Resource recommendation method and apparatus, electronic device, and storage medium

Abstract

The present disclosure provides a resource recommendation method and device, electronic equipment and storage medium, relates to the field of artificial intelligence and big data technology, and particularly relates to the field of intelligent recommendation technology. The specific implementation scheme is: obtaining basic attribute information of a target object, and determining a target object set to which the target object belongs according to the basic attribute information; obtaining an interest confidence of the target object set in at least one interest dimension, wherein the interest confidence of the target object set in each interest dimension is determined according to an effective click rate of the target object set in the interest dimension and a global average effective click rate; obtaining a bias-eliminating click rate of each candidate resource in a plurality of candidate resources, wherein the bias-eliminating click rate is determined according to a click rate relative ranking of a body of the candidate resource in a plurality of bodies; and determining a target resource from the plurality of candidate resources according to the interest confidence and the bias-eliminating click rate.

Description

Technical Field

[0001] This disclosure relates to the fields of artificial intelligence and big data technology, and more particularly to the field of intelligent recommendation technology. More specifically, this disclosure provides a resource recommendation method, apparatus, electronic device, storage medium, and computer program product. Background Technology

[0002] With the continuous development of computer and internet technologies, the content resources available to users have exploded. Traditional recommendation methods can no longer meet users' personalized needs. Current personalized recommendation technologies usually model user interests and make personalized recommendations based on users' historical behavior. However, current recommendation schemes still have shortcomings in cold start user interest identification, accuracy of user profiles, and efficiency of matching content with user interests. Summary of the Invention

[0003] This disclosure provides a resource recommendation method, apparatus, electronic device, storage medium, and computer program product.

[0004] According to the first aspect, a resource recommendation method is provided, the method comprising: obtaining basic attribute information of a target object, and determining the target object set to which the target object belongs based on the basic attribute information; obtaining the interest confidence of the target object set in at least one interest dimension, wherein the interest confidence of the target object set in each interest dimension is determined based on the effective click-through rate of the target object set in that interest dimension and the global average effective click-through rate; obtaining the unbiased click-through rate of each candidate resource among multiple candidate resources, wherein the unbiased click-through rate is determined based on the relative ranking of the click-through rates of the genre to which the candidate resource belongs among multiple genres; and determining the target resource from the multiple candidate resources based on the interest confidence and the unbiased click-through rate.

[0005] According to the second aspect, a resource recommendation device is provided, comprising: a target object set determination module, configured to acquire basic attribute information of target objects and determine the target object set to which the target objects belong based on the basic attribute information; a first acquisition module, configured to acquire the interest confidence of the target object set in at least one interest dimension, wherein the interest confidence of the target object set in each interest dimension is determined based on the effective click-through rate of the target object set in that interest dimension and the global average effective click-through rate; a second acquisition module, configured to acquire the unbiased click-through rate of each candidate resource among multiple candidate resources, wherein the unbiased click-through rate is determined based on the relative ranking of the click-through rates of the genre to which the candidate resource belongs among multiple genres; and a target resource determination module, configured to determine the target resource from the multiple candidate resources based on the interest confidence and the unbiased click-through rate.

[0006] According to a third aspect, an electronic device is provided, 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 perform a method provided according to the present disclosure.

[0007] According to a fourth aspect, a non-transitory computer-readable storage medium is provided that stores computer instructions for causing a computer to perform the methods provided in this disclosure.

[0008] According to a fifth aspect, a computer program product is provided, comprising a computer program stored on at least one of a readable storage medium and an electronic device, wherein the computer program, when executed by a processor, implements the method provided in this disclosure.

[0009] 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

[0010] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0011] Figure 1 This is an exemplary system architecture diagram of a resource recommendation method and apparatus applicable according to an embodiment of the present disclosure;

[0012] Figure 2 This is a flowchart of a resource recommendation method according to an embodiment of the present disclosure;

[0013] Figure 3 This is a flowchart of a method for constructing an interest dictionary according to an embodiment of the present disclosure;

[0014] Figure 4 This is a flowchart of a resource recommendation method according to another embodiment of the present disclosure;

[0015] Figure 5 This is a block diagram of a resource recommendation apparatus according to an embodiment of the present disclosure; and

[0016] Figure 6 This is a block diagram of an electronic device for a resource recommendation method according to an embodiment of the present disclosure. Detailed Implementation

[0017] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and 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.

[0018] Currently, personalized recommendations in content recommendation scenarios typically fall into three categories: recommendations based on user behavior history, recommendations based on collaborative filtering, and recommendations based on content. Recommendations based on user behavior history involve extracting content and features that the user is interested in by analyzing their clicks, favorites, comments, etc., building a user interest profile, and then recommending content based on this profile. Recommendations based on collaborative filtering involve recommending content that the user might be interested in based on the correlation between similar user behaviors or similar items. Recommendations based on content involve analyzing the features of the recommended content, matching them with the user's historical interests, and determining the recommended content based on the matching results.

[0019] However, the methods mentioned above suffer from problems such as cold start, low efficiency in interest matching, insufficient interest mining, and difficulty in interest transfer. The cold start problem arises because new users lack sufficient behavioral data, leading to inaccurate interest profiles and less precise recommendations. Low efficiency in matching content with user interests stems from the latency inherent in real-time calculations across large datasets, impacting user experience. Insufficient interest mining means current methods struggle to capture latent user interests, resulting in insufficient diversity in recommended content. Difficult interest transfer refers to the challenge of effectively migrating and applying user interest data across different platforms.

[0020] The collection, storage, use, processing, transmission, provision, and disclosure of any type of information, such as user personal information, in this technical solution comply with relevant laws and regulations and do not violate public order and good morals.

[0021] In the technical solution disclosed herein, the user's authorization or consent is obtained before acquiring or collecting the user's personal information.

[0022] Figure 1 This is a schematic diagram of an exemplary system architecture for applying a resource recommendation method and apparatus according to an embodiment of this disclosure. It should be noted that... Figure 1 The examples shown are merely examples of system architectures that can be applied to the embodiments of this disclosure, in order to help those skilled in the art understand the technical content of this disclosure, but do not mean that the embodiments of this disclosure cannot be used in other devices, systems, environments or scenarios.

[0023] like Figure 1As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, and 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the terminal devices 101, 102, and 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0024] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Terminal devices 101, 102, and 103 can be various electronic devices, including but not limited to smartphones, tablets, laptops, etc.

[0025] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using terminal devices 101, 102, and 103 (for example only). The backend management server can analyze and process data such as received user requests, and feed back the processing results (determining the target resource from multiple candidate resources based on interest confidence and debiased click-through rate) to the terminal devices.

[0026] The resource recommendation method provided in this disclosure can generally be executed by server 105. Accordingly, the resource recommendation device provided in this disclosure can generally be located in server 105.

[0027] Figure 2 This is a flowchart of a resource recommendation method according to an embodiment of the present disclosure.

[0028] like Figure 2 As shown, the resource recommendation method 200 includes operations S210 to S240.

[0029] In operation S210, the basic attribute information of the target object is obtained, and the target object set to which the target object belongs is determined based on the basic attribute information.

[0030] In this embodiment of the disclosure, the basic attribute information of the target object can be obtained directly from the registration information of the target object, such as the gender, age, region and other information filled in by the user during the registration process. The user is aware of and agrees to the acquisition of this information.

[0031] A target object set can refer to a collection of objects that share the same or similar basic attribute information. Specifically, based on the basic attribute information of different objects, different objects can be divided into multiple historical object sets. For example, objects can be users on a network platform, and users with the same age, gender, and region can be grouped into the same historical object set. After obtaining the basic attribute information of the target object, it can be matched with the common attribute information of each historical object set to determine the target object set to which the target object belongs from multiple historical object sets.

[0032] In operation S220, obtain the interest confidence of the target object set in at least one interest dimension.

[0033] In this embodiment of the disclosure, the confidence level of the target object set in each interest dimension is determined based on the effective click-through rate of the target object set in that interest dimension and the global average effective click-through rate.

[0034] In this embodiment of the disclosure, the interest dimension can refer to the content direction or category that a user or user group is interested in, and can be a hierarchical dimension. For example, the interest points of the first-level interest dimension may be entertainment, sports, food, education, etc. The interest points of the second-level interest dimension, taking education as an example, may be subject education. The interest points of the third-level interest dimension, taking subject education as an example, may be homework tutoring, exams, primary school education, etc.

[0035] Interest confidence can be used to measure whether the preferences of a target set on a certain interest dimension are real and credible. Effective click-through rate can be the click-through rate of a target object on a certain interest dimension after invalid behavior is removed. Global average effective click-through rate is the average effective click-through rate of all objects on all interest dimensions.

[0036] In operation S230, the click-through rate of each candidate resource among multiple candidate resources is obtained.

[0037] In this embodiment of the disclosure, the bias-free click-through rate is determined based on the relative ranking of the click-through rates of the candidate resource's genre among multiple genres.

[0038] In this embodiment of the disclosure, candidate resources can refer to various types of content, goods, services, etc., used for recommendation or display. Genre can refer to the content type of candidate resources, such as short videos, long videos, text and images, and dynamic content. Debiased click-through rate can refer to the click-through rate after debiasing processing. Debiasing processing can refer to eliminating the influence of external factors such as genre, location, and traffic environment on the click-through rate. The debiased click-through rate of each candidate resource is calculated based on the relative ranking of the click-through rate of the genre to which the candidate resource belongs among all genres.

[0039] In operation S240, the target resource is determined from multiple candidate resources based on interest confidence and debiased click-through rate.

[0040] In this embodiment of the disclosure, after obtaining the debiased click-through rate of each candidate resource and the interest confidence of the target object set for each interest dimension, the candidate resources can be evaluated or matched based on the interest confidence and debiased click-through rate to obtain the processing results (e.g., the evaluation value or matching degree of each candidate resource). The target resource is then determined from multiple candidate resources based on the processing results (e.g., sorting according to the evaluation value or matching degree and selecting the candidate resource with the relatively higher ranking as the target resource, or selecting the candidate resource whose evaluation value or matching degree meets the threshold as the target resource).

[0041] This disclosure embodiment can accurately determine the true interest preferences of the target object set by using interest confidence scores of different dimensions. Based on more accurate interest preferences and debiased click-through rates, target resources can be determined, ensuring the accuracy and effectiveness of resource recommendations.

[0042] According to embodiments of this disclosure, the resource recommendation method further includes: determining the effective click-through rate of the target object set in each of the multiple different interest dimensions based on historical behavioral data of each object in the target object set in multiple different interest dimensions; for each interest dimension, determining the interest confidence of the target object set in the current interest dimension based on the effective click-through rate of the target object set in the current interest dimension and the global average effective click-through rate of multiple historical object sets in multiple different interest dimensions, wherein the target object set is any object set in the multiple historical object sets; and storing the interest confidence of the target object set in at least one dimension to an interest dictionary in response to the target object set having an interest confidence greater than a threshold.

[0043] In this embodiment of the disclosure, the historical behavior data of each object in the target object set on multiple different interest dimensions can be quantified to determine the effective click-through rate of the target object set on each of the multiple different interest dimensions. For example, the ratio of the effective click volume (display duration greater than the threshold) to the display volume of the target object on the interest dimension can be used as the effective click-through rate. Then, the effective click-through rate is compared with the global average effective click-through rate of multiple historical object sets on multiple different interest dimensions to determine the interest confidence of the target object set on the current interest dimension.

[0044] Specifically, the confidence level of the target object set for different interest dimensions can be calculated according to formula (1), which can include the confidence level of each interest point in each interest dimension:

[0045] Formula (1)

[0046] in, Represents the i-th set of objects. Represents the j-th point of interest. This represents the confidence level of the interest of the i-th object set at the j-th point of interest. This represents the effective click-through rate of the i-th object set at the j-th point of interest. This represents the average effective click-through rate of all historical object sets across all points of interest.

[0047] After determining the interest confidence level, the interest points with an interest confidence level greater than the threshold can be combined with the identifier of the target object set and stored in the interest dictionary. This indicates that the target object set has a high confidence level in that interest point (higher than the global average level). For example, the threshold can be set to 1.

[0048] For example, the interest dimension can include first-level categories, second-level categories and interest tags. Interest points can include education (first-level category), subject education (second-level category) and homework tutoring (interest tag). Then, the interest confidence of the target object set for education, interest confidence for subject education and interest confidence for homework tutoring can be calculated by formula (1). If all three values ​​meet the threshold, the three interest confidences can be associated with the identifier of the target object set and stored in the interest dictionary.

[0049] This disclosure embodiment calculates the effective click-through rate based on the historical behavioral data of the target object set across various interest dimensions, and then combines the global average effective click-through rate to obtain the interest confidence level that can truly reflect the group's interest preferences, thereby achieving accurate mining of the target object set's interest preferences.

[0050] According to embodiments of this disclosure, determining the effective click-through rate of a target object set across multiple different interest dimensions includes: acquiring historical behavior data of each object in the target object set for historical resources, wherein the historical resources have interest tags across multiple different interest dimensions; dividing the historical behavior data into historical behavior data across multiple different interest dimensions according to the interest tags across multiple different interest dimensions; and determining the effective click-through rate of the target object set across each of the multiple different interest dimensions based on the historical behavior data across multiple different interest dimensions.

[0051] In this embodiment of the disclosure, historical resources can refer to existing content that has been interacted with by users. These contents are all accompanied by interest tags corresponding to multiple different interest dimensions, which are used to identify the category to which the content belongs, such as homework tutoring, exams, primary education, etc. Specifically, historical behavior data can be obtained by performing data collection, data cleaning, and organization based on the user's interaction records on the platform for historical resources.

[0052] After obtaining historical behavior data, the data is divided according to interest tags of different interest dimensions. This division allows for the calculation of the effective click-through rate of the target object set for different interest dimensions.

[0053] This embodiment of the disclosure associates and categorizes user historical behavior with historical resources tagged with interests, accurately splits and statistically analyzes behavioral data according to different interest dimensions, and then calculates a true and reliable effective click-through rate. This achieves quantitative processing of user group interests, which is beneficial for subsequent interest confidence calculation and the accuracy of resource recommendations.

[0054] According to embodiments of this disclosure, obtaining the interest confidence of a target object set in at least one interest dimension includes: obtaining the interest confidence of the target object set in at least one interest dimension from an interest dictionary.

[0055] In this embodiment of the disclosure, since the interest confidence scores of the target object set on at least one interest dimension have been stored in the interest dictionary, the interest confidence scores can be directly obtained from the interest dictionary after the target object set is determined. This facilitates rapid resource recommendation, reduces latency, and ensures a better user experience.

[0056] According to embodiments of this disclosure, the resource recommendation method further includes: acquiring updated behavioral data of target objects; determining updated interest confidence of the target object set in each interest dimension based on the updated behavioral data; and updating the interest dictionary based on the updated interest confidence.

[0057] In this embodiment of the disclosure, the target object can continuously collect the updated interaction records of the target object on the platform for resources, and perform data cleaning, noise reduction and other processing on the interaction records to obtain the updated behavior data of the target object. Then, the interest confidence of the target object set on each interest dimension is updated based on the updated behavior data. Specifically, the updated interest confidence can be calculated based on formula (1).

[0058] After obtaining the updated confidence scores for interests, the interest dictionary is updated. For example, if the updated confidence scores do not meet the threshold, the interest can be deleted from the interest dictionary. Alternatively, if the updated confidence scores meet the threshold but the original confidence scores do not (meaning the interest confidence score does not exist in the interest dictionary), the updated confidence scores can be stored in the interest dictionary. Furthermore, if the updated confidence scores differ from the original confidence scores, and both meet the threshold, then the original confidence scores in the interest dictionary can be replaced with the updated confidence scores.

[0059] This disclosure embodiment acquires the latest behavioral data of the target object in real time and dynamically updates the interest confidence level and interest dictionary of the corresponding group, which enables the user group's interest profile to continuously match the latest behavioral data, avoids interest information from being lagging or solidified, improves the timeliness of resource recommendations, and makes the recommended resources more in line with the user's current real needs.

[0060] According to embodiments of this disclosure, the resource recommendation method further includes: determining the relative ranking of the click-through rate of each genre among multiple genres based on the actual click-through rate of the candidate resources for each genre in the target object set; and determining the debiased click-through rate of each candidate resource based on the actual click-through rate of the candidate resource and the relative ranking of the click-through rate of the genre to which the candidate resource belongs.

[0061] In this embodiment, multiple candidate resources in the candidate resource list can be grouped according to genre, with content of the same genre grouped into one resource to avoid the impact of click-through rate differences between genres on the results. Then, for each genre, the actual click-through rate of the candidate resource subset corresponding to that genre is obtained. After obtaining the actual click-through rates of the candidate resource subsets for all genres, the click-through rate mean quantile (CTR) of that genre can be calculated, which is the relative ranking of the genre's click-through rate among multiple genres.

[0062] For each candidate resource, the unbiased click-through rate (CTR_q) can be generated based on the ratio of the candidate resource's actual click-through rate to the percentage of click-through rate of the genre to which the candidate resource belongs. For example, if the actual click-through rate of a candidate resource is greater than the mean quantile (CTR) of the subset to which the candidate resource belongs, CTR_q will be greater than 1, indicating that the quality of the candidate resource is better than the average level of the subset. If the actual click-through rate of a candidate resource is lower than the mean quantile of the subset to which the candidate resource belongs, CTR_q will be less than 1.

[0063] In this embodiment of the disclosure, by calculating the relative ranking of click rates for different genres, and using this ranking to debias the actual click rates of candidate resources, the biased click rate can be obtained. This solves the bias problem caused by uneven genre distribution in traditional resource recommendations and improves the fairness and diversity of recommendations.

[0064] According to embodiments of this disclosure, the resource recommendation method further includes: determining the relative ranking of the update click-through rate of each genre among multiple genres based on the update click-through rate of candidate resources for each genre in the target object set; and determining the debiased click-through rate of the update of candidate resources based on the relative ranking of the update click-through rate.

[0065] In this embodiment of the disclosure, since user preferences and genre distribution may change over time, the quantile CTR is dynamically updated periodically to ensure the timeliness and accuracy of CTR_q.

[0066] Specifically, the click-through rate (CTR) of the target object set for each genre can be determined by acquiring the updated behavior data of the target object and recalculating based on the updated behavior data. Then, the updated CTR ranking and the updated CTR debiasing can be determined according to the above-mentioned methods for determining the CTR ranking and debiasing CTR.

[0067] This embodiment of the disclosure ensures the timeliness and accuracy of the bias correction click rate by updating the bias correction click rate, avoids information lag or stagnation, improves the timeliness and adaptability of resource recommendations, and makes the recommended resources more in line with the user's current real needs.

[0068] According to embodiments of this disclosure, determining a target resource from multiple candidate resources based on interest confidence and debiased click-through rate includes: determining a weighting coefficient for a candidate resource based on the interest confidence and debiased click-through rate of the interest dimension to which the candidate resource belongs; determining a recommended evaluation value for the candidate resource based on the weighting coefficient; and determining the target resource from multiple candidate resources based on the recommended evaluation value.

[0069] In this embodiment of the disclosure, the adjustment coefficient can be calculated based on formula (2):

[0070] Formula (2)

[0071] in, Represents the weighting coefficient. , , , , , , This represents weight, and pow represents exponentiation. This indicates the confidence level of interest in the primary category to which the candidate resource belongs. This indicates the confidence level of interest in the secondary category to which the candidate resource belongs. This represents the confidence level of the interest tag to which the candidate resource belongs, where, , as well as It can be obtained from an interest dictionary. This indicates the offset click-through rate. This represents the average percentile of the click-through rate.

[0072] After determining the weighting coefficients, the original score of each candidate resource in the candidate resource list is multiplied by its weighting factor to obtain the weighted score, which serves as the recommendation evaluation value. The original score can be either the original score in the candidate resource list or a score obtained through a ranking model. After obtaining the weighted score (recommendation evaluation value), the candidate resources in the candidate resource list are re-ranked according to the weighted score to obtain a re-ranked candidate resource list. Based on this re-ranked list, the target resource is determined from the candidate resources; for example, at least one candidate resource with a relatively high ranking (i.e., a relatively high recommendation evaluation value) is selected as the target resource.

[0073] This embodiment combines the user's interest confidence with the debiased click-through rate of the resource to generate a weighting coefficient, and uses this to calculate the recommendation evaluation value to select the target resource. This ensures that the recommended content closely matches the user's real interests, and effectively avoids the problem of inaccurate ranking caused by genre differences and traffic bias, thereby improving the accuracy and fairness of the resource recommendation results.

[0074] According to embodiments of this disclosure, the interest dimension includes at least one of a primary category, a secondary category, and an interest tag.

[0075] In this embodiment of the disclosure, the interest dimension may include at least one of a primary category, a secondary category, and interest tags. For example, the primary category of the interest dimension may be education, the secondary category may be subject education, and the interest tags may include homework tutoring, exams, primary school education, etc. Alternatively, the interest dimension may only include a primary category and a secondary category.

[0076] Figure 3 This is a flowchart of a method for constructing an interest dictionary according to an embodiment of the present disclosure.

[0077] like Figure 3 As shown, the resource recommendation method 300 includes operations S310 to S340.

[0078] In operation S310, for the target object set, the historical behavior data of each object in the target object set is divided into different interest dimensions.

[0079] In this embodiment of the disclosure, the target object set can be any set of objects from multiple historical object sets, and the behavioral data can refer to the operation data of the target objects in the target object set on historical resources, such as user behavior data on the platform (e.g., clicks, shares, likes, etc.). The interest dimension can include at least one of primary category, secondary category, and interest tag, for example, dividing the historical behavioral data of an object into primary category A, secondary category A1, and interest tag A11.

[0080] In operation S320, based on the historical behavior data of each object in the target object set across different interest dimensions, the effective click-through rate of the target object set in each interest dimension is determined.

[0081] In this embodiment of the disclosure, the effective click-through rate can be determined based on the behavioral data of each object in each interest dimension of the target object set, for example, the ratio of effective clicks to impressions can be used as the effective click-through rate.

[0082] In operation S330, the interest confidence level of the target object set in each interest dimension is determined based on the effective click-through rate of the target object set in each interest dimension and the average effective click-through rate of multiple historical object sets in all interest dimensions.

[0083] In this embodiment of the disclosure, for a certain interest dimension, the ratio of the effective click-through rate of the target object set on that interest dimension to the average effective click-through rate of all historical object sets on all interest dimensions can be used as the interest confidence of the target object set on that interest dimension, as can be found in the above formula (1).

[0084] In operation S340, if the interest confidence score is greater than the threshold, the interest confidence score is stored in the interest dictionary.

[0085] In this embodiment of the disclosure, the interest confidence scores greater than the threshold are stored in the interest dictionary to construct the interest dictionary. When storing, the attribute information of the target object set can be used as an identifier and associated with the confidence scores. For example, the target object set a corresponds to the identifier key1, and key1 is associated with the interest confidence scores of the target object set a for the first-level category A, the interest confidence scores for the second-level category A1, and the interest confidence scores for the interest tag A11.

[0086] Figure 4 This is a flowchart of a resource recommendation method according to another embodiment of the present disclosure.

[0087] like Figure 4 As shown, the resource recommendation method 400 includes operations S410 to S460.

[0088] In operation S410, the target object set to which the target object belongs is determined from multiple historical object sets based on the basic attribute information of the target object.

[0089] In this embodiment of the disclosure, the objects included in each historical object set have the same or similar basic attribute information, i.e., common attribute information, such as age, region, etc. Therefore, the target object set to which the target object belongs can be determined from multiple historical object sets by comparing the basic attribute information of the target object with the common attribute information of each historical object set.

[0090] In operation S420, the interest confidence of the target object set in each interest dimension is obtained from the interest dictionary.

[0091] The interest dictionary can store the interest confidence of the target object set on each interest dimension. The interest dimension can include at least one of the first-level category, second-level category, and interest tag.

[0092] In operation S430, the relative ranking of click-through rates for each genre is determined based on the actual click-through rates of candidate resources for each genre within the target object set.

[0093] In this context, "genre" can refer to the content presentation format of candidate resources, such as videos, text and images, or animations. Specifically, candidate resources can be divided into multiple candidate resource subsets according to genre. Then, the actual click-through rate (CTR) is determined based on the historical behavior data of the target set for each candidate resource subset. This CTR serves as the actual CTR of the target set for each genre of candidate resources. After obtaining the actual CTR of all candidate resource subsets for all genres, the CTR quantile for each genre can be calculated, thereby obtaining the relative CTR ranking of each genre among multiple genres.

[0094] In operation S440, for each candidate resource, the bias-free click-through rate is determined based on the relative ranking of the candidate resource's actual click-through rate and the click-through rate of its genre.

[0095] Specifically, for each candidate resource, it is possible to divide the candidate resource into multiple candidate resource subsets according to genre, and then calculate the average click-through rate quantile of the genre to which each candidate resource subset belongs based on the actual click-through rate of each candidate resource subset, that is, the click-through rate ranking of the genre among multiple genres. Then, the ratio of the actual click-through rate of the candidate resource to the average click-through rate quantile of the genre to which the candidate resource belongs is used as the unbiased click-through rate of the candidate resource.

[0096] In operation S450, for each candidate resource, the recommendation evaluation value is determined based on the interest confidence level and the debiased click-through rate of the candidate resource's interest dimension.

[0097] Specifically, the weighting coefficient can be calculated first based on the interest confidence level and debiased click-through rate of each candidate resource in the interest dimension. Then, the recommendation evaluation value is obtained by multiplying the weighting coefficient by the original score of the candidate resource in the recommended resource list.

[0098] In operation S460, the target resource is determined from multiple candidate resources based on the recommended evaluation value of each candidate resource.

[0099] For example, select the candidate resource with the highest recommended evaluation value as the target resource.

[0100] According to embodiments of this disclosure, this disclosure also provides a resource recommendation device.

[0101] Figure 5 This is a block diagram of a resource recommendation apparatus according to an embodiment of the present disclosure.

[0102] like Figure 5 As shown, the resource recommendation device 500 includes a target object set determination module 510, a first acquisition module 520, a second acquisition module 530, and a target resource determination module 540.

[0103] The target object set determination module 510 is used to obtain the basic attribute information of the target object and determine the target object set to which the target object belongs based on the basic attribute information.

[0104] The first acquisition module 520 is used to acquire the interest confidence of the target object set in at least one interest dimension, wherein the interest confidence of the target object set in each interest dimension is determined based on the effective click-through rate of the target object set in that interest dimension and the global average effective click-through rate.

[0105] The second acquisition module 530 is used to acquire the debiased click rate of each candidate resource among multiple candidate resources. The debiased click rate is determined based on the relative ranking of the click rates of the genre to which the candidate resource belongs among multiple genres.

[0106] The target resource determination module 540 is used to determine the target resource from multiple candidate resources based on interest confidence and debiased click-through rate.

[0107] According to embodiments of this disclosure, the resource recommendation device 500 further includes an effective click-through rate determination module, a first interest confidence determination module, and a storage module. Based on historical behavioral data of each object in the target object set across multiple different interest dimensions, the effective click-through rate determination module determines the effective click-through rate of the target object set in each of the multiple different interest dimensions. The first interest confidence determination module, for each interest dimension, determines the interest confidence of the target object set in the current interest dimension based on the effective click-through rate of the target object set in the current interest dimension and the global average effective click-through rate of multiple historical object sets across multiple different interest dimensions. The target object set is any object set in the multiple historical object sets. The storage module, in response to the target object set's interest confidence in at least one dimension being greater than a threshold, stores the interest confidence of the target object set in at least one dimension into an interest dictionary.

[0108] According to embodiments of this disclosure, the effective click-through rate (CTR) determination module includes a third acquisition module, a segmentation module, and an effective CTR determination submodule. The third acquisition module acquires historical behavior data for each object in the target object set regarding historical resources, wherein the historical resources have multiple interest tags across different interest dimensions. The segmentation module divides the historical behavior data into historical behavior data across multiple interest dimensions based on the interest tags. The effective CTR determination submodule determines the effective CTR of the target object set for each interest dimension based on the historical behavior data across multiple interest dimensions.

[0109] According to an embodiment of this disclosure, the first acquisition module 520 includes a first acquisition submodule, which is used to acquire the interest confidence of the target object set in at least one interest dimension from the interest dictionary.

[0110] According to embodiments of this disclosure, the resource recommendation device 500 further includes a fourth acquisition module, a second interest confidence determination module, and an update module. The fourth acquisition module is used to acquire updated behavioral data of the target object; the second interest confidence determination module is used to determine the updated interest confidence of the target object set in each interest dimension based on the updated behavioral data; and the update module is used to update the interest dictionary based on the updated interest confidence.

[0111] According to embodiments of this disclosure, the resource recommendation device 500 further includes a first ranking determination module and a first debiased click-through rate determination module. The ranking determination module is used to determine the relative ranking of the click-through rate of each genre among multiple genres based on the actual click-through rate of candidate resources for each genre in the target object set; the debiased click-through rate determination module is used to determine the debiased click-through rate of each candidate resource based on the actual click-through rate of the candidate resource and the relative ranking of the click-through rate of the genre to which the candidate resource belongs.

[0112] According to embodiments of this disclosure, the resource recommendation device 500 further includes a second ranking determination module and a second de-biased click-through rate determination module. The second ranking determination module is used to determine the relative ranking of the updated click-through rates of each genre among multiple genres based on the updated click-through rates of candidate resources for each genre in the target object set; the second de-biased click-through rate determination module is used to determine the updated de-biased click-through rate of the candidate resources based on the updated relative click-through rate ranking.

[0113] According to embodiments of this disclosure, the target resource determination module 540 includes a weighting coefficient determination module, a recommended evaluation value determination module, and a target resource determination module. The weighting coefficient determination module determines the weighting coefficient of a candidate resource based on the interest confidence level and debiased click-through rate of the interest dimension to which the candidate resource belongs. The recommended evaluation value determination module determines the recommended evaluation value of the candidate resource based on the weighting coefficient. The target resource determination module determines the target resource from multiple candidate resources based on the recommended evaluation value.

[0114] According to embodiments of this disclosure, the interest dimension includes at least one of a primary category, a secondary category, and an interest tag.

[0115] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0116] 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.

[0117] 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.

[0118] 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.

[0119] 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 resource recommendation method. For example, in some embodiments, the resource recommendation method 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 resource recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the resource recommendation method by any other suitable means (e.g., by means of firmware).

[0120] 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), complex 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.

[0121] 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.

[0122] 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.

[0123] 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).

[0124] 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.

[0125] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

[0126] 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 disclosed in this disclosure can be achieved, and this is not limited herein.

[0127] 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 resource recommendation method, comprising: Obtain the basic attribute information of the target object, and determine the target object set to which the target object belongs based on the basic attribute information; Obtain the interest confidence score of the target object set in at least one interest dimension, wherein the interest confidence score of the target object set in each interest dimension is determined based on the effective click-through rate of the target object set in that interest dimension and the global average effective click-through rate; Obtain the unbiased click-through rate (CTR) for each candidate resource among multiple candidate resources, wherein the unbiased CTR is determined based on the relative ranking of the click-through rates of the genre to which the candidate resource belongs among multiple genres; The target resource is determined from the plurality of candidate resources based on the interest confidence level and the debiased click rate.

2. The method according to claim 1, further comprising: Based on the historical behavior data of each object in the target object set across multiple different interest dimensions, determine the effective click-through rate of the target object set in each of the multiple different interest dimensions. For each interest dimension, the interest confidence of the target object set in the current interest dimension is determined based on the effective click-through rate of the target object set in the current interest dimension and the global average effective click-through rate of multiple historical object sets in multiple different interest dimensions. The target object set is any object set in the multiple historical object sets. In response to the target object set having an interest confidence score greater than a threshold in at least one dimension, the interest confidence score of the target object set in the at least one dimension is stored in the interest dictionary.

3. The method according to claim 2, wherein, Determining the effective click-through rate of the target object set across the multiple different interest dimensions includes: Obtain historical behavior data of each object in the target object set for historical resources, wherein the historical resources have interest tags of multiple different interest dimensions; The historical behavior data is divided into historical behavior data in multiple different interest dimensions according to the interest tags of the multiple different interest dimensions; Based on historical behavioral data across multiple different interest dimensions, determine the effective click-through rate of the target object set across each of the multiple different interest dimensions.

4. The method according to claim 2, wherein, The step of obtaining the interest confidence of the target object set in at least one interest dimension includes: Obtain the interest confidence score of the target object set in at least one interest dimension from the interest dictionary.

5. The method according to claim 2, further comprising: Obtain the updated behavior data of the target object; Based on the updated behavioral data, determine the updated interest confidence level of the target object set in each interest dimension; The interest dictionary is updated based on the updated interest confidence.

6. The method according to claim 1, further comprising: Based on the actual click-through rate of candidate resources for each genre in the target object set, determine the relative ranking of click-through rate of each genre among multiple genres; For each candidate resource, the bias-free click-through rate is determined based on the relative ranking of the candidate resource's actual click-through rate and the click-through rate of the genre to which the candidate resource belongs.

7. The method according to claim 6, further comprising: Based on the click-through rate of the updates of candidate resources for each genre in the target object set, determine the relative ranking of the click-through rate of updates for each genre among multiple genres; Based on the updated relative ranking of click-through rates, the debiased click-through rate of the candidate resource is determined.

8. The method according to claim 1, wherein, The step of determining the target resource from the plurality of candidate resources based on the interest confidence and the debiased click-through rate includes: The weighting coefficient of the candidate resource is determined based on the interest confidence of the interest dimension to which the candidate resource belongs and the debiased click rate. Based on the weighting coefficient, the recommended evaluation value of the candidate resource is determined; Based on the recommended evaluation value, the target resource is determined from the plurality of candidate resources.

9. The method according to any one of claims 1 to 8, wherein, The interest dimension includes at least one of the following: primary category, secondary category, and interest tag.

10. A resource recommendation device, comprising: The target object set determination module is used to obtain basic attribute information of the target object and determine the target object set to which the target object belongs based on the basic attribute information. The first acquisition module is used to acquire the interest confidence of the target object set in at least one interest dimension, wherein the interest confidence of the target object set in each interest dimension is determined based on the effective click-through rate of the target object set in that interest dimension and the global average effective click-through rate; The second acquisition module is used to acquire the unbiased click rate of each candidate resource among multiple candidate resources, wherein the unbiased click rate is determined based on the relative ranking of the click rates of the genre to which the candidate resource belongs among multiple genres; The target resource determination module is used to determine the target resource from the plurality of candidate resources based on the interest confidence and the debiased click rate.

11. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 9.

13. A computer program product comprising a computer program stored on at least one of a readable storage medium and an electronic device, the computer program implementing the method according to any one of claims 1 to 9 when executed by a processor.

Images