Content recommendation method and device and storage medium
By counting the exposure and conversion rates in the information flow recommendation system and using a neural network model to adjust the content weight, the problem of inaccurate recommendation strategies in existing technologies is solved, and the efficiency of content recommendation and user experience are improved.
Patent Information
- Application Number
- CN202410281711.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-12
- Publication Date
- 2025-09-12
AI Technical Summary
In existing information flow recommendation systems, the efficiency and effectiveness of content recommendations are difficult to measure effectively, resulting in inaccurate recommendation strategies, affecting user experience and resource utilization efficiency.
By counting the exposure and conversion rates of the recommended content on the content recommendation page, and using a neural network model trained based on a loss function, we can determine the weight of the content and adjust its display position on the page to improve exposure efficiency.
It has achieved dynamic adjustment of content recommendations based on exposure and conversion ratio, improving content utilization efficiency and user experience, especially in information flow recommendations of shopping apps and news apps.
Smart Images
Figure CN120632192A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of information, and in particular to a content recommendation method, device, and storage medium. Background Art
[0002] In information flow recommendation systems, whether the recommended content has efficient returns is a topic worth studying. Summary of the Invention
[0003] To overcome the problems existing in the related art, the present disclosure provides a content recommendation method, device and storage medium.
[0004] According to a first aspect of an embodiment of the present disclosure, a content recommendation method is provided, comprising: on a content recommendation page, counting the exposure ratio and conversion ratio of content to be recommended; wherein, the content recommendation page includes multiple contents, and the content to be recommended is any one of the multiple contents; according to the exposure ratio and the conversion ratio, determining the weight of the content to be recommended in the total content of the content recommendation page, wherein the higher the weight, the higher the exposure ratio of the content to be recommended; based on the weight, recommending the content to be recommended on the content recommendation page.
[0005] In some embodiments, the weight of the content to be recommended in the total content of the content recommendation page is determined based on the exposure share and the conversion share, including: in response to the exposure share being less than the conversion share, determining that the weight of the content to be recommended in the total content of the content recommendation page is a first weight; in response to the exposure share being greater than the conversion share, determining that the weight of the content to be recommended in the total content of the content recommendation page is a second weight; the first weight is greater than a weight reference value, and the second weight is less than the weight reference value, and the weight reference value is the weight of the content to be recommended in the total content of the content recommendation page when counting the exposure share and the conversion share.
[0006] In some embodiments, determining that the weight of the content to be recommended accounts for the entire content of the content recommendation page is the first weight includes: in response to the larger the difference between the exposure share and the conversion share, determining that the difference between the first weight and the weight reference value is larger; in response to the smaller the difference between the exposure share and the conversion share, determining that the difference between the first weight and the weight reference value is smaller.
[0007] In some embodiments, determining the weight of the content to be recommended as the second weight of the entire content of the content recommendation page includes: in response to the larger the difference between the exposure share and the conversion share, determining that the difference between the second weight and the weight reference value is larger; in response to the smaller the difference between the exposure share and the conversion share, determining that the difference between the second weight and the weight reference value is smaller.
[0008] In some embodiments, the weight of the content to be recommended in the total content of the content recommendation page is determined in the following manner: characteristics of multiple contents in the content recommendation page are obtained; the characteristics are input into a neural network model to obtain weights corresponding to the multiple contents, and the weights corresponding to the multiple contents include the weights corresponding to the content to be recommended; wherein, the neural network model is trained based on a first loss function, and the first loss function is composed of at least two second loss functions, and the at least two second loss functions are respectively determined based on any of the following variables: whether the content is clicked; whether the content is clicked for the first time; whether the content is traded; whether the content belongs to the content to be recommended.
[0009] In some embodiments, the neural network model is trained in the following manner: multiplying the second loss function by a preset value, where different second loss functions correspond to the same or different preset values; the larger the preset value, the greater the difference between the weight output by the neural network model and the weight reference value; adding the results of multiplying the second loss function by the preset value to obtain the first loss function; and using the first loss function to train the neural network model.
[0010] In some embodiments, the preset value satisfies the following conditions: the greater the difference between the exposure share and the conversion share, the greater the preset value; the smaller the difference between the exposure share and the conversion share, the smaller the preset value.
[0011] In some embodiments, the preset value is determined by: obtaining a desired exposure ratio target value; and determining the preset value using the exposure ratio target value and a corresponding relationship, where the corresponding relationship is a corresponding relationship between the target value and the preset value.
[0012] In some embodiments, the content to be recommended is a secondary category of the content recommendation page, and the secondary category means that the content of the content recommendation page is divided twice according to category.
[0013] According to a second aspect of an embodiment of the present disclosure, a content recommendation device is provided, comprising: a statistical unit, configured to count, on a content recommendation page, the exposure ratio and conversion ratio of content to be recommended; wherein, the content recommendation page comprises a plurality of contents, and the content to be recommended is any one of the plurality of contents; a determination unit, configured to determine, based on the exposure ratio and the conversion ratio, the weight of the content to be recommended in the total content of the content recommendation page, wherein the higher the weight, the higher the exposure ratio of the content to be recommended; and a recommendation unit, configured to recommend the content to be recommended on the content recommendation page based on the weight.
[0014] In some embodiments, the determination unit determines the weight of the content to be recommended in the total content of the content recommendation page based on the exposure ratio and the conversion ratio in the following manner: in response to the exposure ratio being less than the conversion ratio, the weight of the content to be recommended in the total content of the content recommendation page is determined to be a first weight; in response to the exposure ratio being greater than the conversion ratio, the weight of the content to be recommended in the total content of the content recommendation page is determined to be a second weight; the first weight is greater than a weight reference value, and the second weight is less than the weight reference value, and the weight reference value is the weight of the content to be recommended in the total content of the content recommendation page when the exposure ratio and the conversion ratio are counted.
[0015] In some embodiments, the determination unit determines that the weight of the content to be recommended as the total content of the content recommendation page is the first weight in the following manner: in response to the larger the difference between the exposure ratio and the conversion ratio, the larger the difference between the first weight and the weight reference value is; in response to the smaller the difference between the exposure ratio and the conversion ratio, the smaller the difference between the first weight and the weight reference value is.
[0016] In some embodiments, the determination unit determines that the weight of the content to be recommended as the total content of the content recommendation page is the second weight in the following manner: in response to the larger the difference between the exposure ratio and the conversion ratio, the larger the difference between the second weight and the weight reference value is; in response to the smaller the difference between the exposure ratio and the conversion ratio, the smaller the difference between the second weight and the weight reference value is.
[0017] In some embodiments, the determination unit determines the weight of the content to be recommended in the total content of the content recommendation page in the following manner: obtaining features of multiple contents in the content recommendation page; inputting the features into a neural network model to obtain weights corresponding to the multiple contents, and the weights corresponding to the multiple contents include the weights corresponding to the content to be recommended; wherein, the neural network model is trained based on a first loss function, and the first loss function is composed of at least two second loss functions, and the at least two second loss functions are respectively determined based on any of the following variables: whether the content is clicked; whether the content is clicked for the first time; whether the content is traded; whether the content belongs to the content to be recommended.
[0018] In some embodiments, the device also includes a training unit for training a neural network model in the following manner: multiplying the second loss function by a preset value, where different second loss functions correspond to the same or different preset values; the larger the preset value, the greater the difference between the weight output by the neural network model and the weight reference value; adding the result of multiplying the second loss function by the preset value to obtain a first loss function; and using the first loss function to train the neural network model.
[0019] In some embodiments, the preset value satisfies the following conditions: the greater the difference between the exposure share and the conversion share, the greater the preset value; the smaller the difference between the exposure share and the conversion share, the smaller the preset value.
[0020] In some embodiments, the determination unit determines the preset value in the following manner: obtaining a required exposure ratio target value; and determining the preset value using the exposure ratio target value and a corresponding relationship, where the corresponding relationship is a corresponding relationship between the target value and the preset value.
[0021] In some embodiments, the content to be recommended is a secondary category of the content recommendation page, and the secondary category means that the content of the content recommendation page is divided twice according to category.
[0022] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a memory for storing instructions; and a processor for calling the instructions stored in the memory to execute the method of the first aspect and any one of the implementations of the first aspect.
[0023] According to a fourth aspect of an embodiment of the present disclosure, a storage medium is provided, in which instructions are stored. When the instructions are executed by a processor, the method in the first aspect or any one of the implementations of the first aspect is executed.
[0024] The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: by counting the exposure ratio and conversion ratio of the content to be recommended on the content recommendation page, the weight of the content to be recommended in the total content of the content recommendation page is determined according to the exposure ratio and conversion ratio. The higher the weight, the higher the exposure ratio, so as to achieve high utilization efficiency of the recommended content by adjusting the exposure ratio of the content to be recommended by determining the weight.
[0025] 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 disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.
[0027] Figure 1 The figure is a flowchart of a content recommendation method according to an exemplary embodiment.
[0028] Figure 2 The figure is a schematic diagram of a content recommendation page according to an exemplary embodiment.
[0029] Figure 3 The figure is a flowchart of a weight determination method according to an exemplary embodiment.
[0030] Figure 4 The figure is a flowchart of a weight determination method according to an exemplary embodiment.
[0031] Figure 5 The figure is a flowchart of a weight determination method according to an exemplary embodiment.
[0032] Figure 6 The figure is a flowchart of a weight determination method according to an exemplary embodiment.
[0033] Figure 7 The figure is a flowchart of a model training method according to an exemplary embodiment.
[0034] Figure 8 It is a schematic diagram of training a neural network model according to an exemplary embodiment.
[0035] Figure 9 The figure is a flowchart of a method for determining a preset value according to an exemplary embodiment.
[0036] Figure 10 The figure is a block diagram of a content recommendation device according to an exemplary embodiment.
[0037] Figure 11 The figure is a block diagram of a content recommendation device according to an exemplary embodiment. DETAILED DESCRIPTION
[0038] Exemplary embodiments are described in detail herein, with examples illustrated in the accompanying drawings. When the following description refers to the drawings, identical numerals in different drawings represent identical or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present disclosure.
[0039] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the corresponding data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.
[0040] In some embodiments, traditional insertion methods can be used to identify high-efficiency categories based on offline metrics. Fixed slots are allocated for inserting these identified categories to achieve traffic distribution goals. However, this approach fails to consider personalization, significantly harming the overall user experience of the information flow.
[0041] Therefore, the present disclosure provides a content recommendation method, which counts the exposure ratio and conversion ratio of the content to be recommended on the content recommendation page, and thus determines the weight of the content to be recommended in the total content of the content recommendation page according to the exposure ratio and conversion ratio. The higher the weight, the higher the exposure ratio, so as to achieve high utilization efficiency of the recommended content by adjusting the exposure ratio of the content to be recommended by determining the weight.
[0042] In some embodiments, the embodiments of the present disclosure can be applied to scenarios of advertising recommendation, for example, it can be applied to the homepage product recommendation scenario of a shopping application (application, APP). It can also be applied to scenarios of news recommendation, for example, it can be applied to the homepage information flow recommendation scenario of a news APP. Or it can also be applied to content recommendation scenarios of search engines, for example, content recommendation when a user searches for content through a search engine. Of course, the present disclosure only lists a few possible examples, but is not limited to them.
[0043] In some embodiments, the embodiments of the present disclosure can be applied to a server, such as a server of various APPs. Of course, it can also be applied to a client. For example, it can be applied to terminal devices such as mobile phones and computers.
[0044] In some embodiments, the terminal may be referred to as a terminal device, a mobile station (MS), a mobile terminal (MT), etc., and is a device that provides voice and / or data connectivity to a user. For example, the terminal may be a handheld device with wireless connection capabilities, a vehicle-mounted device, etc. Currently, some examples of terminals include: a smartphone (mobile phone), a pocket personal computer (PPC), a handheld computer, a personal digital assistant (PDA), a laptop computer, a tablet computer, a wearable device, or a vehicle-mounted device, etc. In addition, when it is a vehicle to everything (V2X) communication system, the terminal device may also be a vehicle-mounted device. It should be understood that the embodiments of the present disclosure do not limit the specific technology and specific device form adopted by the terminal.
[0045] Figure 1 FIG. 1 is a flow chart of a content recommendation method according to an exemplary embodiment. Figure 1As shown, the content recommendation method includes the following steps.
[0046] In step S11, on the content recommendation page, the exposure ratio and conversion ratio of the content to be recommended are counted.
[0047] In some embodiments, the exposure ratio and conversion ratio of the content to be recommended can be counted on the content recommendation page. The content recommendation page includes multiple contents, and the content to be recommended is any one of the multiple contents.
[0048] In some embodiments, for example, when a content recommendation page is displayed on a client, assuming that due to the limitation of the display screen, the content recommendation page is divided into multiple parts, and the user needs to manually scroll down to browse all parts, then the proportion of the recommended content in the first part or the first few parts can be called the exposure proportion. For ease of understanding, the present embodiment provides a schematic diagram of a content recommendation page. Figure 2 FIG. 1 is a schematic diagram of a content recommendation page according to an exemplary embodiment. Figure 2 As shown. Figure 2 In this example, due to the limitations of the mobile phone screen, the content recommendation page is divided into a first part, a second part, and a third part. For example, the first part can be seen by the user as soon as the content recommendation page is opened, without the user having to manually scroll. The ratio of the number of recommended content in the first part to the total number of content in the first part can be the exposure ratio. Alternatively, the ratio of the number of recommended content in the first and second parts to the total number of content in the first and second parts can be the exposure ratio. The content recommendation page can be, for example, a page that recommends products on a shopping app, or a page that recommends news articles on a news app. This disclosure does not provide specific examples. In subsequent embodiments, the example of a page that recommends products on a shopping app will be used as an example, but the examples are not limited to this. For example, the exposure ratio can be the ratio of the recommended content viewed on the homepage of a shopping app after the user opens the app. The conversion ratio represents the ratio of the number of conversions of the recommended content to the total number of conversions of the content. Conversion can be understood as the completion of a desired goal. For example, if the content is a product, the conversion can be the purchase of the product, or if the content is news, the conversion can be the forwarding of the news, etc. The conversion ratio of the content to be recommended can also be represented by the ratio of the gross merchandise volume (GMV) of the content to be recommended to the GMV of all content, that is, the conversion ratio can be the GMV ratio.
[0049] In step S12, the weight of the content to be recommended relative to the entire content of the content recommendation page is determined based on the exposure ratio and the conversion ratio.
[0050] In some embodiments, the weight of the content to be recommended relative to the total content on the content recommendation page can be determined based on the exposure ratio and conversion ratio. The weight of the content to be recommended can determine the exposure ratio of the content to be recommended. For example, the higher the weight of the content to be recommended, the higher the exposure ratio.
[0051] In step S13, based on the weights, the content to be recommended is recommended on the content recommendation page.
[0052] In some embodiments, the content to be recommended can be recommended on the content recommendation page according to the weight of the content to be recommended. Figure 2 As shown, when the content recommendation page is displayed on the client, assuming that the content recommendation page is divided into multiple parts, if the weight of the content to be recommended is higher, the content to be recommended can be placed in the first part or the first few parts to increase the exposure ratio, making it easier for users to see the content they are interested in, that is, to increase the probability of the product being purchased, or to increase the probability of the news being forwarded, etc., that is, to improve the utilization rate of the content recommendation page. For another example, the entire content of the content recommendation page can be sorted according to the weight of the content to be recommended and the weight of other content on the content recommendation page. The higher the weight, the higher the ranking on the content recommendation page, that is, the more likely it is to appear in the first part or the first few parts.
[0053] The present disclosure counts the exposure ratio and conversion ratio of the content to be recommended on the content recommendation page, and thus determines the weight of the content to be recommended in the total content of the content recommendation page according to the exposure ratio and the conversion ratio. The higher the weight, the higher the exposure ratio, so as to achieve high utilization efficiency of the recommended content by adjusting the exposure ratio of the content to be recommended by determining the weight.
[0054] In some embodiments, the present disclosure provides a weight determination method, which specifically describes how to determine the weight of content to be recommended based on the conversion rate and exposure rate of the content to be recommended.
[0055] Figure 3 is a flow chart showing a weight determination method according to an exemplary embodiment. Figure 3 As shown, the weight determination method provided by the present disclosure includes the following steps.
[0056] In step S21 , in response to the exposure ratio being less than the conversion ratio, the weight of the content to be recommended in the entire content of the content recommendation page is determined to be a first weight.
[0057] In step S22 , in response to the exposure ratio being greater than the conversion ratio, the weight of the content to be recommended in the entire content of the content recommendation page is determined to be a second weight.
[0058] In some embodiments, if the exposure ratio is less than the conversion ratio, it can be considered that the content to be recommended can still achieve a larger conversion ratio with a smaller exposure ratio, and the weight of the content to be recommended in the total content of the content recommendation page can be determined to be the first weight. Among them, the first weight is greater than the weight reference value. The weight reference value represents the weight of the content to be recommended in the total content of the content recommendation page when the exposure ratio and conversion ratio are calculated. For example, it can be a fixed value, or it can be the weight value determined last time. Since the higher the weight, the greater the exposure ratio. Determining the weight to be the first weight can increase the exposure ratio of the content to be recommended, making the content to be recommended easier for users to browse.
[0059] In some embodiments, if the exposure ratio is greater than the conversion ratio, it can be considered that the recommended content can only achieve a lower conversion ratio despite a higher exposure ratio. In this case, the weight of the recommended content relative to the total content on the content recommendation page can be determined to be a second weight. The second weight is less than the reference weight value. Since the higher the weight, the greater the exposure ratio. Conversely, the lower the weight, the smaller the exposure ratio. In this case, the second weight is determined, which can reduce the exposure ratio of the recommended content, making it relatively less likely for users to browse the recommended content, while making it easier for users to browse other content.
[0060] In some embodiments, if the exposure ratio is equal to the conversion ratio, the weight of the content to be recommended to the entire content of the content recommendation page can be determined to be the first weight or the second weight or the weight reference value.
[0061] It can be understood that there is no order between step S21 and step S22.
[0062] The present disclosure can improve the utilization rate of the content recommendation page by determining that the weight of the content to be recommended is a first weight when the exposure ratio of the content to be recommended is greater than the conversion ratio, and the first weight is greater than the weight reference value; if the exposure ratio is less than the conversion ratio, then the weight of the content to be recommended is a second weight, and the second weight is less than the weight reference value.
[0063] In some embodiments, for different situations of the difference between the exposure share and the conversion share, the difference between the first weight and the weight reference value can be adjusted, or the difference between the second weight and the weight reference value can be adjusted.
[0064] Figure 4 is a flow chart showing a weight determination method according to an exemplary embodiment. Figure 4 As shown, the present disclosure provides a weight determination method, which includes the following steps.
[0065] In step S31 , in response to a larger difference between the exposure ratio and the conversion ratio, a larger difference between the first weight and the weight reference value is determined.
[0066] In step S32 , in response to the smaller the difference between the exposure share and the conversion share, the smaller the difference between the first weight and the weight reference value is determined.
[0067] In some embodiments, the difference between the exposure ratio and the conversion ratio of the content to be recommended can be calculated. The larger the difference, the larger the difference between the determined weight and the weight reference value. For example, if the exposure ratio is smaller than the conversion ratio, the determined weight is the first weight, and the first weight is larger than the weight reference value. At this time, if the difference between the exposure ratio and the conversion rate is larger, the difference between the first weight and the weight reference value is larger, that is, the first weight is relatively larger, and the exposure ratio is increased more. For example, if the exposure ratio is larger than the conversion ratio, the determined weight is the second weight, and the second weight is smaller than the weight reference value. At this time, if the difference between the exposure ratio and the conversion ratio is larger, the difference between the second weight and the weight reference value is larger, that is, the second weight is relatively smaller, and the exposure ratio is reduced more.
[0068] In some embodiments, the smaller the difference between the exposure share and the conversion share of the content to be recommended, the smaller the difference between the determined weight and the weight reference value. For example, if the exposure share is less than the conversion share, the determined weight is the first weight, which is greater than the weight reference value. In this case, if the difference between the exposure share and the conversion rate is smaller, the difference between the first weight and the weight reference value is smaller. That is, although the first weight is greater than the weight reference value, it is relatively small, and the increased exposure share is small.
[0069] It can be understood that the difference in this embodiment refers to the absolute value of the difference, that is, the difference is a value greater than or equal to 0.
[0070] Figure 5 is a flow chart showing a weight determination method according to an exemplary embodiment. Figure 5 As shown, the present disclosure provides a weight determination method, which includes the following steps.
[0071] In step S41 , in response to a larger difference between the exposure ratio and the conversion ratio, a larger difference between the second weight and the weight reference value is determined.
[0072] In step S42 , in response to the smaller the difference between the exposure share and the conversion share, the smaller the difference between the second weight and the weight reference value is determined.
[0073] In some embodiments, the difference between the exposure share and the conversion share of the content to be recommended can be calculated. The larger the difference, the greater the difference between the determined weight and the weight reference value. For example, if the exposure share is greater than the conversion share, the determined weight is the second weight, which is less than the weight reference value. In this case, the greater the difference between the exposure share and the conversion share, the greater the difference between the second weight and the weight reference value, that is, the smaller the second weight is, and the greater the reduction in exposure share.
[0074] In some embodiments, the difference between the exposure share and the conversion share of the content to be recommended can be calculated. The smaller the difference, the smaller the difference between the determined weight and the weight reference value. For example, if the exposure share is greater than the conversion share, the determined weight is the second weight, and the second weight is less than the weight reference value. In this case, if the difference between the exposure share and the conversion share is smaller, the difference between the second weight and the weight reference value is smaller, that is, although the second weight is less than the weight reference value, it is relatively larger, that is, the exposure share is reduced less.
[0075] The present disclosure determines the size of the first weight or the second weight according to the difference between the exposure ratio and the conversion ratio, thereby making the weight of the content to be recommended more reasonable.
[0076] In some embodiments, the weight of the content to be recommended can be predicted by a neural network model.
[0077] In some embodiments, the weights of various contents on the content recommendation page can be predicted by a multi-task model, but the multi-task model is relatively complex and difficult to train. It also affects the click-through rate (CTR) score and the conversion rate (CVR) score estimation. Among them, CTR is the ratio of the number of times a user clicks on an ad to the number of times the ad is displayed. It is usually expressed as a percentage. CTR is an important indicator for measuring the effect of ad clicks. A high CTR usually means that the ad can attract the user's attention and prompt them to click. CVR is the proportion of users who complete the expected action after clicking on the ad, such as purchase, registration, etc. CVR measures the effect of the actual conversion into the target behavior after the ad is clicked, and is one of the key indicators for evaluating the effectiveness of advertising. The introduction of a multi-task model will cause deviations in the prediction of CTR scores and CVR scores.
[0078] Therefore, the present disclosure provides a method for moving content upward. Figure 6 is a flow chart showing a weight determination method according to an exemplary embodiment. Figure 6 As shown, the method for moving the content position upward includes the following steps.
[0079] In step S51, features of various contents in the content recommendation page are obtained.
[0080] In some embodiments, the CTR score and / or CVR score of the content on the content recommendation page can be predicted in advance and used as a feature of the content. It is understood that the prediction of the CTR score and / or CVR score is performed before the score of each content on the content recommendation page is predicted using the neural network model, and therefore affects the predicted CTR score and / or CVR score.
[0081] In some embodiments, historical data of the content on the content recommendation page may be obtained, such as whether the content has been clicked, the number of clicks, whether it has been forwarded, whether it has been traded, etc. The historical data of the content may be used as a feature of the content.
[0082] It can be understood that the content features acquired in this embodiment may include at least one of the following: a predicted CTR score, a predicted CVR score, and historical data.
[0083] In step S52, the features are input into the neural network model to obtain weights corresponding to the various contents, where the weights corresponding to the various contents include the weights corresponding to the content to be recommended.
[0084] In some embodiments, the acquired features can be input into a neural network model to obtain a weight corresponding to each content. The neural network model is trained based on a first loss function. The first loss function is composed of at least two second loss functions, each of which is determined based on any of the following variables: whether the content has been clicked, whether the content has been clicked for the first time, whether the content has been traded, and whether the content is a candidate for recommendation.
[0085] For example, the second loss function is determined based on whether the content is clicked. The formula of this type of second loss function can refer to the following formula 1.
[0086]
[0087] In formula 1, L ctr represents the second loss function based on whether the content is clicked, N represents the number of samples, that is, the number of training data for training the neural network model, that is, the number of contents. i represents the number of samples, and the size of i ranges from 1 to N. ∑Li represents the sum of Li. Li is equal to -[y i log(p i )+(1-y i )log(1-p i )], where y i Indicates whether the content is clicked. For example, y1 indicates whether the first content in the sample is clicked. If the first content is clicked, y1 is equal to 1. If the first content is not clicked, y1 is equal to 0. iRepresents the output of the model during training, i.e., the predicted weight. For example, p1 represents the output of the neural network model after the feature of the first content in the sample is input. If y1 is 1, the second loss function is used to make p1 as close to 1 as possible. In other words, the first loss function composed of the second loss function adjusts the model parameters so that when the input of the model is the feature of the first content, the output weight is close to 1. If y1 is 0, the second loss function is used to make p1 as close to 0 as possible. log represents the logarithm.
[0088] For another example, the second loss function is determined based on whether the content is clicked for the first time. The formula of this type of second loss function can refer to the following formula 2.
[0089]
[0090] In formula 2, L ipv represents the second loss function based on whether the content is clicked, N represents the number of samples, that is, the number of training data for training the neural network model, that is, the number of contents. i represents the number of samples, and the size of i ranges from 1 to N. ∑Li represents the sum of Li. Li is equal to h i Indicates whether the content is clicked for the first time. For example, h1 indicates whether the first content in the sample is clicked for the first time. If the first content is clicked for the first time, h1 is equal to 1. If the first content is not clicked for the first time, h1 is equal to 0. i Represents the output of the model during training, that is, the predicted score. log represents the logarithm.
[0091] It can be understood that Formula 2 is similar to Formula 1, except that the variable in Formula 2 is h i , indicating whether the content is clicked for the first time, the variable in formula 1 is y i Indicates whether the content has been clicked.
[0092] For another example, the second loss function may be determined based on whether the content is traded. For example, such a second loss function may refer to the following formulas 3 and 4.
[0093]
[0094] In formula 3, L pay The second loss function indicates whether the content is determined by the transaction. i Indicates whether the content is traded. If the content is traded, then z i is equal to 1, if the content is not traded, then z i Equal to 0. The other parameters and variables in the formula can refer to the definitions of Formula 1 and Formula 2, and are not described in detail in this disclosure.
[0095]
[0096] In formula 4, L price Another type of second loss function that indicates whether the content is determined by the transaction. i Indicates whether the content is traded, z i *The price in price indicates the unit price of the content. If the content is traded, then i is equal to 1, if the content is not traded, then z i Equal to 0. The other parameters and variables in the formula can refer to the definitions of Formula 1 and Formula 2, and are not described in detail in this disclosure.
[0097] For another example, the second loss function may be determined based on whether the content belongs to the content to be recommended. Such a second loss function may refer to the following formula 5.
[0098]
[0099] In formula 5, L pv represents the second loss function determined based on whether the content belongs to the content to be recommended. i Indicates whether the content belongs to the content to be recommended. If the content belongs to the content to be recommended, then u i Equal to 1, if the content does not belong to the content to be recommended, then u i Equal to 0.
[0100] In some embodiments, the second loss function can also be determined based on other variables, for example, whether the content is forwarded, etc., which are not listed in this disclosure. The first loss function can be composed of at least two second loss functions. For example, the first loss function can be based on L ctr and L pv The neural network model obtained by training the first loss function, when the characteristics of a certain content are input into the neural network model, the closer the output result of the neural network model is to 1, the more it can be said that the probability of the content being clicked is high and the probability of the content being recommended is high, that is, the weight corresponding to the content is larger. When distributing the contents on the content recommendation page based on the weight, the content is distributed at the front of the content recommendation page due to its larger weight, that is, the exposure ratio is larger. Among them, the front of the content recommendation page, for example Figure 2 The first part or the first few parts in the present disclosure are not repeated in detail. It can be understood that the first loss function can be composed of any two or more second loss functions, which are not listed one by one in the present disclosure, but are not limited to this.
[0101] In some embodiments, the neural network model can be a learning to rank (LTR) model, for example, a two-layer neural network model. For example, it can be a deep neural network (DNN)
[0102] The present disclosure obtains the weight of the content to be recommended through a neural network model trained by a first loss function, so that a simple model, such as an LTR model, can obtain a relatively accurate weight after training with the first loss function without affecting the estimation of the CTR score and / or CVR score.
[0103] In some embodiments, the present disclosure uses the first loss function L ctr 、L ipv 、L pay 、L price and L pv Taking the composition as an example, the training method of the neural network model is introduced.
[0104] Figure 7 FIG. 1 is a flow chart of a model training method according to an exemplary embodiment. Figure 7 As shown, the model training method includes the following steps.
[0105] In step S61, the second loss function is multiplied by a preset value, and the preset values corresponding to different second loss functions are the same or different.
[0106] In step S62, the result of multiplying the second loss function by the preset value is added to obtain the first loss function.
[0107] In step S63, the first loss function is used to train a neural network model.
[0108] In some embodiments, the second loss function may be multiplied by a preset value, and the multiplication results may be added to obtain the first loss function. For example, reference may be made to Formula 6.
[0109] Loss=w1*L ctr +w2*L ipv +w3*L pay +w4*L price +w5*L pv
[0110] Formula 6
[0111] In formula 6, Loss represents the first loss function, w1 represents the loss function with L ctr The corresponding preset value, w2 represents the value of L ipv The corresponding preset value, w3 represents the same as L pay The corresponding preset value, w4 represents the same as Lprice The corresponding preset value, w5 represents the same as L pv Corresponding preset values. It is understood that w1 to w5 can be the same or different. The larger the preset value, the greater the difference between the weight output by the neural network model and the weight reference value. For the output weight being the first weight, the larger the difference, the larger the first weight, and thus the greater the increase in exposure share; for the output weight being the second weight, the larger the difference, the smaller the second weight, and thus the greater the decrease in exposure share.
[0112] In some embodiments, if the preset values are the same, the second loss functions have the same impact on the neural network model. If the preset values are different, the second loss functions have different impacts on the neural network model. The preset value of the second loss function can be determined according to the actual situation. For example, if you want the probability of content being clicked to have a greater impact on the weight of the neural network model output, you can make L ctr The corresponding w1 value is larger. Figure 8 is a training diagram of a neural network model according to an exemplary embodiment. Figure 8 As shown, the first loss function, w1*L ctr 、w2*L ipv 、w3*L pay 、w4*L price and w5*L pv , adjust the parameters of the neural network model until the first loss function converges.
[0113] The present disclosure multiplies the second loss function by a preset value and adds the multiplication results to obtain the first loss function, so that the first loss function can comprehensively consider the influence of various aspects and obtain more reasonable weights. In addition, the model trained by the first loss function composed of the second loss function can be relatively simple, such as a LTR model, without the need for a complex multi-task model, and avoids affecting the estimated CTR score and / or CVR score.
[0114] In some embodiments, the model training method provided by the present disclosure also includes: the preset value satisfies the following conditions: the greater the difference between the exposure ratio and the conversion ratio, the greater the preset value; the smaller the difference between the exposure ratio and the conversion ratio, the smaller the preset value.
[0115] In some embodiments, if the difference between the exposure share and the conversion share is greater, the preset value may be determined to be larger. The larger the preset value, the greater the difference between the weight output by the neural network model and the weight reference value, thereby increasing or decreasing the exposure share more. If the difference between the exposure share and the conversion share is smaller, the preset value may be determined to be smaller. The smaller the preset value, the smaller the difference between the weight output by the neural network model and the weight reference value, thereby decreasing the exposure share.
[0116] In some embodiments, the preset value can be determined based on the required exposure ratio target value and the corresponding relationship.
[0117] Figure 9 FIG. 1 is a flow chart of a method for determining a preset value according to an exemplary embodiment. Figure 9 As shown, the preset value determination method includes the following steps.
[0118] In step S71, the required exposure ratio target value is obtained.
[0119] In some embodiments, a desired exposure share target value may be obtained, for example, an exposure share target value set by a user, or an exposure share target value specified by a client.
[0120] In step S72, the preset value is determined using the exposure ratio target value and the corresponding relationship, where the corresponding relationship is the corresponding relationship between the target value and the preset value.
[0121] In some embodiments, the correspondence between the target exposure share and the preset value can be predetermined. For example, two sets of experiments can be set up to determine the preset values for achieving two exposure shares, respectively. Based on the preset values corresponding to the exposure shares obtained from the two sets of experiments, a curve is fitted to determine the correspondence between the target exposure share and the preset value. The corresponding preset value is then determined using the target exposure share and the corresponding relationship.
[0122] The present disclosure determines the preset value by obtaining the required exposure ratio target value, so as to achieve flexible determination of the preset value and improve the user experience.
[0123] In some embodiments, the content to be recommended may be a secondary category. The secondary category means that the content of the content recommendation page is divided twice according to the category. For example, the first division is the primary category, assuming that the primary category is food. The secondary category is a division of the primary category, that is, the second division, and the secondary category may include bread, milk, vegetables, fruits, etc. The content to be recommended may also be a primary category or a tertiary category. The tertiary category is a division of the secondary category. For example, for the secondary category is vegetables, its corresponding tertiary category may include cabbage, cucumber, etc. Of course, the examples given in this disclosure are only for explanation and are not limited to this.
[0124] Based on the same concept, an embodiment of the present disclosure also provides a content recommendation device.
[0125] It is understandable that the content recommendation device provided by the embodiment of the present disclosure includes hardware structures and / or software modules corresponding to the execution of each function in order to realize the above functions. In combination with the units and algorithm steps of each example disclosed in the embodiment of the present disclosure, the embodiment of the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to realize the described functions for each specific application, but such implementation should not be considered to exceed the scope of the technical solution of the embodiment of the present disclosure.
[0126] It should be noted that those skilled in the art will appreciate that the various implementation methods / embodiments involved in the embodiments of the present disclosure can be used in conjunction with the aforementioned embodiments or can be used independently. Whether used alone or in conjunction with the aforementioned embodiments, the implementation principles are similar. In the implementation of the present disclosure, some embodiments are described in terms of implementation methods used together. Of course, those skilled in the art will appreciate that such examples are not limitations on the embodiments of the present disclosure.
[0127] Figure 10 FIG. 1 is a block diagram of a content recommendation device 100 according to an exemplary embodiment. Figure 10 As shown, the apparatus 100 includes: a statistics unit 101 , a determination unit 102 , a recommendation unit 103 and a training unit 104 .
[0128] The statistical unit 101 is configured to calculate the exposure rate and conversion rate of the content to be recommended on the content recommendation page. The content recommendation page may include multiple types of content, and the content to be recommended is any one of the multiple types of content. The determination unit 102 is configured to determine the weight of the content to be recommended relative to the total content on the content recommendation page based on the exposure rate and conversion rate, wherein a higher weight indicates a higher exposure rate for the content to be recommended. The recommendation unit 103 is configured to recommend the content to be recommended on the content recommendation page based on the weight.
[0129] In some embodiments, the determination unit 102 determines the weight of the content to be recommended relative to the entire content of the content recommendation page based on the exposure ratio and the conversion ratio in the following manner: in response to the exposure ratio being less than the conversion ratio, the weight of the content to be recommended relative to the entire content of the content recommendation page is determined to be a first weight. In response to the exposure ratio being greater than the conversion ratio, the weight of the content to be recommended relative to the entire content of the content recommendation page is determined to be a second weight. The first weight is greater than a weight reference value, and the second weight is less than the weight reference value. The weight reference value is the weight of the content to be recommended relative to the entire content of the content recommendation page when the exposure ratio and the conversion ratio are calculated.
[0130] In some embodiments, the determining unit 102 determines the weight of the content to be recommended relative to the entire content of the content recommendation page as the first weight in the following manner: in response to a larger difference between the exposure ratio and the conversion ratio, the difference between the first weight and the weight reference value is determined to be larger. In response to a smaller difference between the exposure ratio and the conversion ratio, the difference between the first weight and the weight reference value is determined to be smaller.
[0131] In some embodiments, the determining unit 102 determines the weight of the content to be recommended relative to the entire content of the content recommendation page as the second weight in the following manner: in response to a larger difference between the exposure ratio and the conversion ratio, the difference between the second weight and the weight reference value is determined to be larger. In response to a smaller difference between the exposure ratio and the conversion ratio, the difference between the second weight and the weight reference value is determined to be smaller.
[0132] In some embodiments, the determination unit 102 determines the weight of the content to be recommended relative to the total content on the content recommendation page by obtaining features of the various contents on the content recommendation page. The features are input into a neural network model to obtain weights corresponding to the various contents, where the weights corresponding to the various contents include the weight corresponding to the content to be recommended. The neural network model is trained based on a first loss function, and the first loss function is composed of at least two second loss functions. The at least two second loss functions are respectively determined based on any of the following variables: whether the content has been clicked, whether the content has been clicked for the first time, whether the content has been traded, and whether the content belongs to the content to be recommended.
[0133] In some embodiments, the apparatus further includes a training unit 104 configured to train a neural network model by multiplying a second loss function by a preset value, where different second loss functions correspond to the same or different preset values. A larger preset value results in a greater difference between the weight output by the neural network model and the weight reference value. The result of multiplying the second loss function by the preset value is added to obtain a first loss function. The neural network model is trained using the first loss function.
[0134] In some embodiments, the preset value satisfies the following conditions: the greater the difference between the exposure share and the conversion share, the larger the preset value; and the smaller the difference between the exposure share and the conversion share, the smaller the preset value.
[0135] In some embodiments, the determining unit 102 determines the preset value by obtaining a desired exposure ratio target value and determining the preset value using the exposure ratio target value and a corresponding relationship, where the corresponding relationship is a corresponding relationship between the target value and the preset value.
[0136] In some embodiments, the content to be recommended is a secondary category of the content recommendation page, and the secondary category means that the content of the content recommendation page is divided twice according to the category.
[0137] Figure 11is a block diagram of a content recommendation device 200 according to an exemplary embodiment.
[0138] like Figure 11 As shown, apparatus 200 may include one or more of the following components: a processing component 202 , a memory 204 , a power component 206 , a multimedia component 208 , an audio component 210 , an input / output (I / O) interface 212 , a sensor component 214 , and a communication component 216 .
[0139] The processing component 202 generally controls the overall operation of the device 200, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 202 may include one or more processors 220 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 202 may include one or more modules to facilitate interaction between the processing component 202 and other components. For example, the processing component 202 may include a multimedia module to facilitate interaction between the multimedia component 208 and the processing component 202.
[0140] The memory 204 is configured to store various types of data to support operations on the device 200. Examples of such data include instructions for any application or method operating on the device 200, contact data, phone book data, messages, pictures, videos, etc. The memory 204 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.
[0141] The power component 206 provides power to the various components of the device 200. The power component 206 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device 200.
[0142] The multimedia component 208 includes a screen that provides an output interface between the device 200 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can not only sense the boundaries of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 208 includes a front camera and / or a rear camera. When the device 200 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.
[0143] The audio component 210 is configured to output and / or input audio signals. For example, the audio component 210 includes a microphone (MIC) that is configured to receive external audio signals when the device 200 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may be further stored in the memory 204 or transmitted via the communication component 216. In some embodiments, the audio component 210 further includes a speaker for outputting audio signals.
[0144] I / O interface 212 provides an interface between processing component 202 and peripheral interface modules, such as a keyboard, click wheel, buttons, etc. These buttons may include but are not limited to: a home button, volume buttons, a start button, and a lock button.
[0145] The sensor assembly 214 includes one or more sensors for providing various aspects of the status assessment of the device 200. For example, the sensor assembly 214 can detect the open / closed state of the device 200, the relative positioning of components, such as the display and keypad of the device 200. The sensor assembly 214 can also detect changes in the position of the device 200 or a component of the device 200, the presence or absence of user contact with the device 200, the orientation or acceleration / deceleration of the device 200, and temperature changes of the device 200. The sensor assembly 214 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 214 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 214 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
[0146] The communication component 216 is configured to facilitate wired or wireless communication between the device 200 and other devices. The device 200 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 216 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 216 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
[0147] In an exemplary embodiment, the apparatus 200 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.
[0148] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as the memory 204 including instructions, which can be executed by the processor 220 of the apparatus 200 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.
[0149] The present disclosure counts the exposure ratio and conversion ratio of the content to be recommended on the content recommendation page, and thus determines the weight of the content to be recommended in the total content of the content recommendation page according to the exposure ratio and the conversion ratio. The higher the weight, the higher the exposure ratio, so as to achieve high utilization efficiency of the recommended content by adjusting the exposure ratio of the content to be recommended by determining the weight.
[0150] It is understood that in this disclosure, "plurality" refers to two or more than two, and other quantifiers are similar. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the related objects before and after are in an "or" relationship. The singular forms "a", "an", and "the" are also intended to include plural forms, unless the context clearly indicates otherwise.
[0151] It will be further understood that the terms "first," "second," and the like are used to describe various types of information, but such information should not be limited to these terms. These terms are used solely to distinguish information of the same type from one another and do not indicate a particular order or level of importance. In fact, the terms "first," "second," and the like are fully interchangeable. For example, first information could be referred to as second information, and similarly, second information could be referred to as first information without departing from the scope of this disclosure.
[0152] It is further understood that although operations are described in a particular order in the drawings in the embodiments of the present disclosure, this should not be construed as requiring that the operations be performed in the particular order shown or in a serial order, or that all of the operations shown be performed to obtain the desired results. In certain circumstances, multitasking and parallel processing may be advantageous.
[0153] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein.
[0154] Regarding the apparatus 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.
Claims
1. A content recommendation method, characterized in that: The method comprises: On the content recommendation page, count the exposure and conversion rates of the recommended content; Wherein, the content recommendation page includes multiple contents, and the content to be recommended is any one of the multiple contents; Determining the weight of the content to be recommended relative to the entire content of the content recommendation page based on the exposure ratio and the conversion ratio, wherein the higher the weight, the higher the exposure ratio of the content to be recommended; Based on the weight, the content to be recommended is recommended on the content recommendation page.
2. The method according to claim 1, characterized in that Determining the weight of the content to be recommended relative to the entire content of the content recommendation page based on the exposure ratio and the conversion ratio includes: In response to the exposure ratio being less than the conversion ratio, determining that the weight of the to-be-recommended content relative to all content on the content recommendation page is a first weight; In response to the exposure ratio being greater than the conversion ratio, determining that the weight of the to-be-recommended content relative to all content on the content recommendation page is a second weight; The first weight is greater than a weight reference value, and the second weight is less than the weight reference value. The weight reference value is the weight of the content to be recommended in the total content of the content recommendation page when calculating the exposure ratio and the conversion ratio.
3. The method according to claim 2, characterized in that Determining the weight of the content to be recommended in the total content of the content recommendation page as a first weight includes: In response to a larger difference between the exposure share and the conversion share, determining a larger difference between the first weight and the weight reference value; In response to the smaller difference between the exposure share and the conversion share, the smaller the difference between the first weight and the weight reference value is determined.
4. The method according to claim 2, characterized in that The determining that the weight of the content to be recommended in the total content of the content recommendation page is a second weight includes: In response to a larger difference between the exposure share and the conversion share, determining a larger difference between the second weight and the weight reference value; In response to the smaller difference between the exposure share and the conversion share, the smaller the difference between the second weight and the weight reference value is determined.
5. The method according to claim 1, characterized in that The weight of the content to be recommended relative to the entire content of the content recommendation page is determined in the following manner: Obtaining features of multiple contents in the content recommendation page; Inputting the features into a neural network model to obtain weights corresponding to the plurality of contents, wherein the weights corresponding to the plurality of contents include the weight corresponding to the content to be recommended; In which, the neural network model is trained based on a first loss function, and the first loss function is composed of at least two second loss functions, and the at least two second loss functions are respectively determined based on any of the following variables: whether the content is clicked; whether the content is clicked for the first time; whether the content is traded; whether the content belongs to content to be recommended.
6. The method according to claim 5, characterized in that The neural network model is trained in the following way: Multiplying the second loss function by a preset value, where different second loss functions correspond to the same or different preset values; The larger the preset value, the greater the difference between the weight output by the neural network model and the weight reference value; Add the result of multiplying the second loss function by the preset value to obtain the first loss function; The neural network model is trained using the first loss function.
7. The method according to claim 6, characterized in that The preset value meets the following conditions: The greater the difference between the exposure ratio and the conversion ratio, the greater the preset value; The smaller the difference between the exposure ratio and the conversion ratio, the smaller the preset value.
8. The method according to claim 6, characterized in that The preset value is determined in the following manner: Get the desired exposure share target value; The preset value is determined by using the exposure ratio target value and the corresponding relationship, where the corresponding relationship is the corresponding relationship between the target value and the preset value.
9. The method according to claim 1, characterized in that The content to be recommended is a secondary category of the content recommendation page, and the secondary category means that the content of the content recommendation page is divided twice according to the category.
10. A content recommendation device, characterized in that: The device comprises: The statistics unit is used to count the exposure ratio and conversion ratio of the recommended content on the content recommendation page; Wherein, the content recommendation page includes multiple contents, and the content to be recommended is any one of the multiple contents; a determining unit, configured to determine a weight of the content to be recommended relative to the entire content of the content recommendation page according to the exposure ratio and the conversion ratio, wherein a higher the weight, a higher the exposure ratio of the content to be recommended; A recommendation unit is configured to recommend the content to be recommended on the content recommendation page based on the weight.
11. An electronic device, characterized in that: include: a memory for storing instructions; as well as A processor, configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 9.
12. A storage medium, characterized in that: The storage medium stores instructions, and when the instructions are executed by the processor, the method according to any one of claims 1 to 9 is executed.