Methods, apparatus, equipment, storage media, and procedures for determining recommended content

Abstract

This disclosure provides methods, apparatus, devices, storage media, and program products for determining recommended content, relating to artificial intelligence technologies such as personalized recommendation, intelligent recommendation, and information filtering. One specific implementation of the method includes: reading historical tags of previously recommended content in an evaluation window, and the percentage of historical recommended content associated with each historical tag relative to the total number of all historical recommended content; generating adjustment parameters corresponding to the target historical tag based on the target percentage value of the target historical tag; updating a first recommendation reference value for a first candidate recommended content associated with the target historical tag to an updated first recommendation reference value based on the adjustment parameters; and determining target recommended content for recommendation from the first candidate recommended content and the second candidate recommended content based on the updated first recommendation reference value and the second candidate recommended content.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, specifically to the field of artificial intelligence technologies such as personalized recommendation, intelligent recommendation, and information filtering, and particularly to methods, apparatus, electronic devices, computer-readable storage media, and computer program products for determining recommended content. Background Technology

[0002] With the rapid development of the internet and mobile devices, the amount of information and content that users can access online is exploding. Against this backdrop, when faced with massive amounts of information, users often struggle to obtain content that truly matches their interests and needs in a timely and efficient manner, leading to the problem of "information overload."

[0003] To improve users' content acquisition experience and enhance content distribution efficiency, intelligent recommendation technology has emerged and gradually become a core functional module of many online platforms. Through intelligent recommendation technology, users can access the content they expect and are interested in at a lower cost and with a higher probability, thereby significantly improving content acquisition efficiency and user experience.

[0004] Therefore, improving recommendation quality and enhancing the user experience when accessing content through recommendation services is a matter of concern and urgent need. Summary of the Invention

[0005] This disclosure provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for determining recommended content.

[0006] In a first aspect, embodiments of this disclosure propose a method for determining recommended content, comprising: reading historical tags of historical recommended content already recommended in an evaluation window, and the percentage of the number of historical recommended content associated with the historical tags to the total number of all historical recommended content; generating adjustment parameters corresponding to the target historical tags based on the target percentage value of the target historical tags; updating the first recommendation reference value of the first candidate recommended content associated with the target historical tags to an updated first recommendation reference value based on the adjustment parameters; and determining the target recommended content for recommendation from the first candidate recommended content and the second candidate recommended content based on the updated first recommendation reference value and the second recommendation reference value corresponding to the second candidate recommended content.

[0007] Secondly, embodiments of this disclosure propose an apparatus for determining recommended content, comprising: a tag proportion reading unit configured to read historical tags of historical recommended content already recommended in an evaluation window, and the proportion of the number of historical recommended content associated with the historical tags to the total number of all historical recommended content; an adjustment parameter generation unit configured to generate adjustment parameters corresponding to the target historical tag based on the target proportion value of the target historical tag; a reference value updating unit configured to update a first recommendation reference value of a first candidate recommended content associated with the target historical tag to an updated first recommendation reference value based on the adjustment parameters; and a recommended content determination unit configured to determine target recommended content for recommendation from the first candidate recommended content and the second candidate recommended content based on the updated first recommendation reference value and the second recommendation reference value corresponding to the second candidate recommended content.

[0008] Thirdly, embodiments of this disclosure provide an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to implement the method for determining recommended content as described in any implementation of the first aspect.

[0009] Fourthly, embodiments of this disclosure provide a non-transitory computer-readable storage medium storing computer instructions for enabling a computer to perform a method for determining recommended content as described in any implementation of the first aspect.

[0010] Fifthly, embodiments of this disclosure provide a computer program product including a computer program that, when executed by a processor, enables the method for determining recommended content as described in any implementation of the first aspect.

[0011] The method, apparatus, electronic device, computer-readable storage medium, and computer program product for determining recommended content provided in this disclosure first read the historical tags of historical recommended content that has already been recommended in the evaluation window, as well as the percentage of the number of historical recommended content associated with the historical tags to the total number of all historical recommended content; then, based on the target percentage value of the target historical tag, an adjustment parameter corresponding to the target historical tag is generated; next, based on the adjustment parameter, the first recommendation reference value of the first candidate recommended content associated with the target historical tag is updated to an updated first recommendation reference value; finally, based on the updated first recommendation reference value and the second recommendation reference value corresponding to the second candidate recommended content, the target recommended content for recommendation is determined from the first candidate recommended content and the second candidate recommended content.

[0012] This disclosure can dynamically adjust the recommendation weight of future recommended content based on the already recommended content, thereby achieving a balance of diversity in recommended content, avoiding excessive concentration and duplication of recommended content, optimizing the overall content distribution of the recommendation system, and improving the quality of recommended content determination.

[0013] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0014] Other features, objects, and advantages of this disclosure will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is an exemplary system architecture to which this disclosure can be applied; Figure 2 A flowchart illustrating a process for determining recommended content, provided as an embodiment of this disclosure; Figure 3 A flowchart illustrating a process for generating adjustment parameters is provided as an embodiment of this disclosure; Figure 4 A flowchart illustrating the process of determining recommended content in a specific application scenario, as provided in this embodiment of the disclosure; Figure 5 A structural block diagram of an apparatus for determining recommended content provided in an embodiment of this disclosure; Figure 6 This is a schematic diagram of the structure of an electronic device suitable for performing a method for determining recommended content, as provided in an embodiment of this disclosure. Detailed Implementation

[0015] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding; these should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. It should be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0016] Furthermore, the acquisition, storage, use, processing, transportation, provision, and disclosure of any type of information involved in the technical solutions disclosed herein, such as user personal information (e.g., historical recommendation content that has been recommended to users as discussed later in this disclosure), comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0017] Figure 1 An exemplary system architecture 100 is shown, illustrating embodiments of methods, apparatuses, electronic devices, and computer-readable storage media for determining recommended content that can be applied according to this disclosure.

[0018] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0019] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various applications for enabling information communication between the terminal devices 101, 102, and 103 and server 105 can be installed. These applications include content filtering applications, content recommendation applications, and instant messaging applications.

[0020] Terminal devices 101, 102, and 103 and server 105 can be either hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with displays, including but not limited to smartphones, tablets, laptops, and desktop computers. When terminal devices 101, 102, and 103 are software, they can be installed in the aforementioned electronic devices, and can be implemented as multiple software programs or software modules, or as a single software program or software module; no specific limitation is made here. When server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When server 105 is software, it can be implemented as multiple software programs or software modules, or as a single software program or software module; no specific limitation is made here.

[0021] Server 105 can provide various services through its built-in applications. For example, it can provide services to rearrange, filter, and optimize candidate recommendation content to be recommended to users. Taking a recommendation content filtering application as an example, when running this application, server 105 can achieve the following effects: First, server 105 reads the historical tags of previously recommended content (e.g., for users using terminal devices 101, 102, and 103) from the evaluation window, as well as the percentage of historical recommendation content associated with those tags relative to the total number of historical recommendation content. Then... Based on the target proportion value of the target historical tag, adjustment parameters corresponding to the target historical tag are generated. Next, based on the adjustment parameters, the first recommendation reference value of the first candidate recommendation content associated with the target historical tag is updated to the updated first recommendation reference value. Finally, based on the updated first recommendation reference value and the second recommendation reference value corresponding to the second candidate recommendation content, target recommendation content for recommendation is determined from the first candidate recommendation content and the second candidate recommendation content (correspondingly, the server 105 can subsequently recommend and return these target recommendation contents to the terminal devices 101, 102, and 103 to realize the process of providing recommendation content to users).

[0022] It should be noted that the previously recommended historical content can be obtained from terminal devices 101, 102, and 103 via network 104, or it can be pre-stored locally on server 105 through various means. Therefore, when server 105 detects that this data is already stored locally (e.g., a record of historical recommended content pushed to users that is maintained locally), it can choose to retrieve this data directly from the local storage. In this case, the exemplary system architecture 100 may not include terminal devices 101, 102, and 103 and network 104.

[0023] Since managing recommended content (e.g., historical recommended content, candidate recommended content, etc.) and determining and pushing target recommended content often requires significant computing resources and capabilities, the methods for determining recommended content provided in the subsequent embodiments of this disclosure are generally executed by a server 105 with strong computing power and abundant computing resources. Correspondingly, the apparatus for determining recommended content is also generally located in the server 105. However, it should also be noted that when terminal devices 101, 102, and 103 also possess sufficient computing power and resources, they can also perform the aforementioned calculations performed by the server 105 through their installed recommended content filtering applications, thereby outputting the same results as the server 105. Especially when multiple terminal devices with different computing capabilities exist simultaneously, but the content recommendation filtering application determines that the terminal device has strong computing power and sufficient remaining computing resources, the terminal device can perform the aforementioned calculations, thereby appropriately reducing the computing pressure on server 105. Correspondingly, the device for determining the recommended content can also be located in terminal devices 101, 102, and 103. In this case, the exemplary system architecture 100 may also exclude server 105 and network 104.

[0024] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0025] Please refer to Figure 2 , Figure 2 A flowchart of a process for determining recommended content provided for embodiments of this disclosure includes process 200.

[0026] Process 200 specifically includes the following steps: Step 201: Read the historical tags of the historical recommended content that has been recommended in the evaluation window, and the percentage of the number of historical recommended content associated with the historical tags to the total number of all historical recommended content; In embodiments of this disclosure, this step is intended to be performed by the entity executing the method for determining recommended content (e.g., Figure 1 The server 105 shown first reads the historical recommended content that has been recommended to the user in an evaluation window, and then reads the historical tags associated with the historical recommended content, as well as the distribution of the number of these historical tags.

[0027] In practice, historical recommendation content can be combined with different scenarios, and it can be "content" in the form of text, images, audio, video, etc. that has been recommended to or has been recommended to users in the past.

[0028] The evaluation window can be pre-divided based on criteria such as the quantity and duration of the historical recommended content. For example, when using "quantity" as the criterion for dividing and determining the evaluation window, the starting point of the evaluation window can be determined, and then the "quantity" can be used as the length to actually describe and define the evaluation window.

[0029] For example, if the length is determined to be "8 historical recommended content", and the first historical recommended content recommended to the user is the starting point of the evaluation window, the historical recommended content included and associated in the evaluation window can be the first historical recommended content that has been recommended to the user, and the next 7 historical recommended content that are consecutively recommended to the user after that historical recommended content.

[0030] Accordingly, in this step, the executing entity can read the historical recommended content that has been recommended and is associated with an evaluation window, and read the historical tags involved in each of the historical recommended content.

[0031] In some embodiments, historical tags can be predetermined so that the implementing entity can "analyze" and "understand" the historical recommendations already made in the evaluation window based on the dimensions involved in these historical tags. For example, historical tags can be adaptively set in advance based on dimensions that need to be focused on, such as theme, author, style, etc. For example, historical tag 1 can be theme A, historical tag 2 can be theme B, historical tag 3 can be style C, and so on.

[0032] Accordingly, in this step, the implementing entity can determine the historical tags involved and matched by each of the historical recommendations by analyzing the content. For example, the historical tags matched by the recommendations can be determined based on their themes, content, and authors.

[0033] It should be understood that in this process, for the same historical recommended content, due to differences in the standards and methods of tagging, it may be associated with one or at least two historical tags simultaneously, and this disclosure is not intended to restrict this. For example, a historical recommended content may simultaneously match two historical tags: theme A and style C.

[0034] In this step, after the executing entity determines the hit historical tag, for a historical tag, the executing entity can divide the number of times it is hit (i.e., the number of historical recommended contents associated with the historical tag) by the total number of historical recommended contents in the evaluation window to determine the proportion value corresponding to the historical tag.

[0035] For ease of understanding, this "percentage value" can also be interpreted as the proportion of historical recommended content with that historical tag in the evaluation window out of the total number of historical recommended content in the evaluation window. For example, if an evaluation window is associated with 10 historical recommended content items, and the historical tag for topic A is matched by 7 historical recommended content items, then the percentage value for topic A can be 0.7.

[0036] Step 202: Based on the target percentage value of the target historical label, generate adjustment parameters corresponding to the target historical label; Based on step 201, this step aims to have the aforementioned executing entity generate adjustment parameters corresponding to a specific target historical tag, based on the percentage value determined in step 201 corresponding to the target historical tag (for ease of understanding, this corresponding percentage value can be described as "target percentage value" similar to the target historical tag).

[0037] For example, a reference value for the proportion can be predefined. If the proportion is less than or equal to the reference value, the executing entity can generate an adjustment parameter corresponding to the target historical tag to adjust its attention weight, making it easier for candidate recommended content with the target historical tag to be recommended and exposed (for example, the value of the adjustment parameter is preset, and the direction of the adjustment parameter is determined according to the comparison result so that it can be used from the perspective of increasing exposure). This allows the recommendation strategy to be adjusted in the future, making it easier for candidate recommended content with the target historical tag to be recommended.

[0038] For example, depending on the difference in recommendation strategies, in different embodiments, the adjustment parameter may be a parameter used to amplify the reference weight (e.g., in a scenario where recommendations are made based on weight coefficients) or to increase the recommendation reference value (e.g., in a scenario where the recommendation reference value is determined based on the similarity between historical tags and user preference tags, and the recommendation reference value is used for recommendations).

[0039] Similarly, if the percentage value is greater than the reference value, the executing entity can generate adjustment parameters to reduce the likelihood of candidate recommended content with the target historical tag being recommended and exposed. This adjustment parameter can then be used to adjust the recommendation strategy to make candidate recommended content with the target historical tag even less likely to be recommended. This will not be explained again here.

[0040] Additionally, it should be understood that, depending on the scenario, in some scenarios, adjustments can be made only in one direction. For example, only one option can be allowed to increase or decrease the exposure, rather than allowing both increases and decreases simultaneously. For instance, if the percentage value is less than or equal to the reference value, the adjustment parameter can be a specific value used to increase the exposure. However, if the percentage value is greater than the reference value, the adjustment parameter may be a value such as "0" that will not adjust the current recommendation strategy (for example, adding "0" to the existing recommendation reference value will not cause the recommendation reference value to change).

[0041] In practice, the implementing entity can use all the historical labels that appear in the evaluation window involved in step 201 as target historical labels to determine the adjustment parameters corresponding to each historical label, so as to make more comprehensive and specific adjustments to the subsequent recommendation strategy.

[0042] In some embodiments, depending on different scenarios and needs, the implementing entity may also choose to use a subset of historical tags as target historical tags to improve recommendation effectiveness and quality while saving computational resources. For example, by pre-setting which historical tags can be used as target historical tags, constraints can be placed on those historical tags that need attention, allowing the implementing entity to focus more on those historical tags that may be more important and require more attention.

[0043] Step 203: Based on the adjustment parameters, update the first recommendation reference value of the first candidate recommendation content associated with the target historical tag to the updated first recommendation reference value; In the embodiments of this disclosure, as discussed above, after the executing entity generates the adjustment parameters corresponding to the target historical tag based on the above step 202, the executing entity can update the first recommendation reference value of the first candidate recommendation content associated with the target historical tag to an updated first recommendation reference value, as discussed above. For example, the first recommendation reference value can be increased or decreased based on different directions to update it to an updated first recommendation reference value (for example, directly changing it by a value, such as adding or subtracting the adjustment parameter, or using the adjustment parameter to proportionally enlarge or shrink it).

[0044] Typically, the first candidate recommendation content and the second candidate recommendation content (which will be discussed below) together constitute a set of candidate recommendation content to be recommended to the user.

[0045] In practice, the set of candidate recommended content can be a collection of candidate recommended content that will be subsequently recommended to users, determined based on factors such as user preferences and push platform settings. In practice, candidate recommended content can have the same modality as historical recommended content; for example, they may both be videos. In some embodiments, their modalities may also differ; for example, historical recommended content may be audio, while the content to be recommended may be video (e.g., they can be referenced across modalities because they are semantically associated with some of the same historical tags).

[0046] In practice, we can first generate a recommendation reference value for the content based on the similarity between the content tags and the user's corresponding preference tags, and then select candidate recommended content from the content based on the recommendation reference value to form a set of candidate recommended content.

[0047] For example, content with a recommended reference value greater than or equal to the reference value threshold can be used as candidate recommended content to form a set of candidate recommended content. Alternatively, after sorting the recommended reference values ​​in descending order, a preset number of content can be selected as candidate recommended content to form a set of candidate recommended content.

[0048] Accordingly, the first candidate recommendation content in this step can be understood as candidate recommendation content in the candidate recommendation content set that is associated with at least one target historical tag (or, in other words, associated with the target historical tag), and the second candidate recommendation content below can be understood as candidate recommendation content in the candidate recommendation content set that is not associated with the target historical tag.

[0049] Step 204: Based on the updated first recommendation reference value and the second recommendation reference value corresponding to the second candidate recommendation content, determine the target recommendation content for recommendation from the first candidate recommendation content and the second candidate recommendation content.

[0050] In the embodiments of this disclosure, in this step, after updating the first recommendation reference value to an updated first recommendation reference value based on step 203 above, the executing entity can reuse the updated first recommendation reference value and the second recommendation reference value (i.e., the second recommendation reference value corresponding to the second candidate recommendation content that was not updated because it was not associated with the target historical tag) to determine the target recommendation content to be recommended to the user.

[0051] For example, the implementing entity can, according to a prior recommendation strategy, sort and update the first and second recommendation reference values ​​(e.g., in descending order), and based on the sorting result, select the top-ranked candidate recommendation content and the first preset number of candidate recommendations as the target recommendation content. Alternatively, the implementing entity can, as discussed above, select the first and second candidate recommendation content that are greater than or equal to the reference value threshold from the updated first and second recommendation reference values ​​as the target recommendation content.

[0052] In some embodiments, if multiple target historical tags exist, the executing entity can determine the adjustment parameters corresponding to each target historical tag in a similar manner, and then "aggregate" them to determine an overall adjustment parameter for updating the recommendation reference value (e.g., a first recommendation reference value). The executing entity can then use this overall adjustment parameter to refer to the adjustment parameters of at least two or more target historical tags, and adjust the overall push reference value (or the probability of being recommended and exposed) of the candidate recommendation content.

[0053] Subsequently, after determining the target recommendation content, the implementing entity can, in conjunction with the pre-configured recommendation timing, actually recommend the target recommendation content to the user (e.g., the terminal device used by the user) to achieve the purpose of recommending the target recommendation content to the user, which will not be repeated here.

[0054] The method for determining recommended content provided in this embodiment first reads the historical tags of previously recommended content in the evaluation window, as well as the percentage of historical recommended content associated with each historical tag relative to the total number of all historical recommended content. Then, based on the target percentage of the target historical tag, an adjustment parameter corresponding to the target historical tag is generated. Next, based on the adjustment parameter, the first recommendation reference value of the first candidate recommended content associated with the target historical tag is updated to an updated first recommendation reference value. Finally, based on the updated first recommendation reference value and the second recommendation reference value corresponding to the second candidate recommended content, the target recommended content for recommendation is determined from the first and second candidate recommended content. Therefore, the recommendation weight of future recommended content can be dynamically adjusted based on already recommended content, thereby achieving a balance of recommended content diversity, avoiding excessive concentration and duplication of recommended content, optimizing the overall content distribution of the recommendation system, and improving the quality of recommended content determination.

[0055] In some embodiments, as discussed above, depending on different concerns, it is also possible to select which specific historical tags need to be considered in the configuration evaluation window in order to balance the computing resource usage of the execution entity.

[0056] Accordingly, in such cases, a corresponding list of historical tags can be set for a specific evaluation window, and this list of historical tags can be used to constrain the historical tags that need to be followed and read in the evaluation window.

[0057] Accordingly, in such a case, when the executing entity is reading the historical tags of the historical recommended content that has already been recommended in the evaluation window, and the percentage of the number of historical recommended content associated with the historical tags to the total number of all historical recommended content, it may alternatively choose to read the historical tags of the historical recommended content that has already been recommended based on the historical tag list corresponding to the evaluation window, and the percentage of the number of historical recommended content associated with the historical tags to the total number of all historical recommended content.

[0058] Accordingly, the "historical tags" retrieved here should be included in the list of historical tags to be monitored. This allows the implementing entity to focus more on potentially more important historical tags, enabling it to concentrate its attention and computing resources on those tags.

[0059] Furthermore, this approach allows for the differential configuration of multiple evaluation windows (e.g., evaluation windows of different lengths), and the use of different evaluation windows to differentiate the analysis of previously recommended historical content and the distribution of associated historical tags from different dimensional strategies, thereby enhancing the execution entity's ability to mine "historical distribution".

[0060] In some embodiments, considering the limited amount of historical recommendations involved in the evaluation window, to avoid overfitting due to excessive analysis of sparse data, the number of historical tags included in the historical tag list can be positively correlated with the length of the evaluation window. This allows the implementing entity to dynamically adjust the mining strategy based on the specific number of historical recommendations associated with the evaluation window, ensuring mining quality while balancing computational usage and avoiding waste of computing resources.

[0061] To enable a more comprehensive analysis and understanding of historical recommendations already presented to users, and to avoid situations where short-term uneven distribution occurs before the evaluation window is reached, preventing timely discovery and analysis, or where the evaluation window is set too short to provide feedback on long-term recommendations, some embodiments may utilize multiple evaluation windows of different lengths. This approach balances long-term and short-term analysis needs, allowing for a holistic evaluation and analysis from multiple length dimensions to determine whether recommended content exhibits clustering or duplication at historical tags, and whether adjustments are needed to subsequent candidate recommendations associated with historical tags (e.g., increasing or decreasing their likelihood of being recommended).

[0062] Accordingly, in such a case, the executing entity can actually refer to the target percentage values ​​of the target historical labels in at least two different evaluation windows to generate the adjustment parameters corresponding to the target historical labels during the process of generating adjustment parameters.

[0063] In step 202 above, if at least two evaluation windows exist, the executing entity can further select a target historical label and generate an adjustment parameter corresponding to the target historical label based on the target percentage value of the target historical label in each of the at least two evaluation windows of different lengths. For example, the executing entity can determine the corresponding adjustment parameter using the target percentage value determined in each evaluation window as discussed above, and then integrate them into a single adjustment parameter through summation, averaging, weighted averaging, etc., so that the adjustment parameter can better take into account the situation of different evaluation windows.

[0064] In some embodiments, if at least two evaluation windows are used to determine the adjustment parameters, the starting points of the different evaluation windows used and selected by the executing entity can be the same (e.g., all starting from the first recommended historical content). That is, the starting points of each evaluation window are the same, but their lengths are different. This ensures the continuity of the mining logic between them, avoids analysis errors caused by window interruptions or jumps, and improves the mining quality.

[0065] In some embodiments, during the process of determining adjustment parameters using at least two cross-evaluation windows, in order to make adjustments in a more granular manner, in addition to the above-mentioned determination based solely on "direction" (e.g., based on whether it is less than or equal to the percentage reference value, or greater than the percentage reference value, to determine whether to increase, decrease, or not to adjust), the executing entity can also determine the specific value of the adjustment parameter based on the specific circumstances of the target percentage value, so as to manage the magnitude and intensity of the adjustment in a more granular manner, so that the subsequent "first recommended reference value" can be adjusted and updated more effectively and with higher quality.

[0066] In some embodiments, as discussed above, the implementing entity can simultaneously utilize the target percentage values ​​from at least two evaluation windows to determine the adjustment parameters for the target historical label. Therefore, considering that such a process is more complex than using only one evaluation window, and that using only one evaluation window is merely a partial implementation of using multiple evaluation windows, for the sake of brevity, this scenario will be primarily discussed. For easier understanding, please refer to... Figure 3 Let's explain them together.

[0067] In addition, for a similar purpose of ease of understanding, Figure 3 Similarly, the example only focuses on the single direction of wanting to reduce exposure.

[0068] Figure 3 A flowchart of a process for generating adjustment parameters is provided for an embodiment of this disclosure, including process 300. For example, process 300 may be an alternative or alternative implementation of step 202 in process 200 described above, in the presence of at least two evaluation windows.

[0069] Process 300 specifically includes the following steps: Step 301: In each evaluation window, if the target percentage value in response to the target historical label is greater than the percentage value threshold corresponding to that evaluation window, an overflow percentage value is generated based on the difference between the target percentage value and the percentage value threshold. Specifically, a corresponding percentage threshold can be set for each evaluation window (in fact, this "percentage threshold" can be equivalent to the "percentage reference value" discussed above, so that the "percentage threshold" can be used to quantify whether the target historical label appears "too much" or "too frequently"). Then, in each evaluation window used, the executing entity detects the target percentage value of the target historical label and checks whether the target percentage value is greater than the percentage threshold corresponding to the evaluation window.

[0070] If it is greater than this, the executing entity can respond by generating an overflow percentage based on the difference between the target percentage and the percentage threshold (e.g., subtracting the percentage threshold from the target percentage).

[0071] To facilitate understanding, a specific scenario can be used as an example, such as a specific evaluation window. w The historical tags involved v corresponding percentage value It can then be calculated using the following formula (1): (1).

[0072] in, It can be understood as an evaluation window w Chinese and historical tags v The number of associated historical recommended content 。

[0073] Accordingly, if the percentage value Larger than the evaluation window w percentage threshold ， The executing entity can then respond to this, based on ( Generate overflow percentage value .

[0074] Accordingly, since process 300 is executed in a unidirectional manner, such as "reducing the exposure," this step... In practice, it can be further expressed as follows: (2).

[0075] Next, as discussed above, since process 300 targets "at least two evaluation windows," the executing entity can continue to generate adjustment parameters corresponding to the target historical label based on each overflow percentage value. For example, as discussed above, the executing entity can merge the overflow percentage values ​​in step 302, for example, by summing the individual overflow percentage values, to generate a total overflow percentage value. Then, the executing entity can use this total overflow percentage value, either directly or by multiplying it by a pre-defined coefficient aligned with the dimensions of the (first) recommended reference value, as an adjustment parameter to adjust the first recommended reference value.

[0076] For example, in some embodiments, the executing entity may also choose to utilize information entropy to determine the above. : For example, information entropy It can be determined by the following formula (3): (3).

[0077] in, V For the evaluation window w All historical tags in China 。

[0078] Accordingly, in such a case, the implementing entity can base its actions on... The expected decrease in information entropy, as mentioned above 。

[0079] It should be understood that, in the case of using only one evaluation window, the implementing entity can also directly use the overflow percentage value as the "adjustment coefficient" in a similar way, after aligning the dimensions. This will not be repeated here.

[0080] Step 302: Merge the various overflow percentage values ​​to generate the total overflow percentage value; Specifically, as discussed above, the implementing entity can use the following formula (4) to calculate the percentage of each overflow. Integration to generate historical tags v Total overflow percentage : (4).

[0081] in, W It is the set of evaluation windows used.

[0082] In some embodiments, because the lengths of the evaluation windows differ, the reference value provided by the target percentage values ​​provided by the evaluation windows may differ. Therefore, in order to better balance this difference and make more granular use of those evaluation windows that may have higher reference value, it is also possible to configure corresponding reference coefficients for the evaluation windows.

[0083] In some embodiments, the level of the reference coefficient can be positively correlated with, for example, the length of the evaluation window, so that the implementing entity can focus more on the target percentage values ​​provided by those evaluation windows with longer lengths.

[0084] For example, in some embodiments, if it is desired that the implementing entity focus more on "short-term" situations, the value of the reference coefficient can be negatively correlated with the length of the evaluation window. Similarly, for those "high-value evaluation windows," a specific reference coefficient can be set accordingly to stabilize their reference weight, ensuring that they are "valued" in all situations.

[0085] Accordingly, in this case, compared to the above formula (4), the total overflow ratio is... The generation process can be further represented by the following formula (5) to incorporate... Corresponding reference coefficient : (5).

[0086] In some embodiments, in order to make the "adjustment parameter" smoother and avoid the update process of the first recommended reference value becoming staged due to the direct use of discrete values, which would cause the recommended content to tear or jitter, the executing entity can, after obtaining the total overflow percentage value based on step 302, actually generate the adjustment parameter corresponding to the target historical tag based on the exponential function associated with the total overflow percentage value, instead of directly using the total overflow percentage value as the adjustment parameter.

[0087] Therefore, in such a case, process 300 may also include step 303.

[0088] Step 303: Generate adjustment parameters corresponding to the target historical label based on the exponential function associated with the total overflow percentage.

[0089] Specifically, for those associated with historical tags v Adjustment parameters It can be determined by the following formula (6): (6).

[0090] Here, β is a hyperparameter used to manage the magnitude of change.

[0091] In some embodiments, as discussed above, if at least two target historical tags are involved, the executing entity obtains each historical tag separately. v Each of the adjustment parameters Then, the implementing entity can use the following formula (7) to evaluate the candidate recommendation content. i First recommended reference value Adjustments are made (for example, in such cases, the adjustment parameters can be used in a "direct" or "scaled-down" manner) to adjust them to the updated first recommendation reference value ultimately used for sorting and determining the target recommended content. : (7).

[0092] in, Recommended content for candidates i The historical tags that were hit v The combination of .

[0093] Based on any of the above embodiments, if it is permissible to use and divide at least two evaluation windows, then at least two evaluation windows of different lengths can be divided based on the Fibonacci sequence. For example, based on the corresponding quantity, with Five evaluation windows were determined so that the length of the evaluation windows could better conform to the laws of perception and improve the quality of evaluation window division.

[0094] Based on any of the above embodiments, the executing entity can use the tags used when recommending historical content as historical tags. That is, the aforementioned historical content is recommended to the user based on historical tags. Therefore, this adjustment, mining, and rearrangement process is implemented based on the recommendation strategies and logic used when recommending historical content. This avoids serious damage to the original recommendation logic and system anomalies caused by significant differences between the determination logic and standards of historical tags and the original logic. It ensures the stability of the original recommendation system while improving the quality of recommendations.

[0095] To enhance understanding, this disclosure also provides a specific implementation scheme based on a particular application scenario. Please refer to the example below. Figure 4 The process shown is 400. Figure 4This is a flowchart illustrating the process of determining recommended content in a specific application scenario, as provided in an embodiment of this disclosure.

[0096] For example, this process 400 can also be performed by the aforementioned "server 105" ( Figure 4 (Not shown again in the text) is implemented as the "executing entity".

[0097] In process 400, the "content" in the exemplary historical recommended content and candidate recommended content can both be "video".

[0098] Based on this, the historical recommended content in process 400 may include historical recommended content 411, historical recommended content 412, historical recommended content 413, historical recommended content 414, historical recommended content 415... historical recommended content 41N, where N is a positive integer.

[0099] Accordingly, process 400 may include two evaluation windows, for example, evaluation window 421 (e.g., which may include historical recommended content 411, historical recommended content 412, historical recommended content 413) and evaluation window 421 (e.g., which may include historical recommended content 411, historical recommended content 412, historical recommended content 413, historical recommended content 414, historical recommended content 415... historical recommended content 41N).

[0100] It should be understood that the number of evaluation windows, the specific historical recommendations associated with each evaluation window, and the number of historical recommendations in process 400 are merely illustrative examples for ease of understanding and are not intended to impose any limitations.

[0101] In this case, server 105 can first execute S401 to read the historical tags involved in evaluation windows 421 and 422 respectively, as well as the proportion of each historical tag in the corresponding evaluation window.

[0102] For example, both evaluation windows 421 and 422 can be read by server 105 to retrieve historical tags 431, 432 and 433.

[0103] For evaluation window 421, the percentage of historical label 431 is 441, the percentage of historical label 432 is 442, and the percentage of historical label 433 is 443.

[0104] For evaluation window 422, the percentage of historical label 431 is 444, the percentage of historical label 432 is 445, and the percentage of historical label 433 is 446.

[0105] Then, server 105 can continue to execute S402 to generate adjustment parameters corresponding to the target historical label based on the target percentage value of the target historical label.

[0106] For ease of understanding, in process 400, the "target history label" can be defined as history label 431 (of course, in different scenarios, the target history label can be one or more of history label 431, history label 432, and history label 433, which will not be elaborated here).

[0107] In this case, server 105 can generate adjustment parameters 451 for historical label 431 by combining the proportion values ​​441 and 444 corresponding to historical label 431 (for example, by generating adjustment parameters corresponding to the target historical label based on each overflow proportion value, and generating adjustment parameters corresponding to the target historical label based on each overflow proportion value and the reference coefficient of the evaluation window corresponding to the overflow proportion value, which will not be repeated here).

[0108] For example, the subsequent candidate recommended content can specifically be candidate recommended content 461 and candidate recommended content 462. Correspondingly, it can also be exemplarily assumed that in the current state, because the recommendation reference value 471 corresponding to candidate recommended content 461 is greater than the recommendation reference value 472 corresponding to candidate recommended content 462, the server 105 intends to recommend candidate recommended content 461 as the target recommended content to the user (e.g., send it to the terminal devices 101, 102, and 103 used by the user) if it expects to provide only one target recommended content.

[0109] In this case, since the candidate recommended content 461 is associated with the historical tag 431, the server 105 can execute S403 to update the recommended reference value 471 of the candidate recommended content 461 associated with the historical tag 431 to the updated first recommended reference value 471' based on the adjustment parameter 451.

[0110] Then, server 105 executes S404 to determine the target recommended content for recommendation by comparing the updated recommended reference value 471' (which is no longer the recommended reference value 471) and the recommended reference value 472 corresponding to the candidate recommended content 462.

[0111] For example, if updates and adjustments are made to reduce exposure and make it less likely to be recommended, resulting in the updated recommendation reference value 471' being lower than the recommendation reference value 472, then in this case, server 105 can use the candidate recommendation content 462 corresponding to the recommendation reference value 472 as the "target recommendation content". And when it is necessary to actually provide and recommend this "target recommendation content" in the future, the server 105 will recommend and provide the candidate recommendation content 462 to the user.

[0112] Further reference Figure 5 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of an apparatus for determining recommended content, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0113] like Figure 5 As shown, the apparatus 500 for determining recommended content in this embodiment may include: a tag proportion reading unit 501, an adjustment parameter generation unit 502, a reference value updating unit 503, and a recommended content determination unit 504. Specifically, the tag proportion reading unit 501 is configured to read historical tags of historical recommended content already recommended in the evaluation window, and the proportion of the number of historical recommended content associated with each historical tag to the total number of all historical recommended content; the adjustment parameter generation unit 502 is configured to generate adjustment parameters corresponding to the target historical tag based on the target proportion value of the target historical tag; the reference value updating unit 503 is configured to update the first recommendation reference value of the first candidate recommended content associated with the target historical tag to an updated first recommendation reference value based on the adjustment parameters; and the recommended content determination unit 504 is configured to determine the target recommended content for recommendation from the first candidate recommended content and the second candidate recommended content based on the updated first recommendation reference value and the second recommendation reference value corresponding to the second candidate recommended content.

[0114] In this embodiment, the specific processing and technical effects of the tag proportion reading unit 501, the adjustment parameter generation unit 502, the reference value update unit 503, and the recommended content determination unit 504 in the device 500 for determining recommended content can be referred to respectively. Figure 2 The relevant descriptions of steps 201-204 in the corresponding embodiments will not be repeated here.

[0115] In some optional implementations of this embodiment, the adjustment parameter generation unit 502 is further configured to generate adjustment parameters corresponding to the target historical label based on the target proportion values ​​of the target historical label in at least two evaluation windows of different lengths, wherein the different evaluation windows have the same starting point.

[0116] In some optional implementations of this embodiment, the adjustment parameter generation unit 502 includes: an overflow percentage value generation subunit, configured to generate an overflow percentage value based on the difference between the target percentage value and the percentage value threshold in each evaluation window, in response to the target percentage value of the target historical label being greater than the percentage value threshold corresponding to the evaluation window; and an adjustment parameter generation subunit, configured to generate adjustment parameters corresponding to the target historical label based on each overflow percentage value.

[0117] In some optional implementations of this embodiment, the adjustment parameter generation subunit is further configured to generate adjustment parameters corresponding to the target historical label based on each overflow percentage value and the reference coefficient of the evaluation window corresponding to the overflow percentage value.

[0118] In some optional implementations of this embodiment, the value of the reference coefficient is positively correlated with the length of the evaluation window.

[0119] In some optional implementations of this embodiment, the adjustment parameter generation subunit includes: a total overflow percentage value calculation module, configured to merge various overflow percentage values ​​to generate a total overflow percentage value; and an adjustment parameter generation module, configured to generate adjustment parameters corresponding to the target historical label based on an exponential function associated with the total overflow percentage value.

[0120] In some optional implementations of this embodiment, the apparatus 400 further includes an evaluation window division unit configured to divide at least two evaluation windows of different lengths based on the Fibonacci sequence, starting from the same point.

[0121] In some optional implementations of this embodiment, the tag proportion reading unit 401 is further configured to read the historical tags of the historical recommended content that has been recommended, and the proportion of the number of historical recommended content associated with the historical tags to the total number of all historical recommended content, based on the historical tag list of the following window. The historical tags are included in the historical tag list.

[0122] In some optional implementations of this embodiment, the number of historical tags included in the historical tag list is positively correlated with the length of the evaluation window.

[0123] In some optional implementations of this embodiment, historical recommended content is recommended to the user based on historical tags.

[0124] This embodiment exists as a device embodiment corresponding to the above method embodiment. The device for determining recommended content provided in this embodiment can dynamically adjust the recommendation weight of future recommended content based on the already recommended content, thereby achieving a balance of diversity in recommended content, avoiding excessive concentration and repetition of recommended content, optimizing the overall content distribution of the recommendation system, and improving the quality of recommended content determination.

[0125] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0126] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0127] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded into random access memory (RAM) 603 from storage unit 608. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0128] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0129] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the method for determining recommended content. For example, in some embodiments, the method for determining recommended content may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the method for determining recommended content described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the method for determining recommended content by any other suitable means (e.g., by means of firmware).

[0130] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0131] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0132] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0133] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0134] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0135] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is established by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, also known as cloud computing servers or cloud hosts, which are hosting products within the cloud computing service ecosystem to address the management difficulties and weak business scalability inherent in traditional physical hosts and Virtual Private Servers (VPS) services. Servers can also be categorized as distributed system servers or servers incorporating blockchain technology.

[0136] According to the technical solution of this disclosure, the recommendation weight of future recommended content can be dynamically adjusted based on the already recommended content, thereby achieving a balance of diversity in recommended content, avoiding excessive concentration and duplication of recommended content, optimizing the overall content distribution of the recommendation system, and improving the quality of recommended content determination.

[0137] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution provided in this disclosure can be achieved, and this is not limited herein.

[0138] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for determining recommended content, comprising: Read the historical tags of the historical recommended content that has been recommended in the evaluation window, and the percentage of the number of the historical recommended content associated with the historical tags to the total number of all historical recommended content; Based on the target percentage value of the target historical label, generate adjustment parameters corresponding to the target historical label; Based on the adjustment parameters, the first recommendation reference value of the first candidate recommendation content associated with the target historical tag is updated to the updated first recommendation reference value; Based on the updated first recommendation reference value and the second recommendation reference value corresponding to the second candidate recommendation content, target recommendation content for recommendation is determined from the first candidate recommendation content and the second candidate recommendation content.

2. The method according to claim 1, wherein, The process of generating adjustment parameters corresponding to the target historical tags based on the target percentage value includes: For a target historical label, based on the target percentage value of the target historical label in at least two evaluation windows of different lengths, an adjustment parameter corresponding to the target historical label is generated, wherein the different evaluation windows have the same starting point.

3. The method according to claim 2, wherein, The step of generating adjustment parameters corresponding to the target historical label, based on the target percentage value of the target historical label in at least two evaluation windows of different lengths, includes: In each evaluation window, in response to the target percentage value of the target historical label being greater than the percentage value threshold corresponding to the evaluation window, an overflow percentage value is generated based on the difference between the target percentage value and the percentage value threshold. Based on each of the overflow percentage values, adjustment parameters corresponding to the target historical label are generated.

4. The method according to claim 3, wherein, The step of generating adjustment parameters corresponding to the target historical tag based on each of the overflow percentage values ​​includes: Based on each of the overflow percentage values ​​and the reference coefficient of the evaluation window corresponding to the overflow percentage value, adjustment parameters corresponding to the target historical label are generated.

5. The method according to claim 4, wherein, The value of the reference coefficient is positively correlated with the length of the evaluation window.

6. The method according to claim 3, wherein, The step of generating adjustment parameters corresponding to the target historical tag based on each of the overflow percentage values ​​includes: Combine the various overflow percentage values ​​to generate a total overflow percentage value; An adjustment parameter corresponding to the target historical label is generated based on an exponential function associated with the total overflow percentage.

7. The method according to claim 2, further comprising: Starting from the same point, at least two evaluation windows of different lengths are divided based on the Fibonacci sequence.

8. The method according to claim 1, wherein, The historical tags of the historical recommended content already recommended in the reading evaluation window, and the percentage of the number of historical recommended content associated with the historical tags relative to the total number of all historical recommended content, include: Based on the historical tag list corresponding to the evaluation window, the historical tags of the recommended content that has been recommended are read, as well as the percentage of the number of the recommended content associated with the historical tags relative to the total number of all the recommended content. The historical tags are included in the historical tag list.

9. The method according to claim 7, wherein, The number of historical tags included in the historical tag list is positively correlated with the length of the evaluation window.

10. The method according to any one of claims 1-9, wherein, The historical recommended content is recommended to the user based on the historical tags.

11. An apparatus for determining recommended content, comprising: The tag percentage reading unit is configured to read the historical tags of the historical recommended content that has been recommended in the evaluation window, and the percentage of the number of the historical recommended content associated with the historical tags to the total number of all the historical recommended content. The parameter generation unit is configured to generate adjustment parameters corresponding to the target historical label based on the target proportion value of the target historical label. The reference value update unit is configured to update the first recommendation reference value of the first candidate recommendation content associated with the target historical tag to an updated first recommendation reference value based on the adjustment parameters. The recommended content determination unit is configured to determine target recommended content for recommendation from the first candidate recommended content and the second candidate recommended content based on the updated first recommended reference value and the second recommended reference value corresponding to the second candidate recommended content.

12. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method for determining recommended content as described in any one of claims 1-10.

13. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the method of determining recommended content according to any one of claims 1-10.

14. A computer program product comprising a computer program that, when executed by a processor, implements the method for determining recommended content according to any one of claims 1-10.

Images