefficients of the multiple indicators based on the feedback adjustment coefficient to obtain the weight coefficients of the multiple indicators in the first target period.

[0128] According to an embodiment of the present disclosure, the first determining module 720 includes a third determining unit and a fourth determining unit.

[0129] The third determining unit is used to determine a second target time period corresponding to the current moment.

[0130] The fourth determining unit is used to obtain the current weight coefficients of the multiple indicators based on the weight coefficients of the multiple indicators in the second target period.

[0131] According to an embodiment of the present disclosure, the pre-recommendation result also includes display materials of each of the multiple objects to be displayed, and the target page includes multiple display areas.

[0132] According to an embodiment of the present disclosure, the multi-target reordering apparatus 700 further includes a display module.

[0133] The display module is used to respectively set the display materials of the multiple objects to be displayed in multiple display areas based on the reordering result, so as to display the multiple objects to be displayed on the target page.

[0134] According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits, or at least part of the functions of any one of them can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be at least partially implemented as hardware circuits, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems on chips, systems on substrates, systems on packages, application specific integrated circuits (ASICs), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging circuits, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, submodules, units, and subunits can be at least partially implemented as computer program modules, and when the computer program modules are run, the corresponding functions can be performed.

[0135] For example, any multiple of the first acquisition module 710, the first determination module 720, the processing module 730, the parameter adjustment module 740, and the reordering module 750 can be combined in one module / unit / sub-unit for implementation, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of these modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units and implemented in one module / unit / sub-unit. According to an embodiment of the present disclosure, at least one of the first acquisition module 710, the first determination module 720, the processing module 730, the parameter adjustment module 740, and the reordering module 750 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, at least one of the first acquisition module 710, the first determination module 720, the processing module 730, the parameter adjustment module 740 and the reordering module 750 may be at least partially implemented as a computer program module, which may perform corresponding functions when executed.

[0136] It should be noted that the multi-target reordering device part in the embodiment of the present disclosure corresponds to the multi-target reordering method part in the embodiment of the present disclosure. The description of the multi-target reordering device part specifically refers to the multi-target reordering method part, which will not be repeated here.

[0137] Figure 8 A block diagram of an electronic device suitable for implementing a multi-objective reordering method according to an embodiment of the present disclosure is schematically shown. Figure 8 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0138] like Figure 8As shown, the electronic device 800 according to an embodiment of the present disclosure includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 to a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a dedicated microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include an onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0139] In RAM 803, various programs and data required for the operation of electronic device 800 are stored. Processor 801, ROM 802 and RAM 803 are connected to each other via bus 804. Processor 801 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 802 and / or RAM 803. It should be noted that the program can also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0140] According to an embodiment of the present disclosure, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to the bus 804. The electronic device 800 may further include one or more of the following components connected to the input / output (I / O) interface 805: an input portion 806 including a keyboard, a mouse, etc.; an output portion 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 808 including a hard disk, etc.; and a communication portion 809 including a network interface card such as a LAN card, a modem, etc. The communication portion 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that the computer program read therefrom is installed into the storage portion 808 as needed.

[0141] According to an embodiment of the present disclosure, the method flow according to an embodiment of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the processor 801, the above-mentioned functions defined in the system of the embodiment of the present disclosure are executed. According to an embodiment of the present disclosure, the system, equipment, device, module, unit, etc. described above can be implemented by a computer program module.

[0142] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0143] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium. For example, it may include, but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0144] For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the ROM 802 and / or the RAM 803 described above and / or one or more memories other than the ROM 802 and the RAM 803 .

[0145] An embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method provided by the embodiment of the present disclosure. When the computer program product runs on an electronic device, the program code is used to enable the electronic device to implement the multi-target reordering method provided by the embodiment of the present disclosure.

[0146] When the computer program is executed by the processor 801, the above functions defined in the system / device of the embodiment of the present disclosure are executed. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0147] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 809, and / or installed from a removable medium 811. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0148] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0149] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions. It can be understood by those skilled in the art that the features recorded in the various embodiments of the present disclosure can be combined and / or combined in a variety of ways, even if such a combination or combination is not explicitly recorded in the present disclosure. In particular, without departing from the spirit and teaching of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0150] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A multi-objective reordering method, comprising: In response to an access request for a target page, obtaining a pre-recommendation result for the target page, the pre-recommendation result including a plurality of material features of each of a plurality of objects to be displayed, the plurality of material features being respectively related to a plurality of indicators; Determine the current weight coefficient of each of the multiple indicators at the current moment; Processing the current weight coefficients of the multiple indicators respectively by using the hypernetwork to obtain dynamic network parameters; Adjusting the parameters of the initial model based on the dynamic network parameters to obtain a reordering model; as well as The plurality of material features of the plurality of objects to be displayed are respectively input into the reordering model to obtain a reordering result.

2. The method according to claim 1, wherein: The initial model includes a static network module and a dynamic network module; The step of adjusting the parameters of the initial model based on the dynamic network parameters to obtain a reordering model includes: Replacing the network parameters of the dynamic network module in the initial model with the dynamic network parameters to obtain an adjusted dynamic network module; and The static network module and the adjusted dynamic network module are spliced ​​to obtain the reordering model.

3. The method according to claim 1, further comprising: In response to triggering a periodic task, acquiring traffic characteristic data collected within a first target time period, wherein the first target time period is determined based on a preset duration and a triggering time of the periodic task; and Based on the traffic characteristic data, weight coefficients of the respective indicators within the first target time period are determined.

4. The method according to claim 3, wherein: The determining, based on the traffic characteristic data, weight coefficients of the respective indicators within the first target period includes: Based on the preset constraint conditions and the flow characteristic data, obtaining an adjustment error; and Based on the adjustment error, feedback adjustment is performed on the initial weight coefficients of each of the multiple indicators to obtain the weight coefficients of each of the multiple indicators within the first target time period, wherein the initial weight coefficients of each of the multiple indicators are determined based on the flow characteristic data collected in the previous time period of the first target time period.

5. The method according to claim 4, wherein: The step of performing feedback adjustment on the initial weight coefficients of the multiple indicators based on the adjustment error to obtain the weight coefficients of the multiple indicators in the first target period includes: Obtaining a feedback adjustment coefficient based on a preset control coefficient, the adjustment error, and a historical adjustment error determined in a historical period; and Feedback adjustment is performed on the initial weight coefficients of the multiple indicators based on the feedback adjustment coefficient to obtain the weight coefficients of the multiple indicators in the first target period.

6. The method according to claim 3, wherein: The determining of the current weight coefficients of the multiple indicators at the current moment includes: determining a second target time period corresponding to the current moment; and Based on the weight coefficients of the respective indicators in the second target period, current weight coefficients of the respective indicators are obtained.

7. The method according to claim 1, wherein: The pre-recommendation result also includes display materials of the multiple objects to be displayed, and the target page includes multiple display areas; The method further comprises: Based on the reordering result, the display materials of the multiple objects to be displayed are respectively set in the multiple display areas to display the multiple objects to be displayed on the target page.

8. A multi-objective reordering device, comprising: A first acquisition module is used to obtain, in response to an access request for a target page, a pre-recommendation result about the target page, wherein the pre-recommendation result includes a plurality of material features of each of a plurality of objects to be displayed, and the plurality of material features are respectively related to a plurality of indicators; A first determination module is used to determine the current weight coefficient of each of the multiple indicators at the current moment; A processing module, used for processing the current weight coefficients of the multiple indicators respectively by using the hypernetwork to obtain dynamic network parameters; A parameter adjustment module, used to adjust the parameters of the initial model based on the dynamic network parameters to obtain a reordering model; as well as The reordering module is used to input the plurality of material characteristics of the plurality of objects to be displayed into the reordering model to obtain a reordering result.

9. An electronic device, comprising: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the method according to any one of claims 1 to 7.

11. 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 7.

Images