Training method, device and equipment of physical resource allocation model and storage medium

Abstract

The invention provides a physical resource allocation model training method and device, equipment and a storage medium. The method relates to the technical fields of resource distribution, resource recommendation, artificial intelligence and the like. According to the specific implementation scheme, a training data set is collected based on historical consumption information of a user; training a satisfaction evaluation model based on the training data set; the satisfaction degree evaluation model is used for predicting the satisfaction degree of a training user on resources recommended based on the training resource candidate set; based on the satisfaction evaluation model, screening multiple pieces of target training data with the maximum satisfaction from the training data set; and adopting the multiple pieces of target training data to train the physical resource allocation model. According to the technology disclosed by the invention, the physical resource allocation model can be effectively trained, so that the proportion of the physical resources of a plurality of recommended resources can be accurately and efficiently predicted.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, specifically to technical fields such as resource distribution, resource recommendation, and artificial intelligence, and in particular to a training method, device, equipment, and storage medium for a genre resource allocation model. Background Art

[0002] Based on the growing demand for consumption efficiency and the popularity of immersive consumption habits, more and more users like to consume resources in an immersive way.

[0003] In various resource recommendation scenarios represented by immersive recommendation, multiple resources can be recommended for each user request. Specifically, multiple resources can be filtered by the popularity of the resource content. Summary of the invention

[0004] The present disclosure provides a training method, apparatus, device and storage medium for a genre resource allocation model.

[0005] According to one aspect of the present disclosure, a method for training a genre resource allocation model is provided, comprising:

[0006] Based on the user's historical consumption information, a training data set is collected, wherein the training data set includes multiple training data, each of which includes training user characteristics, training resource candidate set characteristics, and training user behavior characteristics; and also includes: characteristics of multiple recommended resources recommended based on the training resource candidate set, consumption information of the multiple recommended resources, and real genre proportions of the multiple recommended resources;

[0007] Based on the training data set, a satisfaction evaluation model is trained; the satisfaction evaluation model is used to predict the satisfaction of the training user with the resources recommended based on the training resource candidate set;

[0008] Based on the satisfaction evaluation model, selecting a plurality of target training data with the highest satisfaction from the training data set;

[0009] Multiple target training data are used to train the genre resource allocation model.

[0010] According to another aspect of the present disclosure, a method for predicting resource proportion is provided, comprising:

[0011] Obtain user features, resource candidate set features, and user behavior features respectively;

[0012] Based on the user characteristics, the resource candidate set characteristics and the user behavior characteristics, a pre-trained genre resource allocation model is used to predict the proportion of various genre resources in a plurality of resources recommended to the user based on the resource candidate set.

[0013] According to another aspect of the present disclosure, a training device for a genre resource allocation model is provided, comprising:

[0014] A collection module is used to collect a training data set based on the user's historical consumption information, wherein the training data set includes multiple training data, each of which includes training user characteristics, training resource candidate set characteristics, and training user behavior characteristics; and also includes: characteristics of multiple recommended resources recommended based on the training resource candidate set, consumption information of the multiple recommended resources, and real genre proportions of the multiple recommended resources;

[0015] A first training module is used to train a satisfaction evaluation model based on the training data set; the satisfaction evaluation model is used to predict the satisfaction of the training user with the resources recommended based on the training resource candidate set;

[0016] A screening module, used for screening a plurality of target training data with the highest satisfaction from the training data set based on the satisfaction evaluation model;

[0017] The second training module is used to train the genre resource allocation model using multiple target training data.

[0018] According to another aspect of the present disclosure, a device for predicting resource proportion is provided, comprising:

[0019] An acquisition module, used to respectively acquire user features, resource candidate set features, and user behavior features;

[0020] The prediction module is used to predict the proportion of various genre resources in multiple resources recommended to the user based on the resource candidate set, using a pre-trained genre resource allocation model based on the user characteristics, the resource candidate set characteristics and the user behavior characteristics.

[0021] According to yet another aspect of the present disclosure, there is provided an electronic device, including:

[0022] at least one processor; and

[0023] a memory communicatively connected to the at least one processor; wherein,

[0024] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any possible implementation manner and the aspects described above.

[0025] According to yet another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method of the above-mentioned aspect and any possible implementation manner.

[0026] According to yet another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the computer program implements the above-mentioned aspects and the method of any possible implementation manner.

[0027] According to the technology disclosed in the present invention, the genre resource allocation model can be effectively trained, and thus the proportion of genre resources of multiple recommended resources can be accurately and efficiently predicted.

[0028] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0030] Figure 1 is a schematic diagram according to a first embodiment of the present disclosure;

[0031] Figure 2 is a schematic diagram according to a second embodiment of the present disclosure;

[0032] Figure 3 This is a schematic diagram of a resource recommendation device provided by an embodiment of the present disclosure.

[0033] Figure 4 is a schematic diagram according to a third embodiment of the present disclosure;

[0034] Figure 5 is a schematic diagram according to a fourth embodiment of the present disclosure;

[0035] Figure 6 is a schematic diagram according to a fifth embodiment of the present disclosure;

[0036] Figure 7 is a schematic diagram according to a sixth embodiment of the present disclosure;

[0037] Figure 8 is a schematic diagram according to a seventh embodiment of the present disclosure;

[0038] Fig. 9 is a schematic diagram according to an eighth embodiment of the present disclosure;

[0039] Fig.10 is a schematic diagram according to a ninth embodiment of the present disclosure;

[0040] Fig.11 is a schematic diagram according to a tenth embodiment of the present disclosure;

[0041] Fig.12 is a schematic diagram according to an eleventh embodiment of the present disclosure;

[0042] Fig.13 The block diagram is a block diagram of an electronic device for implementing the method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0043] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0044] Obviously, the described embodiments are only part of the embodiments of the present disclosure, but not all of them. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present disclosure.

[0045] It should be noted that the terminal devices involved in the embodiments of the present disclosure may include but are not limited to mobile phones, personal digital assistants (PDAs), wireless handheld devices, tablet computers and other smart devices; display devices may include but are not limited to personal computers, televisions and other devices with display functions.

[0046] In addition, the term "and / or" in this article is only a description of the association relationship between the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0047] Figure 1 is a schematic diagram according to the first embodiment of the present disclosure; Figure 1 As shown, this embodiment provides a resource recommendation method, which is applied to an immersive resource recommendation scenario and may specifically include the following steps:

[0048] S101, based on the user's request, respectively recall resources of various genres to obtain recall results of resources of various genres;

[0049] The resource recommendation method of this embodiment may be executed by a resource recommendation device, which may be an electronic entity or an application integrated with software, and is used to make immersive resource recommendations to users. The resource recommendation device may be applied in a resource recommendation system.

[0050] The immersive resource recommendation of this embodiment means that the user does not have specific request content, and the resource recommendation device autonomously recommends resources to the user continuously. For the user, only need to slide the screen up and down to consume different resources and achieve an immersive experience.

[0051] The genre of the resource in this embodiment may refer to the style of the resource or the format of the resource.

[0052] In this embodiment, when a user clicks on an entry resource to enter an immersive recommendation scenario, it can be considered that the user has initiated a resource request; or when the user consumes the next resource by sliding down in the immersive resource recommendation scenario, if the user has consumed a preset proportion of the currently recommended resources, if he slides up again, it can also be considered that the user has initiated a resource request.

[0053] Since the technical solution of this embodiment is applied in an immersive resource recommendation scenario, the user's request may be an implicit request that does not include any specific request content or request limiting conditions.

[0054] S102, sorting multiple resources of corresponding genres based on the recall results of resources of various genres, and obtaining first sorting results of resources of various genres;

[0055] Specifically, the resource sorting in step 102 is to sort multiple resources of each genre within the genre, where the multiple resources can be only some of the relatively high-quality resources in the resource recall result of the genre. After processing in this step, each genre of resources can obtain a first sorting result.

[0056] S103, configuring the proportion of various genre resources based on the pre-trained genre resource allocation model;

[0057] The genre resource allocation model of this embodiment can reasonably and accurately configure the proportion of various genre resources, and can cater to the personalized needs of users, so that immersive resources can be recommended to users more specifically in the future, effectively improving the effect of immersive resource recommendations.

[0058] S104, based on the proportions of resources of various genres and the first ranking results of the resources of various genres, comprehensively ranking a preset number of resources of various genres to obtain a second ranking result;

[0059] The sorting in step S102 is to sort the resources of various genres separately, so as to facilitate the selection of high-quality resources from the resources of various genres. The difference is that the sorting in step S104 is a comprehensive sorting of the resources of various genres mixed together, so as to facilitate the subsequent selection of several resources to be recommended.

[0060] S105. Based on the second sorting result, push several resources to the user's client in an immersive resource recommendation manner.

[0061] The number of resources in this embodiment may be the first number of resources in the second sorting result. The number of the number of resources may be set based on demand, for example, 6, 8 or other number of resources, which is not limited here.

[0062] The resource recommendation method of this embodiment, by adopting the above steps, can recommend resources of multiple genres together, which can effectively make up for the immersive resource recommendation scenario in the prior art that can only recommend resources of the video genre, and can effectively enrich the resource recommendation genres and improve the resource recommendation efficiency. Moreover, in this embodiment, the proportion of resources of various genres can be configured before recommendation, which can effectively optimize the content of resource recommendation, effectively improve the accuracy of resource recommendation, and thus improve the resource recommendation efficiency.

[0063] Figure 2 is a schematic diagram according to the second embodiment of the present disclosure; the resource recommendation method of this embodiment, in the above Figure 1 Based on the technical solutions of the embodiments shown in the figure, the technical solutions of the present disclosure are further described in more detail. Figure 2 As shown, the resource recommendation method of this embodiment may specifically include the following steps:

[0064] S201, based on the user's request, respectively recall resources of various genres to obtain recall results of resources of various genres;

[0065] In this embodiment, in order to improve the matching degree between the recalled resources and the user, when recalling resources of various genres, the recall can be performed based on the user's information. The user's information can include the user's attribute information, such as the user's age, gender, occupation, etc. Further, the user's information can also include the user's interests and hobbies, such as finance, entertainment, sports, etc. The user's interests and hobbies can be identified in the user's attribute information in the form of tags.

[0066] Further optionally, in an embodiment of the present disclosure, when recalling resources of various genres, the popularity of each resource may also be referred to, and resources with high popularity are recalled preferentially.

[0067] Further optionally, in an embodiment of the present disclosure, if a user is a new user and enters the immersive recommendation scene for the first time, the resource recommendation device can recall resources of various genres according to the popularity of each resource based on the user's request to obtain recall results of resources of various genres.

[0068] It should be noted that the above-mentioned resource recall rules are examples in the embodiments of the present disclosure. In actual applications, other rules can also be used to recall resources, such as based on the distance between the location information published by the resource and the user's location, giving priority to recalling resources whose published locations are close to the user's location, etc., and no further examples will be given here.

[0069] The technical solution of this embodiment can be applied to the mixed recommendation of resources of various genres in an immersive recommendation scenario. The various genres in this embodiment may include videos and graphics; wherein the videos may include short videos and / or small videos; wherein the playback time of the small videos is less than the playback time of the short videos. Graphics include dynamic graphics and / or text graphics. Dynamic graphics must include pictures and may include a small amount of text to describe the pictures. Text graphics must include text and may include a small amount of pictures to explain the text.

[0070] S202, respectively obtaining a plurality of resources ranked top from the recall results of resources of various genres;

[0071] In actual applications, the number of resources included in the resource recall results of various genres is very large. In order to improve the efficiency of resource recommendation, the recalled resources of various genres can be preliminarily sorted. This sorting can be called rough sorting. The specific sorting mechanism and sorting method can refer to the rough sorting in the existing resource recommendation architecture, which will not be repeated here. In this embodiment, step S202 can select multiple resources with top rankings based on the results of the rough sorting.

[0072] S203, for each resource among the multiple resources of each genre, obtaining resource characteristics and user characteristics;

[0073] In this embodiment, the characteristics of the resource may include at least one of the length of the resource, the identifier of the resource, the tag of the resource, the genre of the resource, the entry genre of the resource, and the estimated effect characteristics of the resource.

[0074] When the resource genre is a video resource, the resource length refers to the duration of the video; when the resource genre is a graphic resource, the resource length refers to the length of the text included in the resource and the number of images included.

[0075] The entry genre feature of the resource in this embodiment may refer to the genre feature of the entry resource of the resource, which can identify the genre of the entry resource of the resource as video or graphic;

[0076] The estimated effect characteristics of the resources in this embodiment may include at least one of the estimated user playback time of the resource, the estimated like value, the estimated attention value, the estimated forwarding value, the estimated completion rate and the estimated delay value.

[0077] The estimated delayed value is used to indicate whether the user will continue to want to view other resources of the author of the resource after consuming the resource, or whether the user will continue to want to view other related resources of the resource, such as related content resources or resources with related tags.

[0078] The user characteristics of this embodiment may include: at least one of the user's attribute characteristics and the user's historical consumption characteristics.

[0079] The attribute characteristics of a user may include at least one of the basic attribute characteristics of the user and the preference characteristics of the user; the basic attribute characteristics of the user include the user's age, gender, occupation, etc. The preference characteristics of the user include the user's interests and hobbies, etc. The preference characteristics of the user may be identified in the attribute information of the user in the form of a tag, for example, interests and hobbies may include travel, entertainment, football, etc. The attribute characteristics of the user are all related to the resources to a certain extent, and can provide a reference for predicting the user's satisfaction with the resources. For example, different ages may have different resource preferences, for example, the middle-aged and elderly like to consume video resources; middle-aged office workers like to consume graphic resources; middle-aged women like to consume fashion resources, while middle-aged men like to consume financial resources; and the elderly like to consume health resources, and so on.

[0080] The historical consumption characteristics of the user may include at least one of the consumption proportion of resources of various genres consumed by the user in multiple historical time periods before the current period, the duration of consumption of resources of various genres, and the completion rate of resources of various genres. If the genre of the currently predicted resource belongs to a resource genre with a relatively high consumption proportion in multiple historical time periods, the corresponding satisfaction value will be higher. On the contrary, if the genre of the currently predicted resource belongs to a resource genre with the lowest consumption proportion in multiple historical time periods, the corresponding satisfaction value will be lower. Similarly, if the genre of the currently predicted resource belongs to a resource with a relatively high completion rate in multiple historical time periods, the corresponding satisfaction value will be higher. On the contrary, if the genre of the currently predicted resource belongs to a resource with a relatively low completion rate in multiple historical time periods, the corresponding satisfaction value will be lower.

[0081] The length of the current time period can be a preset time length defined in the resource recommendation scenario, such as the current time period can be the current day, the current week, etc. Multiple historical time periods can be configured according to demand, for example, including historical one day, historical three days, historical seven days, etc. The length and number of specific historical time periods are not limited here. The consumption proportion of users consuming resources of various genres, the duration of consumption of resources of various genres, and the completion rate of resources of various genres in each historical time period can be obtained by statistically analyzing the historical consumption information of the user.

[0082] Further optionally, in this embodiment, a scene feature can also be obtained, and the scene feature can identify whether the current immersive resource recommendation scene belongs to a recommendation scene or a discovery scene. The recommendation scene and the discovery scene of this embodiment can be two different sections in the application. Different users may have different scene preferences. Based on the scene feature, the accuracy of the predicted user satisfaction value for the resource can be improved.

[0083] In specific implementation, for each estimated effect feature, a corresponding estimation model can be pre-trained. When in use, at least one of the above-mentioned resource features, user features, scene features, and resource entry genre features can be input into the estimation model, and the estimation model can predict and output the corresponding estimated effect. In this way, the user's estimated playback time, estimated like value, estimated attention value, estimated forwarding value, estimated completion rate, and estimated delayed value of the resource can be estimated. Each feature can be represented by a numerical value, for example, the numerical value can be a value between 0 and 1. The higher the numerical value, the greater the probability of the corresponding effect.

[0084] S204, for each resource among the multiple resources of each genre, based on the characteristics of the resource and the characteristics of the user, using a pre-trained satisfaction prediction model to predict the satisfaction value of the resource;

[0085] When making specific predictions, for each resource, the resource characteristics and user characteristics are input into the satisfaction prediction model, and the satisfaction prediction model can predict and output a satisfaction value. In practical applications, the satisfaction value can be a value between 0 and 1. The larger the value, the higher the satisfaction level of the resource, and the more worthy of recommendation.

[0086] Optionally, if scenario features are obtained at the same time, the scenario features, together with resource features and user features, are input into the satisfaction prediction model to improve the accuracy of the satisfaction value predicted by the satisfaction prediction model.

[0087] S205, sorting multiple resources of various genres based on the satisfaction value of each resource, and obtaining a first sorting result of the resources of various genres;

[0088] For example, when step S205 is specifically implemented, it may include the following steps:

[0089] (a1) For each resource in the plurality of resources of each genre, multiply the satisfaction value of the resource by the score of the acquired resource to obtain the updated score of the resource;

[0090] (b1) For each resource among the plurality of resources of each genre, the resources are sorted based on the updated score of the resource to obtain a first sorting result of the resources of each genre.

[0091] In this embodiment, the obtained resource scores may refer to the scores corresponding to the coarse sorting of the recalled resources after the resources are recalled. The specific acquisition method may refer to the relevant schemes of the coarse and fine sorting layer of the resource recommendation device, which is not limited here.

[0092] In this embodiment, by estimating the satisfaction value of each resource, each resource can be sorted more accurately, thereby effectively improving the accuracy and efficiency of subsequent resource recommendations.

[0093] Steps S202-S205 are as above Figure 1 A specific implementation of step S102 of the illustrated embodiment. By this method, resources of various genres can be sorted more reasonably, accurately and effectively.

[0094] S206, obtaining user attribute characteristics, resource candidate set characteristics, and user behavior characteristics;

[0095] S207, based on the user's attribute characteristics, the resource candidate set characteristics and the user's behavior characteristics, using a genre resource allocation model to predict the proportion of various genre resources in the multiple resources recommended to the user based on the resource candidate set;

[0096] The user characteristics in this step may also include at least one of the user's attribute characteristics and the user's historical consumption characteristics. For details, please refer to the above related records, which will not be repeated here.

[0097] The resource candidate set features of this embodiment include features of a plurality of top-ranked resources obtained from the first ranking results of various genre resources.

[0098] In this embodiment, during the specific implementation, the number of resource candidate sets can be pre-configured, such as 400. In order to enrich the resource genres in the resource candidate set, it is necessary to select some top-ranked resources from the first ranking results of resources of various genres. For example, if the multi-genre resources include small videos, short videos, dynamic graphics and text graphics, the first 100 resources of each genre can be selected. Alternatively, a lower limit can be set for each genre, such as a minimum of 50, and the remaining number can be supplemented by resources of other genres. Regardless of the method used, as long as the resource candidate set is guaranteed to include genres of various resources, the number reaches the number of resource candidate sets.

[0099] The resource candidate set features include feature expressions of multiple resources; the feature expression of each resource is obtained by vector expression and pooling operation based on at least one of the resource title, resource tag, resource genre, and resource length; and also includes the average feature expression of the resource candidate set.

[0100] User behavior characteristics include user's historical consumption characteristics, user's satisfaction sequence characteristics and / or user's resource consumption scenario characteristics;

[0101] The historical consumption characteristics of the user include at least one of the consumption proportion of resources of various genres consumed by the user in multiple historical time periods before the current period, the duration of consumption of resources of various genres, and the completion rate of resources of various genres;

[0102] The user's satisfaction sequence feature includes the feature expressions of multiple resources that the user is most satisfied with within a preset time period before the current period. The number of resources included can be set according to demand or experience, for example, it can be 10, 20, etc., which is not limited here. The feature expression of the specific resource is the same as the feature expression of the resource in the above-mentioned resource candidate set feature.

[0103] Steps S206-S207 are as above Figure 1 A specific implementation of step S103 of the illustrated embodiment. In this way, the proportion of resources of various genres can be accurately predicted, so that resource recommendations can be made more targeted later, thereby effectively improving the accuracy and efficiency of resource recommendations.

[0104] S208, acquiring a second preset number of resources ranked top from the first ranking results of the resources of various genres according to the proportions of the resources of various genres; the second preset number of resources includes resources of various genres;

[0105] S209, comprehensively sorting the second preset number of resources to obtain the second sorting result;

[0106] Specifically, for resources of various genres, the second preset number can be multiplied by the proportion of resources of the genre to obtain the number of resources of the genre. Then, from the first sorting result of the resources of the genre, a corresponding number of resources are selected in order from front to back. In this way, a second preset number of resources can be obtained in total, providing effective resource support for subsequent immersive resource recommendations.

[0107] Further optionally, when step S209 of this embodiment is specifically implemented, it may include the following steps:

[0108] (a2) for each resource in the second preset number of resources, obtaining at least one of a score of the obtained resource, a context relevance factor between the resource and a previously recommended resource, a quality value of the resource, an estimated effect feature of the resource, an estimated decline rate, and an estimated exit rate;

[0109] The contextual relevance factor between the current resource and the previous recommended resource can be calculated using a pre-trained relevance factor prediction model. During the specific calculation, the feature expression of the current resource and the feature expression of the previous resource are input into the relevance factor prediction model. The model can predict and output the contextual relevance factor between the current resource and the previous recommended resource.

[0110] The obtained resource score may refer to the score updated in step S205.

[0111] The quality value of a resource can be used to comprehensively characterize the quality of the resource, and can be obtained by evaluating using a quality assessment strategy or a quality assessment model.

[0112] The estimated effect characteristics of resources can refer to the relevant records of the above embodiments, which will not be repeated here.

[0113] The estimated sliding rate and estimated exit rate of a resource are used to predict the probability that a user will slide down and consume the next resource after clicking on the current resource, as well as the probability that the user will exit the current resource when clicking on it. They can respectively represent the user's preference for the current resource and the delayed feeling it brings.

[0114] Specifically, the estimated decline rate and estimated exit rate of resources can also be predicted by referring to the acquisition method of the estimated effect characteristics of resources and using a pre-trained estimation model. When making specific predictions, accurate predictions can be made based on the characteristics of users and resources.

[0115] (b2) Based on the Evolution Strategies (ES) algorithm, the weight factor of each feature is obtained; different genres of resources have different weight factors;

[0116] Specifically, the ES algorithm is used to optimize and find the optimal weight factor of each feature. For details, please refer to the implementation principle of the ES algorithm.

[0117] Resources of different genres have different focuses on the features considered. In order to improve accuracy, different weight factors are sought for resources of different genres.

[0118] (c2) Based on the weight factors of each feature, as well as the scores of each resource, the contextual relevance factors of each resource, the quality values ​​of each resource, the estimated effect characteristics of each resource, the estimated decline rate and the estimated exit rate, a second preset number of resources are fused and sorted to obtain a second sorting result.

[0119] For example, in a specific implementation, for each resource in the second preset number of resources, the resource score, the context relevance factor of the resource, the quality value of the resource, the estimated effect characteristics of each resource, the estimated decline rate and the estimated exit rate can be multiplied by the weight factor of the corresponding feature respectively as the comprehensive score of the resource; then, based on the comprehensive score of each resource, the second preset number of resources are fused and sorted to obtain the second sorting result.

[0120] Similarly, when fusion sorting is performed, the more features are referenced, the higher the accuracy of the sorting is. Preferably, when fusion sorting is performed, all features including the scores of the above resources, the contextual relevance factors of the resources, the quality values ​​of the resources, the estimated effect characteristics of the resources, the estimated decline rates, and the estimated exit rates are referred to.

[0121] Steps S208-S209 are as above Figure 1 A specific implementation of step S104 of the illustrated embodiment. In this way, the weight factor of each feature of each resource can be accurately obtained, and the second preset number of resources can be sorted more reasonably and accurately.

[0122] S210, obtaining several resources ranked first from the second sorting result;

[0123] S211. Push several resources to the user's client in an immersive resource recommendation manner.

[0124] In this embodiment, since the second sorting result is a comprehensive sorting of resources of multiple genres, the genres of the several resources pushed to the user can be the same or different, which can effectively enrich the types of resources recommended in the immersive resource recommendation scenario.

[0125] The immersive resource recommendation of this embodiment specifically refers to actively and seamlessly recommending several resources to the user without any request from the user. After consuming one resource, the user can directly scroll down and continue to consume the next resource.

[0126] The resource recommendation method of this embodiment takes a request to recommend several resources as an example. In actual application, if the user consumes a preset ratio of several resources in the immersive recommendation scenario and continues to consume resources by sliding down, the next request will be triggered, and the resource recommendation device will continue to recommend resources to the user according to the resource recommendation method of this embodiment. The user is unaware of this and can only feel the immersive resource consumption.

[0127] Figure 3It is an architecture diagram of a resource recommendation device provided by an embodiment of the present disclosure, which may specifically include a recall layer, a coarse and fine ranking layer, a resource proportion configuration layer, and a reordering layer. Specifically, the scheme of the above-mentioned step S201 of this embodiment can be carried out in the recall layer of the resource recommendation device. The scheme of steps S202-S205 can be carried out in the coarse and fine ranking layer of the resource recommendation device. The scheme of steps S208-S209 can be carried out in the reordering layer of the resource recommendation device. Steps S206-S207 can be considered to be based on the resources of the coarse and fine ranking layer, predicting the proportion of various genre resources, and then effectively providing data support for the resources of the reordering layer, which can be specifically carried out in the resource proportion configuration layer. Finally, based on the sorting result of the reordering layer, that is, the second sorting result, several resources with the highest sorting can be selected to push resources to the user.

[0128] The resource recommendation method of this embodiment, by adopting the above technical solution, can effectively enrich the genres of resource recommendations in the immersive recommendation scenario.

[0129] Furthermore, in this embodiment, at the coarse-fine ranking layer, a pre-trained satisfaction model can be used to predict the satisfaction of each resource, so that resources of each genre can be accurately ranked within each genre.

[0130] Furthermore, in this embodiment, a pre-trained genre resource allocation model can also be used to accurately predict the proportion of various genre resources, which can improve the funnel efficiency in the resource allocation stage, provide more accurate resources to the downstream, and effectively improve the efficiency of resource recommendation.

[0131] Furthermore, in this embodiment, at the re-ranking layer, the ES algorithm can also be used to find the optimal feature weight factor, thereby more reasonably and accurately calculating the score of each resource, and further more reasonably and accurately ranking the resources of each genre.

[0132] Figure 4 is a schematic diagram according to the third embodiment of the present disclosure; Figure 4 As shown, this embodiment provides a Figure 1 or Figure 2 The training method of the satisfaction prediction model in the illustrated embodiment may specifically include the following steps:

[0133] S401, configuring a training data group, wherein the training data group includes characteristics of a training user, characteristics of two training resources, and a true satisfaction relationship between the training user and the two training resources;

[0134] S402, based on the characteristics of the training user in the training data group and the characteristics of the two training resources, using a satisfaction prediction model to respectively predict the training user's predicted satisfaction values ​​for the two training resources;

[0135] S403: Adjust parameters of the satisfaction estimation model based on the predicted satisfaction values ​​of the training users for the two training resources and the actual satisfaction relationship of the training users for the two training resources.

[0136] The execution subject of the training method of the satisfaction prediction model in this embodiment is the training device of the satisfaction prediction model.

[0137] In the training process of the satisfaction prediction model of this embodiment, the satisfaction prediction model is trained using a training data group. Specifically, each training data group includes the characteristics of a training user, the characteristics of two training resources, and the real satisfaction relationship between the training user and the two training resources. Moreover, in the training data group, the satisfaction of the training user to the two training resources cannot be at the same level, so the real satisfaction relationship cannot be equal, and must be greater than or less than.

[0138] During the specific training process, for each training resource, the characteristics of the training user and the characteristics of the training resource are input into the satisfaction prediction model, and the satisfaction prediction model can predict and output the predicted satisfaction value of the training user for the training resource. The predicted satisfaction value can be a value between 0 and 1, and the larger the value, the higher the satisfaction; conversely, the smaller the value, the lower the satisfaction.

[0139] In the training of this embodiment, only the actual satisfaction relationship of the training users with the two training resources is used for supervision, which can effectively learn the relationship between the user's satisfaction with different resources, effectively reduce the training difficulty of the satisfaction prediction model, and improve the training efficiency of the model.

[0140] The training method of the satisfaction prediction model of this embodiment configures a training data group and adjusts the parameters of the satisfaction prediction model according to the predicted satisfaction values ​​of the training users for the two training resources and the actual satisfaction relationship between the training users for the two training resources, thereby making the predictions of the satisfaction prediction model more accurate.

[0141] Figure 5 is a schematic diagram according to a fourth embodiment of the present disclosure; Figure 5 As shown, the training method of the satisfaction prediction model in this embodiment is as follows: Figure 4 Based on the technical solution of the illustrated embodiment, the technical solution of the present disclosure is further described in more detail, which may specifically include the following steps:

[0142] S501, collecting characteristics of the training user and the consumption information of the training user from the historical consumption information of the user;

[0143] S502, extracting information of two training resources with different consumption satisfaction levels from the consumption information of the training user based on the pre-configured resource satisfaction strategy table;

[0144] S503: Based on the information of the two extracted training resources, obtain features of the two training resources;

[0145] S504, configuring a real satisfaction relationship between the two training resources based on the consumption satisfaction levels of the two training resources;

[0146] For example, before step S502, it may also include: configuring a resource satisfaction strategy table, the resource satisfaction strategy table including an identification strategy for satisfactory consumption; an identification strategy for insensitive consumption; and an identification strategy for unsatisfactory consumption; the satisfaction level of intentional consumption is greater than the satisfaction level of insensitive consumption; the satisfaction level of insensitive consumption is greater than the satisfaction level of unsatisfactory consumption.

[0147] Optionally, in an embodiment of the present disclosure, configuring the resource satisfaction policy table may specifically include the following steps:

[0148] (a3) Based on the preset satisfaction behavior judgment conditions, configure the identification strategy of satisfied consumption and the identification strategy of indifferent consumption;

[0149] For example, if the user's consumption behavior reaches the preset satisfaction behavior judgment condition, the user's consumption level is determined to be a satisfactory consumption level; if the user's consumption behavior does not reach the preset satisfaction behavior judgment condition, the user's consumption level is determined to be an insensitive consumption level.

[0150] For example, satisfaction behavior judgment conditions may include: if the user's consumption time is greater than a first preset time or a preset time ratio, and if the user has at least one of interactive behaviors, the user is considered satisfied.

[0151] The first preset duration and the preset duration ratio can be set according to the actual scenario, and are not limited here. For example, the time length can be 10s, 15s or 20s. The preset duration ratio can also be set to 60%, 70% or 80% based on experience, and are not limited here.

[0152] The user's interactive behavior may include at least one of liking, commenting, following, and sharing.

[0153] (b3) Based on the preset dissatisfaction behavior judgment conditions, configure the identification strategy of dissatisfied consumption.

[0154] In this embodiment, it is necessary to separately configure a preset unsatisfactory behavior judgment condition. If the condition is met, the user's consumption level can be determined to be an unsatisfactory consumption level. For example, the unsatisfactory behavior judgment condition can be: if the consumption time is greater than the second preset time threshold, or the user exits, it is determined that the user is dissatisfied with the resource; the second preset time threshold is less than the first preset time threshold. For example, the second preset time threshold can be 2s, 3s or other time lengths.

[0155] The characteristics of the training user in this embodiment include: at least one of the attribute characteristics of the training user and the historical consumption characteristics of the training user;

[0156] The attribute characteristics of the training user include: at least one of the basic attribute characteristics of the training user and the preference characteristics of the training user; the historical consumption characteristics of the training user include at least one of the consumption ratio of the training user's consumption of resources of various genres in a plurality of historical time periods before the consumption period of the corresponding training resource and the completion rate of the resources of various genres;

[0157] The characteristics of the training resource include at least one of the length of the training resource, the identification of the training resource, the label of the training resource, the genre of the training resource, the entry genre of the training resource, and the estimated effect characteristics of the training resource;

[0158] When the genre of the training resource is a video resource, the length of the training resource refers to the duration of the video; when the genre of the training resource is a graphic resource, the length of the training resource refers to the length of the text and the number of pictures included in the resource;

[0159] The entry genre of the training resource refers to the genre of the entry resource of the training resource, and the genre of the entry resource of the training resource can be identified as video or graphic;

[0160] The effect characteristics of the training resources include at least one of the playing time, likes, attention, forwarding, completion rate and delayed value of the training resources by the training users.

[0161] Steps S501-S504 are as above Figure 4 A specific implementation of step S401 of the illustrated embodiment.

[0162] S505, based on the characteristics of the training user in the training data group and the characteristics of the two training resources, using a satisfaction estimation model to respectively predict the predicted satisfaction values ​​of the training user for the two training resources;

[0163] S506, obtaining a predicted satisfaction relationship between the training user and the two training resources based on the predicted satisfaction values ​​of the training user for the two training resources;

[0164] S507 : Based on the predicted satisfaction relationship of the training user with respect to the two training resources and the actual satisfaction relationship of the training user with respect to the two training resources, adjust parameters of the satisfaction estimation model.

[0165] Specifically, it is detected whether the predicted satisfaction relationship of the training user with respect to the two training resources is consistent with the actual satisfaction relationship of the training user with respect to the two training resources. If not, the parameters of the satisfaction prediction model are adjusted so that the satisfaction prediction model can learn to predict the actual satisfaction relationship of the user with respect to the two training resources.

[0166] The training method of the satisfaction prediction model of this embodiment configures a training data group and adjusts the parameters of the satisfaction prediction model according to the predicted satisfaction values ​​of the training users for the two training resources and the actual satisfaction relationship between the training users for the two training resources, thereby making the predictions of the satisfaction prediction model more accurate.

[0167] Figure 6 is a schematic diagram according to the fifth embodiment of the present disclosure; Figure 6 As shown, this embodiment provides a method for training a genre resource allocation model, which may specifically include the following steps:

[0168] S601, collecting training data sets based on the user's historical consumption information;

[0169] The training data set includes multiple training data, each of which includes training user features, training resource candidate set features, and training user behavior features; it also includes: features of multiple recommended resources recommended based on the training resource candidate set, consumption information of the multiple recommended resources, and real genre proportions of the multiple recommended resources;

[0170] One piece of training data corresponds to a resource recommendation of one scroll, or the data of the resource recommendation process requested by the user. Each feature in the training data can be calculated or obtained based on the user's historical consumption information.

[0171] S602, training a satisfaction evaluation model based on the training data set; the satisfaction evaluation model is used to predict the satisfaction of the training user with the resources recommended based on the training resource candidate set;

[0172] S603, based on the satisfaction evaluation model, selecting multiple target training data with the highest satisfaction from the training data set;

[0173] S604: Use multiple target training data to train the genre resource allocation model.

[0174] The training method of the genre resource allocation model of this embodiment is implemented based on the structure of the generator-evaluator (GE) model.

[0175] In this embodiment, by using the training data set to train the satisfaction evaluation model, the satisfaction evaluation model can accurately score the satisfaction level of each training data in the training data set. Furthermore, the satisfaction evaluation model is used to filter multiple target training data with the highest satisfaction from the training data set. For example, the 30% or 20% of the training data with the highest satisfaction can be filtered as the target training data. Then, the target training data is used to train the genre resource allocation model, so that the genre resource allocation model can learn the genre resource ratio of high-quality samples.

[0176] By adopting the training method of the genre resource allocation model of this embodiment, it is possible to screen out high-quality training data with very high user satisfaction, train the genre resource allocation model, and further effectively improve the accuracy of the trained genre resource allocation model.

[0177] Further optionally, when step S602 of the above embodiment is specifically implemented, the following steps may be included:

[0178] (a4) for each of the training data in the training data set, based on the training user characteristics, training resource candidate set characteristics, training user behavior characteristics and characteristics of multiple recommended resources in the training data, using the satisfaction evaluation model to predict the predicted satisfaction of the multiple recommended resources;

[0179] (b4) obtaining the real satisfaction of the plurality of recommended resources based on the consumption information of the plurality of recommended resources and the preset satisfaction behavior judgment condition;

[0180] For example, based on the user's consumption information of multiple recommended resources, it can be known whether the user is satisfied with each recommended resource at a single point. Single point satisfaction can refer to the satisfaction behavior judgment condition of a single resource in the aforementioned embodiment, for example, it can include: the user's consumption time is greater than the first preset time or the preset time ratio, and the user has at least one of interactive behaviors, then the user is considered satisfied.

[0181] If the proportion of satisfied consumption of a single point in a user's request, i.e., multiple resources recommended in one swipe, is greater than or equal to a preset ratio threshold, such as 70% or 75%, the user can be considered satisfied with the multiple resources recommended this time. If the proportion of satisfied consumption of a single point in one swipe is less than the preset ratio threshold, the user is considered dissatisfied with the multiple resources recommended this time.

[0182] (c4) Based on the predicted satisfaction and actual satisfaction of multiple recommended resources, the parameters of the satisfaction prediction model are adjusted.

[0183] Further optionally, when step S604 of the above embodiment is specifically implemented, the following steps may be included:

[0184] (a5) For each target training data in the target training data set, based on the training user characteristics, training resource candidate set characteristics and training user behavior characteristics in the target training data, a genre resource allocation model is used to predict the predicted genre proportions of multiple recommended resources;

[0185] (b5) Based on the predicted genre proportions and actual genre proportions of multiple recommended resources, the parameters of the genre resource allocation model are adjusted.

[0186] By adjusting the parameters of the genre resource allocation model, the predicted genre proportion predicted by the genre resource allocation model is made consistent with the actual genre proportion.

[0187] Using multiple training data, the genre resource allocation model is continuously trained in the above manner until a cutoff condition is reached, thereby obtaining a genre resource allocation model.

[0188] Further optionally, in one embodiment of the present disclosure, the training user characteristics include attribute characteristics of the training user; the attribute characteristics of the training user include: at least one of basic attribute characteristics of the training user and preference characteristics of the user.

[0189] Further optionally, in one embodiment of the present disclosure, the training resource candidate set includes feature expressions of multiple training resources; the feature expression of each training resource is obtained by vector expression and pooling operation based on at least one of the title of the training resource, the label of the training resource, the genre of the training resource, and the length of the training resource; and also includes the average feature expression of the training resource candidate set.

[0190] Further optionally, in one embodiment of the present disclosure, the training user behavior characteristics include historical consumption characteristics of the training user, satisfaction sequence characteristics of the training user, and / or scenario characteristics of resource consumption by the training user;

[0191] The historical consumption characteristics of the training user include at least one of the consumption ratio of the user's consumption of resources of various genres, the duration of consumption of resources of various genres, and the completion rate of resources of various genres in multiple historical time periods before the current period;

[0192] The satisfied sequence features of the training user include the feature expressions of multiple resources that the user is most satisfied with within a preset time period before the current period.

[0193] It should be noted that, in this embodiment, although the target training data is also the training data with the highest satisfaction selected from the training data set, in this embodiment, the number of training resources included in the training resource candidate set used in the training satisfaction evaluation model may be smaller than the number of training resources included in the training resource candidate set used in the training of the genre resource allocation model.

[0194] However, the two models are trained using the data corresponding to the same resource recommendation. When training the satisfaction evaluation model, the training resources in the training resource candidate set can be obtained from the results of sorting in the re-ranking layer of the resource recommendation device. When training the genre resource allocation model, the training resources in the training resource candidate set can be obtained from the data sorted in the coarse and fine ranking layer during the resource recommendation process.

[0195] In the above embodiment, by adopting the above rich and comprehensive features, the genre resource allocation model can be effectively trained. In addition, in the above embodiment, the training mode of the GE structure is also adopted to screen out training data with very high user satisfaction, and to train the genre resource allocation model, which can effectively improve the accuracy of the trained genre resource allocation model.

[0196] Figure 7 is a schematic diagram according to the sixth embodiment of the present disclosure; Figure 7 As shown, this embodiment provides a method for predicting resource proportions, which may specifically include the following steps:

[0197] S701, respectively obtaining user characteristics, resource candidate set characteristics and user behavior characteristics;

[0198] S702: Based on user characteristics, resource candidate set characteristics and user behavior characteristics, a pre-trained genre resource allocation model is used to predict the proportion of various genre resources in multiple resources recommended to the user based on the resource candidate set.

[0199] During specific use, user characteristics, resource candidate set characteristics and user behavior characteristics are input into the genre resource allocation model, and the genre resource allocation model can predict and output the proportion of various genre resources in multiple resources recommended to users based on the resource candidate set.

[0200] The solution of this embodiment can be applied after the coarse sorting layer and before the reordering layer of the resource recommendation device. It can effectively predict the proportion of various genres of resources in multiple resources recommended to users, and then based on the proportion, make targeted resource recommendations to users, thereby improving the accuracy and efficiency of resource recommendations.

[0201] Further optionally, in step S701 in the above embodiment, obtaining user characteristics includes:

[0202] Obtain at least one of the user's attribute characteristics and the user's historical consumption characteristics; the user's attribute characteristics include: at least one of the user's basic attribute characteristics and the user's preference characteristics; the user's historical consumption characteristics include the user's consumption ratio of various genres of resources consumed in multiple historical time periods before the current period, and at least one of the completion rate of various genres of resources.

[0203] Further optionally, in step S701 in the above embodiment, obtaining the resource candidate set features includes:

[0204] Acquire feature expressions of multiple resources in the resource candidate set; the feature expression of each resource is obtained by performing vector expression and pooling operations based on at least one of a resource title, a resource tag, a resource genre, and a resource length;

[0205] Based on the feature expressions of multiple resources in the resource candidate set, an average feature expression of the resource candidate set is obtained.

[0206] Further optionally, in step S701 in the above embodiment, obtaining user behavior characteristics includes:

[0207] Obtaining the user's historical consumption characteristics, the user's satisfaction sequence characteristics and / or the scenario characteristics of the user's consumption of resources;

[0208] The historical consumption characteristics of the user include at least one of the consumption proportion of resources of various genres consumed by the user in multiple historical time periods before the current period, the duration of consumption of resources of various genres, and the completion rate of resources of various genres;

[0209] The user's satisfaction sequence characteristics include characteristic expressions of multiple resources that the user is most satisfied with within a preset time period before the current period.

[0210] In the above embodiment, by adopting the above rich and comprehensive features and the genre resource allocation model, the proportion of various genre resources in multiple resources can be accurately and effectively predicted.

[0211] The resource ratio prediction method of this embodiment can be used in addition to the above Figure 1 or Figure 2 The scenario of the embodiment shown can also be applied to other scenarios of recommending multi-genre resources, which is not limited here.

[0212] Figure 8 is a schematic diagram according to the seventh embodiment of the present disclosure; Figure 8 As shown, this embodiment provides a resource recommendation device 800, which is applied in an immersive resource recommendation scenario, including:

[0213] The recall module 801 is used to recall resources of various genres based on user requests and obtain recall results of resources of various genres;

[0214] A first sorting module 802 is used to sort multiple resources of corresponding genres based on the recall results of resources of various genres, and obtain first sorting results of resources of various genres;

[0215] A proportion configuration module 803 is used to configure the proportion of various genre resources based on a pre-trained genre resource allocation model;

[0216] A second sorting module 804 is used to comprehensively sort a preset number of resources of various genres based on the proportions of the resources of various genres and the first sorting results of the resources of various genres to obtain a second sorting result;

[0217] The recommendation module 805 is used to push a number of resources to the client of the user in an immersive resource recommendation manner based on the second sorting result.

[0218] The resource recommendation device 800 of this embodiment implements the implementation principle and technical effect of resource recommendation by adopting the above modules, which is the same as the implementation of the above related method embodiments. For details, please refer to the records of the above related method embodiments, which will not be repeated here.

[0219] Fig. 9 is a schematic diagram according to an eighth embodiment of the present disclosure; Fig. 9 As shown, this embodiment provides a resource recommendation device 900. Figure 8 Based on the technical solutions of the embodiments shown in the figure, the technical solutions of the present disclosure are further described in more detail. Fig. 9 As shown, the resource recommendation device 900 of this embodiment includes the above Figure 8 The modules with the same name and function are shown as follows: a recall module 901 , a first sorting module 902 , a proportion configuration module 903 , a second sorting module 904 , and a recommendation module 905 .

[0220] like Fig. 9 As shown, in the resource recommendation device 900 of this embodiment, the first sorting module 902 includes:

[0221] The first resource acquisition unit 9021 is used to acquire a plurality of resources ranked top from the recall results of resources of various genres;

[0222] A first feature acquisition unit 9022 is used to acquire, for each resource in the plurality of resources of each genre, at least one of a resource feature and a user feature, a scene feature, an entry genre feature of the resource, and a resource estimated effect feature;

[0223] The first prediction unit 9023 is used to predict the satisfaction value of each resource among the multiple resources of each genre by using a pre-trained satisfaction model satisfaction prediction model based on the characteristics of the resource and at least one of the user characteristics, the scene characteristics, the entry genre characteristics of the resource and the estimated effect characteristics of the resource;

[0224] The first sorting unit 9024 is used to sort multiple resources of various genres based on the satisfaction value of each resource, and obtain a first sorting result of the resources of various genres.

[0225] Further optionally, in one embodiment of the present disclosure, the first sorting unit 9024 is used to:

[0226] For each resource among the multiple resources of each genre, multiply the satisfaction value of the resource by the acquired score of the resource to obtain the updated score of the resource;

[0227] For each resource among the multiple resources of each genre, sorting is performed based on the updated score of the resource to obtain a first sorting result of the resources of each genre.

[0228] Further optionally, in an embodiment of the present disclosure, the resource feature includes at least one of a length of the resource, an identifier of the resource, a tag of the resource, and a genre of the resource;

[0229] When the resource genre is a video resource, the resource length refers to the duration of the video; when the resource genre is a graphic resource, the resource length refers to the length of the text and the number of pictures included in the resource;

[0230] The user characteristics include: at least one of the user's attribute characteristics and the user's historical consumption characteristics; the user's attribute characteristics include: at least one of the user's basic attribute characteristics and the user's preference characteristics; the user's historical consumption characteristics include at least one of the user's consumption ratio of resources of various genres consumed in multiple historical time periods before the current period and the completion rate of resources of various genres;

[0231] The scene feature identifier currently belongs to a recommended scene or a discovered scene;

[0232] The entry genre feature of the resource refers to the genre feature of the entry resource of the resource, which can identify the genre of the entry resource of the resource as video or graphic;

[0233] The estimated effect characteristics of the resource include predicting at least one of the estimated playback time, estimated like value, estimated attention value, estimated forwarding value, estimated completion rate and estimated delayed value of the user for the resource.

[0234] Further optionally, if Fig. 9 As shown, in one embodiment of the present disclosure, the proportion configuration module 903 includes:

[0235] The second feature acquisition unit 9031 is used to acquire the user's attribute features, resource candidate set features and user behavior features;

[0236] The second prediction unit 9032 is used to predict the proportion of various genre resources in the multiple resources recommended to the user based on the resource candidate set, using a pre-trained genre resource allocation model based on the user's attribute characteristics, the resource candidate set characteristics and the user behavior characteristics.

[0237] Further optionally, in an embodiment of the present disclosure, the resource candidate set features include feature expressions of multiple resources; the feature expressions of each resource are obtained by performing vector expression and pooling operations based on at least one of a resource title, a resource tag, a resource genre, and a resource length; and also include an average feature expression of the resource candidate set;

[0238] The user behavior characteristics include the user's historical consumption characteristics, the user's satisfaction sequence characteristics and / or the scene characteristics of the user's consumption of resources;

[0239] The historical consumption characteristics of the user include at least one of the consumption proportion of resources of various genres consumed by the user in multiple historical time periods before the current time period, the duration of consumption of resources of various genres, and the completion rate of resources of various genres;

[0240] The user's satisfaction sequence characteristics include characteristic expressions of a plurality of resources that the user is most satisfied with within a preset time period before the current period.

[0241] Further optionally, if Fig. 9 As shown, in one embodiment of the present disclosure, the second sorting module 904 includes:

[0242] The second resource acquisition unit 9041 is used to acquire a second preset number of resources with a top ranking from the first ranking result of the resources of various genres according to the proportion of the resources of various genres; the second preset number of resources includes resources of various genres;

[0243] The second sorting unit 9042 is used to comprehensively sort the second preset number of resources to obtain the second sorting result.

[0244] Further optionally, in one embodiment of the present disclosure, the second sorting unit 9042 is used to:

[0245] For each of the resources in the second preset number of resources, obtaining at least one of a score of the resource that has been obtained, a context relevance factor between the resource and a previously recommended resource, a quality value of the resource, an estimated effect feature of the resource, an estimated decline rate, and an estimated exit rate;

[0246] Based on the evolutionary strategy algorithm, the weight factor of each feature is obtained; different genres of resources have different weight factors;

[0247] Based on the weight factors of each of the characteristics, as well as the scores of each of the resources, the context relevance factors of each of the resources, the quality values ​​of each of the resources, the estimated effect characteristics of each of the resources, the estimated decline rate and at least one of the estimated exit rate, the second preset number of resources are fused and sorted to obtain the second sorting result.

[0248] Further optionally, in one embodiment of the present disclosure, the second sorting unit 9042 is used to:

[0249] For each of the resources in the second preset number of resources, multiply the score of the resource, the context relevance factor of the resource, the quality value of the resource, at least one of the estimated effect feature, the estimated decline rate and the estimated exit rate of each resource by the weight factor of the corresponding feature, respectively, as the comprehensive score of the resource;

[0250] Based on the comprehensive scores of the resources, the second preset number of resources are integrated and sorted to obtain the second sorting result.

[0251] Further optionally, if Fig. 9 As shown, in one embodiment of the present disclosure, the recommendation module 905 includes:

[0252] A third resource acquisition unit 9051 is used to acquire a number of resources ranked top from the second sorting result;

[0253] The push unit 9052 is used to push a number of resources to the client of the user in an immersive resource recommendation manner.

[0254] Further optionally, in an embodiment of the present disclosure, the various genres include videos and graphics; the videos include short videos and / or small videos; the graphics include dynamic graphics and / or text graphics.

[0255] The resource recommendation device 900 of this embodiment implements the implementation principle and technical effect of resource recommendation by adopting the above modules, which is the same as the implementation of the above related method embodiments. For details, please refer to the records of the above related method embodiments, which will not be repeated here.

[0256] Fig.10 is a schematic diagram according to an eighth embodiment of the present disclosure; Fig.10 As shown, this embodiment provides a training device 1000 for a satisfaction prediction model, including:

[0257] Configuration module 1001, configured to configure a training data set, wherein the training data set includes the characteristics of the training user, the characteristics of two training resources, and the real satisfaction relationship of the training user to the two training resources; and the real satisfaction relationship is greater than or less than;

[0258] Prediction module 1002, used to predict the predicted satisfaction value of the training user for the two training resources respectively based on the characteristics of the training user and the characteristics of the two training resources in the training data group and using a satisfaction prediction model;

[0259] The adjustment module 1003 is used to adjust parameters of the satisfaction estimation model based on the predicted satisfaction values ​​of the training users for the two training resources and the actual satisfaction relationship between the training users for the two training resources.

[0260] The training device 1000 of the satisfaction prediction model of this embodiment realizes the implementation principle and technical effect of the training of the satisfaction prediction model by adopting the above-mentioned modules, which is the same as the implementation of the above-mentioned related method embodiments. For details, please refer to the records of the above-mentioned related method embodiments, which will not be repeated here.

[0261] Further optionally, in the above Fig.10 Based on the technical solution of the illustrated embodiment, in one embodiment of the present disclosure, the adjustment module 1003 is used to:

[0262] Based on the predicted satisfaction values ​​of the training user for the two training resources, obtaining the predicted satisfaction relationship of the training user for the two training resources;

[0263] Based on the predicted satisfaction relationship of the training user to the two training resources and the actual satisfaction relationship of the training user to the two training resources, parameters of the satisfaction prediction model are adjusted.

[0264] Further optionally, in one embodiment of the present disclosure, the configuration module 1001 is used to:

[0265] Collecting the characteristics of the training user and the consumption information of the training user from the user's historical consumption information;

[0266] Based on a pre-configured resource satisfaction strategy table, extracting information of two training resources with different consumption satisfaction levels from the consumption information of the training user;

[0267] Based on the information of the two extracted training resources, obtaining features of the two training resources;

[0268] Based on the consumption satisfaction levels of the two training resources, the real satisfaction relationship between the two training resources is configured.

[0269] Further optionally, in one embodiment of the present disclosure, the configuration module 1001 is further configured to:

[0270] The resource satisfaction strategy table is configured, and the resource satisfaction strategy table includes an identification strategy for satisfactory consumption; an identification strategy for insensitive consumption; and an identification strategy for unsatisfactory consumption; the satisfaction level of the intended consumption is greater than the satisfaction level of the insensitive consumption; the satisfaction level of the insensitive consumption is greater than the satisfaction level of the unsatisfactory consumption.

[0271] Further optionally, in one embodiment of the present disclosure, the configuration module 1001 is used to:

[0272] Based on the preset satisfaction behavior judgment conditions, configure the identification strategy of the satisfied consumption and the identification strategy of the insensitive consumption;

[0273] Based on the preset dissatisfaction behavior judgment conditions, the identification strategy of the dissatisfaction consumption is configured.

[0274] Further optionally, in an embodiment of the present disclosure, the characteristics of the training user include: at least one of the attribute characteristics of the training user and the historical consumption characteristics of the training user;

[0275] The attribute characteristics of the training user include: at least one of the basic attribute characteristics of the training user and the preference characteristics of the training user; the historical consumption characteristics of the training user include at least one of the consumption ratio of the training user's consumption of resources of various genres in multiple historical time periods before the corresponding training resource consumption period and the completion rate of the resources of various genres;

[0276] Further optionally, in an embodiment of the present disclosure, the characteristics of the training resource include at least one of the length of the training resource, the identification of the training resource, the label of the training resource, the genre of the training resource, the entry genre of the training resource, and the effect characteristics of the training resource;

[0277] When the genre of the training resource is a video resource, the length of the training resource refers to the duration of the video; when the genre of the training resource is a graphic resource, the length of the training resource refers to the length of the text included in the resource and the number of pictures included;

[0278] The entry genre of the training resource refers to the genre of the entry resource of the training resource, which can identify the genre of the entry resource of the training resource as video or graphic;

[0279] The effect characteristics of the training resources include at least one of the playing time, likes, attention, forwarding, completion rate and delayed value of the training resources by the training users.

[0280] In the above embodiment, the implementation principle and technical effect of training the satisfaction prediction model by adopting the above module are the same as those of the above related method embodiments. For details, please refer to the records of the above related method embodiments, which will not be repeated here.

[0281] Fig.11 is a schematic diagram according to the tenth embodiment of the present disclosure; Fig.11 As shown, this embodiment provides a training device 1100 for a genre resource allocation model, including:

[0282] The collection module 1101 is used to collect a training data set based on the user's historical consumption information, wherein the training data set includes multiple training data, each of which includes training user characteristics, training resource candidate set characteristics, and training user behavior characteristics; and further includes: characteristics of multiple recommended resources recommended based on the training resource candidate set, consumption information of the multiple recommended resources, and real genre proportions of the multiple recommended resources;

[0283] A first training module 1102 is used to train a satisfaction evaluation model based on the training data set; the satisfaction evaluation model is used to predict the satisfaction of the training user with the resources recommended based on the training resource candidate set;

[0284] A screening module 1103 is used to screen multiple target training data with the highest satisfaction from the training data set based on the satisfaction evaluation model;

[0285] The second training module 1104 is used to train the genre resource allocation model using multiple target training data.

[0286] The training device 1100 of the genre resource allocation model of this embodiment, by adopting the above-mentioned modules, realizes the implementation principle and technical effect of the training of the genre resource allocation model, which is the same as the implementation of the above-mentioned related method embodiments. For details, please refer to the records of the above-mentioned related method embodiments, which will not be repeated here.

[0287] Further optionally, in one embodiment of the present disclosure, the first training module 1102 is used to:

[0288] For each of the training data in the training data set, based on the training user characteristics, the training resource candidate set characteristics, the training user behavior characteristics and the characteristics of the multiple recommended resources in the training data, the satisfaction evaluation model is used to predict the predicted satisfaction of the multiple recommended resources;

[0289] Based on the consumption information of the plurality of recommended resources and a preset satisfaction strategy, obtaining the real satisfaction of the plurality of resources;

[0290] Based on the predicted satisfaction levels and the actual satisfaction levels of the multiple recommended resources, parameters of the satisfaction estimation model are adjusted.

[0291] Further optionally, in one embodiment of the present disclosure, the second training module 1104 is used to:

[0292] For each of the target training data in the target training data set, based on the training user characteristics, the training resource candidate set characteristics and the training user behavior characteristics in the target training data, a genre resource allocation model is used to predict the predicted genre proportions of multiple recommended resources;

[0293] Based on the predicted genre proportions and the actual genre proportions of the multiple recommended resources, parameters of the genre resource allocation model are adjusted.

[0294] Further optionally, in an embodiment of the present disclosure, the training user characteristics include attribute characteristics of the training user; the attribute characteristics of the training user include: at least one of basic attribute characteristics of the training user and preference characteristics of the user.

[0295] Further optionally, in an embodiment of the present disclosure, the training resource candidate set includes feature expressions of multiple training resources; the feature expression of each training resource is obtained by performing vector expression and pooling operations based on at least one of the title of the training resource, the label of the training resource, the genre of the training resource, and the length of the training resource; and also includes the average feature expression of the training resource candidate set.

[0296] Further optionally, in an embodiment of the present disclosure, the training user behavior characteristics include the training user's historical consumption characteristics, the training user's satisfaction sequence characteristics and / or the training user's resource consumption scenario characteristics;

[0297] The historical consumption characteristics of the training user include at least one of the consumption proportion of resources of various genres consumed by the user in multiple historical time periods before the current period, the duration of consumption of resources of various genres, and the completion rate of resources of various genres;

[0298] The satisfaction sequence features of the training user include feature expressions of a plurality of resources that the user is most satisfied with within a preset time period before the current period.

[0299] The training device 1100 of the genre resource allocation model in the above embodiment, by adopting the above module to realize the implementation principle and technical effect of the training of the genre resource allocation model, is the same as the implementation of the above related method embodiment. For details, please refer to the records of the above related method embodiment, which will not be repeated here.

[0300] Fig.12 is a schematic diagram according to the eleventh embodiment of the present disclosure; Fig.12 As shown, this embodiment provides a resource proportion prediction device 1200, including:

[0301] An acquisition module 1201 is used to respectively acquire user characteristics, resource candidate set characteristics and user behavior characteristics;

[0302] The prediction module 1202 is used to predict the proportion of various genre resources in multiple resources recommended to the user based on the resource candidate set, using a pre-trained genre resource allocation model based on the user characteristics, the resource candidate set characteristics and the user behavior characteristics.

[0303] The resource ratio prediction device 1200 of this embodiment realizes the prediction of resource ratio by adopting the above-mentioned modules, and the implementation principle and technical effect are the same as those of the above-mentioned related method embodiments. For details, please refer to the records of the above-mentioned related method embodiments, which will not be repeated here.

[0304] Further optionally, in one embodiment of the present disclosure, the acquisition module 1201 is used to:

[0305] Obtain at least one of the user's attribute characteristics and the user's historical consumption characteristics; the user's attribute characteristics include: at least one of the user's basic attribute characteristics and the user's preference characteristics; the user's historical consumption characteristics include the user's consumption ratio of various genres of resources consumed in multiple historical time periods before the current period, and at least one of the completion rate of various genres of resources.

[0306] Further optionally, in one embodiment of the present disclosure, the acquisition module 1201 is used to:

[0307] Acquire feature expressions of multiple resources in the resource candidate set; the feature expressions of each resource are obtained by performing vector expression and pooling operations based on at least one of a resource title, a resource tag, a resource genre, and a resource length;

[0308] Based on the feature expressions of multiple resources in the resource candidate set, an average feature expression of the resource candidate set is obtained.

[0309] Further optionally, in one embodiment of the present disclosure, the acquisition module 1201 is used to:

[0310] Obtaining the user's historical consumption characteristics, the user's satisfaction sequence characteristics and / or the scenario characteristics of the user's consumption of resources;

[0311] The historical consumption characteristics of the user include at least one of the consumption proportion of resources of various genres consumed by the user in multiple historical time periods before the current time period, the duration of consumption of resources of various genres, and the completion rate of resources of various genres;

[0312] The user's satisfaction sequence characteristics include characteristic expressions of a plurality of resources that the user is most satisfied with within a preset time period before the current period.

[0313] The resource ratio prediction device 1200 of the above embodiment realizes the prediction of resource ratio by adopting the above module, and the implementation principle and technical effect are the same as those of the above related method embodiments. For details, please refer to the records of the above related method embodiments, which will not be repeated here.

[0314] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0315] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0316] Fig.13 A schematic block diagram of an example electronic device 1300 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0317] like Fig.13 As shown, the device 1300 includes a computing unit 1301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1302 or a computer program loaded from a storage unit 1308 into a random access memory (RAM) 1303. In the RAM 1303, various programs and data required for the operation of the device 1300 can also be stored. The computing unit 1301, the ROM 1302, and the RAM 1303 are connected to each other via a bus 1304. An input / output (I / O) interface 1305 is also connected to the bus 1304.

[0318] A number of components in the device 1300 are connected to the I / O interface 1305, including: an input unit 1306, such as a keyboard, a mouse, etc.; an output unit 1307, such as various types of displays, speakers, etc.; a storage unit 1308, such as a disk, an optical disk, etc.; and a communication unit 1309, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1309 allows the device 1300 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0319] The computing unit 1301 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1301 performs the various methods and processes described above, such as the above-mentioned methods of the present disclosure. For example, in some embodiments, the above-mentioned methods of the present disclosure may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 13011. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1300 via the ROM 1302 and / or the communication unit 1309. When the computer program is loaded into the RAM 1303 and executed by the computing unit 1301, one or more steps of the above-mentioned methods of the present disclosure described above may be performed. Alternatively, in other embodiments, the computing unit 1301 may be configured to execute the above method of the present disclosure in any other appropriate manner (for example, by means of firmware).

[0320] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0321] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0322] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0323] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0324] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0325] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0326] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0327] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A training method for a genre resource allocation model, comprising: Based on the user's historical consumption information, a training data set is collected, wherein the training data set includes a plurality of training data, each of which includes a training user feature, a training resource candidate set feature, and a training user behavior feature; It also includes: features of multiple recommended resources recommended based on the training resource candidate set, consumption information of the multiple recommended resources, and real genre proportions of the multiple recommended resources; Based on the training data set, a satisfaction evaluation model is trained; the satisfaction evaluation model is used to predict the satisfaction of the training user with the resources recommended based on the training resource candidate set; Based on the satisfaction evaluation model, selecting a plurality of target training data with the highest satisfaction from the training data set; Multiple target training data are used to train the genre resource allocation model.

2. The method according to claim 1, wherein: Based on the training data set, training a satisfaction evaluation model includes: For each of the training data in the training data set, based on the training user characteristics, the training resource candidate set characteristics, the training user behavior characteristics and the characteristics of the multiple recommended resources in the training data, the satisfaction evaluation model is used to predict the predicted satisfaction of the multiple recommended resources; Based on the consumption information of the plurality of recommended resources and the preset satisfaction behavior judgment condition, obtaining the real satisfaction of the plurality of resources; Based on the predicted satisfaction levels and the actual satisfaction levels of the multiple recommended resources, parameters of the satisfaction estimation model are adjusted.

3. The method according to claim 1, wherein: Multiple target training data are used to train the genre resource allocation model, including: For each of the target training data in the target training data set, based on the training user characteristics, the training resource candidate set characteristics and the training user behavior characteristics in the target training data, a genre resource allocation model is used to predict the predicted genre proportions of multiple recommended resources; Based on the predicted genre proportions and the actual genre proportions of the multiple recommended resources, parameters of the genre resource allocation model are adjusted.

4. The method according to any one of claims 1 to 3, wherein: The training user characteristics include attribute characteristics of the training user; The attribute characteristics of the training user include: at least one of the basic attribute characteristics of the training user and the user's preference characteristics.

5. The method according to any one of claims 1 to 3, wherein: The training resource candidate set features include feature expressions of multiple training resources; The feature expression of each training resource is obtained by performing vector expression and pooling operation based on at least one of the title of the training resource, the label of the training resource, the genre of the training resource, and the length of the training resource; and also includes the average feature expression of the candidate set of training resources.

6. The method according to any one of claims 1 to 3, wherein: The training user behavior characteristics include the training user's historical consumption characteristics, the training user's satisfaction sequence characteristics and / or the training user's resource consumption scenario characteristics; The historical consumption characteristics of the training user include at least one of the consumption ratio of the user's consumption of resources of various genres, the duration of consumption of resources of various genres, and the completion rate of resources of various genres in multiple historical time periods before the current period; The satisfaction sequence features of the training user include feature expressions of a plurality of resources that the user is most satisfied with within a preset time period before the current period.

7. A method for predicting resource proportion, comprising: Obtain user features, resource candidate set features, and user behavior features respectively; Based on the user characteristics, the resource candidate set characteristics and the user behavior characteristics, a pre-trained genre resource allocation model is used to predict the proportion of various genre resources in the multiple resources recommended to the user based on the resource candidate set; the genre resource allocation model adopts the genre resource allocation model trained according to any one of claims 1-6 above.

8. The method according to claim 7, wherein: Get user characteristics, including: Acquire the attribute characteristics of the user; the attribute characteristics of the user include at least one of the basic attribute characteristics of the user and the preference characteristics of the user.

9. The method according to claim 7, wherein: Get the resource candidate set features, including: Acquire feature expressions of multiple resources in the resource candidate set; the feature expressions of each resource are obtained by performing vector expression and pooling operations based on at least one of a resource title, a resource tag, a resource genre, and a resource length; Based on the feature expressions of multiple resources in the resource candidate set, an average feature expression of the resource candidate set is obtained.

10. The method according to claim 7, wherein: Obtain user behavior characteristics, including: Obtaining the user's historical consumption characteristics, the user's satisfaction sequence characteristics and / or the scenario characteristics of the user's consumption of resources; The historical consumption characteristics of the user include at least one of the consumption proportion of resources of various genres consumed by the user in multiple historical time periods before the current time period, the duration of consumption of resources of various genres, and the completion rate of resources of various genres; The user's satisfaction sequence characteristics include characteristic expressions of a plurality of resources that the user is most satisfied with within a preset time period before the current period.

11. A training device for a genre resource allocation model, comprising: A collection module, used to collect a training data set based on the user's historical consumption information, wherein the training data set includes a plurality of training data, each of which includes a training user feature, a training resource candidate set feature, and a training user behavior feature; It also includes: features of multiple recommended resources recommended based on the training resource candidate set, consumption information of the multiple recommended resources, and real genre proportions of the multiple recommended resources; A first training module is used to train a satisfaction evaluation model based on the training data set; the satisfaction evaluation model is used to predict the satisfaction of the training user with the resources recommended based on the training resource candidate set; A screening module, used for screening a plurality of target training data with the highest satisfaction from the training data set based on the satisfaction evaluation model; The second training module is used to train the genre resource allocation model using multiple target training data.

12. The device according to claim 11, wherein The first training module is used to: For each of the training data in the training data set, based on the training user characteristics, the training resource candidate set characteristics, the training user behavior characteristics and the characteristics of the multiple recommended resources in the training data, the satisfaction evaluation model is used to predict the predicted satisfaction of the multiple recommended resources; Based on the consumption information of the plurality of recommended resources and the preset satisfaction behavior judgment condition, obtaining the real satisfaction of the plurality of resources; Based on the predicted satisfaction levels and the actual satisfaction levels of the multiple recommended resources, parameters of the satisfaction estimation model are adjusted.

13. The device according to claim 11, wherein: The second training module is used to: For each of the target training data in the target training data set, based on the training user characteristics, the training resource candidate set characteristics and the training user behavior characteristics in the target training data, a genre resource allocation model is used to predict the predicted genre proportions of multiple recommended resources; Based on the predicted genre proportions and the actual genre proportions of the multiple recommended resources, parameters of the genre resource allocation model are adjusted.

14. The device according to any one of claims 11 to 13, wherein: The training user characteristics include attribute characteristics of the training user; The attribute characteristics of the training user include: at least one of the basic attribute characteristics of the training user and the user's preference characteristics.

15. The device according to any one of claims 11 to 13, wherein: The candidate set of training resources includes feature expressions of multiple training resources; The feature expression of each training resource is obtained by performing vector expression and pooling operation based on at least one of the title of the training resource, the label of the training resource, the genre of the training resource, and the length of the training resource; and also includes the average feature expression of the candidate set of training resources.

16. The device according to any one of claims 11 to 13, wherein: The training user behavior characteristics include the training user's historical consumption characteristics, the training user's satisfaction sequence characteristics and / or the training user's resource consumption scenario characteristics; The historical consumption characteristics of the training user include at least one of the consumption ratio of the user's consumption of resources of various genres, the duration of consumption of resources of various genres, and the completion rate of resources of various genres in multiple historical time periods before the current period; The satisfaction sequence features of the training user include feature expressions of a plurality of resources that the user is most satisfied with within a preset time period before the current period.

17. A resource proportion prediction device, comprising: An acquisition module, used to respectively acquire user features, resource candidate set features, and user behavior features; A prediction module is used to predict the proportion of various genre resources in multiple resources recommended to the user based on the resource candidate set based on the user characteristics, the resource candidate set characteristics and the user behavior characteristics, using a pre-trained genre resource allocation model; the genre resource allocation model uses the genre resource allocation model trained according to any one of claims 11-16 above.

18. The device according to claim 17, wherein: The acquisition module is used to: Obtain at least one of the user's attribute characteristics and the user's historical consumption characteristics; the user's attribute characteristics include: at least one of the user's basic attribute characteristics and the user's preference characteristics; the user's historical consumption characteristics include the user's consumption ratio of various genres of resources consumed in multiple historical time periods before the current period, and at least one of the completion rate of various genres of resources.

19. The device according to claim 17, wherein: The acquisition module is used to: Acquire feature expressions of multiple resources in the resource candidate set; the feature expressions of each resource are obtained by performing vector expression and pooling operations based on at least one of a resource title, a resource tag, a resource genre, and a resource length; Based on the feature expressions of multiple resources in the resource candidate set, an average feature expression of the resource candidate set is obtained.

20. The device according to claim 17, wherein: The acquisition module is used to: Obtaining the user's historical consumption characteristics, the user's satisfaction sequence characteristics and / or the scenario characteristics of the user's consumption of resources; The historical consumption characteristics of the user include at least one of the consumption proportion of resources of various genres consumed by the user in multiple historical time periods before the current time period, the duration of consumption of resources of various genres, and the completion rate of resources of various genres; The user's satisfaction sequence characteristics include characteristic expressions of a plurality of resources that the user is most satisfied with within a preset time period before the current period.

21. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6 or 7-10.

22. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-6 or 7-10.

23. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1-6 or 7-10.

Images