Content recommendation method, device and storage medium

Abstract

The present disclosure relates to a content recommendation method and its device and storage medium, and relates to the field of computer technology. The present disclosure at least solves the problem of low accuracy in recommending content to an account in the related art. The method includes: obtaining a predicted value of an account interaction parameter of each content to be played among multiple contents to be played, the account interaction parameter is used to characterize the number of times the account interacts with the content; according to the type of each content to be played, determining a target content set corresponding to each content to be played; the target content set includes multiple historical contents of the same type as each content to be played and sorted according to the historical values ​​of the account interaction parameter; according to the predicted value of the account interaction parameter of each content to be played, determining the distribution position of each content to be played in the target content set; according to the distribution position of each content to be played in the target content set, determining the recommendation order of each content to be played among multiple contents to be played.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a content recommendation method and a device and a storage medium thereof. Background Art

[0002] At present, publishing content in the client and recommending content to accounts has become an important means of corporate publicity and product promotion, and has been widely used in various industries. When there are multiple contents to be played, the recommendation order of the multiple contents to be played is usually determined according to the bid for publishing the contents to be played, that is, the higher the bid for a certain content to be played, the higher its recommendation order.

[0003] However, in the process of determining the recommendation order of the plurality of contents to be recommended, only the revenue issue is considered, resulting in that the determined recommendation order of the plurality of contents to be played may be inaccurate. Summary of the invention

[0004] The present disclosure provides a content recommendation method and device, and a storage medium, to at least solve the problem of low accuracy in recommending content to an account in the related art. The technical solution of the present disclosure is as follows:

[0005] According to a first aspect of an embodiment of the present disclosure, a content recommendation method is provided, comprising: obtaining a predicted value of an account interaction parameter of each content to be played among a plurality of content to be played, the account interaction parameter being used to characterize the number of times an account interacts with the content; determining, according to the type of each content to be played, a target content set corresponding to each content to be played; the target content set comprising a plurality of historical contents of the same type as the content to be played and sorted according to the historical values ​​of the account interaction parameter; determining, according to the predicted value of the account interaction parameter of each content to be played, a distribution position of each content to be played in the target content set; and determining, according to the distribution position of each content to be played in the target content set, a recommendation order of each content to be played among the plurality of contents to be played.

[0006] Optionally, the above-mentioned determining the recommended order of each content to be played among multiple contents to be played according to the distribution position of each content to be played in the target content set includes: determining the scaling factor of each content to be played according to the distribution position of each content to be played in the target content set; scaling the predicted value of the account interaction parameter of each content to be played according to the scaling factor of each content to be played to obtain the scaled predicted value of each content to be played; and determining the recommended order of each content to be played among the multiple contents to be played according to the scaled predicted value.

[0007] Optionally, the above-mentioned determining the scaling factor of each content to be played according to the distribution position of each content to be played in the target content set includes: when the distribution position of each content to be played in the target content set is before a first distribution position threshold, determining the scaling factor of each content to be played as a magnification factor; when the distribution position of each content to be played in the target content set is after a second distribution position threshold, determining the scaling factor of each content to be played as a reduction factor.

[0008] Optionally, the above-mentioned obtaining of the predicted value of the account interaction parameter of each content to be played among the multiple contents to be played includes: inputting the characteristic information of each content to be played and the characteristic information of the account to be pushed into a pre-trained click-through rate prediction model, and predicting the predicted click-through rate of each content to be played; inputting the characteristic information of each content to be played and the characteristic information of the account to be pushed into a pre-trained conversion rate prediction model, and predicting the predicted conversion rate of each content to be played; determining the predicted value of the account interaction parameter of each content to be played according to the predicted click-through rate of each content to be played and the predicted conversion rate of each content to be played.

[0009] Optionally, the above-mentioned determining the predicted value of the account interaction parameter of each content to be played based on the predicted click rate of each content to be played and the predicted conversion rate of each content to be played includes: determining the product of the predicted click rate of each content to be played and the predicted conversion rate of each content to be played as the predicted value of the account interaction parameter of each content to be played.

[0010] Optionally, the above method also includes: after recommending each content to be played to multiple accounts, obtaining each account's click operations and conversion operations on each content to be played, and performing statistics to obtain statistical results of account interaction parameters of each content to be played; based on the statistical results of the account interaction parameters of each content to be played, updating the target content set corresponding to each content to be played.

[0011] According to a second aspect of an embodiment of the present disclosure, a content recommendation device is provided, including: an acquisition unit and a determination unit; the acquisition unit is used to acquire a predicted value of an account interaction parameter of each to-be-played content among a plurality of to-be-played contents, the account interaction parameter being used to characterize the number of times an account interacts with the content; the determination unit is used to determine, according to the type of each to-be-played content, a target content set corresponding to each to-be-played content; the target content set includes a plurality of historical contents of the same type as each to-be-played content and sorted according to the historical values ​​of the account interaction parameter; the determination unit is further used to determine a distribution position of each to-be-played content in the target content set according to the predicted value of the account interaction parameter of each to-be-played content; the determination unit is further used to determine a recommendation order of each to-be-played content among a plurality of to-be-played contents according to the distribution position of each to-be-played content in the target content set.

[0012] Optionally, the above-mentioned determination unit is specifically used to: determine the scaling factor of each content to be played according to the distribution position of each content to be played in the target content set; scale the predicted value of the account interaction parameter of each content to be played according to the scaling factor of each content to be played to obtain the scaled predicted value of each content to be played; determine the recommended order of each content to be played among multiple contents to be played according to the scaled predicted value.

[0013] Optionally, the above-mentioned determination unit is specifically used to: when the distribution position of each content to be played in the target content set is before a first distribution position threshold, determine the scaling factor of each content to be played as a magnification factor; when the distribution position of each content to be played in the target content set is after a second distribution position threshold, determine the scaling factor of each content to be played as a reduction factor.

[0014] Optionally, the above-mentioned acquisition unit is specifically used to: input the characteristic information of each content to be played and the characteristic information of the account to be pushed into a pre-trained click-through rate prediction model to predict the predicted click-through rate of each content to be played; input the characteristic information of each content to be played and the characteristic information of the account to be pushed into a pre-trained conversion rate prediction model to predict the predicted conversion rate of each content to be played; determine the predicted value of the account interaction parameter of each content to be played according to the predicted click-through rate of each content to be played and the predicted conversion rate of each content to be played.

[0015] Optionally, the acquisition unit is specifically configured to: determine the product of the predicted click rate of each to-be-played content and the predicted conversion rate of each to-be-played content as the predicted value of the account interaction parameter of each to-be-played content.

[0016] Optionally, the above-mentioned device also includes an updating unit; an acquisition unit, which is also used to obtain the click operations and conversion operations of each account on each content to be played after recommending each content to be played to multiple accounts, and perform statistics to obtain the statistical results of the account interaction parameters of each content to be played; the updating unit is used to update the target content set corresponding to each content to be played according to the statistical results of the account interaction parameters of each content to be played.

[0017] According to a third aspect of an embodiment of the present disclosure, a content recommendation device is provided, comprising: a processor and a memory for storing processor executable instructions; wherein the processor is configured to execute instructions to implement the content recommendation method provided in the first aspect above.

[0018] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, comprising instructions, which, when executed by a processor, enable the processor to execute the content recommendation method provided in the first aspect above.

[0019] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising instructions, which, when executed by a processor, enable the processor to execute the content recommendation method provided in the first aspect above.

[0020] The technical solution provided by the present disclosure brings at least the following beneficial effects: when sorting the content to be played, the present disclosure determines the order of advertisement playback according to the distribution position of the content to be played in the target content set, because the distribution position of the content in the target content set can represent the ranking of the content in the same type of historical content, that is, it can represent the account's interest in the content compared to multiple historical contents of the same type. In this way, the recommendation order determined according to the distribution position of the content can improve the accuracy of content recommendation.

[0021] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0023] Figure 1 is a schematic diagram of the architecture of a content recommendation system according to an exemplary embodiment;

[0024] Figure 2 This is one of the flowchart diagrams of a content recommendation method according to an exemplary embodiment;

[0025] Figure 3 is a second flowchart of a content recommendation method according to an exemplary embodiment;

[0026] Figure 4 is a third flowchart of a content recommendation method according to an exemplary embodiment;

[0027] Figure 5 is a fourth flowchart of a content recommendation method according to an exemplary embodiment;

[0028] Figure 6 is a fifth flowchart of a content recommendation method according to an exemplary embodiment;

[0029] Figure 7 is a sixth flowchart of a content recommendation method according to an exemplary embodiment;

[0030] Figure 8 is one of the structural schematic diagrams of a content recommendation device according to an exemplary embodiment;

[0031] Fig. 9 This is a second structural diagram of a content recommendation device according to an exemplary embodiment. DETAILED DESCRIPTION

[0032] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0034] In addition, in the description of the embodiments of the present disclosure, unless otherwise specified, " / " means or, for example, A / B can mean A or B. "And / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present disclosure, "multiple" means two or more than two.

[0035] First, the technical terms involved in this disclosure are introduced:

[0036] Content: refers to advertisements, media information, or multimedia resources published on various clients or platforms (such as video websites, shopping websites).

[0037] Content publisher refers to the investor who invests in publishing content on various clients or platforms.

[0038] The click-through rate of content refers to the ratio of the number of times an account clicks on the content to the number of times the content is played. For example, if content A is played to an account 100 times on a certain website page, and the account clicks on content A 23 times, the click-through rate of content A is 23%.

[0039] The conversion rate of content refers to the ratio of the number of times an account clicks on a content page and uses the services of the content publisher to the number of times the account clicks on the content page. For example, if on a certain website page, shopping content B is played to an account 100 times in total, and the account clicks on shopping content B 20 times, after each click, some detailed information of the content promotion content and the order page will be displayed to the account. During the 20 times the order page is displayed, the account purchases the goods in shopping content B 5 times, then the conversion rate of shopping content B is 25%.

[0040] The account interaction parameter of the content may be the rate of return of the content, which is used to characterize the number of times the account interacts with the content. The rate of return refers to the ratio of the number of times the content is played that generates revenue to the total number of times the content is played. In actual application, the rate of return of the content may be specifically reflected in the click-through rate and / or conversion rate of the content, and the number of times the account interacts with the content may be the number of times the account browses, collects or purchases the items corresponding to the content based on the content playback.

[0041] The content recommendation method provided by the embodiment of the present disclosure is introduced below with reference to examples.

[0042] At present, in the process of pushing content to an account, such as when pushing content to an account through a video website, it is very important to choose a suitable content playback order. For example, in order to ensure the revenue of the content publisher, the content fee collected is usually used as an important criterion for arranging the content playback order.

[0043] Exemplarily, for n contents to be played, the n contents may be sorted according to the content fees corresponding to the n contents, and then the n contents may be played in sequence according to the sorting result.

[0044] The embodiments of the present disclosure take into account: On the one hand, when determining the order in which the content is played, not only the revenue of the content publisher should be considered, but also the user experience of the account. For example, in the content recommendation method that determines the order in which the content is played according to the content fee, since the content fee is irrelevant to the viewing experience of the account, although the content in the front position has a higher rate of return in the short term, it may cause the account to be lost due to the account's disgust caused by the played content. Therefore, when determining the order in which the content is played, the order in which the content is played can be determined based on the account's interest in the content. For example, the number of account interactions of the content can be estimated, and the estimated rate of return can be used to reflect the account's interest in the content, and then the order in which the content is played can be determined based on the estimated rate of return.

[0045] On the other hand, when determining the order of content playback based on the account's interest in the content, it is taken into account that the yields of different types of content have objective differences. Taking game content and real estate content as examples, since the yield of game content is naturally higher than that of real estate content, even if the account's yield on a certain real estate content (referred to as content x) is lower than the yield on a certain game content (referred to as content y), it cannot be said that the account is more disgusted with the content of content x. For example, there may be a situation where the account is highly interested in the creative content of content x, but less interested in the creative content of content y, but does not click on content x because there is no need to buy a house, but clicks on content y because he wants to play games.

[0046] Therefore, the disclosed embodiment provides a content recommendation method. Specifically, in a scenario where the order of content playback needs to be determined, taking one of the n contents to be played (hereinafter referred to as the first content) as an example, the method can determine the playback order of the first content among the n contents to be played based on the ranking of the yield of the first content among the same type of content. For example, if the n contents to be played include content a and content b of the same type, and content a ranks 10th in terms of yield among m contents of the same type, and content b ranks 5th in terms of yield among the above m contents, then when determining the playback order of the n contents, the playback position of content b can be appropriately moved forward, and the playback position of content a can be appropriately moved backward to obtain a better account usage experience.

[0047] like Figure 1 , which is a schematic diagram of the architecture of a content recommendation system provided by an embodiment of the present disclosure. The content recommendation system 10 may include: an offline data server 101, a database 102, an offline push server 103, an online sorting server 104, and an account device 105. The following takes the working process of each server in the content recommendation system 10 as an example to introduce the content recommendation method provided by the embodiment of the present disclosure.

[0048] It is understandable that the above-mentioned content recommendation system 10 only exemplarily shows the architecture of a content recommendation system that applies the method provided by the embodiment of the present disclosure. In actual applications, the method provided by the embodiment of the present disclosure can also be applied to content recommendation systems with other architectures. For example, in actual applications, a server can realize the functions of multiple devices in the offline data server 101, the database 102, the offline push server 103 and the online sorting server 104, thereby forming a system different from the content recommendation system 10, and the method provided by the embodiment of the present disclosure can also be applied to the system.

[0049] Specifically, the embodiment of the present disclosure takes the application of the content recommendation method to the above server as an example, and the content recommendation method provided by the embodiment of the present disclosure is described as follows: Figure 2 As shown, the method may include:

[0050] S201. The server obtains a predicted value of an account interaction parameter of each of a plurality of contents to be played.

[0051] Among them, the account interaction parameter is a parameter used to indicate the degree of attention paid to the content, which is determined according to the number of accounts interacting with the content, for example, it can be a click-through rate or a conversion rate, and the click-through rate can be the ratio of the number of times an account clicks on the content to the number of times the content is played. The conversion rate can be the ratio of the number of times an account clicks on the content page and uses the service of the content publisher to the number of times the account clicks on the content page. The content to be played can be, for example, an advertisement, media information, or a multimedia resource. The predicted value can be a value predicted in real time, for example, a value of an account interaction parameter, a click-through rate, or a conversion rate predicted in real time in response to an advertisement push request.

[0052] As a possible implementation method, taking the first content among multiple contents to be played as an example (the first content is any one of the multiple contents to be played), the server determines the predicted value of the account interaction parameter of the first content based on the characteristic information of the first content and the characteristic information of the account to be pushed.

[0053] In one implementation, the account interaction parameter of the content may be the click-through rate of the content. In another implementation, the account interaction parameter of the content may be the conversion rate of the content. In yet another implementation, the account interaction parameter of the content may be the product of the click-through rate and the conversion rate of the content.

[0054] The following mainly takes the product of the click-through rate and conversion rate of the content as the account interaction parameter of the content as an example to introduce the method provided by the embodiment of the present disclosure. It can be understood that when other parameters are used to represent the account interaction parameters of the content, the method provided by the embodiment of the present disclosure is also applicable.

[0055] Specifically, the server predicts the predicted click rate of the first content based on the feature information of the first content, the feature information of the account to be pushed, and a preset click rate prediction model. At the same time, the server also predicts the predicted conversion rate of the first content based on the feature information of the first content, the feature information of the account to be pushed, and a preset conversion rate prediction model. Further, the server determines the product of the predicted click rate and the predicted conversion rate of the first content as the predicted value of the account interaction parameter of the first content.

[0056] S202: The server determines target content sets corresponding to the respective contents to be played according to the types of the contents to be played.

[0057] The target content set includes multiple historical contents of the same type as the contents to be played and sorted according to the historical values ​​of the account interaction parameters. Among the multiple contents to be played, there are different types of contents to be played. For each content to be played, the target content set of the corresponding type is obtained.

[0058] As a possible implementation manner, the server obtains the type of the first content, and obtains a target content set corresponding to the type of the first content from a storage of the server according to the type of the first content.

[0059] It should be noted that the memory of the server stores a plurality of content sets, each of which corresponds to a type of content. The target content set corresponding to the type of the first content includes a plurality of historical contents of the same type as the first content.

[0060] Exemplarily, if the first content is game content, then the multiple historical contents in the target content set are also game content.

[0061] In actual application, each content set in the server is stored in the form of a key-value pair list, wherein the key in the key-value pair can be used to record the identification of the historical content, and the value can be used to record the historical value of the account interaction parameter corresponding to each historical content.

[0062] In some embodiments, in order to store the target content set corresponding to each to-be-played content in the server, the server may pre-classify the played historical content according to the content type. For multiple historical contents of the same type, the server determines the historical click rate and historical conversion rate of each historical content. The server then calculates the product of the historical click rate and historical conversion rate of each historical content. The server then uses the product of the historical click rate and historical conversion rate of each historical content as the historical value of the account interaction parameter of the historical content.

[0063] Furthermore, the server sorts the historical values ​​of the account interaction parameters of each historical content in order from large to small or from small to large, and obtains a target content set corresponding to the content of this type.

[0064] S203: The server determines the distribution position of each to-be-played content in the target content set according to the predicted value of the account interaction parameter of each to-be-played content.

[0065] Specifically, the distribution position of the first content in the target content set may be reflected as the quantile of the first content in the target content set.

[0066] As a possible implementation method, the server compares the predicted value of the account interaction parameter of the first content with the historical value of the account interaction parameter of each historical content in the target content set in turn, and determines the target order of the predicted value of the account interaction parameter of the first content in the target content set. Further, the server calculates the ratio of the target order to the number of multiple historical contents in the target content set, obtains the order ratio of the first content in the target content set, and then determines the quantile of the order ratio of the first content as the distribution position of the first content in the target content set.

[0067] For example, if the number of target content sets is 10 (i.e., there are 10 historical contents of the same type as the first content in the target content set), and the first content is determined to be ranked between the 4th and 5th contents among the 10 historical contents by sorting according to the account interaction parameter, then the quantile of the first content in the first content set is determined to be 0.4.

[0068] S204: The server determines a recommendation order of each to-be-played content among multiple to-be-played contents according to the distribution position of each to-be-played content in the target content set.

[0069] Subsequently, after determining the recommendation order of the multiple contents to be played, the server sends the recommendation order of the multiple contents to be played to the account device where the account to be recommended is located, so that the account device recommends the multiple contents to be played to the account to be recommended according to the recommendation order.

[0070] For example, take three contents to be played (Content 1, Content 2, Content 3) as an example, Content 1 is game content, Content 2 is real estate content, and Content 3 is entertainment content. After the predicted values ​​of the account interaction parameters of Content 1, Content 2, and Content 3 are predicted to be 0.77, 0.69, and 0.58 respectively, it can be understood that the predicted account is more interested in Game Content 1 than Real Estate Content 2, and is more interested in Real Estate Content 2 than Entertainment Content 3, and the ranking result can be [Content 1, Content 2, Content 3].

[0071] In this case, the server determines the target content set corresponding to content 1 (hereinafter referred to as the first target content set, the first target content set includes historical content of multiple game categories), and determines that the distribution position of the predicted value of the account interaction parameter of content 1 in the first target content set is [0.3, 0.4] (also the distribution position of content 1 in the first target content set). When the server determines that the distribution position of content 1 is before the preset distribution position threshold of 0.5, it determines that the scaling factor of content 1 is the magnification factor, and determines that the magnification factor is 1.1 (the scaling factor of each content is used to characterize the relative degree of preference of the account for the content. If it is a magnification factor, it represents that the account likes the content more, and as the magnification factor increases, the account likes the content more. At the same time, if the scaling factor of a content is a reduction factor, it represents that the account likes the content less, and as the reduction factor decreases, the account likes the content less). Similarly, the server determines that the scaling factor of content 2 is 0.8 (reduction factor) and the scaling factor of content 3 is 1.2.

[0072] Furthermore, the server scales the predicted value of the account interaction parameter of each content to be played according to the scaling factor of the content to be played, and obtains the scaled predicted value of the content to be played. The scaled predicted value of content 1 is 0.77*1.1=0.847, the scaled predicted value of content 2 is 0.69*0.8=0.552, and the scaled predicted value of content 3 is 0.25*1.2=0.696.

[0073] Finally, the server sorts the multiple contents to be played according to the predicted values ​​after scaling, and obtains the recommended order of the multiple contents to be played. For the above content 1, content 2 and content 3, the recommended order sorted by the server according to the predicted values ​​after scaling is [content 1, content 3, content 2].

[0074] The specific implementation of this step can refer to the specific description of the subsequent embodiments of the present disclosure, and will not be repeated here.

[0075] The technical solution provided by the present disclosure brings at least the following beneficial effects: when sorting the content to be played, the present disclosure determines the order of advertisement playback according to the distribution position of the content to be played in the target content set, because the distribution position of the content in the target content set can represent the ranking of the content in the same type of historical content, that is, it can represent the account's interest in the content compared to multiple historical contents of the same type. In this way, the recommendation order determined according to the distribution position of the content can improve the accuracy of content recommendation.

[0076] In one design, the server may first determine a preset sorting result of multiple contents to be played, and then use the distribution position of the first content in the target content set to enlarge or reduce the order of the first content in the preset sorting result to determine the recommended order of the first content in the multiple contents to be played. Figure 3 As shown, S204 may include the following S2041-S2043:

[0077] S2041: The server determines a scaling factor of each to-be-played content according to a distribution position of each to-be-played content in a target content set.

[0078] As a possible implementation, when the distribution position of each to-be-played content in the target content set is before the first distribution position threshold, the server determines the scaling factor of each to-be-played content to be a magnification factor. When the distribution position of each to-be-played content in the target content set is after the second distribution position threshold, the server determines the scaling factor of each to-be-played content to be a reduction factor.

[0079] It should be noted that the first distribution position threshold and the second distribution position threshold can be set in advance by the operation and maintenance personnel in the server. The first distribution position threshold is greater than or equal to the second distribution position threshold. The magnification factor is a natural number greater than 1, and the reduction factor is a natural number greater than 0 and less than 1.

[0080] For example, when the distribution position is a quantile, the quartile method can be used to divide the quantile of the first content in the target content set into four intervals, such as (0, 0.25], (0.25, 0.5], (0.5, 0.75] and (0.75, 1]. Among them, the four intervals (0, 0.25], (0.25, 0.5], (0.5, 0.75] and (0.75, 1] ​​respectively represent that the account interaction parameter of the first content ranks in the top 25% of the target content set, the account interaction parameter of the first content ranks in the top 25%, and the account interaction parameter of the first content ranks in the top 25%. The first content's account interaction parameter ranks between the top 25%-50% in the target content set, the first content's account interaction parameter ranks between the top 50%-75% in the target content set, and the first content's account interaction parameter ranks between the bottom 25% in the target content set. The first distribution position threshold and the second distribution position threshold can both be 0.5. When the first content's distribution position is within the intervals of (0,0.25] and (0.25,0.5], and the first content's distribution position is before the first division position threshold, then the scaling of the first content The coefficient is the magnification coefficient. When the distribution position of the first content is in the intervals of (0, 0.25] and (0.25, 0.5], the corresponding magnification coefficients are 1.2 and 1.1 respectively, that is, when the scaling coefficient of the distribution position of the first content is the magnification coefficient, the closer the distribution position of the first content is, the larger the magnification coefficient of the first content is, and the magnification coefficient of the first content is positively correlated with the order of the distribution position of the first content. On the other hand, when the distribution position of the first content is in the intervals of (0.5, 0.75] and (0.75, 1], the distribution position of the first content is after the second distribution position threshold, and the scaling coefficient of the first content is the reduction coefficient. When the distribution position of the first content is in the intervals of (0.5, 0.75] and (0.75, 1], the corresponding magnification coefficients are 0.9 and 0.8 respectively, that is, when the scaling coefficient of the distribution position of the first content is the reduction coefficient, the closer the distribution position of the first content is, the smaller the reduction coefficient of the first content is, and the reduction coefficient of the first content is positively correlated with the order of the distribution position of the first content.

[0081] The technical solution provided by the present disclosure brings at least the following beneficial effects: by setting the magnification factor and the reduction factor, the determined zoom factor can be made more accurate, thereby ensuring that the zoom factor can accurately represent the account's true preference for the played content.

[0082] S2042: The server obtains preset sorting results of multiple contents to be played.

[0083] In one implementation, the server may sort the multiple contents to be played according to the predicted values ​​of the content fees and account interaction parameters respectively corresponding to the multiple contents to be played, and obtain a preset sorting result.

[0084] Specifically, in the process of sorting the multiple contents to be played according to the content fees and account interaction parameters respectively corresponding to each of the multiple contents to be played, the sorting parameters corresponding to each of the multiple contents to be played can be calculated respectively according to the content fees and account interaction parameters of each of the multiple contents to be played. Then, the multiple contents to be played are sorted according to the size of the sorting parameters corresponding to each of the contents to obtain a preset sorting result.

[0085] For example, the product of the content fee of each of the multiple contents to be played and the account interaction parameter (i.e., content fee×account interaction parameter, hereinafter referred to as the first sorting parameter) can be calculated, and then the multiple contents to be played can be sorted according to the first sorting parameters corresponding to each of the multiple contents to be played to obtain a preset sorting result.

[0086] For another example, the content fee and account interaction parameter of each of the multiple contents to be played may be weightedly summed to obtain a second sorting parameter for each content, and then the multiple contents to be played may be sorted according to the second sorting parameters corresponding to each of the multiple contents to be played to obtain a preset sorting result.

[0087] It should be noted that, in actual application, other methods may be used to sort the multiple contents to be played to obtain a preset sorting result. For example, the multiple contents to be played may be sorted only according to the content fees corresponding to the multiple contents to be played to obtain a preset sorting result. The present disclosure embodiment may not limit the sorting method of the preset sorting result.

[0088] In actual applications, S2041 may be executed first and then S2042, or S2042 may be executed first and then S2041, or S2041 and S2042 may be executed simultaneously.

[0089] S2043: The server adjusts the order of each to-be-played content in the preset sorting result according to the scaling factor of each to-be-played content, and obtains a recommended order of each to-be-played content among the multiple to-be-played contents.

[0090] As a possible implementation manner, the server modifies the first sorting parameter of the first content in the preset sorting result (for example, the first sorting parameter may be content fee×account interaction parameter) according to the scaling factor of the first content to obtain a modified sorting parameter corresponding to the first content.

[0091] For example, the server may multiply the scaling factor corresponding to the first content by the ranking parameter of the first content in the preset ranking result to obtain a modified ranking parameter corresponding to the first content.

[0092] Furthermore, the server modifies the order of the first content in the preset sorting result according to the modified sorting parameter corresponding to the first content, and determines the recommendation order of the first content.

[0093] For example, before the correction, according to the first ranking parameter of the first content in the preset ranking result, it is determined that the first content ranks fifth among the multiple contents to be played. After the correction, if the corrected ranking parameter corresponding to the first content is greater than the ranking parameter of the first content in the preset ranking result, the order of the first content will move forward to the third or second place, etc.

[0094] It is understandable that in actual applications, for each of the multiple contents to be played, the distribution position of each content in the corresponding content set can be determined according to the above-mentioned S201-S204 method, and the preset sorting results can be corrected according to the distribution position of each content in the corresponding content set, and the playback order of each content in the multiple contents to be played can be determined according to the correction result.

[0095] In another design, the server may also use the distribution position of the first content in the target content set to enlarge or reduce the predicted value of the account interaction parameter of the first content, and determine the ranking result of the first content among the multiple contents to be played according to the enlarged or reduced predicted value. Figure 4 As shown, S204 may also include the following S2044-S2046:

[0096] S2044: The server determines a scaling factor of each to-be-played content according to a distribution position of each to-be-played content in the target content set.

[0097] The specific implementation method of this step can refer to the specific description of S2041 above in the embodiment of the present disclosure, and will not be repeated here.

[0098] S2045: The server scales the predicted value of the account interaction parameter of each content to be played according to the scaling factor of each content to be played, to obtain the scaled predicted value of each content to be played.

[0099] As a possible implementation manner, the server scales the predicted value of the account interaction parameter of the first content according to the scaling factor of the first content to obtain the scaled predicted value of the first content.

[0100] For example, if the predicted value of the account interaction parameter of the first content is 0.77, the scaling factor of the first content is the magnification factor and the magnification factor is 1.1, the server calculates the product of the predicted value of the account interaction parameter of the first content 0.77 and the scaling factor of the first content 1.1 as 0.847, and determines the determined product 0.847 as the predicted value of the first content after scaling.

[0101] S2046: The server determines a recommended order of each to-be-played content among multiple to-be-played contents according to the scaled prediction value.

[0102] As a possible implementation manner, the server sorts the multiple contents to be played according to the scaled predicted value of each content to be played, and obtains a recommended order of the multiple contents to be played.

[0103] As another possible implementation, the server calculates the product of the scaled predicted value of each to-be-played content and the content fee corresponding to each to-be-played content, and sorts the content according to the calculated product to obtain a recommended order of the multiple to-be-played contents.

[0104] The technical solution provided by the present disclosure brings at least the following beneficial effects: by determining the scaling factor of each content to be played, and adjusting the predicted value of the account interaction parameter of the content to be played according to the scaling factor, the final recommendation order is obtained. Since the scaling factor can represent the relative preference of the account for the played content, the recommendation order obtained based on the predicted value after the scaling factor is adjusted will be more accurate.

[0105] In one design, in order to obtain the predicted value of the account interaction parameter of each content to be played, such as Figure 5 As shown, S201 provided in the embodiment of the present disclosure specifically includes the following S2011-S2013.

[0106] S2011. The server inputs the characteristic information of each content to be played and the characteristic information of the account to be pushed into a pre-trained click rate prediction model to predict the click rate of each content to be played.

[0107] The characteristic information of the content may include: one or more of the duration of the content, the visual style of the content (such as whether the content is animation or real scene), and the story elements of the content (such as whether the content involves story elements such as love, family, friendship, science fiction, fantasy or war). The characteristic information of the account may include: one or more of the account's age, gender, occupation, and hobbies.

[0108] It should be noted that the click rate prediction model can be obtained by the server performing supervised training on the neural network model based on the click rate data. In the click rate data, the characteristic information of the sample content is used as the sample feature, and the historical click rate of the sample content is used as the label.

[0109] S2012: The server inputs the characteristic information of each content to be played and the characteristic information of the account to be pushed into a pre-trained conversion rate prediction model to predict the conversion rate of each content to be played.

[0110] It should be noted that the conversion rate prediction model can be obtained by the server through supervised training of a neural network model based on conversion rate data. In the conversion rate data, feature information of sample content is used as sample feature, and historical conversion rate of sample content is used as label.

[0111] In actual application, S2011 may be executed first and then S2012, or S2012 may be executed first and then S2011, or S2011 and S2012 may be executed simultaneously.

[0112] S2013: The server determines a predicted value of an account interaction parameter of each content to be played according to the predicted click rate of each content to be played and the predicted conversion rate of each content to be played.

[0113] As a possible implementation manner, the server determines the product of the predicted click rate of each to-be-played content and the predicted conversion rate of each to-be-played content as the predicted value of the account interaction parameter of each to-be-played content.

[0114] The technical solution provided by the present disclosure brings at least the following beneficial effects: the product of the predicted click-through rate and the predicted conversion rate of the content to be played is used as the predicted value of the account interaction parameter, so that the predicted value of the account interaction parameter is positively correlated with the predicted click-through rate and the predicted conversion rate, respectively, which can more fully characterize the account's actual preference for the played content and make the predicted value of the account interaction parameter more accurate.

[0115] Specifically, for the first content, the server calculates the product of the predicted click rate of the first content and the predicted conversion rate of the first content as the predicted value of the account interaction parameter of the first content.

[0116] The technical solution provided by the present disclosure brings at least the following beneficial effects: based on the preset click-through rate model and conversion rate model, the click-through rate and conversion rate are predicted respectively, and the product of the predicted click-through rate and conversion rate is used as the predicted value of the account interaction parameter. On the one hand, the neural network can be used to learn the relationship between account characteristics, content characteristics and account interaction behaviors, so that the predicted values ​​of click-through rate and conversion rate are more accurate, so as to achieve the accuracy of the predicted values ​​of account interaction parameters. On the other hand, in the process of calculating the predicted values ​​of account interaction parameters, the influence of content fees on account interaction parameters in related technologies is discarded, so that the predicted account interaction parameters are more accurate.

[0117] In one design, in order to ensure the validity and real-time nature of the data in each content set, such as Figure 6 As shown, the content recommendation method provided by the embodiment of the present disclosure further includes the following S301-S302.

[0118] S301. After recommending each content to be played to multiple accounts, the server obtains click operations and conversion operations of each account on each content to be played, and performs statistics to obtain statistical results of account interaction parameters of each content to be played.

[0119] As a possible implementation, the server obtains click operations and conversion operations of all accounts on the first content, and calculates the actual click rate and actual conversion rate of the first content. Further, the server calculates the product of the actual click rate and the actual conversion rate of the first content, and uses the calculated product as the statistical result of the account interaction parameter of the first content.

[0120] It can be understood that the statistical result of the account interaction parameter of the first content is the historical value of the account interaction parameter of the first content.

[0121] S302: The server updates a target content set corresponding to each to-be-played content according to the statistical result of the account interaction parameter of each to-be-played content.

[0122] As a possible implementation manner, the server updates the statistical result of the account interaction parameter of the first content to the target content set corresponding to the type of the first content.

[0123] It should be noted that if the first content already exists in the target content set, the server updates the historical value of the interaction parameter of the first content in the target content set based on the statistical result of the account interaction parameter of the first content. If the first content does not exist in the target content set, the server updates the first content and the historical value of the account interaction parameter of the first content to the target content set.

[0124] The technical solution provided by the present disclosure brings at least the following beneficial effects: by updating the target content set in real time according to the interactive behavior of the account, it can be ensured that the distribution position of the content to be played each time in the corresponding target content set is real-time and accurate, thereby ensuring the accuracy of the subsequent zoom factor and the recommended order.

[0125] In another embodiment, when the content recommendation method provided in the embodiment of the present disclosure is applied to the offline data server 101, the database 102, the offline push server 103, the online advertisement ranking server 104 and the account device 105 in the content recommendation system, as shown in FIG. Figure 7 As shown, the content recommendation method provided by the embodiment of the present disclosure also includes the following S401-S408.

[0126] S401 . The offline data server 101 obtains historical values ​​of account interaction parameters of a plurality of historical contents of the same type as the first content.

[0127] The specific implementation method of this step can refer to the description of the server obtaining the historical value of the account interaction parameter of each historical content in S202 in the embodiment of the present disclosure, which will not be repeated here.

[0128] S402 : The offline data server 101 sorts the multiple historical contents according to the historical values ​​of the account interaction parameters of the multiple historical contents to obtain a content set (hereinafter referred to as a first content set).

[0129] The specific implementation of this step can refer to the description of the target content set corresponding to each type of content obtained by the server in S202 in the embodiment of the present disclosure, which will not be described in detail here.

[0130] S403 : The offline data server 101 writes the first content set into the database 102 .

[0131] The specific implementation of this step may refer to the description of the server storing each content set in S202 in the embodiment of the present disclosure, which will not be described in detail here.

[0132] S404 , the offline push server 103 accesses the database 102 to obtain a first content set, and pushes the obtained first content set to the online sorting server 104 .

[0133] S405 . The online ranking server 104 obtains a predicted value of an account interaction parameter of the first content.

[0134] The specific implementation method of this step can refer to the description in S201 and S2011-S2013 in the embodiments of the present disclosure, and will not be repeated here.

[0135] S406 : The online ranking server 104 determines a distribution position of the first content in the first content set according to the predicted value of the account interaction parameter of the first content.

[0136] The specific implementation method of this step can refer to the description of S203 in the embodiment of the present disclosure, and will not be repeated here.

[0137] S407: The online sorting server 104 determines a recommendation order of the first content among the multiple contents to be played according to the distribution position of the first content in the first content set.

[0138] The specific implementation method of this step can refer to the description of S204 and S2041-S2046 in the embodiments of the present disclosure, and will not be repeated here.

[0139] S408 . The online sorting server 104 sends the recommended order of the plurality of contents to be played to the account device 105 , so that the account device 105 plays the plurality of contents to be played according to the recommended order.

[0140] The technical solution provided by the present disclosure brings at least the following beneficial effects: the action of generating a content set and the action of sorting are respectively set in different devices, which can reduce the computing resources of the online sorting server in response to content sorting requests of multiple account devices to a certain extent. At the same time, by writing the first content set into the database 102, the online sorting server or other devices can obtain the first content set by accessing the database 102, thereby avoiding the online sorting server or other devices from frequently accessing the offline data server 101, and reducing the load of the offline data server 101.

[0141] In addition, the present disclosure also provides a content recommendation device, which can be used to implement the above Figure 1 The functions of the online ranking server 104 in the content recommendation system 10 shown in the figure can also realize the functions of the servers involved in the above-mentioned embodiments. In some implementations, the content recommendation device can also be used to realize the functions of one or more devices in the offline data server 101, the database 102 and the offline push server 103. Figure 8 As shown, the content recommendation device 50 includes: an acquisition unit 501 and a determination unit 502.

[0142] The acquisition unit 501 is used to acquire a predicted value of an account interaction parameter of each of a plurality of contents to be played, where the account interaction parameter is used to represent the number of times an account interacts with the contents.

[0143] The determination unit 502 is used to determine the target content set corresponding to each to-be-played content according to the type of each to-be-played content. The target content set includes a plurality of historical contents of the same type as each to-be-played content and sorted according to the historical values ​​of the account interaction parameters.

[0144] The determination unit 502 is further configured to determine the distribution position of each to-be-played content in the target content set according to the predicted value of the account interaction parameter of each to-be-played content.

[0145] The determination unit 502 is further configured to determine a recommendation order of each to-be-played content among the multiple to-be-played contents according to the distribution position of each to-be-played content in the target content set.

[0146] Optional, such as Figure 8 As shown, the determination unit 502 provided in the embodiment of the present disclosure is specifically used for:

[0147] The scaling factor of each to-be-played content is determined according to the distribution position of each to-be-played content in the target content set.

[0148] The predicted value of the account interaction parameter of each content to be played is scaled according to the scaling factor of each content to be played to obtain the scaled predicted value of each content to be played.

[0149] The recommended order of each to-be-played content among the multiple to-be-played contents is determined according to the scaled predicted value.

[0150] Optional, such as Figure 8 As shown, the determination unit 502 provided in the embodiment of the present disclosure is specifically used for:

[0151] When the distribution position of each to-be-played content in the target content set is before the first distribution position threshold, the scaling factor of each to-be-played content is determined to be the magnification factor.

[0152] When the distribution position of each to-be-played content in the target content set is after the second distribution position threshold, the scaling factor of each to-be-played content is determined to be a reduction factor.

[0153] Optional, such as Figure 8 As shown, the acquisition unit 501 provided in the embodiment of the present disclosure is specifically used for:

[0154] The characteristic information of each content to be played and the characteristic information of the account to be pushed are input into a pre-trained click rate prediction model to predict the click rate of each content to be played.

[0155] The characteristic information of each content to be played and the characteristic information of the account to be pushed are input into a pre-trained conversion rate prediction model to predict the conversion rate of each content to be played.

[0156] According to the predicted click rate of each to-be-played content and the predicted conversion rate of each to-be-played content, the predicted value of the account interaction parameter of each to-be-played content is determined.

[0157] Optional, such as Figure 8 As shown, the acquisition unit 501 provided in the embodiment of the present disclosure is specifically used for:

[0158] The product of the predicted click rate of each to-be-played content and the predicted conversion rate of each to-be-played content is determined as the predicted value of the account interaction parameter of each to-be-played content.

[0159] Optional, such as Figure 8 As shown, the device 50 provided in the embodiment of the present disclosure further includes an updating unit 503.

[0160] The acquisition unit 501 is further used to acquire, after recommending each content to be played to multiple accounts, click operations and conversion operations of each account on each content to be played, and perform statistics to obtain statistical results of account interaction parameters of each content to be played.

[0161] The updating unit 503 is used to update the target content set corresponding to each to-be-played content according to the statistical result of the account interaction parameter of each to-be-played content.

[0162] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0163] Fig. 9 is a schematic diagram of the structure of another content recommendation device provided by the present disclosure. Fig. 9 The content recommendation device 60 may include at least one processor 601 and a memory 603 for storing processor executable instructions. The processor 601 is configured to execute instructions in the memory 603 to implement part or all of the content in the content recommendation method in the above embodiment.

[0164] For example, in Figure 1 In the implementation environment shown, the content recommendation device 60 can Figure 1 An online ranking server 104 is shown.

[0165] In addition, the content recommendation device 60 may further include a communication bus 602 and at least one communication interface 604 .

[0166] The processor 601 may be a central processing unit (CPU), a microprocessing unit, an ASIC, or one or more integrated circuits for controlling the execution of the program of the disclosed solution.

[0167] Communication bus 602 may include a pathway for transmitting information between the above-mentioned components.

[0168] The communication interface 604 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0169] The memory 603 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory may exist independently and be connected to the processing unit through a bus. The memory may also be integrated with the processing unit.

[0170] The memory 603 is used to store instructions for executing the solution of the present disclosure, and the execution is controlled by the processor 601. The processor 601 is used to execute the instructions stored in the memory 603, so as to realize the functions in the method of the present disclosure.

[0171] In a specific implementation, as an embodiment, the processor 601 may include one or more CPUs, such as Fig. 9 CPU0 and CPU1 in.

[0172] In a specific implementation, as an embodiment, the content recommendation device 60 may include multiple processors, such as Fig. 9 601 and processor 607 in FIG. Each of these processors may be a single-CPU processor or a multi-CPU processor. A processor herein may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0173] In a specific implementation, as an embodiment, the content recommendation device 60 may further include an output device 605 and an input device 606. The output device 605 communicates with the processor 601 and may display information in a variety of ways. For example, the output device 605 may be a liquid crystal display (LCD), a light emitting diode (LED) display device, a cathode ray tube (CRT) display device, or a projector. The input device 606 communicates with the processor 601 and may receive input from an account in a variety of ways. For example, the input device 606 may be a mouse, a keyboard, a touch screen device, or a sensor device.

[0174] Those skilled in the art will understand that Fig. 9 The structure shown in the figure does not constitute a limitation on the content recommendation device 60, and may include more or less components than those shown in the figure, or combine some components, or adopt a different component arrangement.

[0175] In addition, the present disclosure also provides a computer-readable storage medium, including instructions, which, when executed by a processor, enable the processor to execute the content recommendation method provided in the above embodiment.

[0176] In addition, the present disclosure also provides a computer program product, including instructions, which, when executed by a processor, enable the processor to execute the content recommendation method provided in the above embodiment.

[0177] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims of the present disclosure.

Claims

1. A content recommendation method, characterized in that: include: Obtaining a predicted value of an account interaction parameter of each of the plurality of contents to be played, wherein the account interaction parameter is used to represent the number of times an account interacts with the contents; Determining, according to the type of each of the to-be-played contents, a target content set corresponding to each of the to-be-played contents; the target content set includes a plurality of historical contents of the same type as each of the to-be-played contents and sorted according to the historical values ​​of the account interaction parameters; Determining, according to the predicted value of the account interaction parameter of each of the to-be-played contents, a distribution position of each of the to-be-played contents in the target content set; Determining a scaling factor of each of the to-be-played contents according to a distribution position of each of the to-be-played contents in the target content set; Scaling the predicted value of the account interaction parameter of each of the contents to be played according to the scaling coefficient of each of the contents to be played, to obtain the scaled predicted value of each of the contents to be played; A recommended order of each of the to-be-played contents among the multiple to-be-played contents is determined according to the scaled predicted value.

2. The method according to claim 1, characterized in that The step of determining the scaling factor of each of the to-be-played contents according to the distribution position of each of the to-be-played contents in the target content set includes: When the distribution position of each of the to-be-played contents in the target content set is before a first distribution position threshold, determining the scaling factor of each of the to-be-played contents as a magnification factor; When the distribution position of each of the to-be-played contents in the target content set is after the second distribution position threshold, the scaling factor of each of the to-be-played contents is determined to be a reduction factor.

3. The method according to claim 1 or 2, characterized in that: The obtaining of the predicted value of the account interaction parameter of each of the plurality of contents to be played includes: Inputting the characteristic information of each of the to-be-played contents and the characteristic information of the account to be pushed into a pre-trained click rate prediction model to predict the click rate of each of the to-be-played contents; Inputting the characteristic information of each of the to-be-played contents and the characteristic information of the to-be-pushed account into a pre-trained conversion rate prediction model to predict the conversion rate of each of the to-be-played contents; According to the predicted click rate of each of the to-be-played contents and the predicted conversion rate of each of the to-be-played contents, a predicted value of the account interaction parameter of each of the to-be-played contents is determined.

4. The method according to claim 3, characterized in that Determining the predicted value of the account interaction parameter of each of the to-be-played contents according to the predicted click rate of each of the to-be-played contents and the predicted conversion rate of each of the to-be-played contents includes: The product of the predicted click rate of each of the to-be-played contents and the predicted conversion rate of each of the to-be-played contents is determined as the predicted value of the account interaction parameter of each of the to-be-played contents.

5. The method according to claim 1 or 2, characterized in that: The method further comprises: After recommending each of the to-be-played contents to multiple accounts, obtaining click operations and conversion operations of each of the accounts on each of the to-be-played contents, and performing statistics to obtain statistical results of the account interaction parameters of each of the to-be-played contents; The target content set corresponding to each of the to-be-played contents is updated according to the statistical result of the account interaction parameter of each of the to-be-played contents.

6. A content recommendation device, characterized in that: include: Acquiring units and determining units; The acquisition unit is used to acquire a predicted value of an account interaction parameter of each of the plurality of contents to be played, wherein the account interaction parameter is used to represent the number of times an account interacts with the content; The determining unit is used to determine, according to the type of each of the to-be-played contents, a target content set corresponding to each of the to-be-played contents; the target content set includes a plurality of historical contents of the same type as each of the to-be-played contents and sorted according to the historical values ​​of the account interaction parameters; The determining unit is further configured to determine a distribution position of each of the to-be-played contents in the target content set according to a predicted value of the account interaction parameter of each of the to-be-played contents; The determining unit is further configured to: Determining a scaling factor of each of the to-be-played contents according to a distribution position of each of the to-be-played contents in the target content set; Scaling the predicted value of the account interaction parameter of each of the contents to be played according to the scaling coefficient of each of the contents to be played, to obtain the scaled predicted value of each of the contents to be played; A recommended order of each of the to-be-played contents among the multiple to-be-played contents is determined according to the scaled predicted value.

7. The device according to claim 6, characterized in that The determining unit is specifically configured to: When the distribution position of each of the to-be-played contents in the target content set is before a first distribution position threshold, determining the scaling factor of each of the to-be-played contents as a magnification factor; When the distribution position of each of the to-be-played contents in the target content set is after the second distribution position threshold, the scaling factor of each of the to-be-played contents is determined to be a reduction factor.

8. The device according to claim 6 or 7, characterized in that The acquisition unit is specifically used for: Inputting the characteristic information of each of the to-be-played contents and the characteristic information of the account to be pushed into a pre-trained click rate prediction model to predict the click rate of each of the to-be-played contents; Inputting the characteristic information of each of the to-be-played contents and the characteristic information of the to-be-pushed account into a pre-trained conversion rate prediction model to predict the conversion rate of each of the to-be-played contents; According to the predicted click rate of each of the to-be-played contents and the predicted conversion rate of each of the to-be-played contents, a predicted value of the account interaction parameter of each of the to-be-played contents is determined.

9. The device according to claim 8, characterized in that The acquisition unit is specifically used for: The product of the predicted click rate of each of the to-be-played contents and the predicted conversion rate of each of the to-be-played contents is determined as the predicted value of the account interaction parameter of each of the to-be-played contents.

10. The device according to claim 6 or 7, characterized in that The device also includes an updating unit; The acquisition unit is further configured to, after recommending each of the to-be-played contents to multiple accounts, acquire click operations and conversion operations of each of the accounts on each of the to-be-played contents, and perform statistics to obtain statistical results of the account interaction parameters of each of the to-be-played contents; The updating unit is used to update the target content set corresponding to each of the to-be-played contents according to the statistical result of the account interaction parameter of each of the to-be-played contents.

11. A content recommendation device, characterized in that: include: A processor and a memory for storing processor executable instructions; wherein the processor is configured to execute the instructions to implement the content recommendation method provided in any one of claims 1-5.

12. A computer-readable storage medium, characterized in that: The invention comprises instructions, which, when executed by a processor, cause the processor to execute the content recommendation method provided in any one of claims 1 to 5.

13. A computer program product, characterized in that The invention comprises instructions, which, when executed by a processor, cause the processor to execute the content recommendation method provided in any one of claims 1 to 5.

Images