Video processing method and related device
By acquiring and smoothing video distribution efficiency metrics, and combining them with estimated recommendation scores and content quality scores, the problem of insufficient exposure for long-tail videos and new videos in traditional video recommendation systems is solved, thereby improving the accuracy of video exposure and conversion rates.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-24
- Publication Date
- 2026-03-27
AI Technical Summary
In traditional video recommendation systems, long-tail videos and new videos cannot get enough exposure, resulting in low accuracy of exposure results and low conversion rates.
By obtaining the real-time value of the distribution efficiency index of candidate videos, and performing numerical smoothing if the number of exposures is lower than the threshold, a reference index value is obtained. Combined with the estimated recommendation score and content quality score, the exposure judgment result of the video is determined.
It improved the accuracy and conversion rate of video exposure, increased the exposure opportunities for long-tail and new videos, reduced the exposure probability of low-quality videos, and enhanced the diversity and conversion rate of video exposure.
Smart Images
Figure CN121750902A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of video recommendation technology, and in particular to a video processing method and related apparatus. Background Technology
[0002] Currently, video recommendation systems select at least one video from multiple candidate videos for exposure, aiming to improve video conversion rates by exposing high-quality videos. However, when traditional video recommendation systems select videos for exposure, a large amount of attention and exposure opportunities are concentrated on a few top videos, while most long-tail videos or new videos cannot get enough exposure, resulting in an imbalance in exposure. This leads to low accuracy in video exposure results and consequently, low conversion rates. Summary of the Invention
[0003] In view of the above problems, this application provides a video processing method and related apparatus to improve the accuracy of video exposure results. The specific solution is as follows:
[0004] The first aspect of this application provides a video processing method, including:
[0005] For each candidate video, based on the number of times the candidate video is exposed, the real-time index values of each distribution efficiency index of the candidate video are obtained. Different distribution efficiency indices characterize the distribution efficiency of the candidate video from different dimensions.
[0006] If the number of times the video is exposed is lower than a preset threshold, the real-time value of the distribution efficiency index is numerically smoothed based on the average value of the distribution efficiency index to obtain a reference value of the distribution efficiency index. The average value is equal to the real-time value of the distribution efficiency index of all videos to be exposed, divided by the number of all videos to be exposed.
[0007] If the number of video exposures is not lower than the number threshold, the real-time value of the distribution efficiency index will be used as the reference value of the distribution efficiency index.
[0008] Based on the reference index values of the various distribution efficiency indicators of the candidate video, the exposure determination result of the candidate video is determined, and the exposure determination result indicates whether the candidate video should be exposed.
[0009] In one possible implementation, the exposure determination result of the candidate video is determined based on the reference index values of each of the distribution efficiency indicators of the candidate video, including:
[0010] Using a preset fusion algorithm, the reference index values of each of the distribution efficiency indicators are fused to obtain the content quality score of the candidate video;
[0011] Based on the content quality score of the candidate videos, the video recommendation score of the candidate videos is obtained;
[0012] If the candidate video meets the preset exposure conditions, the exposure determination result of the candidate video is determined to be to expose the candidate video. The exposure conditions include the video recommendation degree being greater than the preset recommendation degree threshold.
[0013] In one possible implementation, based on the average value of the distribution efficiency index, the real-time value of the distribution efficiency index is numerically smoothed to obtain a reference value for the distribution efficiency index, including:
[0014] Based on the number of video exposures, a first trial weight, a second trial weight, and a smoothing coefficient are determined. The first trial weight is the ratio of the number of video exposures to the number threshold, and the sum of the first trial weight and the second trial weight is 1.
[0015] The product of the smoothing coefficient and the average index value is obtained as the initial index value;
[0016] The reference index value is obtained by weighting and summing the real-time index value and the initial index value using the first trial weight as the weight coefficient of the real-time index value and the second trial weight as the weight coefficient of the initial index value.
[0017] In one possible implementation, a preset fusion algorithm is used to fuse the reference index values of each of the distribution efficiency indicators to obtain the content quality score of the candidate video, including:
[0018] The distribution efficiency indicators are standardized to obtain the standard indicator values of each distribution efficiency indicator.
[0019] The content quality score of the candidate video is obtained by summing the standard index values of each of the aforementioned distribution efficiency indicators.
[0020] In one possible implementation, the video recommendation score of the candidate video is obtained based on its content quality score, including:
[0021] The estimated recommendation score of the candidate video is obtained. The estimated recommendation score is obtained by the prediction of a pre-trained intelligent recommendation model. The intelligent recommendation model is trained with historical videos as sample videos and the estimated recommendation score as the target output. The estimated recommendation score is positively correlated with the conversion rate.
[0022] The content quality score of the candidate video is normalized to obtain the normalized quality score of the candidate video;
[0023] Based on the estimated recommendation score and normalized quality score of the candidate video, the video recommendation degree of the candidate video is obtained, and the video recommendation degree is positively correlated with the estimated recommendation score and the normalized quality score, respectively.
[0024] In one possible implementation, the video recommendation score is the product of the estimated recommendation score and the normalized quality score. The normalization operation on the content quality score of the candidate videos to obtain the normalized quality score includes:
[0025] Based on the accuracy of the intelligent recommendation model, normalized weights are calculated, and the normalized weights are positively correlated with the accuracy of the intelligent recommendation model.
[0026] Calculate the product of the normalized weight and the highest content quality score to obtain the scaled result of the highest content quality score, where the highest content quality score is the highest value of the content quality scores of all the candidate videos;
[0027] Obtain a first normalized reference value for the candidate video, where the first normalized reference value is equal to the sum of the content quality score and the scaled result of the highest content quality score;
[0028] Obtain a second normalized reference value for the candidate video, the second normalized reference value being equal to the sum of the highest content quality score and the scaling result of the highest content quality score;
[0029] Based on the first normalized reference value and the second normalized reference value, the normalized quality score of the candidate video is obtained, and the normalized quality score is the ratio of the first normalized reference value to the second normalized reference value.
[0030] A second aspect of this application provides a video processing apparatus, comprising:
[0031] The real-time value acquisition unit is used to acquire the real-time index values of each distribution efficiency index of each candidate video based on the number of video exposures of the candidate video. Different distribution efficiency indices characterize the distribution efficiency of the candidate video from different dimensions.
[0032] The reference value acquisition unit is configured to, if the number of video exposures is lower than a preset threshold, perform numerical smoothing on the real-time value of the distribution efficiency index based on the average value of the distribution efficiency index to obtain a reference value of the distribution efficiency index, wherein the average value is equal to the real-time value of the distribution efficiency index for all videos to be exposed, divided by the number of all videos to be exposed; if the number of video exposures is not lower than the threshold, the real-time value of the distribution efficiency index is used as the reference value of the distribution efficiency index.
[0033] An exposure determination unit is used to determine the exposure determination result of the candidate video based on the reference index values of each of the distribution efficiency indicators of the candidate video, wherein the exposure determination result indicates whether the candidate video should be exposed.
[0034] A third aspect of this application provides a computer program product including computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the video processing method described in the first aspect or any implementation thereof.
[0035] A fourth aspect of this application provides an electronic device, including at least one processor and a memory connected to the processor, wherein:
[0036] The memory is used to store computer programs;
[0037] The processor is used to execute the computer program so that the electronic device can implement the video processing method of the first aspect or any implementation thereof.
[0038] By employing the above technical solution, this application provides a video processing method and related apparatus. Based on the number of exposures of candidate videos, it obtains real-time index values of various distribution efficiency indicators for each candidate video. Different distribution efficiency indicators characterize the distribution efficiency of candidate videos from different dimensions. If the number of video exposures is lower than a preset threshold, the real-time index value of the distribution efficiency indicators is numerically smoothed based on the average index value to obtain a reference index value. If the number of video exposures is not lower than the threshold, the real-time index value of the distribution efficiency indicators is used as the reference index value. Based on the reference index values of each distribution efficiency indicator for each candidate video to be exposed, the exposure determination result of the candidate video is determined, indicating whether the candidate video should be exposed. Since the average index value is the average of the real-time index values of the distribution efficiency indicators of all videos to be exposed, for low-exposure videos with fewer than the threshold exposures, the real-time index value is adjusted based on the average value to obtain a reference index value for the distribution efficiency indicators, improving the reliability of the distribution efficiency index values. Therefore, the reference index value is used as a reference feature for whether to expose the candidate video, improving the accuracy and conversion rate of video exposure. Attached Figure Description
[0039] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.
[0040] Figure 1A schematic diagram of the structure of a video recommendation system provided in this application;
[0041] Figure 2 A flowchart illustrating a video processing method provided in this application;
[0042] Figure 3 A schematic diagram illustrating the specific implementation process of a video processing method provided in this application;
[0043] Figure 4 A schematic diagram illustrating the effect of a video processing method provided in this application;
[0044] Figure 5 A schematic diagram of the structure of a video processing device provided in this application;
[0045] Figure 6 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation
[0046] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.
[0047] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.
[0048] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.
[0049] This application may be applied, but is not limited to, to applications with video processing capabilities or cloud services provided by cloud-side servers, which will be described in detail below:
[0050] See Figure 1 , Figure 1 A schematic diagram of a system architecture is shown. The system may include a terminal 100 and a server 200. The server 200 can provide the methods provided in the embodiments of this application to one or more terminals.
[0051] The terminal 100 may have a video processing application installed. The application and webpage can provide an interface. The terminal 100 can receive relevant parameters input by the user on the video processing interface and send the parameters to the server 200. The server 200 can obtain the processing result based on the received parameters and return the processing result to the terminal 100.
[0052] It should be understood that in some optional implementations, the terminal 100 can also complete the action of obtaining the processing result based on the received parameters on its own, without the need for the server to cooperate. This application embodiment is not limited to this.
[0053] The following description Figure 1 The product form of the mid-terminal 100;
[0054] The terminal 100 in this application embodiment can be a mobile phone, tablet computer, wearable device, vehicle device, augmented reality (AR) / virtual reality (VR) device, laptop computer, ultra-mobile personal computer (UMPC), netbook, personal digital assistant (PDA), etc., and this application embodiment does not impose any restrictions on it.
[0055] Terminal 100 may include a radio frequency unit, memory, input unit, display unit, camera (optional), audio circuitry (optional), speaker (optional), microphone (optional), headphone jack (optional), processor, external interface, power supply, and other components. Those skilled in the art will understand that the above-mentioned components are merely examples and do not constitute a limitation on the terminal or multifunctional device; it may include more or fewer components, or a combination of certain components, or different components.
[0056] The input unit can be used to receive input numeric or character information, and to generate key signal inputs related to user settings and function control of the portable multi-functional device. Specifically, the input unit may include a touchscreen (optional) and / or other input devices. Other input devices may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control buttons, power buttons, etc.), trackball, mouse, joystick, etc.
[0057] Among them, the input device can receive input data, etc.
[0058] The display unit can be used to display information input by the user or information provided to the user, various menus of the terminal, interactive interfaces, file display, and / or playback of any multimedia file. In the embodiments of this application, the display unit can be used to display the interface of a video recommendation process, video exposure results, etc.
[0059] The memory can be used to store software code related to the video processing method, the processor can execute the steps of the video processing method, and can also schedule other units (such as the input unit and display unit mentioned above) to achieve the corresponding functions.
[0060] This radio frequency unit (optional) can be used to receive and send signals during information transmission or calls.
[0061] In this embodiment of the application, the radio frequency unit can send data to the server 200 and receive the processing results sent by the server 200.
[0062] It should be understood that this radio frequency unit is optional and can be replaced with other communication interfaces, such as a network port.
[0063] Terminal 100 also includes a power source (such as a battery) for supplying power to the various components.
[0064] Terminal 100 also includes an external interface, which can be a standard Micro USB interface or a multi-pin connector, which can be used to connect terminal 100 to other devices for communication or to connect a charger to charge terminal 100.
[0065] Server 200 includes a bus, a processor, a communication interface, and memory. The processor, memory, and communication interface communicate with each other via the bus.
[0066] The memory can be used to store software code related to video processing methods, the processor can execute the steps of the chip's video processing methods, and it can also schedule other units to achieve corresponding functions.
[0067] Reference Figure 2 , Figure 2 This is a flowchart illustrating a video processing method provided in an embodiment of this application, as shown below. Figure 2 As shown in the embodiment of this application, a data processing method may include steps S201 to S204, which are described in detail below.
[0068] S201. For each candidate video, based on the number of times the candidate video is exposed, obtain the real-time index values of each distribution efficiency index of the candidate video.
[0069] In this embodiment, the distribution efficiency metric is an indicator that can characterize the efficiency of video distribution. Video distribution efficiency refers to the ability and effectiveness of transmitting video content to end-user (viewer) devices quickly, stably, and at low cost. Different types of distribution efficiency metrics are used to measure the effectiveness of the platform in successfully recommending video content and reaching users from different dimensions. For example, distribution efficiency metrics include one or more of the following: video click-through rate, average video playback duration, and video play rate.
[0070] In this embodiment, the video exposure count of a candidate video refers to the number of times the candidate video is displayed to the user. One video exposure means that at least one frame of the video is displayed on the display interface. At this time, the video is recorded as one "exposure", that is, the video exposure count is incremented by 1.
[0071] S202. If the number of video exposures is lower than the preset threshold, the real-time value of the distribution efficiency index is numerically smoothed based on the average value of the distribution efficiency index to obtain a reference value for the distribution efficiency index.
[0072] In this embodiment, the average index value is the average of the real-time index values of the distribution efficiency index of all videos to be exposed.
[0073] In this embodiment, when the number of video exposures is lower than the threshold, it indicates that the candidate video is a long-tail video or a new video. The reliability of the real-time index values of the various distribution efficiency indicators of the candidate video is low. Therefore, the real-time index values of the distribution efficiency indicators are numerically smoothed based on the average index value of the distribution efficiency indicators in order to correct the distribution efficiency indicators of the candidate video and improve the reliability.
[0074] In one optional embodiment, the method for numerically smoothing the real-time value of the distribution efficiency index is as follows: multiply the average value of the distribution efficiency index by a preset smoothing coefficient to obtain a reference value for the distribution efficiency index. The smoothing coefficient can be preset to a value between 0 and 1 based on factors such as the video type, importance, and revenue of the candidate video. For example, the higher the importance and the higher the revenue, the larger the smoothing coefficient.
[0075] In one optional embodiment, the method for numerically smoothing the real-time value of the distribution efficiency index is as follows: a weighted sum of a preset initial index value and a real-time index value is performed, with the weight coefficient of the real-time index value being higher the closer the number of video exposures is to a threshold. The initial index value is determined based on the average index value; for example, the initial index value is equal to the average index value, or, for another example, the initial index value is equal to the average index value multiplied by a preset smoothing coefficient.
[0076] S203. If the number of video exposures is not lower than the number threshold, the real-time value of the distribution efficiency index will be used as the reference value of the distribution efficiency index.
[0077] S204. Based on the reference index values of various distribution efficiency indicators of the candidate videos, determine the exposure judgment result of the candidate videos.
[0078] In this embodiment, the exposure determination result indicates whether to expose the candidate video.
[0079] As can be seen from the above technical solution, the video processing method provided in this application obtains the real-time index values of various distribution efficiency indicators of candidate videos based on the number of video exposures. If the number of video exposures is lower than a preset threshold, the real-time index value of the distribution efficiency indicator is numerically smoothed based on the average index value of the distribution efficiency indicator to obtain a reference index value of the distribution efficiency indicator. If the number of video exposures is not lower than the threshold, the real-time index value of the distribution efficiency indicator is used as the reference index value of the distribution efficiency indicator. Based on the reference index values of each distribution efficiency indicator of each candidate video to be exposed, the exposure determination result of the candidate video is determined, and the exposure determination result indicates whether the candidate video should be exposed. Since the average index value is the average of the real-time index values of the distribution efficiency indicators of all videos to be exposed, for low-exposure videos with a number of video exposures lower than the threshold, the real-time index value is adjusted based on the average value to obtain the reference index value of the distribution efficiency indicator, thereby improving the reliability of the index value of the distribution efficiency indicator. Thus, the reference index value is used as a reference feature for whether to expose the candidate video, improving the accuracy and conversion rate of video exposure.
[0080] Furthermore, the video processing method provided in this application can be specifically applied to a video processing system that recommends videos based on an intelligent recommendation model. Traditional video processing systems are affected by the low timeliness of intelligent recommendation models, the small number of training samples, and the uneven distribution of training samples, resulting in inaccurate prediction of recommendation scores for videos with few samples and high timeliness. This leads to videos with low conversion rates receiving more exposure than videos with high conversion rates, resulting in a serious imbalance in video exposure, reducing the overall distribution efficiency of videos, and degrading the user experience.
[0081] For example, in scenarios with a small user base, the training sample size will be small, resulting in insufficient model training and inaccurate scoring. Another example is the low update frequency of the intelligent recommendation model; if the model hasn't been trained on newly released or unexposed videos during certain periods, its learning ability will be poor, leading to inaccurate predictions of recommendation scores. Furthermore, for time-sensitive videos, such as animated videos, fewer people watch them during weekdays and late at night; distributing too many animated videos during these times will result in lower conversion rates.
[0082] Based on the aforementioned technical issues, this application provides a video processing method that enables all videos to receive more reasonable exposure. By combining the estimated recommendation score, it increases the exposure of high-quality videos, new videos, and long-tail videos, while reducing the exposure of low-quality videos, thereby improving the video exposure conversion rate and the diversity of video exposure.
[0083] Reference Figure 3 , Figure 3 A flowchart illustrating a specific implementation of a video processing method provided in this application embodiment is shown below. Figure 3 As shown, this method includes:
[0084] S301. In response to reaching the preset real-time statistical timing, based on the number of times the candidate video is exposed, obtain the real-time index values of each distribution efficiency index of the candidate video.
[0085] In this embodiment, the real-time statistical timing includes the data update time that reaches the data update frequency indication. It should be noted that, unlike the intelligent recommendation model update which requires high computing resources and a large amount of sample data, the computational load of indicator statistics and calculation is low. Therefore, the frequency and delay of data updates can reach the second level.
[0086] In this embodiment, the candidate video is any one of the videos to be exposed from the set of videos to be exposed. Distribution efficiency metrics include video click-through rate (CTR), average video playback time (PPUI), and video playback rate (effic_vv_ratio).
[0087] The real-time metric for video click-through rate (CTR) is the ratio of the number of video clicks to the number of video impressions, i.e.:
[0088] CTR = Video clicks / Video impressions.
[0089] The real-time metric for video playback duration is the ratio of the total video playback duration to the number of video exposures, that is:
[0090] ppui = Total video playback time / Number of video exposures.
[0091] The real-time metric for video play rate is the ratio of the number of times a video is played to the number of times the video is exposed, that is:
[0092] effic_vv_ratio = Number of video plays / Number of video exposures.
[0093] S302. Determine whether the number of times the candidate video is exposed is lower than the preset number threshold.
[0094] In this embodiment, the number of exposures threshold is used to evaluate whether the real-time index value of the distribution efficiency index is reliable. The fewer the number of exposures, the more unstable the real-time index value of the distribution efficiency index is. The number of exposures threshold is used to determine whether the number of exposures in a stable state has been reached.
[0095] S303. If the number of video exposures is lower than the preset threshold, the real-time value of the distribution efficiency index is numerically smoothed based on the average value of the distribution efficiency index to obtain a reference value for the distribution efficiency index.
[0096] In this embodiment, the average index value is the average of the real-time index values of the distribution efficiency index of all videos to be exposed.
[0097] In this embodiment, the specific method for numerical smoothing of the real-time value of the distribution efficiency index includes:
[0098] 1. Based on the number of video exposures, determine the first trial weight, the second trial weight, and the smoothing coefficient.
[0099] The first trial weight is the ratio of the number of video exposures n to the number of exposures threshold N, n / N. The sum of the first trial weight and the second trial weight is 1, that is, the second trial weight is (1-n / N). The smoothing coefficient K is related to the video type, importance, expected benefits, etc. of the candidate video. For example, it is positively correlated with the importance or expected benefits, or K is a dynamic coefficient that is inversely correlated with n.
[0100] 2. Obtain the product of the smoothing coefficient K and the average index value avg_x, and use it as the initial index value.
[0101] 3. Using the first trial weight as the weighting coefficient for the real-time indicator value and the second trial weight as the weighting coefficient for the initial indicator value, the real-time indicator value and the initial indicator value are weighted and summed to obtain the reference indicator value.
[0102] In summary, let the real-time indicator value be x. The calculation method for the reference indicator value x_smooth after numerical smoothing is shown in the following formula:
[0103] x_smooth= (1-n / N)×K×avg_x+n / N×x.
[0104] Specifically, the calculation methods for the reference metric values CTR_smooth (video click-through rate), ppui_smooth (video average playback duration), and effic_vv_ratio_smooth (video playback ratio) are shown in the following formulas:
[0105] CTR_smooth = (1 - n / N) × (K × avg_CTR) + n / N × CTR, where K × avg_CTR represents the initial value of the video click-through rate, and CTR represents the real-time value of the video click-through rate.
[0106] ppui_smooth = (1-n / N) × (K × avg_ppui) + n / N × ppui, where K × avg_ppui represents the initial index value of the average video playback duration, and ppui represents the real-time index value of the average video playback duration.
[0107] effic_vv_ratio_smooth = (1 - n / N) × (K × avg_effic_vv_ratio) + n / N × effic_vv_ratio, where K × avg_effic_vv_ratio represents the initial value of the video playback rate, and effic_vv_ratio represents the real-time value of the video playback rate.
[0108] Understandably, the closer the video exposure count is to the threshold, the more reliable the real-time metric value, the greater the impact of the real-time metric value on the reference metric value, and the closer the reference metric value is to the real-time metric value. Conversely, the fewer the video exposure counts, the less reliable the real-time metric value, the smaller the impact of the real-time metric value on the reference metric value, and the closer the reference metric value is to the initial metric value. Furthermore, the initial metric value is related to the overall real-time metric value of the video. The smoothing coefficient is used to adjust the initial metric value based on the average metric value. The initial metric value of a candidate video is determined by combining the video attributes of the candidate video (such as type, importance, or expected return). For example, a new candidate video with high importance can be assigned a higher initial metric value even if the exposure count is very low, thereby improving the reference metric value of the new candidate video under low exposure conditions.
[0109] S304. If the number of video exposures is not lower than the number threshold, the real-time value of the distribution efficiency index will be used as the reference value of the distribution efficiency index.
[0110] It should be noted that for high-exposure videos with a number of exposures not less than the threshold, the real-time value of the distribution efficiency index is highly reliable. The real-time value of the distribution efficiency index can be used as a reference value for the distribution efficiency index. That is, when n is greater than N, x_smooth=x.
[0111] S305. Based on the reference index value and average index value of each distribution efficiency index, standardize each distribution efficiency index to obtain the standard index value of each distribution efficiency index.
[0112] In this embodiment, standardization is performed to unify the various distribution efficiency indicators to the same dimension. Logarithmic transformation is used to unify the dimensions of each distribution efficiency indicator based on its reference and average values. The standard indicator value is denoted as Wx. The standardization method is shown in the following formula:
[0113] Wx= log(1.0 + x_smooth / avg_x);
[0114] Specifically, the calculation methods for the standard metrics of video click-through rate (CTR) (WCTR), average video playback duration (PPUI) (WPPUI), and video playback rate (effic_vv_ratio) (Weffic_vv_ratio) are described in the following embodiments:
[0115] WCTR= log(1.0 + CTR_smooth / avg_CTR);
[0116] Wppui= log(1.0 + ppui_smooth / avg_ppui);
[0117] Weffic_vv_ratio= log(1.0 + effic_vv_ratio_smooth / avg_effic_vv_ratio).
[0118] S306. Add the standard index values of each distribution efficiency index to obtain the content quality score of the candidate video.
[0119] In this embodiment, when different distribution efficiency indicators are fused, logarithmic transformation is used to unify the dimensions of each distribution efficiency indicator, thereby ensuring that the distribution efficiency indicators at each scale are uniform.
[0120] In this embodiment, the content quality score WX of the candidate video is calculated using the following formula:
[0121] WX=WCTR+Wppui+Weffic_vv_ratio.
[0122] S307. Obtain the estimated recommendation score of the candidate video. The estimated recommendation score is obtained by the prediction of the pre-trained intelligent recommendation model.
[0123] In this embodiment, the intelligent recommendation model is trained using historical videos as sample videos and with the estimated recommendation score as the target output, wherein the estimated recommendation score is positively correlated with the conversion rate.
[0124] S308. Obtain the normalized weights. Based on the normalized weights, obtain the scaled result of the highest content quality score.
[0125] In this embodiment, the scaling result is the product of the normalized weight and the highest content quality score. The normalized weight is positively correlated with the accuracy of the intelligent recommendation model.
[0126] S309. Obtain the first normalized reference value and the second normalized reference value of the candidate video.
[0127] In this embodiment, the first normalized reference value is equal to the sum of the content quality score and the scaled result of the highest content quality score. The second normalized reference value is equal to the sum of the highest content quality score and the scaled result of the highest content quality score.
[0128] S310. Based on the first normalized reference value and the second normalized reference value, obtain the normalized quality score of the candidate video.
[0129] In this embodiment, the normalized quality is defined as the ratio of the first normalized reference value to the second normalized reference value.
[0130] S308~S310 provides a specific method for scaling and normalizing the content quality score. Based on S308~S310, the calculation method for the normalized quality score Wr of the candidate video is shown in the following formula:
[0131] Wr = (WX + L ×max_WX) / ( (1+ L) × max_WX);
[0132] Where L is the normalization weight, max_WX is the highest content quality score, which is the maximum content quality score of all videos, and L × max_WX is the scaling result of the highest content quality score. WX + L × max_WX represents the first normalization reference value, (1+ L) × max_WX represents the second normalization reference value, and Wr is equal to the ratio of the first normalization reference value and the second normalization reference value.
[0133] As can be seen from the above formula, the larger L is, the closer the maximum and minimum values of the normalized mass score are.
[0134] S311. Obtain the product of the estimated recommendation score and the normalized quality score of the candidate video, and use it as the video recommendation score of the candidate video.
[0135] In this embodiment, the video recommendation score W is the product of the estimated recommendation score Wy and the normalized quality score Wr, that is, W = Wy × Wr. The video recommendation score is positively correlated with the estimated recommendation score and the normalized quality score, respectively.
[0136] It should be noted that the estimated recommendation score and content quality score are balanced by adjusting the normalization weight L. The larger L is, the greater the role of the estimated recommendation score, and the smaller L is, the greater the role of the content quality score.
[0137] S312. Based on the video recommendation rate of candidate videos, determine the exposure judgment result of candidate videos.
[0138] In this embodiment, if a candidate video meets the preset exposure conditions, the exposure determination result of the candidate video is determined to be an exposed candidate video. The exposure conditions include that the video recommendation degree is greater than the preset recommendation degree threshold.
[0139] As can be seen from the above technical solutions, the data processing method provided in this application addresses the issue that low-exposure videos (those with fewer than a threshold exposure count) have lower reliability in real-time distribution efficiency metrics. Therefore, by averaging the metric values, the real-time distribution efficiency metrics are numerically smoothed to obtain reference metric values, thereby improving the reliability of each distribution efficiency metric. Furthermore, the closer the video exposure count is to the threshold, the closer the reference metric value is to the real-time metric value; conversely, the lower the video exposure count, the closer the reference metric value is to the initial metric value. The initial metric value is related to factors such as the video's importance and revenue, thus improving the reliability of the metric values for low-exposure videos and further enhancing the accuracy of the content quality score in representing the quality of candidate videos. Therefore, increasing the exposure probability of high-quality videos and decreasing the exposure probability of low-quality videos improves the overall scene effect.
[0140] Furthermore, compared to the daily or even lower update frequency of intelligent recommendation models, the content quality score in this application can be calculated in real time, thereby improving the real-time perception of video changes, enabling the reasonable distribution of long-tail videos and new videos, and improving the conversion rate and diversity of video exposure.
[0141] In conclusion, Figure 4 This is a schematic diagram illustrating the effect of a video processing method provided in an embodiment of this application, such as... Figure 4As shown, when candidate videos are long-tail content or new content that has not been exposed before, i.e., the video exposure count is 0, based on the above embodiment, the content quality score is calculated based on various initial indicator values, i.e., the initial content quality score. The initial indicator values of each distribution efficiency indicator are greater than low-quality content and lower than high-quality content. Therefore, based on the estimated recommendation score and content quality score, it is further determined whether to expose the video, thereby increasing the exposure probability of long-tail content and new content. When the video exposure count reaches the threshold, the content quality score is calculated based on the real-time indicator value. Based on the estimated recommendation score and content quality score, it is determined whether the candidate video becomes high-quality content due to a high content quality score or low-quality content due to a low content quality score. If it is high-quality content, a low video recommendation degree is obtained by combining the estimated recommendation score, increasing exposure; if it is low-quality content, a high video recommendation degree is obtained by combining the estimated recommendation score, decreasing exposure. It can be understood that by merging the content quality score of frequently updated candidate videos with the estimated recommendation score, the exposure probability of new videos and long-tail videos will increase, and the number of video exposures will increase, relative to the estimated recommendation score.
[0142] In summary, ensuring higher exposure probability for high-efficiency videos can improve the real-time performance of scene effects and alleviate scoring inaccuracies in scenarios where model timeliness is low or sample size is small.
[0143] The above describes a video processing method provided by an embodiment of this application. The following describes the apparatus for performing the above video processing method.
[0144] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a video processing apparatus provided in an embodiment of this application. Figure 5 As shown, the video processing apparatus 500 includes:
[0145] The real-time value acquisition unit 501 is used to acquire, for each candidate video, the real-time index value of each distribution efficiency index of the candidate video based on the number of video exposures of the candidate video, wherein different distribution efficiency indices are indices that characterize the distribution efficiency of the candidate video from different dimensions.
[0146] The reference value acquisition unit 502 is used to, if the number of video exposures is lower than a preset threshold, perform numerical smoothing on the real-time value of the distribution efficiency index based on the average value of the distribution efficiency index to obtain a reference value of the distribution efficiency index, wherein the average value is equal to the real-time value of the distribution efficiency index of all videos to be exposed, divided by the number of all videos to be exposed; if the number of video exposures is not lower than the threshold, the real-time value of the distribution efficiency index is used as the reference value of the distribution efficiency index.
[0147] Exposure determination unit 503 is used to determine the exposure determination result of the candidate video based on the reference index values of each of the distribution efficiency indicators of the candidate video, wherein the exposure determination result indicates whether the candidate video should be exposed.
[0148] This application also provides an electronic device in its embodiments. (See reference...) Figure 6 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 6 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.
[0149] like Figure 6 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. When the electronic device is powered on, the RAM 603 also stores various programs and data required for the operation of the electronic device. The processing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.
[0150] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 608 including, for example, memory cards, hard drives, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.
[0151] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the video processing methods provided in this application.
[0152] This application also provides a computer-readable storage medium that carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the video processing methods provided in this application.
[0153] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.
[0154] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.
[0155] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.
[0156] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).
Claims
1. A method of video processing, the method comprising: The method comprises the following steps: For each candidate video, based on the video exposure times of the candidate video, real-time indicator values of various distribution efficiency indicators of the candidate video are obtained, and different distribution efficiency indicators represent the distribution efficiency of the candidate video from different dimensions; If the video exposure times are lower than a preset number threshold, based on an average indicator value of the distribution efficiency indicators, the real-time indicator values of the distribution efficiency indicators are subjected to numerical smoothing processing to obtain reference indicator values of the distribution efficiency indicators, and the average indicator value is equal to the real-time indicator values of the distribution efficiency indicators of all to-be-exposed videos divided by the number of all the to-be-exposed videos; If the video exposure times are not lower than the number threshold, the real-time indicator values of the distribution efficiency indicators are taken as the reference indicator values of the distribution efficiency indicators; Based on the reference indicator values of the various distribution efficiency indicators of the candidate video, an exposure determination result of the candidate video is determined, and the exposure determination result indicates whether the candidate video is exposed.
2. The video processing method of claim 1, wherein, The method comprises the following steps: A preset fusion algorithm is used to fuse the reference indicator values of the various distribution efficiency indicators to obtain a content quality score of the candidate video; Based on the content quality score of the candidate video, a video recommendation degree of the candidate video is obtained; If the candidate video meets a preset exposure condition, it is determined that the exposure determination result of the candidate video is to expose the candidate video, and the exposure condition comprises that the video recommendation degree is greater than a preset recommendation degree threshold.
3. The video processing method of claim 1, wherein, The method comprises the following steps: Based on the video exposure times, a first exploration weight, a second exploration weight and a smoothing coefficient are determined, the first exploration weight is a ratio of the video exposure times to the number threshold, and the sum of the first exploration weight and the second exploration weight is 1; A product of the smoothing coefficient and the average indicator value is obtained as an initial indicator value; The first exploration weight is taken as a weight coefficient of the real-time indicator value, the second exploration weight is taken as a weight coefficient of the initial indicator value, and the real-time indicator value and the initial indicator value are subjected to weighted summation to obtain the reference indicator value.
4. The video processing method of claim 2, wherein, The method comprises the following steps: Each distribution efficiency indicator is subjected to standardization processing to obtain a standard indicator value of each distribution efficiency indicator; The standard indicator values of the various distribution efficiency indicators are added to obtain the content quality score of the candidate video.
5. The video processing method of claim 2, wherein, The method comprises the following steps: obtain an estimated recommendation score of the candidate video, the estimated recommendation score being predicted by a pre-trained intelligent recommendation model, the intelligent recommendation model being trained with historical videos as sample videos and with the estimated recommendation score as a target output, wherein the estimated recommendation score is positively correlated with a conversion rate; normalize the content quality score of the candidate video to obtain a normalized quality score of the candidate video; obtain a video recommendation degree of the candidate video based on the estimated recommendation score and the normalized quality score of the candidate video, the video recommendation degree being positively correlated with the estimated recommendation score and the normalized quality score respectively.
6. The video processing method of claim 5, wherein, The video recommendation degree is a product of the estimated recommendation score and the normalized quality score. The normalization operation on the content quality score of the candidate video to obtain the normalized quality score of the candidate video includes: calculate a normalization weight based on an accuracy of the intelligent recommendation model, the normalization weight being positively correlated with the accuracy of the intelligent recommendation model; calculate a product of the normalization weight and a highest content quality score to obtain a scaling result of the highest content quality score, the highest content quality score being a highest value of the content quality scores of all the candidate videos; obtain a first normalization reference value of the candidate video, the first normalization reference value being equal to a sum of the content quality score and the scaling result of the highest content quality score; obtain a second normalization reference value of the candidate video, the second normalization reference value being equal to a sum of the highest content quality score and the scaling result of the highest content quality score; obtain the normalized quality score of the candidate video based on the first normalization reference value and the second normalization reference value, the normalized quality score being a ratio of the first normalization reference value to the second normalization reference value.
7. A video processing apparatus, comprising: The method comprises: a real-time value obtaining unit configured to, for each candidate video, obtain real-time indicator values of respective distribution efficiency indicators of the candidate video based on a video exposure number of the candidate video, different distribution efficiency indicators representing the distribution efficiency of the candidate video from different dimensions; a reference value obtaining unit configured to, if the video exposure number is lower than a preset number threshold, perform numerical smoothing processing on the real-time indicator values of the distribution efficiency indicators based on average indicator values of the distribution efficiency indicators to obtain reference indicator values of the distribution efficiency indicators, the average indicator values being equal to the real-time indicator values of the distribution efficiency indicators of all to-be-exposed videos divided by a number of all the to-be-exposed videos; and if the video exposure number is not lower than the number threshold, taking the real-time indicator values of the distribution efficiency indicators as the reference indicator values of the distribution efficiency indicators; an exposure determination unit configured to determine an exposure determination result of the candidate video based on the reference indicator values of the respective distribution efficiency indicators of the candidate video, the exposure determination result indicating whether to expose the candidate video.
8. A computer program product, characterised in that, The computer readable instructions, when executed on an electronic device, cause the electronic device to implement the video processing method of any one of claims 1 to 6.
9. An electronic device, comprising: An electronic device comprising at least one processor and a memory connected with the processor, wherein: the memory is configured to store a computer program; the processor is configured to execute the computer program to enable the electronic device to implement the video processing method according to any one of claims 1 to 6.
10. A computer storage medium, characterized in that, The storage medium has one or more computer programs carried thereon, and when the one or more computer programs are executed by an electronic device, the electronic device is enabled to implement the video processing method according to any one of claims 1 to 6.