Video recommendation method and device based on actor preference degree, equipment and medium

Abstract

The invention relates to a video recommendation method and device based on the preference degree of actors, equipment, a medium and a product. The method comprises the following steps: in response to a video acquisition request sent by a client, acquiring multiple types of target behavior data of a user for a video containing a target actor in a historical time period; determining a target strength conversion parameter of each type of target behavior data; according to a piecewise linear mapping rule, mapping the numerical value of the target behavior data and the target strength conversion parameter to obtain a behavior strength score; accumulating the behavior intensity scores of the user under the various target behavior data, and determining the preference score of the user for the target actor; generating a target interest portrait of the user based on the preference score of the user for the target actor; and based on the target interest portrait, screening out an actor recommendation video corresponding to the user from a preset actor database, and performing video recommendation on the user based on the actor recommendation video. The method can improve the integrating degree of the recommended content and the real preference of the user.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a video recommendation method, apparatus, device, and medium based on actor preference. Background Technology

[0002] Currently, the preferred level of a user for a particular actor is typically determined based on a combination of various user interactions, and videos featuring that actor are then recommended. In this approach, each interaction has numerous adjustable parameters; for example, each interaction has its own defined score range and a threshold for each range. When calculating the preference score for multiple interactions, combinations of different interactions require different weights, and these weights influence each other. Therefore, this approach suffers from an excessively large number of adjustable parameters, leading to excessively high time costs in calculating the final preference score, which in turn affects the accuracy of subsequent video recommendations. Summary of the Invention

[0003] To address the aforementioned technical issues, this disclosure provides a video recommendation method, apparatus, device, and medium based on actor preference.

[0004] According to one aspect of this disclosure, a video recommendation method based on actor preference is provided, the method being applied to a video recommendation server, the method comprising: In response to a video retrieval request sent by the client, acquire target behavior data of the user for videos containing the target actor in multiple categories within a historical time period; Based on the one-to-one correspondence between behavioral data and intensity conversion parameters, the target intensity conversion parameters for each type of target behavioral data are determined. According to the preset piecewise linear mapping rule, the numerical value of the target behavior data is mapped to the target intensity conversion parameter to obtain the behavior intensity score; The user's behavior intensity scores under various target behavior data are accumulated to determine the user's preference score for the target actor; Based on the user's preference rating for the target actor, a target interest profile of the user is generated; Based on the target interest profile, recommended videos of actors corresponding to the user are selected from a preset actor database, and video recommendations are made to the user based on the recommended videos of actors.

[0005] According to another aspect of this disclosure, a video recommendation device based on actor preference is also provided, the device being applied to a video recommendation server, the device comprising: The data acquisition module is used to respond to video acquisition requests sent by the client and acquire target behavior data of users in various categories for videos containing target actors within a historical time period; The conversion parameter determination module is used to determine the target intensity conversion parameter for each type of target behavior data based on the one-to-one correspondence between behavior data and intensity conversion parameters. The intensity mapping module is used to map the numerical values ​​of the target behavior data to the target intensity conversion parameters according to a preset piecewise linear mapping rule, so as to obtain a behavior intensity score. The preference rating determination module is used to accumulate the user's behavior intensity rating under multiple target behavior data to determine the user's preference rating for the target actor; The profile generation module is used to generate a target interest profile of the user based on the user's preference rating of the target actor; The video recommendation module is used to filter out recommended videos of actors corresponding to the user from a preset actor database based on the target interest profile, so as to recommend videos to the user based on the recommended videos of actors.

[0006] According to another aspect of this disclosure, an electronic device is also provided, the electronic device comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the video recommendation method based on actor preference.

[0007] According to another aspect of this disclosure, a computer-readable storage medium is also provided, the storage medium storing a computer program for performing the above-described video recommendation method based on actor preference.

[0008] According to another aspect of this disclosure, there is also a computer program product characterized in that, when the computer program product is run on a computer, the computer implements the video recommendation method based on actor preference as described above.

[0009] The technical solution provided in this disclosure has the following advantages compared with the prior art: The technical solution provided in this disclosure is applied to a video recommendation server, including: responding to a video acquisition request sent by a client, acquiring target behavior data of users for multiple categories of videos containing target actors within a historical time period; determining the target intensity conversion parameter for each type of target behavior data based on the one-to-one correspondence between the behavior data and the intensity conversion parameter; mapping the values ​​of the target behavior data to the target intensity conversion parameter according to a preset piecewise linear mapping rule to obtain a behavior intensity score; accumulating the user's behavior intensity scores under multiple target behavior data to determine the user's preference score for the target actor; generating a user's target interest profile based on the user's preference score for the target actor; and filtering recommended videos of actors corresponding to the user from a preset actor database based on the target interest profile, so as to recommend videos to the user based on the recommended videos of actors.

[0010] In this technical solution, behavioral data has matching target intensity transformation parameters. For each type of behavioral data, the parameters that need to be adjusted only include the matching target intensity transformation parameters. The parameters are independent of each other, and their number is linearly related to the category of the behavioral data, effectively reducing the number of parameters and the difficulty of parameter adjustment. Compared with the traditional exponentially growing weight combination design, this solution significantly reduces the number of parameters and debugging complexity, while facilitating the rapid integration of new behavioral data. Based on piecewise linear mapping rules and target intensity transformation parameters, behavioral data of different dimensions can be quickly converted into quantifiable behavioral intensity scores that accurately reflect the degree of behavioral data preference for target actors. By accumulating behavioral intensity scores under multiple target behavioral data using a monotonically non-decreasing algorithm, it is ensured that the richer the user interaction behavior, the higher the final determined preference score, fundamentally avoiding the logical paradox that may occur in traditional weighted methods, making preference evaluation more consistent with common sense. The preference score obtained based on the above process objectively reflects the user's interest intensity, and the target interest profile constructed based on this has high credibility. Furthermore, selecting matching videos from the actor database can significantly improve the fit between recommended content and the user's true preferences, enhancing the targeting of recommendations and user satisfaction. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0012] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of the video recommendation method based on actor preference as described in the embodiments of this disclosure; Figure 2 This is a schematic diagram of the structure of the video recommendation device based on actor preference as described in an embodiment of this disclosure; Figure 3 This is a schematic diagram of the structure of the electronic device described in an embodiment of this disclosure. Detailed Implementation

[0014] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0015] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0016] In video recommendation schemes based on actor preferences, it is necessary to analyze various user behaviors related to actors. These user behaviors include watching videos, making reservations, posting comments, following, searching for names, and searching for albums.

[0017] In some implementations, the process typically begins by drawing a histogram of the probability distribution of each behavior across all accounts, thereby defining segment thresholds and the corresponding scores within each interval, with score ranges such as [0,1]. Then, based on business requirements, different weights are assigned to each behavior, resulting in a weighted average of the account's scores for the target actor across all behaviors, calculating the user's score indicating their preference for the target actor.

[0018] In this approach, the number of different user behaviors can lead to discrepancies in score calculation. For example, assuming a weight of 0.5 for watching a movie and 0.1 for searching a person's name, if an account only watches movies for actor A but engages in both watching movies and searching for the name of actor B, then the highest score this account can achieve for actor A is only 0.5 × 1 = 0.5, while for actor B it can reach a maximum of 0.5 × 1 + 0.1 × 1 = 0.6. This difference in score ceilings can result in significant issues with the final score.

[0019] Therefore, in order to mitigate the impact of the limited variety of user behaviors, some approaches can design different weight combinations for different types of behaviors, so that the upper limit of the final score will not deviate significantly regardless of the number of behaviors.

[0020] This method can alleviate the problem of different score caps to some extent, but it still has some shortcomings. On the one hand, this method requires tiered scoring for each behavior, and different weight combinations need to be designed for different behavior combinations. Therefore, the parameter values ​​that need to be adjusted include: the threshold for tiers, the score for each tier, and different weight combinations. From the perspective of the long-term preference of an account for an actor, the number of parameters that need to be adjusted is already very large. If we consider it from the perspective of actual business, the large amount of data in the behavior table means that the time cost required for each parameter adjustment and verification is very high. On the other hand, since the scores for tiered scoring and weight combinations are all within the range of [0,1], the final multiplication of the values ​​will further narrow the preference score. The theoretical range of the preference score is [0,1], but the condition for the preference score to reach 1 is that the user's behavior towards the target actor must rank at the top among all users' behavior scores towards all actors. This results in the preference scores of most (over 99%) accounts being in the range of [0,0.3], making it difficult to determine the threshold to classify the categories of those with or without preferences.

[0021] Furthermore, this method may lead to a situation where the total preference score decreases as the number of behaviors used to calculate the actor's preference score increases. For example, the design might assign a weight of 1 to behavior A when only behavior A is used; and a weight of 0.8 to behavior A and 0.2 to behavior B when both behavior A and behavior B are used. Suppose user X scores the target actor 10 points for behavior A and 1 point for behavior B; user Y only scores 10 points for behavior A for the target actor. Obviously, user X's preference for the target actor is stronger than Y's. However, from a scoring perspective, user X's preference score for the target actor is 8.2 points, while user Y's preference score is 10 points, which contradicts common sense logic.

[0022] Based on the above, since each behavior has two possible outcomes, "present" or "absent," and new types of user behaviors may be added at any time, it is necessary to design weights for each outcome; if there are multiple... Such behavior requires design. With various weight combinations, the number of parameters increases exponentially with the number of behaviors. The parameter adjustment process is very cumbersome, time-consuming, and prone to logical flaws.

[0023] In short, the above-mentioned solutions suffer from the problem of requiring an excessive number of adjustable parameters and poor interpretability. Therefore, to simplify parameter adjustment, reduce time costs, and enhance the interpretability of parameter adjustment and preference score results, thereby improving the accuracy and practicality of video recommendations, this disclosure provides a video recommendation method, apparatus, device, medium, and product based on actor preference levels. For ease of understanding, the embodiments of this disclosure are described below.

[0024] Figure 1 This flowchart illustrates a video recommendation method based on actor preference, applicable to scenarios such as precision marketing, content recommendation, and user operations, to target specific audiences based on actor preferences. This actor preference-based video recommendation method can be executed by a video recommendation device, which can be implemented using software and / or hardware, specifically such as an electronic device or a server. The electronic device can include devices with storage and computing capabilities, such as tablets, desktop computers, laptops, and smartphones. The server can be a cloud server or server cluster, specifically a video recommendation server.

[0025] like Figure 1 As shown in the embodiments of this disclosure, the video recommendation method based on actor preference can be applied to a video recommendation server, and includes the following steps.

[0026] S102, in response to the video acquisition request sent by the client, acquire the user's target behavior data for videos containing the target actor in multiple categories within a historical time period.

[0027] In this embodiment, when a user opens a video app through a client and refreshes the video, the client sends a video retrieval request containing the user's identifier (such as a user account or user ID) to the video recommendation server. Based on the video retrieval request, the video recommendation server can first evaluate the user's preferred actors in order to accurately recommend personalized videos that match the user's preferences.

[0028] Therefore, to accurately quantify and evaluate users' preferences for target actors, comprehensive data collection of user behavior is necessary. This involves collecting factual data on users' multi-dimensional behaviors towards target actors within a preset historical timeframe (e.g., the past 30 days), yielding at least one type of behavioral data. This behavioral data represents users' interactive behaviors with videos containing the target actor across various interactive scenarios, covering all types of user interactions with the aforementioned videos. Behavioral data may include at least one of the following: number of views, viewing duration, number of searches, reservations, number of bullet comments sent, number of comments, following behavior, number of likes, and number of favorites. Of course, the above are just examples of behavioral data; in actual interactive scenarios, there may be much more behavioral data, which will not be listed here.

[0029] This embodiment avoids the limitations of a single behavioral indicator by acquiring at least one behavioral data point within a historical time period, comprehensively covering all user interactions and providing basic data support for subsequent preference calculations, thus ensuring the objectivity and accuracy of the final preference score.

[0030] S104. Based on the one-to-one correspondence between behavioral data and intensity conversion parameters, determine the target intensity conversion parameters for each type of target behavioral data.

[0031] In real-world interaction scenarios, the behavioral data generated by different interactive behaviors are not consistent. For example, behavioral data such as the number of views and searches are statistically measured by the number of views, while viewing duration is measured by the time spent. Moreover, the strength of a user's preference for a target actor is reflected differently depending on the type of interactive behavior. For instance, the viewing behavior corresponds to 3 views, while the following behavior corresponds to 1 view. However, clearly, a single following action strongly indicates a user's preference for a target actor compared to 3 viewing actions.

[0032] Based on this, in order to quantify and standardize the degree of preference of behavioral data for target actors, this embodiment can pre-set its own intensity conversion parameters for each interactive behavior (or behavioral data for each interactive behavior). Intensity conversion parameters can be understood as preset benchmark parameters for different behavioral data, used to eliminate statistical dimension differences (such as frequency, duration, etc.) among different behavioral data. The specific value of the intensity conversion parameter for each type of behavioral data can be set based on the industry general rules or business scenario requirements of that interactive behavior. It is a key benchmark for converting behavioral data into behavioral intensity scores and has interpretability. Through intensity conversion parameters, behavioral data with multiple statistical dimensions are converted into a unified and standardized behavioral intensity score.

[0033] In some weighted averaging schemes for behavioral data, the sum of the weights of each behavioral data point is 1. This results in the weights of each behavioral data point not being independent, increasing the number of parameters and the difficulty of weight design. To address this issue, in this embodiment, the intensity conversion parameters of different behavioral data points are independent and uncorrelated. The intensity conversion parameters corresponding to interactive behaviors and their data are only related to the interactive behaviors and their data themselves. In this way, the intensity conversion parameters are not only interpretable, but the number of parameters is also linearly related to the number of types of behavioral data, greatly reducing the number of parameters.

[0034] In this embodiment, when setting the intensity conversion parameters for each type of behavioral data, only the impact of the current interactive behavior on the target actor's preference needs to be considered, without concern for other interactive behaviors, allowing the intensity conversion parameters to be adjusted independently. In particular, for new interactive behaviors, new intensity conversion parameters can be added at any time to match them without needing to adjust the intensity conversion parameters of other interactive behaviors, which is very convenient.

[0035] One implementation may include: obtaining a pre-defined one-to-one correspondence between behavioral data and intensity conversion parameters; determining the target intensity conversion parameter associated with each type of target behavioral data based on the correlation; and then determining the behavioral intensity of the target behavioral data based on the target intensity conversion parameters.

[0036] Specifically, for various interactive behaviors between users and target actors, a pre-defined correlation is established between interactive behaviors, behavioral data, and intensity conversion parameters. This correlation represents the correspondence between behavioral data and intensity conversion parameters for different interactive behaviors. Referring to the example provided in Table 1 below, the interactive behaviors, behavioral data, intensity conversion parameters, and the basis for setting the intensity conversion parameters in the correlation are described.

[0037] Table 1: Relationships

[0038] Then, based on the user's actual behavioral data, the target intensity conversion parameter corresponding to each target behavioral data is determined according to the aforementioned preset correlation. For example, when it is found that the user has watched the target actor's video 8 times, the corresponding target intensity conversion parameter is determined to be 3 times based on the correlation; when it is found that the user has made 3 valid comments on the target actor's dynamics (such as articles or videos), the corresponding target intensity conversion parameter is determined to be 2 times based on the correlation.

[0039] S106. According to the preset piecewise linear mapping rule, the numerical value of the target behavior data is mapped to the target intensity conversion parameter to obtain the behavior intensity score.

[0040] After determining the target intensity transformation parameters associated with each target behavior data according to the above embodiments, the behavior intensity of the target behavior data is determined according to the preset piecewise linear mapping rules and the target intensity transformation parameters to obtain the behavior intensity score.

[0041] In one example, the target behavior data can be used as a multiple of its target intensity transformation parameter as a behavior intensity score, referring to the following formula (1).

[0042] (1) Among them, S i x represents the behavioral intensity score of target behavioral data i. i w represents the numerical value of target behavior data i. i This represents the target intensity transformation parameter associated with target behavior data i.

[0043] For example, if a user watches a target actor's video 8 times (x1=8), and the matched target intensity transformation parameter is 5 times (w1=5), then the standardized behavioral intensity score S1 corresponding to the number of views is 8÷5=1.6. This behavioral intensity score is greater than 1, reflecting that the user has a high degree of viewing intention for the target actor and a relatively high level of preference. As another example, if a user comments on a target actor's post 1 time (x2=1), and the matched target intensity transformation parameter is 2 times (w2=2), then the standardized behavioral intensity score S2 corresponding to the number of comments is 1÷2=0.5. This behavioral intensity score is less than 1, reflecting that the user's engagement with the target actor is low and their preference is relatively low. The above embodiments convert target behavior data into quantifiable and standardized behavior intensity scores by using target intensity conversion parameters. This allows target behavior data with different statistical dimensions to be converted into behavior intensity scores with a unified dimension. The behavior intensity scores can not only accurately reflect the differences in the degree of preference of target behavior data for target actors, but also perform comprehensive intensity calculations on different types of behavior data, providing accumulative basic data for subsequent comprehensive calculations of the user's overall preference for target actors.

[0044] In another example, the numerical values ​​of the target behavior data are mapped to the target intensity conversion parameters according to the preset piecewise linear mapping rules to obtain the behavior intensity score, which can be referred to as the following formula (2), including the following contents.

[0045] (2) Among them, S i x represents the behavioral intensity score of target behavioral data i. i w represents the numerical value of target behavior data i. i The parameter represents the target intensity transformation parameter associated with the target behavior data i, and k represents the adjustment coefficient.

[0046] Based on the piecewise linear mapping rule shown in the above formula (2), if the value of the target behavior data is not greater than the target intensity conversion parameter, the ratio of the value of the target behavior data to the target intensity conversion parameter is calculated to obtain the behavior intensity score.

[0047] Alternatively, if the value of the target behavior data is greater than the target intensity transformation parameter, the ratio of the target behavior data value to the target intensity transformation parameter is used as the baseline intensity score; this baseline intensity score can be understood as equal to 1.

[0048] And, the portion of the target behavior data that exceeds the value of the target intensity transformation parameter (i.e., x) i -w i ), and the magnified target intensity conversion parameters (i.e. The ratio is calculated to obtain the incremental intensity score; then the sum of the baseline intensity score and the incremental intensity score is determined as the behavioral intensity score.

[0049] The adjustment coefficient k used to amplify the target intensity conversion parameter can control the growth rate exceeding the baseline intensity score, and k is usually ≥ 1.

[0050] When the value of the target behavior data is not greater than the target intensity conversion parameter ( When the target behavior data does not reach the target intensity transformation parameter, the behavior intensity score is a multiple of the target behavior data relative to its target intensity transformation parameter, reflecting the degree of preference when the target behavior data does not reach the target intensity transformation parameter, which is the baseline intensity.

[0051] When the value of the target behavior data is greater than the target intensity conversion parameter ( When the behavior intensity score continues to increase based on the baseline intensity score 1, the growth rate slows down, reflecting diminishing marginal effects, which is consistent with the diminishing marginal effect of behavior intensity in actual business.

[0052] This embodiment, through a piecewise linear mapping rule, can intuitively reflect the degree of preference when the user's target behavior data does not reach the baseline strength, and can reasonably control the increase in score after the target behavior data exceeds the baseline strength, avoiding the excessive influence of a single high-frequency behavior on the overall preference judgment. This makes the behavior intensity score more consistent with the marginal diminishing law of user interest, improving the rationality and stability of preference assessment, while maintaining the efficiency of calculation and the interpretability of parameters.

[0053] S108, accumulate the user's behavior intensity scores under multiple target behavior data to determine the user's preference score for the target actor.

[0054] This embodiment may include: performing non-normalized accumulation of user behavior intensity scores under multiple target behavior data based on a monotonically increasing scoring algorithm to determine the user's preference score for the target actor; wherein, the monotonically increasing scoring algorithm is an algorithm in which the preference score increases with the increase of the number of categories of target behavior data.

[0055] In one possible embodiment, the monotonically increasing rating algorithm may include the algorithm shown in the following formula (3): take the square root of the sum of squares of each behavior intensity rating to obtain the user's preference rating for the target actor.

[0056] For example, there are now a total of This is a type of target behavior data used to construct target actor preference ratings, whereby various target behavior data of users towards target actors are denoted as follows: The target intensity conversion parameter for matching is The monotonically increasing scoring algorithm is as follows: (3) Among them, S i denoted by , i represents the intensity score of the target behavior data i, n represents the number of target behavior intensity scores, or the total number of types of target behavior data, and score represents the user's preference score for the target actor.

[0057] This embodiment uses the above-mentioned monotonically increasing scoring algorithm. When the final preference score is greater than or equal to 1, it is considered that the user has a strong preference for the target actor.

[0058] This embodiment integrates at least one behavior intensity score based on the scoring algorithm shown in the above formula. Each behavior intensity score is considered as a component of a multi-dimensional vector. The preference score is obtained by calculating the vector magnitude (the square root of the sum of squares). Geometrically, this scoring algorithm can be understood as the distance from a point in a Cartesian coordinate system to the origin. Within a sphere with a radius of 1 and outside the sphere, a preference for the target actor is considered. Compared to a direct linear increase, this scoring algorithm slows down the final... The rate at which the total number of behaviors increases can be explained with the following example: Suppose a user performs 5 behaviors... Both scores are 0.2. If they are directly added linearly, the score is 1, which is considered to indicate a preference. However, in reality, the user's behavior intensity scores for each behavior are not very high. Using a scoring algorithm that calculates the "arithmetic square root of the sum of squares", the preference score is about 0.447, which is more in line with cognition and also has a geometric interpretation.

[0059] In another possible embodiment, the user's behavioral intensity scores across multiple target behavioral data are accumulated to determine the user's preference score for the target actor, including: The intensity scores of user behavior under multiple target behavior data are accumulated to obtain an accumulated intensity score; a reward factor matching the number of categories of target behavior data is determined; the accumulated intensity score is adjusted according to the reward factor to determine the user's preference score for the target actor.

[0060] Specifically, refer to the following formula (4): (4) First, the intensity scores of user behavior across multiple target behavioral data are accumulated. Adding a new type of behavioral data will inevitably increase the accumulated intensity score. Based on this, to reinforce the perception that a greater number of behavioral data types equates to a stronger preference, this embodiment introduces a reward factor R(n) that matches the number of categories of target behavioral data. As an example, the reward factor could be: R(n) = 1 + α (n 1) (5) Where n is the number of categories of the target behavior data, and α is the preset category reward coefficient, for example, α=0.1.

[0061] By using reward factors, we can ensure that for every additional type of target behavioral data, the user's overall preference rating for the target actor will receive an additional percentage increase. Therefore, by adjusting the cumulative intensity rating based on the reward factors, we can determine the user's preference rating for the target actor.

[0062] In the above-described embodiment for determining preference scores, combining linear accumulation with a reward factor ensures a strict positive correlation between preference scores and the number of target behavior data categories. This not only avoids the logical contradiction in traditional weighted summation where the redistribution of weights leads to an increase in behavior categories but a decrease in the total score, but also, based on the accumulated intensity score, further amplifies the amplification effect of multiple target behavior data categories on preference intensity through the reward factor, improving the accuracy of preference scores and thus more comprehensively and reasonably reflecting the user's true interests.

[0063] In the above embodiment, after obtaining the user's behavioral intensity rating for at least one target actor, since the behavioral intensity rating itself can reflect the user's preference for the target actor, there is no need to set weights or other parameters for the behavioral intensity. Instead, all behavioral intensity ratings are directly used to comprehensively calculate the user's overall preference rating for the target actor. This method reduces the number of weights and other parameters in the calculation process. More importantly, compared to related solutions, in this embodiment, the final preference rating is always non-decreasing as behavioral data increases. For example, there are three interactive behaviors: watching a movie, searching for a name, and watching a live stream. User A and User B have the same behavioral intensity rating for watching a movie and searching for a name, but User B also watches a live stream. In this case, the final preference rating of User A, who only has the behaviors of watching a movie and searching for a name, will definitely be lower than the final preference rating of User B, who has all three behaviors.

[0064] S110 generates a user's target interest profile based on the user's preference rating for the target actor.

[0065] After determining the user's preference rating for the target actor based on the above embodiments, a binary classification of the user's preference for and non-preference for the target actor can be implemented, which can then be applied to Facebook user profiling to support application scenarios such as recommendation, search, and tagging.

[0066] In practice, it can be determined whether the user's preference rating for the target actor is greater than a preset preference threshold; if it is, the target actor is marked as a preferred actor; and a user's target interest profile is generated based on the preferred actor. Conversely, if the rating is not greater than the preset threshold, the target actor is marked as a non-preferred actor.

[0067] In this embodiment, the user's preference rating for the target actor is compared with a preset preference threshold. When the user's preference rating for the target actor is greater than the preference threshold, the target actor is marked as the user's preferred actor, indicating that the user has a significant interest in him. When the user's preference rating for the target actor is not greater than the preference threshold, the target actor is marked as the user's unpreferred actor, indicating that the user has a low interest in him or no clear preference.

[0068] Furthermore, a user's target interest profile can be generated based on preferred actors. For example, feature aggregation and weight ranking can be performed on all preferred actors of a user to generate a structured target interest profile.

[0069] S112, Based on the target interest profile, filter out the actor recommendation videos corresponding to the user from the preset actor database, and recommend videos to the user based on the actor recommendation videos.

[0070] Taking video platforms as an example, based on target interest profiles, videos of actors that match the target interest profiles are selected from the actor database on the video platform, and content is pushed to users based on the recommended videos to ensure that the pushed content is highly consistent with the user's interests.

[0071] This embodiment calculates a user's preference score for a target actor by using multiple target behavior intensity scores. It then further transforms the quantitative results into practical business value through preference threshold determination, labeling of preferred or unpreferred actors, profile generation, and video recommendation. This effectively achieves accurate identification of user interests and personalized content delivery.

[0072] In summary, the video recommendation method based on actor preference provided in this disclosure is applied to a video recommendation server, including: responding to a video acquisition request sent by a client, acquiring target behavior data of multiple categories of videos containing target actors within a historical time period; determining the target intensity conversion parameter for each type of target behavior data based on the one-to-one correspondence between the behavior data and the intensity conversion parameter; mapping the numerical values ​​of the target behavior data to the target intensity conversion parameter according to a preset piecewise linear mapping rule to obtain a behavior intensity score; accumulating the user's behavior intensity scores under multiple target behavior data to determine the user's preference score for the target actor; generating a user's target interest profile based on the user's preference score for the target actor; and filtering recommended videos of the actor corresponding to the user from a preset actor database based on the target interest profile, so as to recommend videos to the user based on the recommended videos of the actor.

[0073] In this technical solution, behavioral data has matching target intensity transformation parameters. For each type of behavioral data, the parameters that need to be adjusted only include the matching target intensity transformation parameters. The parameters are independent of each other, and their number is linearly related to the category of the behavioral data, effectively reducing the number of parameters and the difficulty of parameter adjustment. Compared with the traditional exponentially growing weight combination design, this solution significantly reduces the number of parameters and debugging complexity, while facilitating the rapid integration of new behavioral data. Based on piecewise linear mapping rules and target intensity transformation parameters, behavioral data of different dimensions can be quickly converted into quantifiable behavioral intensity scores that accurately reflect the degree of behavioral data preference for target actors. By accumulating behavioral intensity scores under multiple target behavioral data using a monotonically non-decreasing algorithm, it is ensured that the richer the user interaction behavior, the higher the final determined preference score, fundamentally avoiding the logical paradox that may occur in traditional weighted methods, making preference evaluation more consistent with common sense. The preference score obtained based on the above process objectively reflects the user's interest intensity, and the target interest profile constructed based on this has high credibility. Furthermore, selecting matching videos from the actor database can significantly improve the fit between recommended content and the user's true preferences, enhancing the targeting of recommendations and user satisfaction.

[0074] In actual testing of the target actors, it was found that during the long-term historical period, the "user-actor" pairs with actor preferences accounted for about 15% of the total. In contrast, during the short-term historical period, the "user-actor" pairs with actor preferences accounted for about 30% of the total, but the total number was less than that of the long-term period, which is consistent with the actual logic.

[0075] Figure 2 This is a schematic diagram of a video recommendation device based on actor preference, provided in an embodiment of this disclosure. This device can be applied to a video recommendation server to implement the video recommendation method based on actor preference provided in the above embodiment. The video recommendation device based on actor preference may include the following modules: The data acquisition module 210 is used to respond to the video acquisition request sent by the client and acquire target behavior data of the user for multiple categories of videos containing target actors within a historical time period. The conversion parameter determination module 220 is used to determine the target intensity conversion parameter for each type of target behavior data based on the one-to-one correspondence between behavior data and intensity conversion parameters. The intensity mapping module 230 is used to map the numerical value of the target behavior data to the target intensity conversion parameter according to a preset piecewise linear mapping rule to obtain a behavior intensity score; The preference rating determination module 240 is used to accumulate the behavior intensity rating of the user under multiple target behavior data to determine the user's preference rating for the target actor; The profile generation module 250 is used to generate a target interest profile of the user based on the user's preference rating of the target actor; The video recommendation module 260 is used to filter out recommended videos of actors corresponding to the user from a preset actor database based on the target interest profile, so as to recommend videos to the user based on the recommended videos of actors.

[0076] In one embodiment, the preference scoring determination module 240 is further configured to: According to a monotonically increasing scoring algorithm, the user's behavior intensity scores under various target behavior data are accumulated in a non-normalized manner to determine the user's preference score for the target actor; wherein, the monotonically increasing scoring algorithm is an algorithm in which the preference score increases as the number of categories of the target behavior data increases.

[0077] In one embodiment, the preference scoring determination module 240 is further configured to: The intensity scores of the user's behavior under various target behavior data are accumulated to obtain an accumulated intensity score; Determine a reward factor that matches the number of categories in the target behavior data; The cumulative intensity score is adjusted based on the reward factor to determine the user's preference score for the target actor.

[0078] In one embodiment, the monotonically increasing scoring algorithm includes:

[0079] Among them, S i The score represents the intensity score of the target behavior data i, n represents the number of categories of the target behavior data or the number of intensity scores, and score represents the user's preference score for the target actor.

[0080] In one embodiment, the portrait generation module 250 is further configured to: Determine whether the user's preference rating for the target actor is greater than a preset preference threshold; If the value is greater than the target actor, then the target actor is marked as the preferred actor. Based on the preferred actors, a target interest profile of the user is generated.

[0081] In one embodiment, the behavioral data is used to represent the user's interactive behavior in various interactive scenarios for videos containing the target actor; the behavioral data includes at least one of the following: number of views, viewing duration, number of searches, reservation behavior, number of bullet screen messages sent, number of comments, following behavior, number of likes, and number of favorites.

[0082] In one embodiment, the intensity mapping module 230 is further configured to: If the value of the target behavior data is not greater than the target intensity conversion parameter, then the ratio of the value of the target behavior data to the target intensity conversion parameter is calculated to obtain the behavior intensity score; If the value of the target behavior data is greater than the target intensity conversion parameter, then the ratio of the target behavior data value to the target intensity conversion parameter is used as the baseline intensity score; and, The portion of the target behavior data that exceeds the value of the target intensity conversion parameter is compared with the amplified target intensity conversion parameter to obtain the incremental intensity score. The sum of the baseline intensity score and the incremental intensity score is determined as the behavioral intensity score.

[0083] The device provided in this embodiment has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0084] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Figure 3 As shown, the electronic device 300 includes one or more processors 301 and memory 302.

[0085] The processor 301 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 300 to perform desired functions.

[0086] The memory 302 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 301 may execute the program instructions to implement the video recommendation method based on actor preference described in the embodiments of this disclosure above, and / or other desired functions. Various contents such as input signals, signal components, and noise components may also be stored in the computer-readable storage medium.

[0087] In one example, the electronic device 300 may also include an input device 303 and an output device 304, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0088] In addition, the input device 303 may also include, for example, a keyboard, a mouse, etc.

[0089] The output device 304 can output various information to the outside, including determined distance information, direction information, etc. The output device 304 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0090] Of course, for the sake of simplicity, Figure 3 Only some of the components of the electronic device 300 relevant to this disclosure are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device 300 may include any other suitable components depending on the specific application.

[0091] Furthermore, this embodiment also provides a computer-readable storage medium storing a computer program for executing the above-described video recommendation method based on actor preference.

[0092] The present disclosure provides a computer program product for a video recommendation method, apparatus, electronic device, and medium based on actor preference, including a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0093] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0094] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A video recommendation method based on actor preference, characterized in that, The method is applied to a video recommendation server, and the method includes: In response to a video retrieval request sent by the client, acquire target behavior data of the user for videos containing the target actor in multiple categories within a historical time period; Based on the one-to-one correspondence between behavioral data and intensity conversion parameters, the target intensity conversion parameters for each type of target behavioral data are determined. According to the preset piecewise linear mapping rule, the numerical value of the target behavior data is mapped to the target intensity conversion parameter to obtain the behavior intensity score; The user's behavior intensity scores under various target behavior data are accumulated to determine the user's preference score for the target actor; Based on the user's preference rating for the target actor, a target interest profile of the user is generated; Based on the target interest profile, recommended videos of actors corresponding to the user are selected from a preset actor database, and video recommendations are made to the user based on the recommended videos of actors.

2. The method according to claim 1, characterized in that, The step of accumulating the user's behavior intensity scores under various target behavior data to determine the user's preference score for the target actor includes: According to a monotonically increasing scoring algorithm, the user's behavior intensity scores under various target behavior data are accumulated in a non-normalized manner to determine the user's preference score for the target actor; wherein, the monotonically increasing scoring algorithm is an algorithm in which the preference score increases as the number of categories of the target behavior data increases.

3. The method according to claim 1, characterized in that, The step of accumulating the user's behavior intensity scores under various target behavior data to determine the user's preference score for the target actor includes: The intensity scores of the user's behavior under various target behavior data are accumulated to obtain an accumulated intensity score; Determine a reward factor that matches the number of categories in the target behavior data; The cumulative intensity score is adjusted based on the reward factor to determine the user's preference score for the target actor.

4. The method according to claim 2, characterized in that, The monotonically increasing scoring algorithm includes: Among them, S i The term "i" represents the behavior intensity score of target behavior data i, and "n" represents the number of categories of the target behavior data or the number of behavior intensity scores. This indicates the user's preference rating for the target actor.

5. The method according to claim 1, characterized in that, The step of generating a target interest profile for the user based on the user's preference rating for the target actor includes: Determine whether the user's preference rating for the target actor is greater than a preset preference threshold; If the value is greater than the target actor, then the target actor is marked as the preferred actor. Based on the preferred actors, a target interest profile of the user is generated.

6. The method according to claim 1, characterized in that, The behavioral data is used to represent the user's interactive behavior in various interactive scenarios for videos containing the target actor; the behavioral data includes at least one of the following: number of views, viewing duration, number of searches, reservation behavior, number of bullet screen messages sent, number of comments, following behavior, number of likes, and number of favorites.

7. The method according to claim 1, characterized in that, The step of mapping the numerical values ​​of the target behavior data to the target intensity transformation parameters according to a preset piecewise linear mapping rule to obtain a behavior intensity score includes: If the value of the target behavior data is not greater than the target intensity conversion parameter, then the ratio of the value of the target behavior data to the target intensity conversion parameter is calculated to obtain the behavior intensity score; If the value of the target behavior data is greater than the target intensity conversion parameter, then the ratio of the target behavior data value to the target intensity conversion parameter is used as the baseline intensity score; and, The portion of the target behavior data that exceeds the value of the target intensity conversion parameter is compared with the amplified target intensity conversion parameter to obtain the incremental intensity score. The sum of the baseline intensity score and the incremental intensity score is determined as the behavioral intensity score.

8. A video recommendation device based on actor preference, characterized in that, The device is used in a video recommendation server, and the device includes: The data acquisition module is used to respond to video acquisition requests sent by the client and acquire target behavior data of users in various categories for videos containing target actors within a historical time period; The conversion parameter determination module is used to determine the target intensity conversion parameter for each type of target behavior data based on the one-to-one correspondence between behavior data and intensity conversion parameters. The intensity mapping module is used to map the numerical values ​​of the target behavior data to the target intensity conversion parameters according to a preset piecewise linear mapping rule, so as to obtain a behavior intensity score. The preference rating determination module is used to accumulate the user's behavior intensity rating under multiple target behavior data to determine the user's preference rating for the target actor; The profile generation module is used to generate a target interest profile of the user based on the user's preference rating of the target actor; The video recommendation module is used to filter out recommended videos of actors corresponding to the user from a preset actor database based on the target interest profile, so as to recommend videos to the user based on the recommended videos of actors.

9. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a terminal device, cause the terminal device to perform the method as described in any one of claims 1-7.

11. A computer program product, characterized in that, When the computer program product is run on a computer, it causes the computer to perform the method as described in any one of claims 1 to 7.

Images