Method, apparatus, device, and storage medium for content recommendation

Abstract

To provide a method, apparatus, device for content recommendation, and a storage medium.SOLUTION: A method includes: determining a first estimated value metric obtained by delivering a first recommended content item, for the first recommended content item in a recommended content set; determining a first resource efficiency value based on a historical resource allocation amount and a historical value metric associated with the first recommended content item; determining a first resource allocation amount to be allocated, based on the first resource efficiency value and a total resource allocation amount for the recommended content set; determining a first target estimated value metric, based on the first estimated value metric and the first resource allocation amount; and determining a ranking result of the first recommended content item in the recommended content set based on the first target estimated value metric, and distributing the first recommended content item based on the ranking result.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] (Reference to Related Application) This application claims priority to a Chinese invention patent application filed on October 30, 2023, entitled "Method, apparatus, device, and storage medium for recommending content" and bearing application number CN202311423731.2.

[0002] (Technical field) FIELD OF THE DISCLOSURE Exemplary embodiments of the present invention relate generally to the field of computer technology, and more particularly to a method, apparatus, device, and computer-readable storage medium for recommending content. [Background technology]

[0003] The Internet provides access to a wide variety of resources. For example, various applications, products, audio / video content, etc. can be accessed through the Internet. The accessible content also includes specific recommended content items related to various objects / resources, including, for example, advertisements. A content recommendation system provides content distributors with distribution of the recommended content items, for example, by distributing the recommended content items to specific distribution opportunities, thereby obtaining corresponding content clicks, conversions, etc. Distribution of the recommended content items is usually done on a competitive basis. It is necessary to specially design a recommendation strategy for a new recommended content item in the recommendation system.

[0004] Take advertising as an example. In the process of promotion, advertisers often encounter new distribution needs, and a typical operation is to create a new distribution plan to achieve the target distribution effect. From the customer's perspective, the new plan often encounters the cold start problem, which is reflected in the sales volume increase effect and unstable cost, affecting the advertiser's distribution experience and reducing the customer's willingness to invest in the budget. From the perspective of the content recommendation system, a new advertisement is completely new to various modules in the distribution chain. For example, if the amount of samples is insufficient, the estimation module cannot learn sufficiently, and the chain reaction of inaccurate estimation will easily cause the platform to bid insufficiently, affecting the content distribution effect and revenue. Summary of the Invention

[0005] In a first aspect of the present invention, a content recommendation method is provided, the method including: determining, for a first recommended content item in a recommended content set, a first estimated value measure obtained by delivering the first recommended content item, determining a first resource efficiency value of the first recommended content item based on a historical resource allocation amount and a historical value measure associated with the first recommended content item, determining a first resource allocation amount to be allocated to the first recommended content item based on the first resource efficiency value of the first recommended content item and a total resource allocation amount for the recommended content set, determining a first target estimated value measure for the first recommended content item based on the first estimated value measure and the first resource allocation amount, and determining a ranking result of the first recommended content item in the recommended content set based on the first target estimated value measure, where the first recommended content item is delivered based on the ranking result.

[0006] In a second aspect of the present invention, there is provided an apparatus for recommending content, comprising: a value scale module for determining, for a first recommended content item in a recommended content set, a first estimated value scale obtained by delivering the first recommended content item, an efficiency determination module for determining a first resource efficiency value of the first recommended content item based on a historical resource allocation amount and a historical value scale associated with the first recommended content item, a resource allocation amount determination module for determining a first resource allocation amount to be allocated to the first recommended content item based on the first resource efficiency value of the first recommended content item and a total resource allocation amount for the recommended content set, a target value scale module for determining a first target estimated value scale for the first recommended content item based on the first estimated value scale and the first resource allocation amount, and a ranking determination module for determining a ranking result of the first recommended content item in the recommended content set based on the first target estimated value scale, where the first recommended content item is delivered based on the ranking result.

[0007] In a third aspect of the present invention there is provided an electronic device comprising at least one processing unit and at least one memory coupled to the at least one processing unit for use in storing instructions executed by the at least one processing unit which, when executed by the at least one processing unit, cause the device to perform the method of the first aspect.

[0008] In a fourth aspect of the present invention, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of the first aspect.

[0009] It should be understood that the contents described in the summary of the present invention are not intended to limit the main or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will be readily understood from the following description. [Brief description of the drawings]

[0010] The above and other features, advantages, and aspects of the embodiments of the present invention will become more apparent from the following detailed description taken in conjunction with the drawings, in which like or similar symbols indicate like or similar elements, and in which:

[0011] [Figure 1] 1 illustrates a schematic diagram of an exemplary environment in which embodiments of the present invention may be implemented; [Diagram 2] 1 illustrates a flowchart of a process for recommending content according to some embodiments of the present invention; [Figure 3A] 1 illustrates an example value function according to some embodiments of the present invention; [Figure 3B] 1 illustrates examples of value functions and approximation functions thereof according to some embodiments of the present invention; [Figure 4] 1 illustrates an example architecture for recommending content according to some embodiments of the present invention; [Diagram 5] FIG. 2 shows a block diagram of an apparatus for recommending content according to some embodiments of the present invention; [Figure 6] 1 illustrates a block diagram of an electronic device in which one or more embodiments of the present invention may be implemented. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] In the following, the embodiments of the present invention will be described in more detail with reference to the drawings. Although the drawings show specific embodiments of the present invention, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather, these embodiments are provided for a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes and are not used to limit the protection scope of the present invention.

[0013] In describing embodiments of the present invention, the term "comprising" and similar terms are intended to be open-ended inclusions such as "including, but not limited to." The term "based on" should be understood as "based at least in part on." The terms "one embodiment" or "the embodiment" should be understood as "at least one embodiment." The term "some embodiments" should be understood as "at least some embodiments." Other explicit and implicit definitions may be included below.

[0014] It is understood that any data related to the technical solution (including but not limited to the data itself, the acquisition of the data, or the use of the data) should comply with the corresponding laws and regulations and related specified requirements.

[0015] It is understood that before using the technical solutions disclosed in each embodiment of the present invention, the types, scope of use, usage scenarios, etc. of personal information related to the present invention should be notified to users in an appropriate manner in accordance with relevant laws and regulations, and consent from users should be obtained.

[0016] For example, in response to receiving an unsolicited request from a user, presenting information is sent to the user to explicitly present to the user that the requested operation requires the acquisition and use of the user's personal information, so that the user can independently choose whether or not to provide the personal information to software or hardware, such as an electronic device, application, server, or storage medium, that performs the operation of the technical solution of the present invention, based on the presenting information.

[0017] As an optional, but non-limiting implementation, the method of sending the presented information to the user in response to receiving the user's unsolicited request may be, for example, by utilizing a pop-up window in which the presented information may be displayed in the form of text, and the pop-up window may further include a selection control for the user to select "agree" or "disagree" to providing the personal information to the electronic device.

[0018] It will be appreciated that the notification and user authorization process described above is merely a general outline and is not intended to limit the implementation of the present invention, and other means that comply with relevant laws and regulations may be applied to the implementation of the present invention.

[0019] 1 illustrates a schematic diagram of an exemplary environment 100 in which embodiments of the present invention may be implemented. A content recommendation system 110 is configured to present recommended content items for distribution opportunities. The recommended content items may include, for example, one or more recommended content items 122-1, 122-2, ..., 122-M (collectively or individually referred to as recommended content items 122 for ease of explanation) in a content database 120.

[0020] Recommended content items may be presented to each user. As shown in FIG. 1, one or more terminal devices 130-1, 130-2, 130-3, etc. (collectively or individually referred to as terminal devices 130 for ease of description) may be associated with the content recommendation system 110 and may access various content provided on the content recommendation system 110 based on, for example, corresponding audiences 132-1, 132-2, 132-3, etc. (collectively or individually referred to as audiences 132 for ease of description). As an example, the content recommendation system 110 may be an application, a website, a netpage, and other accessible platform. The terminal device 130 may have the content recommendation system 110 installed for accessing the application or may access the content recommendation system 110 in an appropriate manner.

[0021] The content recommendation system 110 may be configured to deliver (eg, provide or deliver to the terminal device 130) one or more particular recommended content items during a corresponding distribution opportunity or content exhibition opportunity based on a corresponding strategy.

[0022] As used herein, a recommended content item is content that is presented to recommend a corresponding resource. Examples of recommended content items may include advertisements. As used herein, a delivery opportunity or content display opportunity may be categorized at the granularity of a user (or audience), with each audience corresponding to one delivery opportunity. In addition to a user, a delivery opportunity may be categorized at the granularity of a single delivery time, a particular display location on a page, etc.

[0023] In some embodiments, the content recommendation system 110 can recommend corresponding recommended content items in the content recommendation system 110 based on a request of the content provider. In an advertisement distribution scenario, the content provider 150 may also be referred to as an advertiser. In some embodiments, the content provider may pay a fee to the content distributor based on the submission and subsequent conversion of the recommended content item, etc. In some embodiments, the content recommendation system 110 can determine the recommended content items 122 to submit to the corresponding distribution opportunity based on the bidding results.

[0024] In the environment 100, the terminal device 110 may be any kind of mobile, fixed, or portable terminal, including a mobile cell phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a flat-panel computer, a media computer, a multimedia flat-panel, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), a personal navigation device, assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio receiver, an e-book device, a gaming device, or any combination of the above, including accessories and peripherals of these devices or any combination thereof. In some embodiments, the terminal device 110 may also support any kind of interface to a user (such as "wearable" circuitry). The content recommendation system 110 may be, for example, any kind of computing system / server capable of providing computing capabilities, including but not limited to a mainframe, an edge computing node, a computing device in a cloud environment, and the like.

[0025] It should be understood that the structure and functionality of each element in environment 100 is described for illustrative purposes only and does not imply any limitation to the scope of the invention.

[0026] As mentioned above, cold start is a necessary stage of the recommended content item delivery plan. A content cold start means that new content is unfamiliar to the recommendation system without past information accumulation, and needs to accumulate a certain amount of exposure and interaction to collect basic data, and the process of accumulating this basic data is a cold start. In the present invention, a content cold start may include an advertisement cold start. In an advertisement scenario, a cold start refers to a learning stage from the start of advertisement delivery to stable delivery, and advertisement delivery data continues to be accumulated in the cold start stage so that the content recommendation system can obtain enough data samples to perform stable content recommendation. The recommended content items or advertisements in the cold start stage may also be called cold recommended content items or cold advertisements.

[0027] How to rapidly accumulate conversion samples in a short period of time using newly created recommended content items to get through the cold start period is a guarantee of the long-term stable operation and budget introduction of the content distribution platform. This problem can be optimized from multiple stages of the content distribution chain, including the recall stage, the estimation stage, and the fine-tuning stage. In a content recommendation system, the recall stage refers to recalling a certain number of recommended content items from global recommended content item candidates. The estimation stage refers to estimating the cost, click-through rate, conversion rate, etc. of the recommended content items, which can usually be implemented by training an estimation model. The fine-tuning stage refers to ranking the recalled recommended content items and delivering the top or top-ranked recommended content items to a specific distribution opportunity.

[0028] The cold start problem is a common research topic in the content recommendation field, and currently, much research and optimization has been done on the problem to provide a better experience to content providers. Some optimization strategies include cold path recall, generalized estimation, etc. Some optimization strategies focus on the assistance strategies in the fine-tuning phase of the recommended content items. In competitive content distribution, for a particular distribution opportunity, the highest ranked recommended content item is usually selected as the bidder for the distribution opportunity by ranking it depending on the estimated revenue for the recommended content item. The estimated revenue refers to the revenue obtained by submitting the recommended content item to a particular distribution opportunity.

[0029] In advertising scenarios, we typically use ECPM (Expected Cost Per Mille) as a value scale metric for estimated revenue. For a distribution opportunity, the ECPM of a recommended content item can be calculated using the following formula: ECPM I = bId I *pctr I *pcvr I Among them, bId I represents the bid price of the recommended content item for the delivery opportunity, and pctr I represents the expected click-through rate of the recommended content item when presented in a distribution opportunity, and pcvr I represents the expected conversion rate of the recommended content item when it is presented to the delivery opportunity. In the fine-tuning phase, there are N ads competing for a delivery opportunity, and the estimated revenue of the jth ad for delivery opportunity I is ecpm Ij Assuming this, the method for determining which advertisements can compete for a delivery opportunity can be expressed as follows:

number

[0030] For ads in the cold start phase, the content recommendation system applies support strategies to increase the exposure of the ads. Cold start support strategies are generally implemented by the system as "add-ons." (Outside 0001) The goal of this strategy is to improve the ranking of "cold" ads through TIFF2025075023000003.tif6170. This can be used to optimize the rate of increase in the amount of cold ads and improve the cold start pass rate. In this context, the bid for the above delivery opportunity can be expressed as follows:

number

[0031] From the perspective of content recommendation systems, (Outside 0003) TIFF2025075023000006.tif6170 is the revenue loss that the content recommendation system must bear, i.e., the advertisement that was given the opportunity to be delivered in competition. (Outside 0004) TIFF2025075023000007.tif7170 participate in the ranking, but ultimately, the revenue that the content recommendation platform gets from content providers is still ECPM Ij That is, the cost paid by the content provider is ecpm IjObviously, cold start support can improve the cold ad delivery experience of advertisers, but the content recommendation platform must bear a certain revenue loss. In order to avoid the revenue loss caused by support being too large, the total resource allocation budget of the support strategy is usually given after comprehensively considering the cold start revenue and revenue loss. Therefore, the cold start support strategy can be simply expressed as follows:

number

[0032] Regarding content recommendation support under the constraint of total resource allocation B, the solution for the current scenario can be summarized as follows.

[0033] One solution is the fixed backed solution. For each delivery opportunity, (Outside 0005) Fixed coefficient ( TIFF2025075023000009.tif6170) Then, until the total revenue loss (i.e., the total resource allocation) of the content recommendation system in a certain period of time reaches the predetermined budget B of the institution, the resource allocation of cold advertising, i.e. (Outside 0006) This solution is relatively simple and easy to implement, but has the disadvantages that it is difficult to determine the fixed coefficient, and the consumption of the total resource allocation of the budget is not smoothed.

[0034] Another solution is the dynamic support solution. For each delivery opportunity, we multiply the cold ad support revenue by a dynamic factor, i.e. (Outside 0007) Determine TIFF2025075023000011.tif7170. Here (Outside 0008) TIFF2025075023000012.tif6170 is usually a dynamic coefficient calculated by observing the consumption progress of the total resource allocation in real time and using an adjustment algorithm, for example, using a classical PID adjustment algorithm, (Outside 0009) TIFF2025075023000013.tif9170 can be determined. This solution can consume the total resource allocation more smoothly compared to the above solution. However, at any given time, all advertisements are allocated the same factor. (Outside 0010) Because TIFF2025075023000014.tif6170 is shared, the value difference in the cold start phase of different advertisements is not taken into account, and the overall support resource efficiency of cold start is lower.

[0035] Although the above has been described in an example advertising scenario, similar issues remain for other recommendation scenarios and other types of recommended content items.

[0036] Regarding the problem of passing the recommended content items in the cold start phase, it is necessary to consider how to allocate the total resource allocation amount for support to each recommended content item in the cold start phase in order to maximize the value of the entire cold start of the content recommendation system. It is also necessary to further consider how each recommended content item consumes the allocated budget to quickly pass the cold start phase in order to improve the delivery experience of customers.

[0037] In the embodiments of the present invention, a cold start support solution based on resource efficiency is proposed, which can dynamically sense resource efficiency in real time and self-adapt to distribution resources, and thus can better cope with changes in bidding environment such as content provider operations and changes in distribution opportunities, and greatly improve cold start distribution effect. In some embodiments, the solution can be implemented in the fine-tuning stage of the content recommendation process, thereby optimizing the ranking result of each recommended content, and further affecting whether the recommended content can be presented to a specific distribution opportunity.

[0038] In the following, with continuing reference to the drawings, some exemplary embodiments of the present invention will be described.

[0039] 2 shows a flow chart of a content recommendation process 200 according to some embodiments of the present invention. The process 200 may be implemented in the content recommendation system 110 of FIG.

[0040] At block 210, for a first recommended content item in the recommended content set, the content recommendation system 110 determines a first estimated measure of value that would result from the first recommended content item being delivered.

[0041] In some embodiments, the recommended content set includes a plurality of recommended content items, which refer to recommended content items that are delivered to a particular delivery opportunity or content exhibition opportunity. For example, the delivery opportunity or content exhibition opportunity in the content recommendation system 110 can be classified with a granularity of a user, a particular time period, or a particular exhibition position in a page, etc. In some embodiments, the recommended content set includes at least the recommended content items that are determined to be in a cold start phase in the content recommendation system 110. In this specification, the recommended content set includes at least the "cold" recommended content items that share the total resource allocation of the content recommendation system 110 in a certain period of time.

[0042] An embodiment of the present invention relates to selecting and delivering a recommended content item from a recommended content set. Providing a cold start support of certain resources for a recommended content item in a cold start phase improves the probability of the recommended content item being delivered. Assume that the "first recommended content item" described herein is a recommended content item in a cold start phase, and the recommended content set further includes one or more recommended content items in the cold start phase. Although a single recommended content item is described below, it is understood that the content recommendation system 110 can determine cold start resource allocation and delivery for each recommended content item in the cold start phase in a similar manner.

[0043] In some implementations, for a first recommended content item, its estimated value measure can be estimated, for example, in expected cost per thousand exposures (ECPM). From the content provider's (e.g., advertiser's) perspective, ECPM refers to the expected cost the content provider (e.g., advertiser) should pay for the delivery of the recommended content item. From the content recommendation system's perspective, ECPM refers to the expected revenue that the content provider will receive after the delivery of the recommended content item.

[0044] At block 220, the content recommendation system 110 determines a first resource efficiency value for the first recommended content item based on the historical resource shares and the historical value measure associated with the first recommended content item.

[0045] As mentioned above, the support problem of the recommended content items in the cold start phase is divided into two sub-problems. The first sub-problem is the resource allocation problem for support, i.e., how to allocate the total resource allocation to each recommended content item in the cold start phase in order to maximize the value of the entire cold start of the content recommendation system should be considered. The second sub-problem is the resource allocation consumption problem of the recommended content items, i.e., how each recommended content item should consume the allocated resource budget to quickly pass the cold start phase in order to improve the delivery experience of the customer. Here, resource allocation refers to any resource that improves the ranking position of the recommended content item. For the entire recommended content set, each recommended content item therein shares one resource allocation budget, i.e., the total resource allocation.

[0046] The budget allocation problem is defined as follows: assuming that the total resource allocation amount given by the content recommendation system is B, and it is expected to optimize the cold start effect of the content provider in order to make the content provider more willing to distribute content, how to allocate the total resource allocation amount B to the granularity of the recommended content items to maximize the cold start value is considered. This problem can be expressed by the following value function:

number

[0047] among them, (Outside 0011) TIFF2025075023000016.tif5170 represents the total resource allocation amount allocated to the Kth recommended content item among all K recommended content items in the recommended content set. If the sum of the resource allocation amounts of all K recommended content items does not exceed the total resource allocation amount B, the value function corresponding to the K recommended content items is (Outside 0012) Figure 3A shows the expected maximization of the value function (Outside 0013) Here is an example of TIFF2025075023000018.tif5170, which can be thought of as a two-parameter function with diminishing marginal returns. To solve the problem in equation (4) above, dynamic programming is required.

[0048] According to another aspect, the resource allocation budget division problem for each recommended content item is defined as follows: given the Kth recommended content item, its total resource allocation budget is: (Outside 0014) Given a TIFF2025075023000019.tif5170, we hope to optimize the conversion situation after the recommended content item is delivered so that the recommended content item passes the cold start period faster and improves the delivery experience of the customer. In terms of advertising bid prices, the problem is to find the incremental revenue of the Kth recommended content item for each delivery opportunity to maximize the number of conversions of the recommended content item. (Outside 0015) TIFF2025075023000020.tif5170 (The revenue of the relevant part is the total resource allocation amount This problem can be expressed as follows:

number

[0049] In the above formula (5), (Outside 0016) The value of TIFF2025075023000023.tif4170 is 1 or 0, which respectively indicates that the Kth recommended content item is delivered at the delivery opportunity or is not delivered at the delivery opportunity. That is, the Kth recommended content item consumes a part of the resource allocation amount only when it is delivered. (Outside 0017) The calculation for TIFF2025075023000024.tif5170 is: (Outside 0018) TIFF2025075023000025.tif5170, which is an approximation of the cost-per-conversion (CPA) cost, and can directly optimize the conversion of the recommended content item. In a real recommendation scenario, the parameters By adjusting TIFF2025075023000026.tif4170 (Outside 0019) According to the above formula (5), for the K-th recommended content item, the total resource allocation (Outside 0020) If the number of ads does not exceed TIFF2025075023000028.tif5170, we expect to select and deliver appropriate delivery opportunities in order to maximize the number of conversions, i.e., the expected click-through rate (PCTR) and expected conversion rate (PCVR).

[0050] In the resource allocation for the recommended content items, a resource efficiency metric is further taken into consideration. The resource efficiency metric is (Outside 0021) TIFF2025075023000029.tif5170, which is the resource allocation amount. (Outside 0022) This is the target revenue target obtained directly from TIFF2025075023000030.tif5170. The higher the efficiency, the more revenue can be brought to the cold start of the recommended content item with the same resource allocation. An example of the calculation of the target revenue target as a value measure cost for a single delivery opportunity is as follows: (Table 0001) Table 1 TIFF2025075023000031.tif33170

[0051] In the above table, assume that the estimated value scale ecpm for each of the three recommended content items is 3, 2, and 1, respectively. In the case where cold start support is not provided, the recommended content item with ecpm of 3 will be delivered in the current delivery opportunity. However, if support is provided (i.e., a certain amount of resources are allocated) to the recommended content item with ecpm of 1, i.e., ∇ecpm=3, then the total value scale of the recommended content item will be 4, which is higher than the remaining two recommended content items, and therefore it can be delivered in the current delivery opportunity. From the perspective of the content recommendation system, the direct revenue obtained from resource allocation of the recommended content item with ecpm of 1 is 1, which means that (Outside 0023) There is a revenue loss of TIFF2025075023000032.tif5170, and the resource efficiency is k=1 / 3.

[0052] The value function describes the resource efficiency of the recommended content item and defines the functional relationship between the "resource allocation amount" invested in the recommended content item and the "direct profit". In general, the target growth profit tends to decrease as the resource allocation amount is invested, and is usually defined as a two-parameter convex function as shown in FIG. 3A. However, in practical applications, the resource allocation amount is often finite, not infinite. Therefore, in some embodiments, based on the needs of practical applications and taking computing efficiency into consideration, the value function is approximated as a development function of the resource allocation amount and the value scale. That is, when the resource allocation amount is finite, the resource allocation amount and the value scale may be approximated as being in a linear relationship.

[0053] FIG. 3B illustrates an example value function 310 according to some embodiments of the invention. When TIFF2025075023000033.tif4170 increases, the value scale TIFF2025075023000034.tif4170 will also increase gradually. Supported resource allocation amount When TIFF2025075023000035.tif4170 is infinite, the value scale is maximized. However, considering that the amount of resource distribution is finite, the amount of resource distribution and the value scale can be approximated as being in a linear relationship, as shown in the approximation function 312 of the value function in FIG. 3B. In this case, (Outside 0024) TIFF2025075023000036.tif5170, in which the target revenue target can be expressed as a value scale ecpm, and k represents the resource efficiency value.

[0054] Considering the linear relationship between resource allocation and value scale, in practical application, the historical resource allocation and historical value scale of the first recommended content item can be collected, namely: (Outside 0025) TIFF2025075023000037.tif6170 constitutes one point on the value function. (Outside 0026) TIFF2025075023000038.tif6170 allows for the determination of the line shape of the value function and further allows for the determination of a resource efficiency value k of the first recommended content item, where the historical resource allocation amount may include the resource allocation amount allocated to the first recommended content item when the first recommended content item was successfully delivered for a particular delivery opportunity, and the historical value measure may include the direct revenue generated from the first recommended content item during this delivery, where the revenue may be calculated as ecpm, for example.

[0055] According to the above discussion, the preferred solution is to use the recommended content item as the granularity and assign each recommended content item its own "total resource allocation" based on its resource efficiency value. TIFF2025075023000039.tif5170, and adjust the support for the recommended content item based on the total resource allocation amount. However, when the number of recommended content items processed by the content recommendation system 110 is larger, the overall workload and complexity of this solution will be large, and the granularity of a single recommended content item may face the dilemma that the data is sparse and the value function is unreliable. In some embodiments of the present invention, by classifying the recommended content items in the cold start stage, the same type of recommended content items can share the same value function, i.e., the same resource efficiency value. In this way, the resource efficiency values ​​of the same type are combined to approximate the theoretical optimal solution based on the efficiency allocation budget and dynamic adjustment.

[0056] Specifically, in some embodiments, the content recommendation system 110 classifies the recommended content items in the recommended content set into multiple categories. Recommended content items under the same category can be approximated as the same recommended content item for the calculation of resource efficiency value, which can greatly reduce the complexity of solving strategies and applications. There are various options for the distribution solution of recommended content items, such as rule-based classification, model-based clustering, etc. For example, rule-based classification may include classifying recommended content items with the same conversion definition into the same category. The same conversion definition may include regarding the activation, download, registration, or form entry of the recommended content item as a conversion action.

[0057] For each category among the multiple categories, the content recommendation system 110 can determine a resource efficiency value for the category based on the historical resource allocation amount already allocated to the recommended content items of the category and the historical value measure obtained for the submission of the recommended content items of the category. Since the same category may include multiple recommended content items, more historical resource allocation amounts and historical value measures can be obtained. As mentioned above, the historical resource allocation amount and the historical value measure satisfy a value function having a linear relationship. More historical sampling points are helpful for fitting a more accurate resource efficiency value k in the linear function.

[0058] Assuming that the first recommended content item is classified into a first category of a plurality of categories, the resource efficiency value for the first category may be determined as the first resource efficiency value of the first recommended content item. For other recommended content items, their resource efficiency values ​​may also be determined similarly.

[0059] Continuing to refer to Figure 2, in block 230, the content recommendation system 110 determines a first resource allocation amount to be allocated to the first recommended content item based on the first resource efficiency value of the first recommended content item and the total resource allocation amount for the recommended content set.

[0060] Assume that the total resource allocation amount currently to be allocated is B, and the recommended content set is classified into N categories, whose corresponding resource efficiency values ​​are denoted as type_1, type_2, ..., type_N. The resource efficiency function for each category is (Outside 0027) TIFF2025075023000040.tif5170, in which (Outside 0028) TIFF2025075023000041.tif5170 represents the target revenue, e.g., value scale, for the recommended content items in category I; (Outside 0029) TIFF2025075023000042.tif5170 represents the resource efficiency value of the I category, (Outside 0030) TIFF2025075023000043.tif5170 represents the resource allocation amount of the recommended content item in the Ith category.

[0061] According to the above formula (4), the resource allocation problem of the recommended content items of N categories can be defined as follows:

number

[0062] According to the above formula (6), in some embodiments, the optimal solution is to allocate the total resource allocation B to the recommended content items in the category with the highest resource efficiency value k, i.e. (Outside 0031) The solution is to distribute it to TIFF2025075023000045.tif5170.

[0063] In some embodiments, another method can be applied to consider the stability of recommended content items of different categories in the system and tilt the resources for support towards recommended content items with higher resource efficiency values. That is, a larger resource allocation amount can be allocated from the total resource allocation amount B to recommended content items of categories with higher resource efficiency values, and conversely, a smaller resource allocation amount can be allocated from the total resource allocation amount B to recommended content items of categories with lower resource efficiency values. In some embodiments, each allocation factor allocateCoef[type_I] can be determined based on the resource efficiency values ​​of each of the multiple categories, and the allocation factor is determined based on the resource efficiency value of the corresponding category and the sum of the resource efficiency values ​​of all the categories. For example, for a first recommended content item currently under consideration, the content recommendation system can determine a first allocation factor for the first category by comparing the resource efficiency value of the first category with the resource efficiency values ​​of other categories among the multiple categories. For example, for the first category, the allocation factor for the first category can be determined based on the ratio of the resource efficiency value of the first category to the sum of the resource efficiency values ​​of the multiple categories. In some embodiments, to avoid a calculated ratio for a particular category being too low, if the ratio is less than the partition coefficient lower limit, the partition coefficient for the first category may be determined as the partition coefficient lower limit. If the ratio is equal to or greater than the partition coefficient lower limit, the partition coefficient for the first category may be determined as the ratio. In some embodiments, the partition coefficient lower limit may be configured for each category.

[0064] For example, the partition coefficient calculation is as follows:

number

[0065] In the above equation (7), (Outside 0032) TIFF2025075023000047.tif5170 is the lower limit of the distribution coefficient for category type_I.

[0066] In this way, for each category of recommended content items, at least the distribution coefficient corresponding to that category is calculated. (Outside 0033) Based on TIFF2025075023000048.tif5170, it is possible to determine the resource allocation amount to be allocated to the recommended content items under the category from the total resource allocation amount B. For example, the total resource allocation amount that can be allocated to the category type_I can be calculated as follows:

number

[0067] In some embodiments, in the delivery of delivery opportunity I, if the total resource allocation amount of the category has already been determined, the resource efficiency value of each category for a certain period of time is (Outside 0034) Get TIFF2025075023000050.tif5170 and calculate the distribution coefficient (Outside 0035) TIFF2025075023000051.tif5170 can be updated in real time. In some embodiments, in order to achieve the goal of smooth consumption, a dynamic adjustment strategy PID is combined to adjust the total resource allocation of each category. (Outside 0036) The consumption progress of TIFF2025075023000052.tif5170 can be controlled. Such a distribution solution can be called a dynamic adjustment solution combined with resource efficiency. Such a distribution solution can be expressed as follows:

number

number

[0068] According to the above formula (9), for any category, the sum of the resource allocation amounts of the recommended content items of the category that have already been distributed at each distribution opportunity during the history period (Outside 0037) TIFF2025075023000055.tif5170 and the total resource allocation amount for the category (Outside 0038) Based on TIFF2025075023000056.tif5170, the adjustment coefficient for the category in the current period (Outside 0039) Then, according to formula (10) above, for a particular recommended content item and a particular delivery opportunity, the adjustment factor for that category can be determined as (Outside 0040) and an expected click-through rate and an expected conversion rate for the recommended content item for the delivery opportunity. (Outside 0041) For each recommended content item under the same category, the resource allocation amount allocated during the delivery of each delivery opportunity can be calculated by the above formula (9) and the above formula (10).

[0069] In some embodiments, to reduce the adjustment complexity, an adjustment factor pacIngCoef for the recommended content set or the total resource allocation amount B for the current period can be calculated based on the sum of the total resource allocation amount and the resource allocation amount allocated to the already delivered recommended content items in the recommended content set in the history period, and such dynamic adjustment can be expressed as follows:

number

[0070] The adjustment factor pacIngCoef calculated by the above formula (11) can be used to adjust the resource allocation consumption of the recommended content items of all categories to determine the resource allocation amount to be allocated to each recommended content item under each category during the delivery of the current delivery opportunity I. Such an adjustment factor pacIngCoef may be referred to as a dynamic global adjustment factor. In some embodiments, for a recommended content item under a particular category, the dynamic global adjustment factor pacIngCoef and the allocation factor allocateCoef[type] for the category are used together to determine the resource allocation amount of each recommended content item under the category, which can be expressed as follows:

number

[0071] As you can see, in the above equation (12), (Outside 0042) TIFF2025075023000062.tif5170 is the theoretically optimal solution of the above equation (9). (Outside 0043) TIFF2025075023000063.tif5170 can be approximated. Similarly, for the first recommended content item currently under consideration, which falls into the first category, the first resource allocation amount to be allocated to the first recommended content item during the delivery of delivery opportunity I can be determined using the above equations (11) and (12).

[0072] Continuing with reference to FIG. 2, in block 240, the content recommendation system 110 determines a first target estimated value measure for the first recommended content item based on the first estimated value measure and the first resource allocation amount. During delivery for delivery opportunity I: (Outside 0044) If TIFF2025075023000064.tif4170 is the first estimated value scale, the first target estimated value scale is (Outside 0045) TIFF2025075023000065.tif5170, in which (Outside 0046) TIFF2025075023000066.tif5170 is the resource share allocated to the first recommended content item during delivery for delivery opportunity I, which can be thought of as an incremental value of the estimated value scale for the first recommended content item.

[0073] In block 250, the content recommendation system 110 determines a ranking result of the first recommended content item in the recommended content set based on the first target estimated value measure. The ranking result influences whether the first recommended content item is selected from the recommended content set to be used for distribution. Specifically, the content recommendation system 110 calculates a target estimated value measure of each recommended content item in the recommended content set. According to the ranking of the target estimated value measures, the recommended content item to be finally distributed from the recommended content set can be selected according to a predetermined rule, for example, the recommended content item with the highest rank can be selected to be distributed for the current distribution opportunity.

[0074] FIG. 4 illustrates an example of an architecture 400 for recommending content according to some embodiments of the present invention. For distribution to distribution opportunity I, a set of recommended content is ad1, ad2, ..., ad K For these recommended content items, we use estimated value measures ecpm1, ecpm2, ..., ecpm K The content recommendation system 110 may also perform content classification 430 on the K recommended content items, for example dividing them into N categories: (Outside 0047) Calculate the resource efficiency 440 of each category shown in TIFF2025075023000067.tif5170. Based on the content classification result and the resource efficiency value of each category, the distribution coefficient of each category is calculated as (Outside 0048) TIFF2025075023000068.tif5170. Also, the content recommendation system 110 can dynamically calculate an adjustment coefficient pacIngCoef for the current period according to the total resource allocation amount B 410 and the total resource allocation amount already allocated in the history period 420 through a PID adjustment strategy.

[0075] Adjustment coefficient pacIngCoef and distribution coefficient of each category (Outside 0049) TIFF2025075023000069.tif5170, the expected click-through rate for each recommended content item for delivery opportunity I TIFF2025075023000070.tif4170 and expected conversion rate (Outside 0050) Combining TIFF2025075023000071.tif4170, the resource allocation amount for each recommended content item (Outside 0051) In this way, the target estimated value measures of the K recommended content items can be determined. (Outside 0052) TIFF2025075023000073.tif6170 can be calculated. Suppose that based on the ranking results of the target estimated value measures of the K recommended content items, the highest ranked recommended content item ad2 is selected to be delivered to the delivery opportunity I.

[0076] The content recommendation process of FIG. 4 can be iteratively performed and updated. After completing the delivery of delivery opportunity I, the adjustment factor for the next period (Outside 0053) The amount of resources already allocated to the recommended content item ad2 to update TIFF2025075023000074.tif5170 (Outside 0054) By collecting TIFF2025075023000075.tif5170, the total amount of resource allocation already distributed, 420, can be updated. In addition, the amount of resource allocation for the recommended content item ad2 (Outside 0055) TIFF2025075023000076.tif5170 and value scale ecpm2 are also collected and used to update the resource efficiency value of the corresponding category. In some embodiments, the adjustment factor and the resource efficiency value may be updated in real time for each delivery opportunity, or may be updated periodically. For example, in an embodiment that is updated periodically, a period length may be set, and the resource allocation amount already allocated in the period may be collected and used to update the adjustment factor, and the resource allocation amount already allocated and the obtained value scale may be collected and used to update the resource efficiency value.

[0077] According to embodiments of the present invention, in order to achieve value maximization in a cold start assistance scenario given a total resource allocation, resource efficiency is introduced to maximize the resource efficiency for recommended content items, and in some embodiments, a practical solution for the optimal solution to maximize value is provided to reduce the computational complexity of the content recommendation process and facilitate implementation.

[0078] 5 shows a block diagram of an apparatus for recommending content according to some embodiments of the present invention. The apparatus 500 can be implemented as or included in the content recommendation system 110. Each module / component of the apparatus 500 can be implemented by hardware, software, firmware, or any combination thereof. As shown in the drawing, the apparatus 500 comprises a value scale module 510 for determining, for a first recommended content item in a recommended content set, a first estimated value scale obtained by delivering the first recommended content item. The apparatus 500 comprises an efficiency determination module 520 for determining a first resource efficiency value of the first recommended content item based on a historical resource allocation amount and a historical value scale associated with the first recommended content item. The apparatus 500 further comprises a resource allocation amount determination module 530 for determining a first resource allocation amount to be allocated to the first recommended content item based on the first resource efficiency value of the first recommended content item and a total resource allocation amount for the recommended content set. The apparatus 500 further comprises a target value scale module 540 for determining a first target estimated value scale for the first recommended content item based on the first estimated value scale and the first resource allocation amount, and a ranking determination module 550 for determining a ranking result of the first recommended content item in the recommended content set based on the first target estimated value scale, where the first recommended content item is delivered based on the ranking result.

[0079] In some embodiments, the distribution determination module 530 includes a distribution coefficient determination module for determining a first distribution coefficient for the first recommended content item based on a first resource efficiency value of the first recommended content item, and a distribution coefficient-based distribution determination module for determining a first resource distribution amount to be allocated to the first recommended content item based on at least the first distribution coefficient sum of an expected click-through rate and an expected conversion rate for the first recommended content item.

[0080] In some embodiments, the distribution determination module 530 further comprises an adjustment factor determination module for determining an adjustment factor for the current period based on the sum of the total resource distribution amount and the resource distribution amount distributed to already delivered recommended content items in the recommended content set during the historical period, and an adjustment factor based distribution determination module for further determining a first resource distribution amount to be distributed to the first recommended content item based on the adjustment factor.

[0081] In some embodiments, the efficiency determination module 520 comprises: a content classification module for classifying the recommended content items in the recommended content set into a plurality of categories, wherein a first recommended content item is classified into a first category of the plurality of categories; a category efficiency determination module for determining, for each category of the plurality of categories, a resource efficiency value for the category based on historical resource allocation amounts already allocated to the recommended content items of that category and the historical value measure obtained for the submission of the recommended content items of that category; and a content item efficiency determination module for determining the resource efficiency value for the first category as the first resource efficiency value of the first recommended content item.

[0082] In some embodiments, the historical resource allocation amount and the historical value measure satisfy a value function having a linear relationship.

[0083] In some embodiments, the distribution determination module 530 includes a distribution coefficient determination module for determining a first distribution coefficient for the first category by comparing the resource efficiency value of the first category with the resource efficiency values ​​of other categories among the plurality of categories, and a distribution coefficient based distribution determination module for determining a first resource distribution amount to be distributed to the first recommended content item from the total resource distribution amount based at least on the first distribution coefficient.

[0084] In some embodiments, the distribution coefficient based distribution determination module is configured to determine a distribution coefficient for the first category based on a ratio of a resource efficiency value of the first category to a sum of the resource efficiency values ​​of the multiple categories.

[0085] In some embodiments, the distribution coefficient based distribution determination module is further configured to determine a distribution coefficient for a first category as a distribution coefficient lower limit value when the ratio is less than a distribution coefficient lower limit value, and to determine a distribution coefficient for the first category as the ratio when the ratio is equal to or greater than the distribution coefficient lower limit value.

[0086] Fig. 6 illustrates a block diagram of an electronic device 600 capable of implementing one or more embodiments of the present invention. It should be understood that the electronic device 600 illustrated in Fig. 6 is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein. The electronic device 600 illustrated in Fig. 6 may be used to implement the content recommendation system 110 of Fig. 1 or the apparatus 500 of Fig. 5.

[0087] As shown in Fig. 6, the electronic device 600 is in the form of a general-purpose electronic device. The components of the electronic device 600 may include, but are not limited to, one or more processors or processing units 610, memory 620, storage device 630, one or more communication units 640, one or more input devices 650, and one or more output devices 660. The processing unit 610 may be a real or virtual processor and may perform various processes based on programs stored in the memory 620. In a multiprocessor system, multiple processing units execute computer-executable instructions in parallel, thereby improving the parallel processing capabilities of the electronic device 600.

[0088] The electronic device 600 typically includes a number of computer storage media. Such media may be any obtainable media accessible by the electronic device 600, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 620 may be volatile memory (e.g., registers, cache, random access memory (RAM)), non-volatile memory (e.g., read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory), or some combination thereof. The storage device 630 may be removable or non-removable media and may include machine-readable media, such as a flash memory drive, a magnetic disk, or any other media, that may be used to store information and / or data (e.g., training data for training) and may be accessible within the electronic device 600.

[0089] The electronic device 600 may further include other removable / non-removable, volatile / non-volatile storage media. Although not shown in FIG. 6, a magnetic disk drive for reading from or writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to a path (not shown) by one or more data media interfaces. The memory 620 may include a computer program product 625 having one or more program modules, which are configured to perform various methods or operations of various embodiments of the invention.

[0090] The communication unit 640 implements communication with other computing devices over a communication medium. Additionally, the functionality of the components of the electronic device 600 may be implemented as a single computing cluster or multiple computing machines that can communicate over a communication connection. Thus, the electronic device 600 can operate in a networked environment using logical connections to one or more other servers, network personal computers (PCs), or other network nodes.

[0091] The input device 650 may be one or more input devices such as a mouse, a keyboard, a trackball, etc. The output device 660 may be one or more output devices such as a display, a speaker, a printer, etc. The electronic device 600 may further communicate, as necessary, with one or more external devices (not shown) such as a storage device, a display device, etc., via the communication unit 640, one or more devices that allow a user to interact with the electronic device 600, or any device (e.g., a netbook card, a modem, etc.) that allows the electronic device 600 to communicate with one or more other computing devices. Such communication may be performed via an input / output (I / O) interface (not shown).

[0092] According to an exemplary implementation of the present invention, a computer-readable storage medium having one or more computer instructions stored thereon is provided, the one or more computer instructions being executed by a processor to implement the above-mentioned method. According to an exemplary implementation of the present invention, a computer program product is further provided, the computer program product being tangibly stored on a non-transitory computer-readable medium and including computer-executable instructions executed by a processor to implement the above-mentioned method.

[0093] Aspects of the present invention have been described herein with reference to flowchart and / or block diagrams of methods, apparatus (systems) and computer program products implemented by the present invention. It will be understood that each box in the flowchart and / or block diagrams, and combinations of boxes in the flowchart and / or block diagrams, can be implemented by computer readable program instructions.

[0094] These computer readable program instructions may be provided to a processing unit of a general purpose computer, special purpose computer, or other programmable data processing apparatus to generate a machine such that, when the instructions are executed by a processing unit of the computer or other programmable data processing apparatus, they generate an apparatus for implementing the functions / operations specified in one or more boxes in the flowcharts and / or block diagrams. These computer readable program instructions may be stored on a computer readable storage medium such that the instructions cause the computer, programmable data processing apparatus, and / or other device to operate in a particular manner such that the computer readable medium on which the instructions are stored constitutes an article of manufacture including instructions that implement each aspect of the functions / operations specified in one or more boxes in the flowcharts and / or block diagrams.

[0095] Loading the computer readable program instructions into a computer, other programmable data processing apparatus, or other device causes the computer, other programmable data processing apparatus, or other device to perform a series of operational steps to produce a computer-implemented process, such that the instructions executing on the computer, other programmable data processing apparatus, or other device implement the functions / acts specified in one or more boxes in the flowcharts and / or block diagrams.

[0096] The flowcharts and block diagrams in the figures illustrate possible architectures, functions, and operations of some possible systems, methods, and computer program products according to the present invention. In this regard, each box in the flowcharts or block diagrams may represent a module, program fragment, or part of instructions, which includes one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions depicted in the boxes may occur in a different order than depicted in the figures. For example, two consecutive boxes may in fact be executed substantially in parallel, or may be executed in reverse order depending on the functionality involved. It should also be noted that each box in the block diagrams and / or flowcharts, and combinations of boxes in the block diagrams and / or flowcharts, may be implemented by a special purpose hardware-based system that performs the specified functions or operations, or by a combination of special purpose hardware and computer instructions.

[0097] Although each implementation of the present invention has been described above, the above description is illustrative, not exhaustive, and is not limited to each disclosed implementation. Many modifications and changes will be apparent to those skilled in the art without departing from the scope and spirit of each described implementation. The selection of terms used in this specification is intended to best interpret the principles, practical applications, or improvements to technology in the marketplace of each implementation, or to enable those skilled in the art to understand each implementation disclosed in this specification.

Claims

1. determining, for a first recommended content item in the set of recommended content, a first estimated measure of value that would be obtained from delivery of the first recommended content item; determining a first resource efficiency value for the first recommended content item based on a historical resource allocation and a historical value measure associated with the first recommended content item; determining a first resource allocation amount to be allocated to the first recommended content item based on a first resource efficiency value of the first recommended content item and a total resource allocation amount for the recommended content set; determining a first target estimated value metric for the first recommended content item based on the first estimated value metric and the first resource allocation amount; determining a ranking result of the first recommended content item in the recommended content set based on the first target estimated value measure, and delivering the first recommended content item based on the ranking result; Content recommendation methods.

2. Determining a first resource allocation amount to be allocated to the first recommended content item includes: determining a first distribution factor for the first recommended content item based on a first resource efficiency value of the first recommended content item; determining a first resource allocation amount to be allocated to the first recommended content item based at least on the first allocation coefficient and an expected click-through rate and an expected conversion rate for the first recommended content item. The content recommendation method of claim 1.

3. Determining a first resource allocation amount to be allocated to the first recommended content item includes: determining an adjustment factor for a current time period based on the sum of the total resource allocation and the resource allocation allocated to already-delivered recommended content items in the recommended content set during a historical time period; and further determining a first resource allocation amount to be allocated to the first recommended content item based on the adjustment factor. The content recommendation method according to claim 2.

4. Determining a first resource efficiency value for the first recommended content item comprises: classifying the recommended content items in the recommended content set into a plurality of categories, the first recommended content item being classified into a first category of the plurality of categories; determining, for each category of the plurality of categories, a resource efficiency value for the category based on historical resource allocations already allocated to recommended content items in that category and historical value measures obtained for submissions of recommended content items in that category; determining a resource efficiency value for the first category as a first resource efficiency value for the first recommended content item. The content recommendation method of claim 1.

5. The content recommendation method according to claim 4 , wherein the history resource allocation amount and the history value scale satisfy a value function having a linear relationship.

6. Determining a first resource allocation amount to be allocated to the first recommended content item includes: determining a first distribution factor for the first category by comparing a resource efficiency value of the first category with resource efficiency values ​​of other categories of the plurality of categories; determining a first resource allocation amount to be allocated to the first recommended content item in the distribution from the total resource allocation amount based on at least the first allocation coefficient. The content recommendation method according to claim 4.

7. Determining a first partition coefficient for the first category comprises: determining a distribution coefficient for the first category based on a ratio of the resource efficiency value of the first category to a sum of the resource efficiency values ​​of the plurality of categories. The content recommendation method according to claim 6.

8. Determining a first partition coefficient for the first category based on the ratio includes: If the ratio is less than a partition coefficient lower limit, determining a partition coefficient for the first category as the partition coefficient lower limit; If the ratio is equal to or greater than the lower limit of the distribution coefficient, determining a distribution coefficient for the first category as the ratio. The content recommendation method according to claim 7.

9. a value measure module for determining, for a first recommended content item in the set of recommended content, a first estimated value measure resulting from delivery of said first recommended content item; an efficiency determination module for determining a first resource efficiency value of the first recommended content item based on a historical resource allocation amount and a historical value measure associated with the first recommended content item; a resource allocation amount determination module for determining a first resource allocation amount to be allocated to the first recommended content item based on a first resource efficiency value of the first recommended content item and a total resource allocation amount for the recommended content set; a target value measure module for determining a first target estimated value measure for the first recommended content item based on the first estimated value measure and the first resource allocation amount; a ranking determination module for determining a ranking result of the first recommended content item in the recommended content set based on the first target estimated value measure, the first recommended content item being delivered based on the ranking result; A device for recommending content.

10. 1. An electronic device comprising: At least one processing unit; and at least one memory coupled to said at least one processing unit and adapted to store instructions to be executed by said at least one processing unit, said instructions, when executed by said at least one processing unit, causing said electronic device to execute the method according to any one of claims 1 to 8. Electronic devices.

11. A computer program is stored, the computer program being executed by a processor to implement the method according to any one of claims 1 to 8. A computer-readable storage medium.

Images