Multimedia object recall method, device, equipment, storage medium and product

Abstract

The present application discloses a multimedia object recall method, apparatus, device, storage medium and product, which belongs to the field of computer technology. The method comprises: obtaining multimedia object request information corresponding to a target object; performing a first recall process based on the multimedia object request information to obtain at least two multimedia objects; obtaining resource consumption deviation information corresponding to at least two multimedia objects; performing a second recall process based on the resource consumption deviation information to obtain a target multimedia object. The technical solution provided by the embodiment of the present application preliminarily recalls multimedia objects through the request of the target object and obtains its corresponding resource consumption deviation information, and then determines the degree of deviation between the actual consumption resources of the multimedia object and the target consumption resources to reflect its display effect, and further recalls it based on this to obtain the target multimedia object, which can effectively improve the recall volume and recall efficiency of multimedia objects without using the personalized characteristics of the target object.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a multimedia object recall method, apparatus, device, storage medium, and product. Background Art

[0002] With the development of digital technology and Internet technology, various forms of information content services have become popular among the public, and the number of multimedia objects on the Internet has increased dramatically. Therefore, multimedia object recall has become more important.

[0003] In related technologies, on the one hand, user features can be obtained through acquisition, and on the other hand, multimedia features corresponding to multimedia objects can be obtained by performing feature extraction processing on multimedia objects or through the label information corresponding to multimedia objects. The acquired user features are then matched with the multimedia features to achieve the recall of multimedia objects.

[0004] The related technology is highly dependent on the user's characteristics. When the user turns off the personalized features, the recall amount of multimedia objects is small and the recall efficiency is low. Summary of the Invention

[0005] The embodiments of the present application provide a multimedia object recall method, apparatus, device, storage medium, and product, which can effectively improve the recall volume and recall efficiency of multimedia objects without using personalized features of the target objects.

[0006] According to one aspect of an embodiment of the present application, a method for recalling a multimedia object is provided, the method comprising:

[0007] Obtain multimedia object request information corresponding to the target object;

[0008] performing a first recall process based on the multimedia object request information to obtain at least two multimedia objects;

[0009] Obtaining resource consumption deviation information corresponding to the at least two multimedia objects, where the resource consumption deviation information is used to represent a degree of deviation between target resource consumption and actual resource consumption corresponding to the at least two multimedia objects, and the degree of deviation is used to represent a display effect corresponding to the at least two multimedia objects;

[0010] A second recall process is performed based on the resource consumption deviation information to obtain a target multimedia object.

[0011] According to one aspect of an embodiment of the present application, a multimedia object recall device is provided, the device comprising:

[0012] A request acquisition module is used to obtain multimedia object request information corresponding to the target object;

[0013] A first recall module, configured to perform a first recall process based on the multimedia object request information to obtain at least two multimedia objects;

[0014] a consumption deviation determining module, configured to obtain resource consumption deviation information corresponding to the at least two multimedia objects, wherein the resource consumption deviation information is used to represent a degree of deviation between target resource consumption and actual resource consumption corresponding to the at least two multimedia objects, and the degree of deviation is used to represent a display effect corresponding to the at least two multimedia objects;

[0015] The second recall module is configured to perform a second recall process based on the resource consumption deviation information to obtain a target multimedia object.

[0016] According to one aspect of an embodiment of the present application, a computer device is provided, which includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the above-mentioned multimedia object recall method.

[0017] According to one aspect of an embodiment of the present application, a computer-readable storage medium is provided, in which at least one instruction, at least one program, a code set or an instruction set is stored. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the above-mentioned multimedia object recall method.

[0018] According to one aspect of an embodiment of the present application, a computer program product is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to implement the multimedia object recall method.

[0019] The technical solutions provided in the embodiments of the present application can bring the following beneficial effects:

[0020] Through the multimedia object request of the target object, the multimedia object is preliminarily recalled, and the resource consumption deviation information corresponding to the recalled multimedia object is obtained, so as to determine the degree of deviation between the actual resource consumption corresponding to the multimedia object and the target resource consumption. Since the above-mentioned deviation degree can characterize the display effect of the multimedia object, further multimedia object recall processing can be performed based on the above-mentioned resource consumption deviation information to obtain the recalled target multimedia object. The entire recall process does not depend on the feature information of the target object, and realizes the decoupling of multimedia object recall and target object feature information, greatly reducing the dependence of multimedia recall on target object feature information, and can effectively improve the recall volume and recall efficiency of multimedia objects without using the personalized features of the target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0022] Figure 1 is a schematic diagram of an application program operating environment provided by an embodiment of the present application;

[0023] Figure 2 This is the process of the multimedia object recall method provided by an embodiment of the present application Figure 1 ;

[0024] Figure 3 This is the process of the multimedia object recall method provided by an embodiment of the present application Figure 2 ;

[0025] Figure 4 This is the process of the multimedia object recall method provided by an embodiment of the present application Figure 3 ;

[0026] Figure 5 This is the process of the multimedia object recall method provided by an embodiment of the present application Figure 4 ;

[0027] Figure 6 A schematic diagram illustrating an approximate nearest neighbor search process is shown as an example;

[0028] Figure 7 The flowchart of constructing the directional model is shown as an example;

[0029] Figure 8 The following is a schematic diagram showing an exemplary label-based recall process;

[0030] Figure 9 A schematic diagram of a recall data display page is shown as an example;

[0031] Figure 10 A schematic diagram of the technical architecture of a multimedia recall system is shown as an example;

[0032] Figure 11 A schematic diagram of a multimedia object recall branch is exemplarily shown;

[0033] Figure 12 The following is an exemplary diagram of the recall process of the non-personalized feature recall branch;

[0034] Figure 13 A schematic diagram illustrating an exemplary advertising recall process;

[0035] Figure 14 is a block diagram of a multimedia object recall device provided by one embodiment of the present application;

[0036] Figure 15 This is a structural block diagram of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0037] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0038] Please refer to Figure 1 , which shows a schematic diagram of an application program running environment provided by an embodiment of the present application. The application program running environment may include: a terminal 10 and a server 20.

[0039] Terminal 10 includes, but is not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, in-vehicle terminals, game consoles, e-book readers, multimedia playback devices, wearable devices, aircraft, and other electronic devices. Application clients can be installed in terminal 10. Embodiments of the present invention can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, and assisted driving.

[0040] In an embodiment of the present application, the above-mentioned application can be any application that can provide a display of multimedia objects. Typically, the application is an information content service application. Of course, in addition to information content service applications, multimedia objects can also be displayed in other types of applications. For example, video applications, news applications, social applications, interactive entertainment applications, browser applications, shopping applications, content sharing applications, virtual reality (VR) applications, augmented reality (AR) applications, etc., which are not limited in this embodiment of the present application. In addition, for different applications, the multimedia objects displayed will also be different, and the corresponding functions will also be different. This can be pre-configured according to actual needs, which is not limited in this embodiment of the present application. Optionally, a client of the above-mentioned application is running in the terminal 10. Optionally, the above-mentioned multimedia objects include multimedia advertisements.

[0041] The server 20 is used to provide background services for the client of the application in the terminal 10. For example, the server 20 can be the background server of the above-mentioned application. The server 20 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. Optionally, the server 20 provides background services for applications in multiple terminals 10 at the same time.

[0042] Optionally, the terminal 10 and the server 20 may communicate with each other via a network 30. The terminal 10 and the server 20 may be directly or indirectly connected via wired or wireless communication, which is not limited in this application.

[0043] Before introducing the method embodiments provided in the present application, a brief introduction is first given to the application scenarios, relevant terms or nouns that may be involved in the method embodiments of the present application to facilitate understanding by technical personnel in the field of the present application.

[0044] Smart Targeting: In the advertising world, targeting refers to the demographics to which an ad is delivered. Naturally, if these targeted demographics are handled well, the ad delivery will be effective; otherwise, it will result in a poor advertising experience for the advertiser. Given the importance of targeting, both advertisers and platforms naturally attach great importance to it. Advertisers desperately want their ads to reach the right targeted demographics, but due to capacity constraints, advertisers cannot precisely specify the targeted demographics. If the advertising system relies heavily on advertiser-defined targeting, it would be unfriendly to advertisers and difficult to achieve effective results. Therefore, advertising platforms need to achieve effective advertising results without relying on advertisers to select precise targeting (e.g., a broad targeting setting, such as for males, or even no targeting at all). To achieve this, advertising platforms must devise a strategy to find the right target demographic for each ad. This process of identifying the right target demographic for each ad is known as smart targeting.

[0045] Original targeting: Advertisers independently set the targeting combination conditions in addition to using the intelligent targeting function (including automatic expansion and system optimization).

[0046] Non-breakable targeting: Also known as non-ignorable targeting, this is a targeting condition that advertisers set independently and must be met no matter what.

[0047] Automatic expansion: Based on the advertiser's manual selection of precise narrow targeting, the system performs intelligent expansion.

[0048] System optimization: Based on the advertiser's manual selection of broad-based targeting, the system makes intelligent adjustments.

[0049] LookAlike (similar audience targeting): Based on a small number of high-quality groups of people (old customers) provided by advertisers, the system automatically expands to a larger number of similar groups of people (potential new customers).

[0050] Advertising concentration problem: also known as the head advertising problem, mainly manifests itself in the fact that different users tend to retrieve and recall the same small set of advertisements, resulting in other advertisements having no exposure opportunities.

[0051] Please refer to Figure 2 , which shows the process of the multimedia object recall method provided by an embodiment of the present application Figure 1 This method can be applied to computer equipment, which refers to electronic equipment with data calculation and processing capabilities. For example, the execution subject of each step can be Figure 1 The server 20 in the application running environment is shown. The method may include the following steps (210-240).

[0052] Step 210: Acquire multimedia object request information corresponding to the target object.

[0053] Optionally, the multimedia object request information includes request start address data. Optionally, the request start address data is an IP (Internet Protocol) address. The multimedia object request information refers to request information sent by a client corresponding to the multimedia object, requesting the server to push the multimedia object.

[0054] Step 220: Perform a first recall process based on the multimedia object request information to obtain at least two multimedia objects.

[0055] Optionally, the at least two multimedia objects are multimedia objects that match the target region where the target object is located.

[0056] The first recall process is a recall process performed on a multimedia object request.

[0057] In an exemplary embodiment, the multimedia object request information includes request start address data. Figure 3 As shown, the implementation process of the above step 220 may include the following steps (221-223): Figure 3 The process of the multimedia object recall method provided by an embodiment of the present application is shown Figure 2 .

[0058] Step 221: Determine the region information corresponding to the requested starting address data.

[0059] Optionally, the above-mentioned region information includes a region identifier corresponding to the target region where the target object is located.

[0060] Step 222: Acquire the regional feature information corresponding to the multimedia object set.

[0061] Optionally, the regional feature information includes a delivery region identifier corresponding to each multimedia object in the multimedia object set.

[0062] Step 223: Match the region information with the region feature information to obtain at least two multimedia objects.

[0063] The region identifier corresponding to the target object is matched with the delivery region identifiers corresponding to the multimedia objects in the multimedia object set to obtain at least two multimedia objects whose delivery region identifiers are consistent with or associated with the region identifiers.

[0064] For example, the region identifier corresponding to the target object is "S", and all multimedia objects delivered to "S" are obtained.

[0065] The embodiment of the present application performs preliminary advertisement recall by parsing the requested IP address, which can improve the recall capability of multimedia objects without using the personalized features of the target object, and recall multimedia objects that match the region where the target object is located, thereby ensuring the recall quality.

[0066] Step 230: Obtain resource consumption deviation information corresponding to at least two multimedia objects.

[0067] Optionally, the resource consumption deviation information is used to represent a deviation degree between target resource consumption and actual resource consumption corresponding to at least two multimedia objects, and the deviation degree is used to represent a display effect corresponding to the at least two multimedia objects.

[0068] The following describes the process of obtaining resource consumption deviation information.

[0069] In an exemplary embodiment, as Figure 3 As shown, the implementation process of the above step 230 may include the following steps (231-233).

[0070] Step 231: Obtain target resource consumption data corresponding to at least two multimedia objects.

[0071] Optionally, the target resource consumption data is used to represent the target resource consumption amount. The target resource consumption data may be pre-set resource consumption data.

[0072] The target resource consumption data reflects the total amount of resources consumed by pushing the at least two multimedia objects, including but not limited to traffic, computing resources, fees, etc. Optionally, the target resource consumption can be simply understood as the estimated amount of resources required to achieve a predetermined multimedia object push effect.

[0073] In one possible implementation, Figure 4 As shown, the implementation process of the above step 231 may include the following steps (2311-2312): Figure 4 The process of the multimedia object recall method provided by an embodiment of the present application is shown Figure 3 .

[0074] Step 2311: Obtain historical operation data and resource consumption preset data corresponding to at least two multimedia objects.

[0075] Optionally, the historical operation data is used to represent an operation event triggered by at least two multimedia objects. Optionally, the resource consumption data is used to represent a preset amount of resource consumption corresponding to the operation event.

[0076] Optionally, the resource consumption preset data reflects the expected value of consumption corresponding to the triggering operation event on the multimedia object. For example, the multimedia publishing object sets the resource consumption preset data to represent the planned resource consumption of the triggering operation event.

[0077] In a typical application scenario, such as an advertising scenario, the above-mentioned resource consumption preset data includes a target resource consumption preset value set by the publishing object. Optionally, the target resource consumption preset value is used to characterize the preset resource consumption corresponding to a single target operation event triggered on a multimedia object. Optionally, the target resource consumption preset value is the single conversion target cost data (target_cpa) corresponding to the conversion operation event. The above-mentioned single conversion target cost data reflects the cost of a single conversion. For example, the above-mentioned single conversion target cost data is the budget expenditure corresponding to a single advertising conversion set by the advertiser.

[0078] Step 2312: fuse the historical operation data with the preset resource consumption data to obtain target resource consumption data.

[0079] Optionally, the number of operation events triggered on the at least two multimedia objects can be determined based on the historical operation data, and the preset resource consumption data includes a preset resource consumption amount corresponding to each operation event. Therefore, the target resource consumption data can be obtained by summing the preset resource consumption amounts corresponding to each operation event.

[0080] In one possible implementation, the target operation event includes a conversion operation event, and the preset resource consumption data includes a preset resource consumption value corresponding to the conversion operation event. The target resource consumption data can be obtained by combining the quantity corresponding to the conversion operation event (i.e., conversion volume) with the preset resource consumption value (such as the target_cpa described above). Optionally, the target resource consumption data is a target gross merchandise volume (GMV).

[0081] It can be seen that this embodiment calculates the target consumption by using the target conversion fee and the conversion volume, thereby improving the accuracy of the cost deviation data.

[0082] Step 232: Acquire actual resource consumption data corresponding to at least two multimedia objects.

[0083] Optionally, the actual resource consumption data is used to represent the actual resource consumption. Optionally, the actual resource consumption can be simply understood as: the actual resource consumption to achieve a predetermined multimedia object push effect.

[0084] Optionally, the above actual resource consumption data can be acquired by the system in real time during the multimedia object pushing process.

[0085] Step 233 : Compare the actual resource consumption data with the target resource consumption data to obtain resource consumption deviation data.

[0086] Optionally, the resource consumption deviation information includes resource consumption deviation data.

[0087] Optionally, the resource consumption deviation data corresponding to any multimedia object is the ratio between the actual resource consumption data corresponding to the multimedia object and the target resource consumption data, that is, resource consumption deviation data = actual resource consumption data / target resource consumption data.

[0088] If the actual resource consumption data exceeds the target resource consumption data, it indicates excessive resource consumption, resulting in a poor display quality for the multimedia object. If the actual resource consumption data is less than the target resource consumption data, it indicates that there is idle resources, and the display quality for the multimedia object is good, but there is resource waste. Therefore, the resource consumption deviation data can represent the display quality of the multimedia object and the difference in resource consumption from the preset resource consumption amount.

[0089] In an example, as shown in Table 1 below, it reflects the process of determining resource consumption deviation data corresponding to a multimedia object.

[0090] Table 1

[0091]

[0092]

[0093] Among them, target resource consumption data GMV = target_cpa * conversion volume; resource consumption deviation data = actual resource consumption data / GMV.

[0094] The embodiment of the present application can determine resource consumption deviation data that can reflect cost deviation information by comparing actual resource consumption data and target consumption data (GMV), and the resource consumption deviation data can accurately reflect the display effect of the multimedia object. Subsequent further recall based on the data can improve the recall quality, thereby improving the display effect of the recalled multimedia object on the target object side.

[0095] Step 240: Perform a second recall process based on the resource consumption deviation information to obtain a target multimedia object.

[0096] Optionally, the above steps 210 to 240 are performed according to a target period. Optionally, the target period is 1 hour.

[0097] In an exemplary embodiment, the resource consumption deviation information includes resource consumption deviation data corresponding to at least two multimedia objects. Figure 3As shown, the implementation process of the above step 240 may include the following steps (241-244).

[0098] Step 241 : performing classification processing on the resource consumption deviation data to obtain at least two deviation levels.

[0099] Optionally, the resource consumption deviation data corresponding to each multimedia object is sorted to obtain a deviation data sorting result. The deviation data sorting result may be a resource consumption deviation data sequence. The sorting may be in ascending or descending order. Optionally, in the embodiment of the present application, the sorting is performed in ascending order.

[0100] The resource consumption deviation data sequence is graded to obtain at least two deviation grades. Optionally, the resource consumption deviation data in the resource consumption deviation data sequence can be graded into a preset number of deviation grades, such as 100 grades.

[0101] The following example illustrates the grading process. For example, if the resource consumption deviation data sequence is [0.2, 0.23, 0.24, 0.3, 0.31, 0.32, 0.34, 0.36, 0.4, 0.5, 0.54, 0.56, 0.64, 0.7, 0.9, 1.0, 1.11, 1.23, 1.25, 1.31], dividing the resource consumption deviation data into four levels will result in three thresholds: 0.31, 0.5, and 0.9. 0.31 is the threshold for the first deviation level. Data less than or equal to this threshold belongs to the first deviation level, and so on.

[0102] Step 242: Obtain target level parameter data.

[0103] The target level parameter data is used to characterize the order information corresponding to the target deviation level.

[0104] Optionally, the target level parameter data is an order identifier corresponding to the target deviation level. For example, an order identifier 1 corresponding to the target deviation level indicates that the target deviation level is the first deviation level among at least two deviation levels.

[0105] Optionally, the target level parameter data includes first level parameter data and second level parameter data. Optionally, the first level parameter data is data corresponding to the first level parameter (which may be denoted as level 1), and the second level parameter data is data corresponding to the second level parameter (which may be denoted as level 2). The first level parameters and the second level parameters have initial data.

[0106] In an example, the number of deviation levels is 100, the initial level 1 = 20, and the initial level 2 = 80, which means that the first deviation level corresponding to level 1 is the 20th level among the above 100 levels, and the second deviation level corresponding to level 2 is the 80th level among the above 100 levels.

[0107] The above target level parameter data can be updated dynamically. The update process is as follows:

[0108] In an exemplary embodiment, as Figure 4 As shown, the implementation process of the above step 242 may include the following steps (2421-2424).

[0109] Step 2421: Determine a multimedia object whose resource consumption deviation data is smaller than a first deviation data threshold as a first multimedia object.

[0110] The first multimedia object refers to a multimedia object whose resource consumption deviation data is smaller than a first deviation data threshold.

[0111] Optionally, the first deviation data threshold is 0.8.

[0112] Optionally, in an advertising scenario, the first multimedia object may be a poorly-performing advertisement. For example, a poorly-performing advertisement is an advertisement whose actual resource consumption data is less than the target resource consumption data and a first deviation data threshold.

[0113] Step 2422: Determine the multimedia object whose resource consumption deviation data is greater than the second deviation data threshold as the second multimedia object.

[0114] The second multimedia object resource consumption deviation data is greater than the second deviation data threshold value of the multimedia object.

[0115] Optionally, the second deviation data threshold is 1.2.

[0116] Optionally, in the advertising field, the second multimedia object may be an overpaid advertisement. For example, an overpaid advertisement is an advertisement with actual resource consumption data greater than target resource consumption data*a second deviation data threshold.

[0117] Step 2423: Acquire first resource remaining data corresponding to the first multimedia object and first resource excess data corresponding to the second multimedia object.

[0118] Optionally, the first resource margin data is used to represent the remaining degree of the target resource consumption corresponding to the first multimedia object. Optionally, the first resource margin data is the under-yield rate data corresponding to this operation.

[0119] In one possible implementation, the actual resource consumption data corresponding to the first multimedia object is compared with the sum of the actual resource consumption data corresponding to at least two multimedia objects to obtain the first resource margin data. Specifically, the first resource margin data (deficit rate data) = the actual resource consumption data corresponding to the first multimedia object / the sum of the actual resource consumption data.

[0120] Optionally, the first resource excess data is used to represent the degree of excess of actual resource consumption corresponding to the second multimedia object. Optionally, the first resource excess data is excess yield rate data corresponding to this operation.

[0121] In one possible implementation, the actual resource consumption data corresponding to the second multimedia object is compared with the sum of the actual resource consumption data corresponding to at least two multimedia objects to obtain the first resource excess data. Specifically, the first resource excess data (excess yield data) = actual resource consumption data corresponding to the second multimedia object / sum of the actual resource consumption data.

[0122] Step 2424: Based on the first resource margin data and the first resource excess data, the target level parameter data determined last time is updated to obtain the target level parameter data.

[0123] Optionally, the target level parameter data is preset data when it is acquired for the first time.

[0124] In a possible implementation, the target level parameter data includes first level parameter data and second level parameter data. Figure 5 As shown, the implementation process of the above step 2424 may include the following steps (2424a to 2424d): Figure 5 The process of the multimedia object recall method provided by an embodiment of the present application is shown Figure 4 .

[0125] Step 2424a: Obtain the second resource remaining data corresponding to the first multimedia object determined last, and the second resource excess data corresponding to the second multimedia object determined last.

[0126] Optionally, the second resource remainder data is resource remainder data corresponding to the first multimedia object determined last time. Optionally, the second resource excess data is resource excess data corresponding to the second multimedia object determined last time.

[0127] Step 2424b: Determine first difference data between the first resource margin data and the second resource margin data, and second difference data between the first resource excess data and the second resource excess data.

[0128] Optionally, the first difference data includes difference change data between the first resource margin data and the second resource margin data. Optionally, the second difference data includes difference change data between the first resource excess data and the second resource excess data.

[0129] Step 2424c: Based on the first difference data, update the first-level parameter data determined last time to obtain the first-level parameter data.

[0130] Optionally, the process of updating the first-level parameter data determined last time is as follows:

[0131] If the first resource margin data is smaller than the second resource margin data, the first level parameter data determined last time is subtracted from the first difference data to obtain the first level parameter data.

[0132] If the first resource margin data is greater than or equal to the second resource margin data, the first level parameter data determined last time is added to the first difference data to obtain the first level parameter data.

[0133] The above process is illustrated below using a specific example. In one example, the resource margin data is the yield reduction rate. Using the aforementioned method, the yield reduction rate corresponding to the current operation (the first resource margin data) is determined. This yield reduction rate is then compared with the previous calculation result from the previous hour (i.e., the second resource margin data). If the yield reduction rate has decreased by m%, then level 1 = level 1 – m; if the yield reduction rate has increased by m%, then level 1 = level 1 + m, where m is an integer.

[0134] Step 2424d: Based on the second difference data, update the second level parameter data determined last time to obtain the second level parameter data.

[0135] Optionally, the process of updating the second-level parameter data determined last time is as follows:

[0136] If the first resource excess data is less than the second resource excess data, subtracting the second level parameter data determined last time from the second difference data to obtain the second level parameter data;

[0137] If the first resource excess data is greater than or equal to the second resource excess data, the second level parameter data determined last time is added to the second difference data to obtain the second level parameter data.

[0138] The above process is illustrated below using a specific example. In one example, the resource excess data is the over-recovery rate. Using the aforementioned method, the over-recovery rate corresponding to the current operation (the first resource excess data) is determined. This over-recovery rate is then compared with the previous calculation result from the previous hour (the second resource excess data). If the over-recovery rate decreases by n%, level 2 equals level 2 – n; if the over-recovery rate increases by n%, level 2 equals level 2 + n.

[0139] Step 243 : determining a target deviation level based on the target level parameter data and at least two deviation levels.

[0140] Step 244: Determine the target multimedia object according to the target deviation level.

[0141] In an exemplary embodiment, the target level parameter data includes first level parameter data and second level parameter data. Figure 5 As shown, the above step 243 can be replaced by the following step 2431, and the above step 244 can be replaced by the following step 2441.

[0142] Step 2431 : Determine a first deviation level corresponding to the first-level parameter data among the at least two deviation levels, and a second deviation level corresponding to the second-level parameter data among the at least two deviation levels.

[0143] Step 2441 : Determine multimedia objects with a deviation level less than or equal to a first deviation level, and multimedia objects with a deviation level greater than a second deviation level, as target multimedia objects.

[0144] The above-mentioned multimedia objects with a deviation level less than or equal to the first deviation level, and the multimedia objects with a deviation level greater than the second deviation level are all multimedia objects with large resource consumption cost deviations. By recalling them, in addition to improving the recall quality, the resources consumed by pushing multimedia objects can be controlled within a reasonable range, thereby improving the controllability of resource consumption and the conversion efficiency of multimedia objects.

[0145] To sum up, the technical solution provided by the embodiment of the present application performs a preliminary recall of multimedia objects through the multimedia object request of the target object, and obtains the resource consumption deviation information corresponding to the recalled multimedia object, and then determines the degree of deviation between the actual resource consumption corresponding to the multimedia object and the target resource consumption. Since the above-mentioned deviation degree can characterize the display effect of the multimedia object, further multimedia object recall processing can be performed based on the above-mentioned resource consumption deviation information to obtain the recalled target multimedia object. The entire recall process does not depend on the feature information of the target object, and realizes the decoupling of multimedia object recall and target object feature information, greatly reducing the degree of dependence of multimedia recall on target object feature information, and can effectively improve the recall volume and recall efficiency of multimedia objects without using the personalized features of the target object.

[0146] The technical solutions and beneficial effects provided by the embodiments of the present application are described below in conjunction with specific application fields.

[0147] Exemplarily, the field of multimedia object push is a typical application field corresponding to the embodiments of the present application. In the above-mentioned multimedia object push field, multimedia objects in the multimedia object library can be intelligently oriented. The concept of intelligent orientation has been explained above. After confirming the function of intelligent orientation, it is necessary to consider how intelligent orientation should be implemented. The essence of intelligent orientation is to find the right people for multimedia objects. Its goal is to recall target multimedia objects with a high degree of matching for the target object when a multimedia object push request for the target object is received. Optionally, the above-mentioned multimedia object is an advertisement, and the recalled target multimedia object is an advertisement that matches the interests of the target object.

[0148] The above-mentioned matching degree can be determined in different ways, which are not limited in the embodiment of the present application. Schematically, the embodiment of the present application provides two different types of matching methods to determine the matching degree between the target object and the multimedia object.

[0149] In one possible implementation, an ANN (Approximate Nearest Neighbor) algorithm is used to trigger a branch for matching. The original purpose of ANN is to simulate the multimedia publishing objects corresponding to multimedia objects and mine audience packages. For example, advertisers typically use historical data to identify potential high-conversion audiences, which have relatively good advertising effectiveness. ANNs can automatically analyze historical data to identify potential high-conversion audiences, resulting in better multimedia object delivery results.

[0150] The following describes ANN with a specific example. Optionally, ANN can be implemented based on an artificial neural network, that is, an artificial neural network model is established to predict the matching degree between multimedia objects and account objects. For the specific process, please refer to Figure 6 , which exemplifies a schematic diagram of an approximate nearest neighbor search process. The general process is as follows:

[0151] Obtain multiple data sources. Optionally, the multimedia object is an advertisement. The data sources include account object data corresponding to the multimedia object, delivery data corresponding to the multimedia object (including converted object data), and platform data (including feature data corresponding to the target object and the multimedia object).

[0152] The above-mentioned multi-party data sources (where X1, X2, and X3 respectively represent different feature data corresponding to multimedia objects) are input into a directional model automatically constructed based on a machine learning model. This directional model is used to output the matching score between the multimedia object and each account object. The matching score is used to represent the above-mentioned matching degree.

[0153] Optionally, the above-mentioned directional model can generally be abstracted as a probability estimation problem of P(1|user (account object), context (context information), ad (multimedia object identifier)), that is, under a pre-defined target type, when the target object sees a multimedia object in a certain context environment, determine the probability of this target type occurring. This target can characterize the degree of matching between the multimedia object and the target object. Optionally, the target data that can characterize the above-mentioned matching degree include but are not limited to the click-through rate, click-through conversion rate, exposure conversion rate, etc. corresponding to the multimedia object. Optionally, different predefined target types can be modeled separately, constructed into different ANN models, and multiple trigger branches for matching are formed, and then multiple branches act together. For example, directional models such as click-through rate model, click-through conversion rate model, exposure conversion rate model, etc. are constructed, and then these directional models are used to determine whether the target object matches the multimedia object. Optionally, the target data is positively correlated with the degree of matching.

[0154] In one example, if Figure 7As shown, it exemplifies a flowchart for constructing a directional model. During the offline training phase, on the one hand, the object features of the sample object (e.g., the sample account object) and the context features corresponding to the context information are input into the object representation model, and the object features fused with the context information can be output; on the other hand, multimedia object information is input into the multimedia representation model, and the multimedia object features corresponding to the multimedia object can be output. By utilizing these features, label data (such as the above-mentioned click-through rate, click-through conversion rate, and exposure conversion rate), and related constraints, a corresponding directional model can be constructed. The trained directional model can output the object features fused with the context information corresponding to the target object based on the feature information corresponding to the target, such as the object features and context features corresponding to the target object, and perform matching retrieval with the multimedia object feature library to output the above-mentioned target data.

[0155] From the above Figure 7 The following conclusions can be drawn:

[0156] 1. Usually the model structure is a double tower structure, as shown above Figure 7 As shown, object features and context features are in one tower, and multimedia object information is in a separate tower.

[0157] Contextual information appears only on the account object side, not on the multimedia object side. This means that by default, the information on the account object side is context-dependent and may be a dynamic feature that changes with context. However, the information on the multimedia object side is context-independent and is therefore static.

[0158] 3. There is no limitation on the method of determining context features, and context features can be determined by various methods.

[0159] 4. In the offline phase, through training, the model structure and parameters of the account object tower are obtained; as well as the multimedia object features of each multimedia object on the multimedia object side, such as the embedding vector.

[0160] 5. In the online prediction stage, a context-based embedding vector is obtained based on the object features and context features of the target object of a certain request. This vector is used to find multimedia objects corresponding to several vectors similar to this vector from the embedding vectors corresponding to the multimedia objects as the target multimedia objects to be recalled.

[0161] The above content describes the ANN, and another possible implementation method is described below.

[0162] In a possible implementation, the branch is triggered to perform matching through TAG (tag-based recall).

[0163] In the previous implementation, more request details can be captured by refining and modifying contextual features, but there may be issues with multimedia object concentration. That is, many of the recalled multimedia objects resulting from requests are similar, and some multimedia objects do not receive reasonable exposure opportunities. Another issue is the issue of newly stored multimedia objects. The aforementioned model-based training essentially seeks patterns from historical data. However, for newly stored multimedia objects, historical data is relatively scarce, resulting in relatively inaccurate model training and estimation processes for these newly stored multimedia objects. Consequently, newly stored multimedia objects do not receive appropriate exposure opportunities.

[0164] The TAG branch adopts a new matching method that is different from the ANN branch, namely the tag recall method. The TAG branch is designed to simulate the process of selecting targeted tags for multimedia publishing objects. In short, in the tag recall method, you can assign tags to account objects, then assign tags to multimedia objects, and then match the account object tags with the multimedia object tags, and recall the matching ones. In an example, Figure 8 As shown, it exemplifies a schematic diagram of the label-based recall process. Among them, the label categories corresponding to the account object include behavior-oriented class, interest-oriented class, intention-oriented class, application installation-oriented class, and custom population class. Optionally, the label of the behavior-oriented class includes category label and keyword label; the label of the interest-oriented class includes category label and keyword label; the label of the intention-oriented class includes category label; the label of the application installation-oriented class includes the application label in the installed list; the label of the custom population class includes the label in the population package list. The labels corresponding to the multimedia object include creative label, application ID, related conversion object and account ID. The matching model matches according to the different labels on both sides to determine the recalled multimedia object.

[0165] As can be seen from the above description, the quality of account object tags and multimedia object tags, as well as the matching method, can affect the recall effect of the TAG branch. Therefore, the construction of the TAG branch can be considered from the following perspectives.

[0166] 1. Behavioral interest targeting tags corresponding to the account object.

[0167] In some potential application scenarios, multimedia object ads can be assigned tags based on the behavioral interest targeting tag system on the account object side. This allows directly matching multimedia objects to be recalled when a request occurs. For example, if a multimedia object is tagged with "Martial Arts Game Enthusiast" and the account object in a request also has this tag, the ad will be recalled for that request.

[0168] 2. Automatically bind application installation targeting tags.

[0169] In some possible application scenarios, similar applications corresponding to the target application are determined. If the application identifier corresponding to the target account object includes the application identifier corresponding to the target application, multimedia objects corresponding to the similar applications can be recalled when the target account object requests.

[0170] 3. Automatically bind custom audience targeting tags.

[0171] In some possible application scenarios, the account object cluster package corresponding to the publication object is determined as a label of the multimedia object corresponding to the publication object.

[0172] 4. Automatically bind the high TGI (Target Group Index) targeting tag of the converted object.

[0173] In some possible application scenarios, if we only look at the number of converted objects for a multimedia object, we'll find it quite sparse, so we'll roll back according to a specific pattern. Generally speaking, the rollback priority is as follows: multimedia object (e.g., advertisement) -> associated multimedia objects (e.g., product) -> publisher (e.g., advertiser) -> industry. This means we first check to see if there are sufficient converted objects for the same multimedia object. If not, we'll roll back to the converted objects for the associated multimedia object, and so on. A high TGI targeted tag refers to a high proportion of that targeted tag among the converted objects. This is a process of using results as feedback, attempting to strengthen the results through observed, relatively good tags.

[0174] As mentioned above, whether matching and recalling through the ANN branch or the TAG branch, it involves the use of the feature information corresponding to the target object. The multimedia object recall process is highly dependent on the feature information corresponding to the target object. For some users who have turned off personalized recommendations, if they still use the above two methods to match and recall multimedia objects, the amount of multimedia object recall will be sharply reduced. For details, please refer to Figure 9 , which exemplarily shows a schematic diagram of a recall data display page. Figure 9 The data shows that for requests for objects from accounts with personalized recommendations turned off, the average recall queue length is less than 500, while for requests from accounts with personalized recommendations turned on, the recall queue length is over 36,000. This shows that the recall volume of multimedia objects decreases dramatically after personalized recommendations are turned off.

[0175] The following is an analysis of the reasons for the sharp decrease in the recall of multimedia objects. Figure 10, which exemplifies a technical architecture diagram of a multimedia recall system. Among them, the digital marks "1", "2", "3", "4", and "5" represent the same positions respectively. As mentioned above, after the personalized recommendation setting is turned off, the recall volume of multimedia objects is sharply reduced, that is, the recall volume at position 5 is sharply reduced. For analysis, Table 2 below lists the recall volumes corresponding to positions 1 to 5 under different requests at different time points, and with the personalized recommendation setting turned on; Table 3 below lists the recall volumes corresponding to positions 1 to 5 under different requests at different time points, and with the personalized recommendation setting turned on.

[0176] Table 2 (Personalized recommendation is set to on)

[0177] ask Position 1 Position 2 Position 3 Position 4 Position 5 Request 1 24067 38901 47309 174909 36890 Request 2 28381 40910 54810 189032 37991 Request 3 23921 35438 42672 149843 32704 Request 4 27419 41341 52873 193671 36308 Request 5 29304 46937 50900 164793 38927

[0178] Table 3 (Personalized recommendation is set to off)

[0179] ask Position 1 Position 2 Position 3 Position 4 Position 5 Request 1 24182 15837 27093 98578 20904 Request 2 22048 11492 23802 74910 19560 Request 3 19387 21603 27789 84091 21807 Request 4 25712 14745 28801 100851 18938 Request 5 27361 23771 30027 117892 19765

[0180] From the above records, we can see that the recalls at positions 2 and 4 are significantly affected. The recall at position 2 is the recall of the TAG branch; the recall at position 4 is the recall corresponding to the non-breakthrough direction.

[0181] The technical solution provided in the embodiment of the present application can preliminarily recall the multimedia objects associated with the location requested by the multimedia object request of the target object according to the sending location, and then obtain the resource consumption deviation information corresponding to the recalled multimedia objects, and then determine the degree of deviation between the actual resource consumption corresponding to the multimedia objects and the target resource consumption. Since the above-mentioned deviation degree can characterize the display effect of the multimedia object, further multimedia object recall processing can be performed based on the above-mentioned resource consumption deviation information to obtain the recalled target multimedia object. The entire recall process does not depend on the characteristic information of the target object, and realizes the decoupling of multimedia object recall and target object characteristic information, which can effectively improve the recall volume and recall efficiency of multimedia objects without using the personalized characteristics of the target object.

[0182] The technical solution provided in the embodiment of the present application can be used as a separate branch to recall multimedia objects. Figure 11 As shown, it exemplifies a schematic diagram of a multimedia object recall branch. Among them, the part circled by the dotted box 111 is a non-personalized feature recall branch. The non-personalized feature recall branch can apply the technical solution provided by the implementation of this application to recall multimedia objects based on the request location and resource consumption deviation data. Please refer to Figure 12, which exemplarily shows a recall process diagram of the non-personalized feature recall branch. When the personalized feature setting status is off, it is determined to use the above-mentioned non-personalized feature recall branch; in the above-mentioned non-personalized feature recall branch, for the target object side, the system can determine the regional information corresponding to the target object based on the request start address data in the multimedia object request information, and then obtain the regional list; for the multimedia object side, the system can obtain the delivery region information (a kind of regional feature information) corresponding to the multimedia object and the historical statistical information corresponding to the multimedia object. The matching model can recall multimedia objects in the matching region for the target object based on the regional list and the delivery region information. Furthermore, the matching model can determine the above-mentioned resource consumption deviation information based on the above-mentioned historical statistical information, and then recall the multimedia objects in the matching region again based on the above-mentioned resource consumption deviation information to obtain the recalled target multimedia object.

[0183] In a specific application scenario, the multimedia object is an advertisement, and its corresponding recall process can be referred to Figure 13 , which exemplifies a schematic diagram of an advertisement recall process. The general process includes: obtaining an advertisement set; filtering the advertisements in the advertisement set by location based on the target object's geographical information to obtain the filtered advertisements; determining the resource consumption deviation data corresponding to the filtered advertisements and sorting them; taking the first level 1 and the last level 2 advertisements and recalling them, and sending the advertisements to the user based on the recall results. After a preset period, the resource consumption deviation data, as well as the above-mentioned level 1 and level 2, are adjusted based on the feedback data of the advertisements. Among them, level 1 and level 2 represent the above-mentioned first deviation level and second deviation level, respectively.

[0184] For the target object with personalized features set to off, multimedia objects are recalled through the above non-personalized feature recall branch, and the recall data is shown in Table 4 below.

[0185] Table 4 (Personalized recommendation is set to off)

[0186] ask Position 1 Position 2 Position 3 Position 4 Position 5 Request 1 24182 29805 35834 98578 29420 Request 2 22048 20940 32988 74910 27310 Request 3 19387 27642 32800 84091 22872 Request 4 25712 24941 34761 100851 26908 Request 5 27361 28813 39744 117892 28840

[0187] As can be seen, compared to Table 3 above, the recall amount at position 2 has been further enriched, thereby enriching the total length of the final recall queue. Therefore, the multimedia object recall based on the technical solution provided in the embodiment of the present application can effectively improve the recall amount and recall efficiency of multimedia objects without using the personalized features of the target object.

[0188] The following are device embodiments of the present application, which can be used to implement the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0189] Please refer to Figure 14 , which shows a block diagram of a multimedia object recall device provided by one embodiment of the present application. The device has the function of implementing the above-mentioned multimedia object recall method, and the function can be implemented by hardware or by hardware executing corresponding software. The device can be a computer device or can be set in a computer device. The device 1400 can include: a request acquisition module 1410, a first recall module 1420, a consumption deviation acquisition module 1430, and a second recall module 1440.

[0190] The request acquisition module 1410 is configured to acquire multimedia object request information corresponding to a target object.

[0191] The first recall module 1420 is configured to perform a first recall process based on the multimedia object request information to obtain at least two multimedia objects.

[0192] The consumption deviation acquisition module 1430 is used to obtain the resource consumption deviation information corresponding to the at least two multimedia objects. The resource consumption deviation information is used to characterize the degree of deviation between the target resource consumption and the actual resource consumption corresponding to the at least two multimedia objects. The degree of deviation is used to characterize the display effect corresponding to the at least two multimedia objects.

[0193] The second recall module 1440 is configured to perform a second recall process based on the resource consumption deviation information to obtain a target multimedia object.

[0194] In an exemplary embodiment, the resource consumption deviation information includes resource consumption deviation data corresponding to the at least two multimedia objects, and the second recall module 1440 includes: a deviation level determination unit, a target parameter acquisition unit, a target level determination unit, and a multimedia object determination unit.

[0195] The deviation level determination unit is used to perform hierarchical processing on the resource consumption deviation data to obtain at least two deviation levels.

[0196] The target parameter acquisition unit is used to acquire target level parameter data, where the target level parameter data is used to represent order information corresponding to the target deviation level.

[0197] A target level determination unit is configured to determine the target deviation level based on the target level parameter data and the at least two deviation levels.

[0198] The multimedia object determining unit is configured to determine the target multimedia object according to the target deviation level.

[0199] In an exemplary embodiment, the target level parameter data includes first level parameter data and second level parameter data, and the target level determination unit is specifically used to determine the first deviation level corresponding to the first level parameter data among the at least two deviation levels, and the second deviation level corresponding to the second level parameter data among the at least two deviation levels.

[0200] The multimedia object determination unit is specifically configured to determine, as the target multimedia objects, the multimedia objects whose deviation level is less than or equal to the first deviation level and the multimedia objects whose deviation level is greater than the second deviation level.

[0201] In an exemplary embodiment, the target parameter acquisition unit includes: a first object determination subunit, a second object determination subunit, a resource quantity data acquisition subunit, and a target parameter update subunit.

[0202] The first object determining subunit is configured to determine the multimedia object whose resource consumption deviation data is smaller than a first deviation data threshold as the first multimedia object.

[0203] The second object determining subunit is configured to determine the multimedia object whose resource consumption deviation data is greater than a second deviation data threshold as a second multimedia object.

[0204] The resource quantity data acquisition subunit is used to obtain the first resource surplus data corresponding to the first multimedia object and the first resource excess data corresponding to the second multimedia object. The first resource surplus data is used to characterize the remaining degree of the target resource consumption corresponding to the first multimedia object, and the first resource excess data is used to characterize the excess degree of the actual resource consumption corresponding to the second multimedia object.

[0205] The target parameter updating subunit is configured to update the target level parameter data determined last time based on the first resource margin data and the first resource excess data to obtain the target level parameter data, which is preset data when first obtained.

[0206] In an exemplary embodiment, the target level parameter data includes first level parameter data and second level parameter data, and the resource quantity data acquisition subunit is also used to obtain the second resource surplus data corresponding to the first multimedia object determined last, and the second resource excess data corresponding to the second multimedia object determined last.

[0207] The target parameter updating subunit includes: a difference data determining subunit, a first parameter updating subunit, and a second parameter updating subunit.

[0208] The difference data determining subunit is configured to determine first difference data between the first resource margin data and the second resource margin data, and second difference data between the first resource excess data and the second resource excess data.

[0209] The first parameter updating subunit is configured to update the first-level parameter data determined last time based on the first difference data to obtain the first-level parameter data.

[0210] The second parameter updating subunit is configured to update the second-level parameter data determined last time based on the second difference data to obtain the second-level parameter data.

[0211] In an exemplary embodiment, the first parameter updating subunit is specifically configured to:

[0212] If the first resource margin data is less than the second resource margin data, subtracting the first level parameter data determined last time from the first difference data to obtain the first level parameter data;

[0213] If the first resource margin data is greater than or equal to the second resource margin data, the first level parameter data determined last time is added to the first difference data to obtain the first level parameter data.

[0214] In an exemplary embodiment, the second parameter updating subunit is specifically configured to:

[0215] If the first resource excess data is less than the second resource excess data, subtracting the second level parameter data determined last time from the second difference data to obtain the second level parameter data;

[0216] If the first resource excess data is greater than or equal to the second resource excess data, the second level parameter data determined last time is added to the second difference data to obtain the second level parameter data.

[0217] In an exemplary embodiment, the consumption deviation acquisition module 1430 includes: a target data acquisition unit, an actual data acquisition unit, and a consumption data comparison unit.

[0218] The target data acquisition unit is configured to acquire target resource consumption data corresponding to the at least two multimedia objects, wherein the target resource consumption data is used to represent the target resource consumption amount.

[0219] The actual data acquisition unit is configured to acquire actual resource consumption data corresponding to the at least two multimedia objects, where the actual resource consumption data is used to represent the actual resource consumption amount.

[0220] The consumption data comparison unit is used to compare the actual resource consumption data with the target resource consumption data to obtain resource consumption deviation data, and the resource consumption deviation information includes the resource consumption deviation data.

[0221] In an exemplary embodiment, the target data acquisition unit includes: a historical data acquisition subunit and a consumption data fusion subunit.

[0222] The historical data acquisition subunit is used to obtain the historical operation data and resource consumption preset data corresponding to the at least two multimedia objects, the historical operation data is used to characterize the operation events triggered based on the at least two multimedia objects, and the resource consumption data is used to characterize the resource consumption preset amount corresponding to the operation event.

[0223] The consumption data fusion subunit is used to fuse the historical operation data with the resource consumption preset data to obtain the target resource consumption data.

[0224] In an exemplary embodiment, the multimedia object request information includes request start address data, and the first recall module 1420 includes: a region information determination unit, a region feature acquisition unit, and a region matching unit.

[0225] The region information determining unit is configured to determine the region information corresponding to the request start address data.

[0226] The regional feature acquisition unit is used to acquire regional feature information corresponding to the multimedia object set.

[0227] The region matching unit is configured to match the region information with the region feature information to obtain the at least two multimedia objects.

[0228] To sum up, the technical solution provided by the embodiment of the present application performs a preliminary recall of multimedia objects through the multimedia object request of the target object, and obtains the resource consumption deviation information corresponding to the recalled multimedia object, and then determines the degree of deviation between the actual resource consumption corresponding to the multimedia object and the target resource consumption. Since the above-mentioned deviation degree can characterize the display effect of the multimedia object, further multimedia object recall processing can be performed based on the above-mentioned resource consumption deviation information to obtain the recalled target multimedia object. The entire recall process does not depend on the feature information of the target object, and realizes the decoupling of multimedia object recall and target object feature information, greatly reducing the degree of dependence of multimedia recall on target object feature information, and can effectively improve the recall volume and recall efficiency of multimedia objects without using the personalized features of the target object.

[0229] It should be noted that the apparatus provided in the above embodiments, when implementing its functions, is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0230] Please refer to Figure 15 , which shows a structural block diagram of a computer device provided by an embodiment of the present application. The computer device may be a server for executing the above-mentioned multimedia object recall method. Specifically:

[0231] Computer device 1500 includes a central processing unit (CPU) 1501, a system memory 1504 including a random access memory (RAM) 1502 and a read-only memory (ROM) 1503, and a system bus 1505 connecting system memory 1504 and CPU 1501. Computer device 1500 also includes a basic input / output system (I / O system) 1506 that facilitates information transfer between various components within the computer, and a mass storage device 1507 for storing an operating system 1513, application programs 1514, and other program modules 1515.

[0232] The basic input / output system 1506 includes a display 1508 for displaying information and an input device 1509, such as a mouse and keyboard, for user input. Both the display 1508 and the input device 1509 are connected to the central processing unit 1501 via an input / output controller 1510 connected to the system bus 1505. The basic input / output system 1506 may also include an input / output controller 1510 for receiving and processing input from a variety of other devices, such as a keyboard, mouse, or electronic stylus. Similarly, the input / output controller 1510 also provides output to a display screen, printer, or other types of output devices.

[0233] The mass storage device 1507 is connected to the central processing unit 1501 via a mass storage controller (not shown) connected to the system bus 1505. The mass storage device 1507 and its associated computer-readable media provide non-volatile storage for the computer device 1500. In other words, the mass storage device 1507 may include computer-readable media (not shown) such as a hard disk or a CD-ROM (Compact Disc Read-Only Memory) drive.

[0234] Without loss of generality, computer-readable media may include computer storage media and communication media. Computer storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information such as computer-readable instructions, data structures, program modules or other data. Computer storage media include RAM, ROM, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), flash memory or other solid-state storage technologies, CD-ROM, DVD (Digital Video Disc) or other optical storage, tape cassettes, magnetic tape, disk storage or other magnetic storage devices. Of course, those skilled in the art will appreciate that computer storage media are not limited to the above-mentioned ones. The above-mentioned system memory 1504 and mass storage device 1507 can be collectively referred to as memory.

[0235] According to various embodiments of the present application, the computer device 1500 may also be connected to a remote computer on a network such as the Internet for operation. That is, the computer device 1500 may be connected to a network 1512 via a network interface unit 1511 connected to the system bus 1505, or the network interface unit 1511 may be used to connect to other types of networks or remote computer systems (not shown).

[0236] The memory further includes a computer program, which is stored in the memory and configured to be executed by one or more processors to implement the multimedia object recall method.

[0237] In an exemplary embodiment, a computer-readable storage medium is also provided, in which at least one instruction, at least one program, a code set or an instruction set is stored. When the at least one instruction, the at least one program, the code set or the instruction set is executed by a processor, the above-mentioned multimedia object recall method is implemented.

[0238] Optionally, the computer-readable storage medium may include: ROM (Read Only Memory), RAM (Random Access Memory), SSD (Solid State Drives), or an optical disk, etc. Among them, the random access memory may include ReRAM (Resistance Random Access Memory) and DRAM (Dynamic Random Access Memory).

[0239] In an exemplary embodiment, a computer program product or computer program is also provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the multimedia object recall method described above.

[0240] It should be understood that the "multiple" mentioned in this article refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the previous and subsequent associated objects are in an "or" relationship. In addition, the step numbers described in this article only illustrate a possible execution sequence between the steps. In some other embodiments, the above steps may not be executed in the order of the numbers, such as two steps with different numbers are executed at the same time, or two steps with different numbers are executed in the opposite order to the diagram. The embodiments of the present application do not limit this.

[0241] In addition, in the specific implementation of this application, related data such as user information is involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0242] The above description is merely an exemplary embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A multimedia object recall method, characterized in that: The method comprises: Obtain multimedia object request information corresponding to the target object; performing a first recall process based on the multimedia object request information to obtain at least two multimedia objects; Obtaining resource consumption deviation information corresponding to the at least two multimedia objects, the resource consumption deviation information being used to characterize a degree of deviation between target resource consumption and actual resource consumption corresponding to the at least two multimedia objects, the degree of deviation being used to characterize a display effect corresponding to the at least two multimedia objects; the resource consumption deviation information including resource consumption deviation data corresponding to the at least two multimedia objects; A second recall process is performed based on the resource consumption deviation information to obtain a target multimedia object, including: grading the resource consumption deviation data to obtain at least two deviation levels; obtaining target level parameter data, the target level parameter data being used to characterize order information corresponding to the target deviation level; determining the target deviation level based on the target level parameter data and the at least two deviation levels; and determining the target multimedia object based on the target deviation level.

2. The method according to claim 1, characterized in that The target level parameter data includes first level parameter data and second level parameter data, and determining the target deviation level based on the target level parameter data and the at least two deviation levels includes: determining a first deviation level corresponding to the first-level parameter data among the at least two deviation levels, and a second deviation level corresponding to the second-level parameter data among the at least two deviation levels; The determining the target multimedia object according to the target deviation level includes: The multimedia objects whose deviation level is less than or equal to the first deviation level and the multimedia objects whose deviation level is greater than the second deviation level are determined as the target multimedia objects.

3. The method according to claim 1, characterized in that The obtaining of target level parameter data includes: Determine the multimedia object whose resource consumption deviation data is smaller than the first deviation data threshold as the first multimedia object; Determine the multimedia object whose resource consumption deviation data is greater than a second deviation data threshold as a second multimedia object; Obtaining first resource remainder data corresponding to the first multimedia object and first resource excess data corresponding to the second multimedia object, where the first resource remainder data is used to indicate a remaining degree of target resource consumption corresponding to the first multimedia object, and the first resource excess data is used to indicate a degree of excess of actual resource consumption corresponding to the second multimedia object; Based on the first resource margin data and the first resource excess data, the target level parameter data determined last time is updated to obtain the target level parameter data. The target level parameter data is preset data when it is obtained for the first time.

4. The method according to claim 3, characterized in that The target level parameter data includes first level parameter data and second level parameter data, and the target level parameter data is updated based on the first resource margin data and the first resource excess data to obtain the target level parameter data, including: Acquire second resource remaining data corresponding to the last determined first multimedia object, and second resource excess data corresponding to the last determined second multimedia object; determining first difference data between the first resource margin data and the second resource margin data, and second difference data between the first resource excess data and the second resource excess data; Based on the first difference data, updating the first-level parameter data determined last time to obtain the first-level parameter data; Based on the second difference data, the second level parameter data determined last time is updated to obtain the second level parameter data.

5. The method according to claim 4, characterized in that The updating of the first-level parameter data determined last time based on the first difference data to obtain the first-level parameter data includes: If the first resource margin data is less than the second resource margin data, subtracting the first level parameter data determined last time from the first difference data to obtain the first level parameter data; If the first resource margin data is greater than or equal to the second resource margin data, the first level parameter data determined last time is added to the first difference data to obtain the first level parameter data.

6. The method according to claim 4, characterized in that The updating of the second-level parameter data determined last time based on the second difference data to obtain the second-level parameter data includes: If the first resource excess data is less than the second resource excess data, subtracting the second level parameter data determined last time from the second difference data to obtain the second level parameter data; If the first resource excess data is greater than or equal to the second resource excess data, the second level parameter data determined last time is added to the second difference data to obtain the second level parameter data.

7. The method according to claim 1, characterized in that The obtaining of resource consumption deviation information corresponding to the at least two multimedia objects includes: Acquire target resource consumption data corresponding to the at least two multimedia objects, where the target resource consumption data is used to represent the target resource consumption amount; Acquire actual resource consumption data corresponding to the at least two multimedia objects, where the actual resource consumption data is used to represent the actual resource consumption; The actual resource consumption data is compared with the target resource consumption data to obtain resource consumption deviation data, and the resource consumption deviation information includes the resource consumption deviation data.

8. The method according to claim 7, characterized in that The acquiring target resource consumption data corresponding to the at least two multimedia objects includes: Obtaining historical operation data and resource consumption preset data corresponding to the at least two multimedia objects, the historical operation data being used to represent an operation event triggered based on the at least two multimedia objects, and the resource consumption data being used to represent a preset amount of resource consumption corresponding to the operation event; The historical operation data and the resource consumption preset data are fused to obtain the target resource consumption data.

9. The method according to any one of claims 1 to 8, characterized in that The multimedia object request information includes request start address data, and the first recall processing is performed based on the multimedia object request information to obtain at least two multimedia objects, including: Determining the region information corresponding to the request starting address data; Obtaining regional feature information corresponding to a multimedia object set; The region information is matched with the region feature information to obtain the at least two multimedia objects.

10. A multimedia object recall device, characterized in that: The device comprises: A request acquisition module is used to obtain multimedia object request information corresponding to the target object; A first recall module, configured to perform a first recall process based on the multimedia object request information to obtain at least two multimedia objects; a consumption deviation determination module, configured to obtain resource consumption deviation information corresponding to the at least two multimedia objects, the resource consumption deviation information being used to characterize a degree of deviation between target resource consumption and actual resource consumption corresponding to the at least two multimedia objects, the degree of deviation being used to characterize a display effect corresponding to the at least two multimedia objects; the resource consumption deviation information including resource consumption deviation data corresponding to the at least two multimedia objects; A second recall module, configured to perform a second recall process based on the resource consumption deviation information to obtain a target multimedia object; The second recall module includes a deviation level determination unit, a target parameter acquisition unit, a target level determination unit, and a multimedia object determination unit; The deviation level determination unit is configured to perform hierarchical processing on the resource consumption deviation data to obtain at least two deviation levels; The target parameter acquisition unit is used to acquire target level parameter data, wherein the target level parameter data is used to represent order information corresponding to the target deviation level; The target level determination unit is configured to determine the target deviation level based on the target level parameter data and the at least two deviation levels; The multimedia object determination unit is configured to determine the target multimedia object according to the target deviation level.

11. The device according to claim 10, characterized in that The target level parameter data includes first level parameter data and second level parameter data, and the target level determination unit is specifically configured to determine a first deviation level corresponding to the first level parameter data among the at least two deviation levels, and a second deviation level corresponding to the second level parameter data among the at least two deviation levels; The multimedia object determination unit is specifically configured to determine, as the target multimedia objects, the multimedia objects whose deviation level is less than or equal to the first deviation level and the multimedia objects whose deviation level is greater than the second deviation level.

12. The device according to claim 10, characterized in that The target parameter acquisition unit includes: a first object determination subunit, a second object determination subunit, a resource quantity data acquisition subunit, and a target parameter update subunit; The first object determining subunit is configured to determine the multimedia object whose resource consumption deviation data is less than a first deviation data threshold as the first multimedia object; The second object determining subunit is configured to determine the multimedia object whose resource consumption deviation data is greater than a second deviation data threshold as a second multimedia object; The resource quantity data acquisition subunit is configured to acquire first resource remaining data corresponding to the first multimedia object, and first resource excess data corresponding to the second multimedia object, wherein the first resource remaining data is used to indicate a remaining degree of a target resource consumption corresponding to the first multimedia object, and the first resource excess data is used to indicate a degree of excess of an actual resource consumption corresponding to the second multimedia object; The target parameter updating subunit is configured to update the target level parameter data determined last time based on the first resource margin data and the first resource excess data to obtain the target level parameter data, which is preset data when first obtained.

13. The device according to claim 12, characterized in that The target level parameter data includes first level parameter data and second level parameter data, and the resource quantity data acquisition subunit is further configured to acquire second resource remaining data corresponding to the last determined first multimedia object, and second resource excess data corresponding to the last determined second multimedia object; The target parameter updating subunit includes: a difference data determining subunit, a first parameter updating subunit, and a second parameter updating subunit; The difference data determining subunit is configured to determine first difference data between the first resource margin data and the second resource margin data, and second difference data between the first resource excess data and the second resource excess data; The first parameter updating subunit is configured to update the first-level parameter data determined last time based on the first difference data to obtain the first-level parameter data; The second parameter updating subunit is configured to update the second level parameter data determined last time based on the second difference data to obtain the second level parameter data.

14. The device according to claim 13, characterized in that The first parameter updating subunit is specifically configured to: If the first resource margin data is less than the second resource margin data, subtracting the first level parameter data determined last time from the first difference data to obtain the first level parameter data; If the first resource margin data is greater than or equal to the second resource margin data, the first level parameter data determined last time is added to the first difference data to obtain the first level parameter data.

15. The device according to claim 13, characterized in that The second parameter updating subunit is specifically configured to: If the first resource excess data is less than the second resource excess data, subtracting the second level parameter data determined last time from the second difference data to obtain the second level parameter data; If the first resource excess data is greater than or equal to the second resource excess data, the second level parameter data determined last time is added to the second difference data to obtain the second level parameter data.

16. The device according to claim 10, characterized in that The consumption deviation acquisition module includes: a target data acquisition unit, an actual data acquisition unit, and a consumption data comparison unit; The target data acquisition unit is configured to acquire target resource consumption data corresponding to the at least two multimedia objects, wherein the target resource consumption data is used to represent the target resource consumption amount; The actual data acquisition unit is configured to acquire actual resource consumption data corresponding to the at least two multimedia objects, wherein the actual resource consumption data is used to represent the actual resource consumption amount; The consumption data comparison unit is configured to compare the actual resource consumption data with the target resource consumption data to obtain resource consumption deviation data, and the resource consumption deviation information includes the resource consumption deviation data.

17. The device according to claim 16, characterized in that The target data acquisition unit includes: a historical data acquisition subunit and a consumption data fusion subunit; The historical data acquisition subunit is configured to acquire historical operation data and resource consumption preset data corresponding to the at least two multimedia objects, wherein the historical operation data is used to represent operation events triggered by the at least two multimedia objects, and the resource consumption data is used to represent the resource consumption preset amount corresponding to the operation event; The consumption data fusion subunit is used to fuse the historical operation data with the resource consumption preset data to obtain the target resource consumption data.

18. The device according to any one of claims 10 to 17, characterized in that The multimedia object request information includes request start address data, and the first recall module includes: a region information determination unit, a region feature acquisition unit, and a region matching unit; The regional information determining unit is used to determine the regional information corresponding to the request starting address data; The regional feature acquisition unit is used to obtain regional feature information corresponding to the multimedia object set; The region matching unit is configured to match the region information with the region feature information to obtain the at least two multimedia objects.

19. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the multimedia object recall method as described in any one of claims 1 to 9.

20. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the multimedia object recall method as described in any one of claims 1 to 9.

21. A computer program product, characterized in that The computer program product includes computer instructions, which are stored in a computer-readable storage medium. The processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes to implement the multimedia object recall method according to any one of claims 1 to 9.

Images