Object sorting method, device and electronic device

Abstract

The present disclosure relates to an object sorting method, device and electronic device, and relates to the field of Internet technology. The method is: determining the object to be sorted, and the basic information of each account that operates the object to be sorted; inputting the basic information of each account into the attribute labeling model, determining the attribute labeling information of each account and the first embedded feature of each account; the first embedded feature is used to characterize the account characteristics determined based on the basic information of the account; based on the attribute labeling information of each account and the feedback information of each account operating the object to be sorted, determining the second embedded feature of the object to be sorted; based on the first embedded feature and the second embedded feature, determining the click probability of the object to be sorted; and based on the click probability, determining the display order of the object to be sorted.

Description

Technical Field

[0001] The present disclosure relates to the field of Internet technology, and in particular to an object sorting method, device and electronic device. Background Art

[0002] At present, short videos have gradually become the main source of information for people. People search for short videos based on short video platforms, and short video platforms make recommendations based on searches. In addition, accounts registered on short video platforms can publish short videos, and short video platforms determine the recommended ranking of short videos based on feedback from other accounts on the short videos. Based on this, in order to improve the ranking or popularity of short video recommendations, video publishers may take some abnormal measures, such as: finding water armies to add comments, likes, and attention, or come up with eye-catching titles, using covers that do not match the text and pictures, etc.

[0003] However, short video platforms handle accounts that use the aforementioned abnormal means in an untimely and low-accuracy manner, resulting in low accuracy in video sorting, which in turn leads to poor video recommendation effects and reduces user experience. Summary of the invention

[0004] The embodiments of the present disclosure provide an object sorting method, device and electronic device to enhance the accuracy of object sorting, thereby reducing the occurrence of behaviors of using accounts to manipulate rankings, thereby having a positive impact on the overall publishing and browsing environment in the video platform.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for sorting objects is provided, the method comprising:

[0006] Determine the objects to be sorted and basic information of each account that operates the objects to be sorted;

[0007] Inputting the basic information of each of the accounts into an attribute labeling model, determining the attribute labeling information of each of the accounts and a first embedded feature of each of the accounts; the first embedded feature is used to characterize the account characteristics determined based on the basic information of the account;

[0008] Determining a second embedding feature of the object to be sorted based on the attribute tag information of each account and feedback information of each account operating the object to be sorted;

[0009] Based on the first embedded feature and the second embedded feature, a click probability of the objects to be sorted is determined; and based on the click probability, a display order of the objects to be sorted is determined.

[0010] In a possible implementation, the attribute labeling model is trained based on the following method:

[0011] Determine a training sample; wherein the training sample includes a plurality of accounts and attribute labeling information corresponding to the plurality of accounts determined based on prior knowledge; the attribute labeling information is any one of high-quality, abnormal, and ordinary;

[0012] Inputting the training sample into the account embedding feature model to obtain the embedding feature corresponding to the account;

[0013] Determine statistical features corresponding to the account, wherein the statistical features are obtained based on usage statistics of the account on the video platform;

[0014] Inputting the embedded features and the statistical features corresponding to the account into a preset attribute labeling model to determine the predicted attribute labeling information corresponding to the account;

[0015] The predicted attribute labeling information corresponding to the account is compared with the corresponding attribute labeling information to determine a first comparison result. When it is determined that the adjusted preset attribute labeling model corresponding to the first comparison result has converged, an attribute labeling model is obtained.

[0016] In a possible implementation, the account embedding feature model is trained based on the following method:

[0017] Inputting a preset one-hot vector corresponding to the first account in the training sample into an input vector matrix in a preset account embedding feature model to obtain an embedding feature of the first account;

[0018] The first account embedding feature is input into the output vector matrix in the preset account embedding feature model to determine the output one-hot vector of the second account in the training sample; wherein the first account and the second account operate on the same object;

[0019] Determine the difference between the output one-hot vector of the second account and the preset one-hot vector corresponding to the second account. When it is determined that the difference belongs to a preset range, adjust the input vector matrix and the output vector matrix based on the difference to obtain an account embedding feature model.

[0020] In a possible implementation, determining the second embedding feature of the object to be sorted based on the attribute tag information of each of the accounts and feedback information of each of the accounts operating the object to be sorted includes:

[0021] Determine the weight corresponding to each of the accounts based on the attribute tag information of each of the accounts and the preset level and weight mapping relationship;

[0022] The first embedded feature of each of the accounts is multiplied by its corresponding weight to obtain a plurality of first results; and the plurality of first results are added to obtain a second embedded feature of the object to be sorted.

[0023] In a possible implementation, determining the click probability of the object to be sorted based on the first embedded feature and the second embedded feature includes:

[0024] Inputting the first embedded feature and the second embedded feature into a click prediction model to determine the click probability of the object to be sorted;

[0025] The click prediction model is trained based on the following method:

[0026] Inputting the first embedded feature into a preset click prediction model to obtain a first processing result;

[0027] Inputting the second embedded feature into a preset click prediction model to obtain a second processing result;

[0028] Inputting the first processing result and the second processing result into the preset click prediction model to obtain a third processing result; the third processing result is obtained by combining the first processing result and the second processing result;

[0029] Inputting the third processing result into the preset click prediction model to obtain a predicted click result;

[0030] The predicted click result is compared with the actual click result of the account on the object to be sorted to determine a second comparison result. When it is determined that the adjusted preset predicted click model corresponding to the second comparison result has converged, a click prediction model is obtained.

[0031] In a possible implementation, determining the display order of the objects to be sorted based on the click probability includes:

[0032] Determining a preset interval to which the click probability belongs;

[0033] Based on the mapping relationship between the preset intervals and the display order, the display order of the objects to be sorted is determined.

[0034] According to a second aspect of an embodiment of the present disclosure, a device for sorting objects is provided, the device comprising:

[0035] A determination unit, configured to determine the objects to be sorted and basic information of each account that operates the objects to be sorted;

[0036] A first processing unit is configured to input the basic information of each of the accounts into an attribute labeling model, determine the attribute labeling information of each of the accounts and a first embedded feature of each of the accounts; the first embedded feature is used to characterize the account characteristics determined based on the basic information of the account;

[0037] A second processing unit is configured to determine a second embedding feature of the object to be sorted based on the attribute tag information of each account and feedback information of each account operating the object to be sorted;

[0038] A third processing unit is configured to determine the click probability of the object to be sorted based on the first embedded feature and the second embedded feature;

[0039] The sorting unit is configured to determine the display order of the objects to be sorted based on the click probability.

[0040] In a possible implementation, the apparatus further includes a training unit, which is configured to perform training on the attribute labeling model based on the following method:

[0041] Determine a training sample; wherein the training sample includes a plurality of accounts and attribute labeling information corresponding to the plurality of accounts determined based on prior knowledge; the attribute labeling information is any one of high-quality, abnormal, and ordinary;

[0042] Inputting the training sample into the account embedding feature model to obtain the embedding feature corresponding to the account;

[0043] Determine statistical features corresponding to the account, wherein the statistical features are obtained based on usage statistics of the account on the video platform;

[0044] Inputting the embedded features and the statistical features corresponding to the account into a preset attribute labeling model to determine the predicted attribute labeling information corresponding to the account;

[0045] The predicted attribute labeling information corresponding to the account is compared with the corresponding attribute labeling information to determine a first comparison result. When it is determined that the adjusted preset attribute labeling model corresponding to the first comparison result has converged, an attribute labeling model is obtained.

[0046] In a possible implementation, the training unit is configured to perform training on the account embedding feature model based on the following method:

[0047] Inputting a preset one-hot vector corresponding to the first account in the training sample into an input vector matrix in a preset account embedding feature model to obtain an embedding feature of the first account;

[0048] The first account embedding feature is input into the output vector matrix in the preset account embedding feature model to determine the output one-hot vector of the second account in the training sample; wherein the first account and the second account operate on the same object;

[0049] Determine the difference between the output one-hot vector of the second account and the preset one-hot vector corresponding to the second account. When it is determined that the difference belongs to a preset range, adjust the input vector matrix and the output vector matrix based on the difference to obtain an account embedding feature model.

[0050] In a possible implementation, the second processing unit is configured to execute:

[0051] Determine the weight corresponding to each of the accounts based on the attribute tag information of each of the accounts and the preset level and weight mapping relationship;

[0052] The first embedded feature of each of the accounts is multiplied by its corresponding weight to obtain a plurality of first results; and the plurality of first results are added to obtain a second embedded feature of the object to be sorted.

[0053] In a possible implementation, the third processing unit is configured to input the first embedded feature and the second embedded feature into a click prediction model to determine the click probability of the object to be sorted;

[0054] The apparatus further includes a training unit, which is configured to perform training on the click prediction model based on the following manner:

[0055] Inputting the first embedded feature into a preset click prediction model to obtain a first processing result;

[0056] Inputting the second embedded feature into a preset click prediction model to obtain a second processing result;

[0057] Inputting the first processing result and the second processing result into the preset click prediction model to obtain a third processing result; the third processing result is obtained by combining the first processing result and the second processing result;

[0058] Inputting the third processing result into the preset click prediction model to obtain a predicted click result;

[0059] The predicted click result is compared with the actual click result of the account on the object to be sorted to determine a second comparison result. When it is determined that the adjusted preset predicted click model corresponding to the second comparison result has converged, a click prediction model is obtained.

[0060] In a possible implementation, the sorting unit is configured to execute:

[0061] Determining a preset interval to which the click probability belongs;

[0062] Based on the mapping relationship between the preset intervals and the display order, the display order of the objects to be sorted is determined.

[0063] According to a third aspect of the embodiments of the present disclosure, there is provided an electronic device, including:

[0064] processor;

[0065] a memory for storing instructions executable by the processor;

[0066] The processor is configured to execute the instructions to implement any one of the methods provided in the first aspect of the present disclosure.

[0067] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute any method provided in the first aspect of the present disclosure.

[0068] In a fifth aspect, an embodiment of the present disclosure provides a computer program product, including a computer program, which, when executed by a processor, implements any method provided in the first aspect of the present disclosure.

[0069] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0070] In the disclosed embodiment, the attribute tag information of the account that operates the object to be sorted and the first embedded feature of the account can be determined based on the attribute tag model, and then the second embedded feature of the object to be sorted can be determined based on the attribute tag information. Furthermore, the embedded features of the object to be sorted and the embedded features of the account can be input into the click prediction model to determine the click probability of the object to be sorted, and then the display order corresponding to the object to be sorted can be determined based on the click probability and the preset sorting strategy.

[0071] It can be seen that in the embodiment of the present disclosure, the account is attribute-tagged and the attribute tag information of the account is obtained, so that the embedded features of the objects to be sorted are determined based on the attribute tag information and combined with the embedded features of the account to comprehensively determine the sorting basis, namely the click probability. In this way, the influence of accounts with abnormal attribute tag information on the sorting of objects to be sorted can be minimized, and the influence of feedback from accounts with ordinary attribute tag information on the sorting of objects to be sorted can be retained, so that the display order of objects to be sorted determined based on the click probability is as accurate as possible, thereby improving the publishing and browsing environment of the video platform and enhancing the user experience.

[0072] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or be understood by practicing the present disclosure. The purpose and other advantages of the present disclosure can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0074] Figure 1 is a schematic diagram of an application scenario according to an exemplary embodiment;

[0075] Figure 2 It is a flowchart of a method for sorting objects according to an exemplary embodiment;

[0076] Figure 3 is a structural block diagram of an object sorting device according to an exemplary embodiment;

[0077] Figure 4 is a schematic structural diagram of an electronic device according to an exemplary embodiment;

[0078] Figure 5 It is another structural schematic diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0079] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0080] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the images used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0081] The following is a brief introduction to the design concept of the embodiment of the present disclosure:

[0082] At present, in order to improve the ranking or popularity of videos on video platforms, video publishers may adopt some abnormal means, such as: hiring water armies to increase comments, likes, attention or come up with eye-catching titles, etc. If the aforementioned abnormal means are not controlled, on the one hand, the creative enthusiasm of video publishers will be reduced, that is, the aforementioned low-cost, abnormal means will be used to improve the ranking of videos. On the other hand, the low-quality comments, cheap likes, attention and other behaviors of the water army will drown out the behaviors of high-quality accounts, causing video consumers, that is, accounts that operate videos, to reduce the number of behavioral operations, affecting the development of the entire video platform.

[0083] In view of this, in order to solve the above problems, the embodiments of the present disclosure provide an object sorting method, through which the sorting order of objects can be determined more accurately, thereby improving the browsing and publishing environment of the video platform.

[0084] After introducing the design concept of the embodiment of the present disclosure, the following briefly introduces the application scenarios to which the technical solution of the embodiment of the present disclosure can be applied. It should be noted that the application scenarios introduced below are only used to illustrate the embodiment of the present disclosure and are not limited. In specific implementation, the technical solution provided by the embodiment of the present disclosure can be flexibly applied according to actual needs.

[0085] refer to Figure 1 , which is a schematic diagram of an application scenario of the object sorting method provided by an embodiment of the present disclosure. The application scenario includes multiple terminal devices 101 (including terminal device 101-1, terminal device 101-2, ... terminal device 101-n) and a server 102. Each terminal device 101 is connected to the server 102 via a wireless or wired network.

[0086] The server 102 may be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. The terminal device 101 may be a smart phone, a tablet computer, a laptop, a desktop computer, a smart TV, a smart wearable device, etc., but is not limited thereto.

[0087] In this scenario, the user can publish and browse videos based on the account registered in the video platform deployed on the terminal device 101, and can also operate on the video, such as commenting, liking, and forwarding the video. Then, the terminal device 101 can send the videos published by the user based on the account and the information on the video operations to the server 102, so that the server 102 can process the videos, determine the display order corresponding to each video, and feed back the display order corresponding to each video to the video platform deployed in the terminal device 101, so that each video can be displayed according to the display order corresponding to each video.

[0088] Of course, the method provided in the embodiment of the present disclosure is not limited to Figure 1 The application scenarios shown in the figure can also be used in other possible application scenarios, and the embodiments of the present disclosure are not limited thereto. Figure 1 The functions that can be implemented by each device in the application scenario shown will be described in the subsequent method embodiments, and will not be described in detail here.

[0089] To further illustrate the scheme of the object sorting method provided by the embodiment of the present disclosure, this is described in detail below in conjunction with the accompanying drawings and specific implementation methods. Although the embodiment of the present disclosure provides the method operation steps as shown in the following embodiments or drawings, more or fewer operation steps may be included in the method based on routine or no creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided by the embodiment of the present disclosure. The method can be executed in the actual processing process or when the device is executed, it can be executed in the order of the method shown in the embodiment or drawings or in parallel (for example, an application environment of a parallel processor or multi-threaded processing).

[0090] The following combination Figure 2 The method flow chart shown in the figure illustrates the object sorting method in the embodiment of the present disclosure. Figure 2 The steps shown can be performed as follows Figure 1 The electronic device shown performs.

[0091] Step 201: Determine the objects to be sorted and the basic information of each account that operates the objects to be sorted.

[0092] In the embodiments of the present disclosure, objects can be obtained from a database corresponding to the video platform, or from data sent in real time by a terminal device, which is not limited in the embodiments of the present disclosure. In addition, all the acquired objects can be used as objects to be sorted, or a part of the acquired objects can be selected as objects to be processed, which is not limited in the embodiments of the present disclosure. It should be noted that in the embodiments of the present disclosure, in order to facilitate a more accurate description of the sorting scheme of objects, the following text will be described by sorting an object.

[0093] In the embodiment of the present disclosure, after determining the objects to be sorted, the basic information of the account that operates the objects to be sorted can also be determined, wherein the basic information of the account includes but is not limited to the following information: the number of times the video platform is logged in and the number of times within the statistical period, the statistical period is, for example, one day or one week, the length of time logged in to the video platform each day, statistical information of operations such as clicks, favorites, comments, searches and forwarding on videos on the video platform, and newly added accounts followed by the account.

[0094] For example, the object to be sorted is video 1, and the accounts that like video 1 are account A and account F, the accounts that comment on video 1 are account M and account N, and the accounts that favorite video 1 are account C and account S. Then, it can be determined that the accounts that operate on video 1 are account A, account F, account M, account N, account C, and account S, thereby determining the basic information of account A, account F, account M, account N, account C, and account S.

[0095] It should be noted that in the embodiment of the present disclosure, the identifier of the account can be one or more of numbers, letters, Chinese characters, and other characters, and there is no limitation in the embodiment of the present disclosure.

[0096] Step 202: Input the basic information of each account into the attribute labeling model to determine the attribute labeling information of each account and the first embedded feature of each account; the first embedded feature is used to characterize the account characteristics determined based on the basic information of the account.

[0097] In an embodiment of the present disclosure, after the basic information of the account is determined, the basic information of the account can be input into a pre-trained attribute labeling model, so that the attribute labeling information of the account and the first embedded feature of the account can be obtained.

[0098] In the embodiment of the present disclosure, the attribute labeling model can be trained by but not limited to the following steps:

[0099] Step a: Determine a training sample; wherein the training sample includes a plurality of accounts and attribute labeling information corresponding to the plurality of accounts determined based on prior knowledge; the attribute labeling information is any one of high-quality, abnormal, and ordinary.

[0100] In the disclosed embodiment, multiple accounts can be obtained from a database corresponding to the video platform, and attribute tag information of the multiple accounts determined based on prior knowledge can be determined, wherein the attribute tag information is any one of high-quality, abnormal, and ordinary.

[0101] For example, the attribute tag information of the account may be determined by comprehensively considering the number of times the account logs in every day in a week and the number of like operations, favorite operations, comment operations, and search operations on the video.

[0102] For example, if the number of logins of account 1 every day in a week is not 0 and does not exceed 60 times, and the number of likes, favorites, comments and searches on videos is between 10-80, then the attribute marking information of account 1 is determined to be high-quality; if account 3 does not log in every day in a week, and the number of likes, favorites, comments and searches on videos on a certain day is much greater than 80 times, then the attribute marking information of account 3 is determined to be abnormal; if the number of logins of account 7 for 6 days in a week is not 0 and does not exceed 60 times, and the number of likes, favorites, comments and searches on videos is between 0-10, then the attribute marking information of account 7 is determined to be ordinary.

[0103] Step b: Input the training samples into the account embedding feature model to obtain the embedding features corresponding to the account; the account embedding feature model is used to determine the embedding features corresponding to the account based on the characteristics of the account.

[0104] In an embodiment of the present disclosure, a training sample may be input into an account embedding feature model to obtain embedding features corresponding to the account.

[0105] Exemplarily, the account embedding feature model can be trained in the following manner: a preset one-hot vector corresponding to the first account in the training sample, wherein the preset one-hot vector can be used to indicate the characteristics corresponding to the first account that operates the first object; an input vector matrix in the preset account embedding feature model is input to obtain the embedding features of the first account; the embedding features of the first account are then input into the output vector matrix in the preset account embedding feature model to determine the output one-hot vector of the second account in the training sample; wherein the output one-hot vector can be used to indicate the characteristics corresponding to the second account that operates the first object, that is, the first account and the second account operate on the same object; for example, any operation such as liking, commenting, and playing the same video is performed.

[0106] Furthermore, the difference between the output one-hot vector of the second account and the preset one-hot vector corresponding to the second account can be determined, that is, the different characteristics between the accounts operating the same object are determined. When it is determined that the difference belongs to the preset range, that is, the characteristics between different accounts operating the same object are not much different, the preset account embedding feature model obtained by adjusting the input vector matrix and the output vector matrix based on the difference is determined, so that the preset account embedding feature model is used as the trained account embedding feature model.

[0107] In the embodiment of the present disclosure, after the account embedding feature model is determined, the one-hou vector corresponding to the account may be input into the account embedding feature model to obtain the first embedding feature of the account.

[0108] Step c: Determine the statistical features corresponding to the account, wherein the statistical features are obtained based on usage statistics of the account on the video platform.

[0109] In the disclosed embodiment, the statistical features corresponding to the account may include but are not limited to one or more of the following features: the number of logins in the previous 7 days before sorting, the number of logins in the previous 30 days before sorting, the average daily online time, the average number of video clicks per day, the average number of live broadcast clicks per day, the ratio of video and live broadcast consumption time, the average number of likes, attentions, comments, long broadcasts, and searches per day. Among them, long broadcasts can be understood as the video playback time being longer than a preset time, such as 18 seconds.

[0110] Step d: Input the embedded features and statistical features corresponding to the account into a preset tag information model to determine the predicted tag information corresponding to the account.

[0111] In the embodiments of the present disclosure, the predicted attribute tag information corresponding to the account can be determined based on the characteristics determined by the account's operations on a specific object, as well as the feedback and preset attribute level model generated by its operations on other objects in the entire video platform. That is, the characteristics generated by the account's operations on a specific object and other objects can be comprehensively combined to comprehensively judge the predicted attribute tag information corresponding to the account, so that the predicted attribute tag information corresponding to the account can be determined more accurately.

[0112] Step f: Compare the predicted attribute labeling information corresponding to the account with the corresponding attribute labeling information to determine a first comparison result. When it is determined that the adjusted preset attribute labeling model corresponding to the first comparison result has converged, an attribute labeling model is obtained.

[0113] In the disclosed embodiment, the statistical features corresponding to the account are used to predict the attribute tag information of the account, so that the division of the attribute tag information of the account can be more generalized, that is, even for an account without too many features, the corresponding attribute tag information can be determined.

[0114] Step 203: Based on the attribute tag information of each account and the feedback information of each account operating on the object to be sorted, determine the second embedding feature of the object to be sorted.

[0115] In the embodiment of the present disclosure, the weight corresponding to the account can be determined. For example, the weight corresponding to the account can be determined based on the attribute tag information of the account. For example, if the attribute tag information of account # is high-quality, the weight corresponding to account # is determined to be 2, if the attribute tag information of account * is ordinary, the weight corresponding to account * is determined to be 1, and if the attribute tag information of account s is abnormal, the weight corresponding to account s is determined to be 0.2. Further, the first embedded feature of the account can be multiplied by its corresponding weight to obtain multiple first results, and then the multiple first results are added to obtain the second embedded feature of the object to be sorted.

[0116] Step 204: Determine the click probability of the objects to be sorted based on the first embedded feature and the second embedded feature; and determine the display order of the objects to be sorted based on the click probability.

[0117] In the disclosed embodiment, after the first embedded feature and the second embedded feature are obtained, the first embedded feature and the second embedded feature may be input into a click prediction model to determine the click probability of the object to be sorted.

[0118] Exemplarily, the click prediction model can be trained in the following manner: input the first embedded feature into the preset click prediction model to obtain the first processing result; input the second embedded feature into the preset click prediction model to obtain the second processing result; input the first processing result and the second processing result into the preset click prediction model to obtain the third processing result; the third processing result is obtained by combining the first processing result and the second processing result; wherein the preset click prediction model predicts the actual click degree of the account on the object based on the operation of the account on the object. Specifically, the preset click prediction model includes a fully connected layer network and an activation layer network, and the first processing result and the second processing result obtained through the fully connected layer network and the activation layer network are corresponding vectors.

[0119] Furthermore, the third processing result can be input into a preset click prediction model to obtain a predicted click result; then the predicted click result and the actual click result of the account for the object to be sorted are compared to determine a second comparison result. When it is determined that the adjusted preset prediction click model corresponding to the second comparison result has converged, the click prediction model is obtained.

[0120] In the embodiment of the present disclosure, after the click probabilities of the objects to be sorted are obtained, the display order of the objects to be sorted may be determined based on the click probabilities.

[0121] Exemplarily, the preset interval to which the click probability belongs may be determined, and then the display order of the objects to be sorted may be determined based on the mapping relationship between the preset interval and the display order.

[0122] For example, if it is determined that the click probability of object A to be sorted is 0.78, and it is determined that the preset interval to which the click probability belongs is the second interval, then based on the mapping relationship between the preset interval and the display order, it can be determined that the display order of object A belongs to the second echelon. If the multiple objects to be displayed where object A is currently located do not include the first echelon, object A will be displayed at the first place in the search video and video list. If the multiple objects to be displayed where object A is currently located include the first echelon, object A will be displayed at the first place after the first echelon in the search video and video list.

[0123] It can be seen that in the embodiment of the present disclosure, the attribute tag information corresponding to the account is determined by the attribute tag model, that is, the attributes of the user who logs into the account are judged, and the new account is identified by generalizing the first embedded feature corresponding to the account, ensuring the accurate judgment of the attribute tag information of the account, and the true effectiveness of the operation behavior of the objects to be sorted is predicted by accounts with different attribute tag information, and the click probability of the objects to be sorted is obtained, so as to sort the objects to be sorted based on the click probability. In this way, not only can high-quality sorting results be displayed to users, but also the influence of accounts with abnormal attribute tag information on the sorting results can be reduced, and the positive influence of new accounts without attribute tag information and accounts with ordinary attribute tag information on the sorting results can be ensured, thereby enhancing the accuracy and effectiveness of the sorting results, thereby reducing the occurrence of behaviors of using accounts to brush rankings, and further having a positive impact on the overall publishing and browsing environment in the video platform.

[0124] Based on the same inventive concept, the disclosed embodiment provides an object sorting device, which can implement the functions corresponding to the aforementioned object sorting method. The object sorting device can be a hardware structure, a software module, or a hardware structure plus a software module. The object sorting device can be implemented by a chip system, which can be composed of chips or include chips and other discrete devices. See Figure 3 As shown, the object sorting device includes a determination unit 301, a first processing unit 302, a second processing unit 303, a third processing unit 304 and a sorting unit 305. Wherein:

[0125] A determining unit 301 is configured to determine the objects to be sorted and basic information of an account that operates the objects to be sorted;

[0126] The first processing unit 302 is configured to input the basic information of each of the accounts into an attribute labeling model, determine the attribute labeling information of each of the accounts and a first embedded feature of each of the accounts; the first embedded feature is used to characterize the account characteristics determined based on the basic information of the account;

[0127] The second processing unit 303 is configured to determine a second embedding feature of the object to be sorted based on the attribute tag information of each account and feedback information of each account operating the object to be sorted;

[0128] The third processing unit 304 is configured to determine the click probability of the object to be sorted based on the first embedded feature and the second embedded feature;

[0129] The sorting unit 305 is configured to determine the display order of the objects to be sorted based on the click probability.

[0130] In a possible implementation, the apparatus further includes a training unit, which is configured to perform training on the attribute labeling model based on the following method:

[0131] Determine a training sample; wherein the training sample includes a plurality of accounts and attribute labeling information corresponding to the plurality of accounts determined based on prior knowledge; the attribute labeling information is any one of high-quality, abnormal, and ordinary;

[0132] Inputting the training sample into the account embedding feature model to obtain the embedding feature corresponding to the account;

[0133] Determine statistical features corresponding to the account, wherein the statistical features are obtained based on usage statistics of the account on the video platform;

[0134] Inputting the embedded features and the statistical features corresponding to the account into a preset attribute labeling model to determine the predicted attribute labeling information corresponding to the account;

[0135] The predicted attribute labeling information corresponding to the account is compared with the corresponding attribute labeling information to determine a first comparison result. When it is determined that the adjusted preset attribute labeling model corresponding to the first comparison result has converged, an attribute labeling model is obtained.

[0136] In a possible implementation, the training unit is configured to perform training on the account embedding feature model based on the following method:

[0137] Inputting a preset one-hot vector corresponding to the first account in the training sample into an input vector matrix in a preset account embedding feature model to obtain an embedding feature of the first account;

[0138] The first account embedding feature is input into the output vector matrix in the preset account embedding feature model to determine the output one-hot vector of the second account in the training sample; wherein the first account and the second account operate on the same object;

[0139] Determine the difference between the output one-hot vector of the second account and the preset one-hot vector corresponding to the second account. When it is determined that the difference belongs to a preset range, adjust the input vector matrix and the output vector matrix based on the difference to obtain an account embedding feature model.

[0140] In a possible implementation manner, the second processing unit 303 is configured to execute:

[0141] Determine the weight corresponding to each of the accounts based on the attribute tag information of each of the accounts and the preset level and weight mapping relationship;

[0142] The first embedded feature of each of the accounts is multiplied by its corresponding weight to obtain a plurality of first results; and the plurality of first results are added to obtain a second embedded feature of the object to be sorted.

[0143] In a possible implementation, the third processing unit 304 is configured to input the first embedded feature and the second embedded feature into a click prediction model to determine the click probability of the object to be sorted;

[0144] The apparatus further includes a training unit, which is configured to perform training on the click prediction model based on the following manner:

[0145] Inputting the first embedded feature into a preset click prediction model to obtain a first processing result;

[0146] Inputting the second embedded feature into a preset click prediction model to obtain a second processing result;

[0147] Inputting the first processing result and the second processing result into the preset click prediction model to obtain a third processing result; the third processing result is obtained by combining the first processing result and the second processing result;

[0148] Inputting the third processing result into the preset click prediction model to obtain a predicted click result;

[0149] The predicted click result is compared with the actual click result of the account on the object to be sorted to determine a second comparison result. When it is determined that the adjusted preset predicted click model corresponding to the second comparison result has converged, a click prediction model is obtained.

[0150] In a possible implementation, the sorting unit 305 is configured to execute:

[0151] Determining a preset interval to which the click probability belongs;

[0152] Based on the mapping relationship between the preset intervals and the display order, the display order of the objects to be sorted is determined.

[0153] As mentioned above Figure 2 All relevant contents of each step involved in the embodiment of the object sorting method can be referred to the functional description of the functional unit corresponding to the object sorting device in the embodiment of the present disclosure, and will not be repeated here.

[0154] The division of units in the embodiments of the present disclosure is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processor, or may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0155] Based on the same inventive concept, the present disclosure also provides an electronic device, such as Figure 4 As shown, the electronic device in the embodiment of the present disclosure includes at least one processor 401, and a memory 402 and a communication interface 403 connected to the at least one processor 401. The specific connection medium between the processor 401 and the memory 402 is not limited in the embodiment of the present disclosure. Figure 4 In the example, the processor 401 and the memory 402 are connected via a bus 400. Figure 4 The connections between other components are shown in bold lines, and are not intended to be limiting. The bus 400 can be divided into an address bus, an image bus, a control bus, etc. For ease of illustration, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0156] In the embodiment of the present disclosure, the memory 402 stores instructions that can be executed by at least one processor 401 , and the at least one processor 401 can execute the steps included in the aforementioned object sorting method by executing the instructions stored in the memory 402 .

[0157] Among them, the processor 401 is the control center of the electronic device, and can use various interfaces and lines to connect various parts of the entire fault detection device, and through running or executing instructions stored in the memory 402 and calling images stored in the memory 402, various functions of the computing device and processing images, the computing device is monitored as a whole. Optionally, the processor 401 may include one or more processing units, and the processor 401 may integrate an application processor and a modem processor, wherein the processor 401 mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 401. In some embodiments, the processor 401 and the memory 402 may be implemented on the same chip, and in some embodiments, they may also be implemented separately on independent chips.

[0158] The processor 401 may be a general-purpose processor, such as a central processing unit, i.e., a CPU, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and may implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present disclosure. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present disclosure may be directly embodied as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0159] The memory 402 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs and modules. The memory 402 may include at least one type of storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory, a random access memory (Random Access Memory, RAM), a static random access memory (Static Random Access Memory, SRAM), a programmable read-only memory (Programmable Read Only Memory, PROM), a read-only memory (Read Only Memory, ROM), an electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), a magnetic memory, a disk, an optical disk, etc. The memory 402 is any other medium that can be used to carry or store a desired program code in the form of an instruction or image structure and can be accessed by a computer, but is not limited thereto. The memory 402 in the embodiment of the present disclosure can also be a circuit or any other device that can realize a storage function, for storing program instructions and / or images.

[0160] The communication interface 403 is a transmission interface that can be used for communication, and data can be received or sent through the communication interface 403 .

[0161] See also Figure 5 The electronic device shown is a further structural diagram, which also includes a basic input / output system, i.e., an I / O system 501, for helping to transmit information between various devices within the electronic device, and a large-capacity storage device 505 for storing an operating system 502, application programs 503 and other program modules 504.

[0162] The basic input / output system 501 includes a display 506 for displaying information and an input device 507 such as a mouse and a keyboard for user inputting information. The display 506 and the input device 507 are connected to the processor 401 through the basic input / output system 501 connected to the system bus 400. The basic input / output system 501 may also include an input / output controller for receiving and processing inputs from a plurality of other devices such as a keyboard, a mouse, or an electronic stylus. Similarly, the input / output controller also provides output to a display screen, a printer, or other types of output devices.

[0163] The mass storage device 505 is connected to the processor 401 through a mass storage controller (not shown) connected to the system bus 400. The mass storage device 505 and its associated computer readable medium provide non-volatile storage for the server package. That is, the mass storage device 505 may include a computer readable medium (not shown) such as a hard disk or a CD-ROM drive.

[0164] According to various embodiments of the present disclosure, the computing device package can also be connected to a remote computer on the network through a network such as the Internet. That is, the computing device can be connected to the network 508 through the communication interface 403 connected to the system bus 400, or the communication interface 403 can be used to connect to other types of networks or remote computer systems (not shown).

[0165] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory 402 including instructions, and the instructions can be executed by a processor 401 of the device to perform the above method. Optionally, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical image storage device, etc.

[0166] In an exemplary embodiment, a computer program product is also provided, which includes a program code. When the program product is run on an electronic device, the program code is used to enable the electronic device to execute the steps of the object sorting method according to various exemplary embodiments of the present disclosure described above in this specification.

[0167] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present disclosure may take the form of a computer program product implemented in one or more computer-usable storage media containing computer-usable program codes, and the computer-usable storage media include but are not limited to disk storage and optical storage, etc.

[0168] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, electronic devices, computer-readable storage media, and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable object sorting device to generate a machine, so that the instructions executed by the processor of the computer or other programmable object sorting device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0169] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable object sorting device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0170] These computer program instructions may also be loaded onto a computer or other programmable object sequence device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0171] Obviously, those skilled in the art can make various changes and modifications to the present disclosure without departing from the spirit and scope of the present disclosure. Thus, if these modifications and variations of the present disclosure fall within the scope of the claims of the present disclosure and their equivalents, the present disclosure is also intended to include these modifications and variations.

Claims

1. A method for sorting objects, It is characterized in that The method comprises: Determine the objects to be sorted and basic information of each account that operates the objects to be sorted; Inputting the basic information of each of the accounts into an attribute labeling model, determining the attribute labeling information of each of the accounts and a first embedded feature of each of the accounts; the first embedded feature is used to characterize the account characteristics determined based on the basic information of the account; Determining the second embedding feature of the object to be sorted based on the attribute tag information of each account and the feedback information of each account operating the object to be sorted includes: Determine the weight corresponding to each of the accounts based on the attribute tag information of each of the accounts and the preset level and weight mapping relationship; Multiplying the first embedded features of each of the accounts by their corresponding weights to obtain a plurality of first results; and adding the plurality of first results to obtain a second embedded feature of the object to be sorted; wherein the feedback information includes the weights corresponding to each of the accounts and the corresponding first results; Based on the first embedded feature and the second embedded feature, a click probability of the objects to be sorted is determined; and based on the click probability, a display order of the objects to be sorted is determined.

2. The method according to claim 1, It is characterized in that The attribute labeling model is trained based on the following method: Determine a training sample; wherein the training sample includes a plurality of accounts and attribute labeling information corresponding to the plurality of accounts determined based on prior knowledge; the attribute labeling information is any one of high-quality, abnormal, and ordinary; Inputting the training sample into the account embedding feature model to obtain the embedding feature corresponding to the account; Determine statistical features corresponding to the account, wherein the statistical features are obtained based on usage statistics of the account on the video platform; Inputting the embedded features and the statistical features corresponding to the account into a preset attribute labeling model to determine the predicted attribute labeling information corresponding to the account; The predicted attribute labeling information corresponding to the account is compared with the corresponding attribute labeling information to determine a first comparison result. When it is determined that the adjusted preset attribute labeling model corresponding to the first comparison result has converged, an attribute labeling model is obtained.

3. The method according to claim 2, It is characterized in that The account embedding feature model is trained based on the following method: Inputting a preset one-hot vector corresponding to the first account in the training sample into an input vector matrix in a preset account embedding feature model to obtain an embedding feature of the first account; The first account embedding feature is input into the output vector matrix in the preset account embedding feature model to determine the output one-hot vector of the second account in the training sample; wherein the first account and the second account operate on the same object; Determine the difference between the output one-hot vector of the second account and the preset one-hot vector corresponding to the second account. When it is determined that the difference belongs to a preset range, adjust the input vector matrix and the output vector matrix based on the difference to obtain an account embedding feature model.

4. The method according to claim 1, It is characterized in that Determining the click probability of the object to be sorted based on the first embedded feature and the second embedded feature includes: Inputting the first embedded feature and the second embedded feature into a click prediction model to determine the click probability of the object to be sorted; The click prediction model is trained based on the following method: Inputting the first embedded feature into a preset click prediction model to obtain a first processing result; Inputting the second embedded feature into a preset click prediction model to obtain a second processing result; Inputting the first processing result and the second processing result into the preset click prediction model to obtain a third processing result; the third processing result is obtained by combining the first processing result and the second processing result; Inputting the third processing result into the preset click prediction model to obtain a predicted click result; The predicted click result and the actual click result of the account on the object to be sorted are compared to determine a second comparison result. When it is determined that the adjusted preset click prediction model corresponding to the second comparison result has converged, a click prediction model is obtained.

5. The method according to claim 1, It is characterized in that Determining the display order of the objects to be sorted based on the click probability includes: Determining a preset interval to which the click probability belongs; Based on the mapping relationship between the preset intervals and the display order, the display order of the objects to be sorted is determined.

6. An object sorting device, It is characterized in that The device comprises: A determination unit, configured to determine the objects to be sorted and basic information of each account that operates the objects to be sorted; A first processing unit is configured to input the basic information of each of the accounts into an attribute labeling model, determine the attribute labeling information of each of the accounts and a first embedded feature of each of the accounts; the first embedded feature is used to characterize the account characteristics determined based on the basic information of the account; The second processing unit is configured to determine the second embedding feature of the object to be sorted based on the attribute tag information of each account and the feedback information of each account operating the object to be sorted, including: Determine the weight corresponding to each of the accounts based on the attribute tag information of each of the accounts and the preset level and weight mapping relationship; Multiplying the first embedded features of each of the accounts by their corresponding weights to obtain a plurality of first results; and adding the plurality of first results to obtain a second embedded feature of the object to be sorted; wherein the feedback information includes the weights corresponding to each of the accounts and the corresponding first results; A third processing unit is configured to determine the click probability of the object to be sorted based on the first embedded feature and the second embedded feature; The sorting unit is configured to determine the display order of the objects to be sorted based on the click probability.

7. The device according to claim 6, It is characterized in that The apparatus further includes a training unit, which is configured to perform training on the attribute labeling model based on the following manner: Determine a training sample; wherein the training sample includes a plurality of accounts and attribute labeling information corresponding to the plurality of accounts determined based on prior knowledge; the attribute labeling information is any one of high-quality, abnormal, and ordinary; Inputting the training sample into the account embedding feature model to obtain the embedding feature corresponding to the account; Determine statistical features corresponding to the account, wherein the statistical features are obtained based on usage statistics of the account on the video platform; Inputting the embedded features and the statistical features corresponding to the account into a preset attribute labeling model to determine the predicted attribute labeling information corresponding to the account; The predicted attribute labeling information corresponding to the account is compared with the corresponding attribute labeling information to determine a first comparison result. When it is determined that the adjusted preset attribute labeling model corresponding to the first comparison result has converged, an attribute labeling model is obtained.

8. The device according to claim 7, It is characterized in that The training unit is configured to perform training on the account embedding feature model based on the following method: Inputting a preset one-hot vector corresponding to the first account in the training sample into an input vector matrix in a preset account embedding feature model to obtain an embedding feature of the first account; The first account embedding feature is input into the output vector matrix in the preset account embedding feature model to determine the output one-hot vector of the second account in the training sample; wherein the first account and the second account operate on the same object; Determine the difference between the output one-hot vector of the second account and the preset one-hot vector corresponding to the second account. When it is determined that the difference belongs to a preset range, adjust the input vector matrix and the output vector matrix based on the difference to obtain an account embedding feature model.

9. The device according to claim 6, It is characterized in that The third processing unit is configured to input the first embedded feature and the second embedded feature into a click prediction model to determine the click probability of the object to be sorted; The apparatus further includes a training unit, which is configured to perform training on the click prediction model based on the following manner: Inputting the first embedded feature into a preset click prediction model to obtain a first processing result; Inputting the second embedded feature into a preset click prediction model to obtain a second processing result; Inputting the first processing result and the second processing result into the preset click prediction model to obtain a third processing result; the third processing result is obtained by combining the first processing result and the second processing result; Inputting the third processing result into the preset click prediction model to obtain a predicted click result; The predicted click result and the actual click result of the account on the object to be sorted are compared to determine a second comparison result. When it is determined that the adjusted preset click prediction model corresponding to the second comparison result has converged, a click prediction model is obtained.

10. The device according to claim 6, It is characterized in that The sorting unit is configured to perform: Determining a preset interval to which the click probability belongs; Based on the mapping relationship between the preset intervals and the display order, the display order of the objects to be sorted is determined.

11. An electronic device, It is characterized in that include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the object sorting method according to any one of claims 1 to 5.

12. A computer-readable storage medium, It is characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the object sorting method as claimed in any one of claims 1 to 5.

13. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the object sorting method according to any one of claims 1 to 5 is implemented.

Images