Multi-target fusion processing method and device, electronic equipment and storage medium
By combining a two-stage processing approach with the prediction and correction of basic and non-basic objectives, the problems of insufficient accuracy and resource diversity in multi-objective fusion scoring in intelligent recommendation systems are solved, and efficient and accurate resource ranking and recommendation are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING BAIDU NETCOM SCI & TECH CO LTD
- Filing Date
- 2023-08-07
- Publication Date
- 2026-07-31
AI Technical Summary
Existing intelligent recommendation systems struggle to effectively combine the differences in the value ranges of different predicted values during multi-objective fusion processes, resulting in poor accuracy of fusion scoring and potential issues such as insufficient resource diversity and reverse order in fine-grained ranking.
A two-stage processing approach is adopted. First, candidate resources are screened based on the basic objectives. Then, the estimated value is corrected by combining the bias value of non-basic objectives. The first fusion score is calculated by formula (3) to ensure resource diversity and accuracy.
It improves the accuracy of ranking results and resource diversity of recommendation systems, reduces implementation costs, and increases user retention rates.
Smart Images
Figure CN117271882B_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to multi-objective fusion processing methods, apparatuses, electronic devices and storage media in the fields of intelligent recommendation, big data processing and cloud computing. Background Technology
[0002] In intelligent recommendation scenarios, the purpose of multi-objective fusion is to aggregate information from multiple objectives to obtain a more comprehensive and accurate recommendation evaluation result. For recommendation systems, multi-objective fusion mainly occurs during the fine-grained ranking stage. Summary of the Invention
[0003] This disclosure provides a method, apparatus, electronic device, and storage medium for multi-target fusion processing.
[0004] A multi-target fusion processing method includes:
[0005] Obtain N estimated values corresponding to each candidate resource to be recommended to the user. Different estimated values correspond to different fusion targets. The fusion targets include: basic targets and non-basic targets, where N is a positive integer greater than one.
[0006] Based on the estimated value of the basic objective of each candidate resource, a portion of the candidate resources are selected as target resources.
[0007] Based on the N estimated values of each target resource and the bias values corresponding to different estimated values, the first fusion score of each target resource is determined.
[0008] A multi-target fusion processing device includes: an acquisition module, a filtering module, and a processing module;
[0009] The acquisition module is used to acquire N estimated values corresponding to each candidate resource to be recommended to the user. Different estimated values correspond to different fusion targets. The fusion targets include: basic targets and non-basic targets, where N is a positive integer greater than one.
[0010] The filtering module is used to select a portion of the candidate resources as target resources based on the estimated value of the basic target of each candidate resource.
[0011] The processing module is used to determine the first fusion score of each target resource based on the N estimated values of each target resource and the bias values corresponding to the different estimated values.
[0012] An electronic device, comprising:
[0013] At least one processor; and
[0014] A memory communicatively connected to the at least one processor; wherein,
[0015] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described above.
[0016] A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the methods described above.
[0017] A computer program product includes a computer program / instructions that, when executed by a processor, implement the method described above.
[0018] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description
[0019] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:
[0020] Figure 1 This is a flowchart of an embodiment of the multi-target fusion processing method described in this disclosure;
[0021] Figure 2 This is a schematic diagram of the composition structure of Embodiment 200 of the multi-target fusion processing device described in this disclosure;
[0022] Figure 3 A schematic block diagram of an electronic device 300 that can be used to implement embodiments of the present disclosure is shown. Detailed Implementation
[0023] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.
[0024] Furthermore, it should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.
[0025] Figure 1 This is a flowchart illustrating an embodiment of the multi-target fusion processing method described in this disclosure. Figure 1 As shown, the specific implementation methods are as follows.
[0026] In step 101, N estimated values corresponding to each candidate resource to be recommended to the user are obtained. Different estimated values correspond to different fusion targets. The fusion targets include: basic targets and non-basic targets, where N is a positive integer greater than one.
[0027] In step 102, based on the estimated value of the basic objective of each candidate resource, a portion of the candidate resources are selected as target resources.
[0028] In step 103, the first fusion score of each target resource is determined based on the N estimated values of each target resource and the bias values corresponding to the different estimated values.
[0029] As can be seen, the above-described method embodiment proposes a two-stage processing approach. In the first stage, candidate resources are initially screened based on basic objectives to obtain target resources. That is, based on basic objectives, some candidate resources that are unlikely to be recommended to users are first eliminated, thereby reducing the workload of subsequent processing while retaining the required target resources as much as possible. In the second stage, non-basic objectives are further integrated to obtain the final required fusion score for each target resource. The whole process is simple and convenient to implement, thereby reducing the implementation cost. Moreover, when generating the fusion score for each target resource, not only are the estimated values of different targets to be integrated combined, but also the bias values corresponding to the different estimated values are combined, thereby improving the accuracy of the obtained fusion score.
[0030] To distinguish it from other fusion scores, the fusion score of each target resource is referred to as the first fusion score. For any recommendation system, the processing includes different stages such as recall, coarse ranking, and fine ranking. As mentioned earlier, multi-target fusion mainly occurs in the fine ranking stage. After obtaining the first fusion score for each target resource, the resources can be sorted in descending order of the first fusion score, thereby improving the accuracy of the sorting results.
[0031] For any candidate resource, N estimated values can be obtained. Different estimated values correspond to different targets to be merged. The targets to be merged include: basic targets and non-basic targets, where N is a positive integer greater than one.
[0032] The specific value of N can be determined according to actual needs. In addition, the specific targets included in the basic target and the non-basic target can also be determined according to actual needs.
[0033] The basic objectives typically refer to those related to the estimated user browsing time, such as click-through rate (CTR) and playthrough rate (PCR). Conversely, non-basic objectives refer to objectives other than the basic objectives. Basic objectives effectively reflect users' preferences for candidate resources and are important metrics referenced by recommendation systems.
[0034] The basic and non-basic objectives can be collectively referred to as the objectives to be merged. For each candidate resource, the estimated value corresponding to different objectives to be merged can be obtained separately. For example, for candidate resource a, the estimated value corresponding to objective 1 to be merged can be obtained, assuming it is the estimated value of CTR, that is, the estimated value of the user's CTR for candidate resource a. The estimated value corresponding to objective 2 to be merged can be obtained, assuming it is the estimated value of PCR, that is, the estimated value of the user's PCR for candidate resource a, and so on.
[0035] Preferably, in the first stage of the scheme described in this disclosure, some candidate resources can be selected from each candidate resource based on the estimated value of the basic target of each candidate resource and the category to which each candidate resource belongs, and the selected candidate resources can be used as target resources.
[0036] The categories may include food, travel, sports, etc., and the classification method can be determined according to actual needs.
[0037] Preferably, specifically, for any candidate resource, the second fusion score of the candidate resource can be determined based on the estimated value of each basic objective of the candidate resource, and the category to which each candidate resource belongs can be taken as the category to be processed. Then, at least one candidate resource with the highest second fusion score can be selected from the candidate resources under each category to be processed.
[0038] For example, assuming there are two basic objectives, the second fusion score for each candidate resource can be calculated as follows:
[0039]
[0040] Among them, res score '' indicates the second fusion score. Predicted value 1 and predicted value 2 represent the predicted values of the two basic targets, respectively. w1 and w2 represent the power exponents of predicted value 1 and predicted value 2, respectively. The specific values can be determined according to actual needs.
[0041] After obtaining the second fusion score of each candidate resource, if the target resource is selected based solely on the second fusion score, the selected target resources may be concentrated in one or a few categories, resulting in poor diversity of resources recommended to the user.
[0042] Therefore, the solution described in this disclosure proposes that the target resource can be selected by combining the category to which each candidate resource belongs. That is, the category to which each candidate resource belongs can be taken as the category to be processed, and at least one candidate resource with the highest second fusion score can be selected from the candidate resources under each category to be processed.
[0043] For example, assuming there are 10 categories to be processed, 100 candidate resources, and more than 5 candidate resources under each category, then 5 candidate resources with the highest second fusion scores can be selected from each category, and the selected 50 candidate resources can be used as the target resources.
[0044] In practical applications, the number of candidate resources selected from different categories to be processed can be the same or different.
[0045] The above processing increases the diversity of the target resources obtained, which in turn increases the diversity of the final recommendation results.
[0046] In the second phase of the scheme described in this disclosure, the first fusion score of each target resource can be determined based on N estimated values of each target resource and the bias values corresponding to different estimated values.
[0047] Preferably, for any target resource, the following processing can be performed: based on the bias values corresponding to different estimated values, the N estimated values of the target resource are corrected respectively, and the first fusion score of the target resource is determined based on the corrected N estimated values.
[0048] In the traditional method, the first fusion score can be calculated as follows:
[0049]
[0050] Among them, res score ' represents the first fusion score obtained in the traditional way, and N represents the estimated value, i.e. the number of targets to be fused.
[0051] For different targets to be fused, the range of their corresponding estimated values is usually different and may vary greatly. For example, for dense targets such as CTR, the range may be around 0.4 to 0.8, while for sparse targets such as interaction, the range may be in the thousandths. If the different estimated values are directly calculated according to the method in formula (2), there may be a problem of incomparability between multiple targets, resulting in poor accuracy of the obtained first fusion score, which may lead to the problem of reverse order in fine ranking.
[0052] Therefore, the scheme described in this disclosure proposes that each estimated value can be corrected using the corresponding bias value, and the first fusion score can be calculated accordingly as follows:
[0053]
[0054] Among them, res score The first fusion score obtained in accordance with the method described in this disclosure is represented by the bias value 1, the bias value N is represented by the bias value corresponding to the expected value N.
[0055] By increasing the bias value, the range of different estimated values can be corrected as much as possible, thereby overcoming the problems existing in the traditional method and improving the accuracy of the obtained first fusion score.
[0056] The specific values of the biases corresponding to different estimated values can be determined according to actual needs. For example, they can be empirical values, or they can be calculated based on relevant historical data through a predetermined calculation method. In addition, the bias value can be 0 or greater than 0. For example, a bias value of 1 can be 0.
[0057] Preferably, for any target resource, when correcting the N estimated values of the target resource according to the bias values corresponding to the different estimated values, it can be first determined whether the category to which the target resource belongs is a user's confidence point of interest. In response to the fact that the category to which the target resource belongs is a confidence point of interest, the N estimated values of the target resource can be corrected according to the bias values corresponding to the different estimated values.
[0058] That is, for any target resource, before calculating its first fusion score according to formula (3), it can be determined whether the category it describes is a user's trusted interest point. If so, then its first fusion score can be calculated according to formula (3).
[0059] Preferably, for any target resource, in response to the fact that the category to which the target resource belongs is not a user's trusted point of interest, an initial correction and a second correction can be performed sequentially. The initial correction may include: correcting the estimated value of the non-basic target of the target resource based on constraint rules. The second correction may include: correcting the estimated value of the basic target of the target resource and the estimated value of the non-basic target after the initial correction based on the bias values corresponding to different estimated values, and determining the first fusion score of the target resource based on the N estimated values after the second correction.
[0060] As can be seen, the above processing methods can be tailored to the user’s confidence and interest level based on whether the category to which the target resource belongs. This allows for targeted processing and further improves the accuracy of the first fusion score.
[0061] Preferably, the categories to which the resources browsed by the user within the most recent predetermined time period belong can be determined as target categories, and the user's satisfaction with each target category can be obtained, thereby obtaining the average value of each satisfaction. Furthermore, the target categories corresponding to the satisfaction values that are higher than the average value can be taken as the user's confidence points of interest.
[0062] The specific value of the recently scheduled duration can be determined according to actual needs. Furthermore, there are no restrictions on how to obtain user satisfaction with each target category; it can also be determined according to actual needs. For example, for each target category, user satisfaction with that target category can be determined based on the number of resources viewed, the viewing time, and the number of views.
[0063] Furthermore, assuming a user has browsed multiple resources belonging to 10 different categories within the past 6 months, these 10 categories can be designated as target categories, namely target category 1 to target category 10. Additionally, the user's satisfaction level for each target category can be obtained, and the average of the 10 satisfaction levels can be calculated. Further, each satisfaction level can be compared with the average. If the satisfaction level for target category 1 is greater than the average, then target category 1 can be identified as the user's confident point of interest. If the satisfaction level for target category 2 is less than or equal to the average, then target category 2 can be identified as the user's unconfident point of interest, and so on.
[0064] Through the above processing, the user's confidence points of interest can be determined efficiently and accurately, thus laying a good foundation for calculating the first fusion score of different target resources.
[0065] Thus, for any target resource, it can be determined whether its category is a user's trusted interest point. If so, the first fusion score of the target resource can be calculated according to formula (3). If not, the estimated value of the non-basic target of the target resource can be first corrected, and the corrected estimated value can replace the estimated value in formula (3). Then, the first fusion score of the target resource can be calculated according to formula (3).
[0066] The initial correction may include: correcting the estimated value of the non-basic target of the target resource based on constraint rules. Specifically, preferably, for any non-basic target of any target resource, the following processing may be performed respectively: obtaining the average value of the estimated value of the non-basic target of each target resource; in response to the estimated value of the non-basic target of the target resource being greater than the average value, correcting the estimated value of the non-basic target of the target resource to the average value.
[0067] For example, suppose there are 50 target resources, including two non-basic targets, namely non-basic target a and non-basic target b. For any target resource among these 50 target resources, let's say target resource 2, during the initial adjustment, we can first obtain the average estimated value of non-basic target a among the 50 target resources, and then compare the estimated value of non-basic target a of target resource 2 with the average value. If it is greater than the average value, the estimated value of non-basic target a of target resource 2 can be adjusted to the average value; otherwise, the estimated value of non-basic target a of target resource 2 can be left unchanged. In addition, we can obtain the average estimated value of non-basic target b among the 50 target resources, and compare the estimated value of non-basic target b of target resource 2 with the average value. If it is greater than the average value, the estimated value of non-basic target b of target resource 2 can be adjusted to the average value; otherwise, the estimated value of non-basic target b of target resource 2 can be left unchanged.
[0068] Through the above processing, the estimated value of non-basic targets of target resources that do not belong to confidence points of interest can be constrained, that is, adjusted to be less than or equal to the mean, thereby reducing the first fusion score of such target resources and further improving the accuracy of the ranking results.
[0069] Preferably, the candidate resources can be candidate videos, and correspondingly, the target resource can be a target video, and the browsing duration can be the viewing duration, etc.
[0070] It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of actions. However, those skilled in the art should understand that this disclosure is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this disclosure. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this disclosure.
[0071] The above is an introduction to the method embodiments. The following describes the solution described in this disclosure further through device embodiments.
[0072] Figure 2 This is a schematic diagram of the structural composition of Embodiment 200 of the multi-target fusion processing apparatus described in this disclosure. Figure 2 As shown, it includes: an acquisition module 201, a filtering module 202, and a processing module 203.
[0073] The acquisition module 201 is used to acquire N estimated values corresponding to each candidate resource to be recommended to the user. Different estimated values correspond to different fusion targets. The fusion targets include: basic targets and non-basic targets, where N is a positive integer greater than one.
[0074] The screening module 202 is used to select a portion of the candidate resources as target resources based on the estimated value of the basic target of each candidate resource.
[0075] The processing module 203 is used to determine the first fusion score of each target resource based on N estimated values of each target resource and the bias values corresponding to different estimated values.
[0076] The above-described device embodiment proposes a two-stage processing method. In the first stage, candidate resources are initially screened based on basic objectives to obtain target resources. That is, some candidate resources that are unlikely to be recommended to users are eliminated based on basic objectives, thereby reducing the workload of subsequent processing while retaining the required target resources as much as possible. In the second stage, non-basic objectives are further integrated to obtain the final required fusion score for each target resource. The whole process is simple and convenient to implement, thereby reducing the implementation cost. Moreover, when generating the fusion score for each target resource, not only are the estimated values of different targets to be integrated combined, but also the bias values corresponding to the different estimated values are combined, thereby improving the accuracy of the obtained fusion score.
[0077] Preferably, in the first stage of the scheme described in this disclosure, the screening module 202 can select some candidate resources from each candidate resource based on the estimated value of the basic target of each candidate resource and the category to which each candidate resource belongs, and can use the selected candidate resources as target resources.
[0078] Preferably, for any candidate resource, the screening module 202 can determine the second fusion score of the candidate resource based on the estimated value of each basic target of the candidate resource, and can take the category to which each candidate resource belongs as the category to be processed, and then select at least one candidate resource with the highest second fusion score from the candidate resources under each category to be processed.
[0079] In the second stage of the scheme described in this disclosure, the processing module 203 can determine the first fusion score of each target resource based on N estimated values of each target resource and the bias values corresponding to different estimated values.
[0080] Preferably, for any target resource, the processing module 203 may perform the following processing: according to the bias values corresponding to different estimated values, the N estimated values of the target resource are corrected respectively, and the first fusion score of the target resource is determined according to the corrected N estimated values.
[0081] In addition, preferably, for any target resource, when the processing module 203 corrects the N estimated values of the target resource according to the bias values corresponding to the different estimated values, it can first determine whether the category to which the target resource belongs is a user's confidence point of interest. In response to the fact that the category to which the target resource belongs is a user's confidence point of interest, the N estimated values of the target resource can be corrected according to the bias values corresponding to the different estimated values.
[0082] Preferably, for any target resource, in response to the fact that the category to which the target resource belongs is not a user's trusted point of interest, the processing module 203 may sequentially perform an initial correction and a second correction. The initial correction may include: correcting the estimated value of the non-basic target of the target resource based on constraint rules. The second correction may include: correcting the estimated value of the basic target of the target resource and the estimated value of the non-basic target after the initial correction based on the bias values corresponding to different estimated values, and determining the first fusion score of the target resource based on the N estimated values after the second correction.
[0083] Preferably, the processing module 203 can determine the category to which the resources browsed by the user within the most recent predetermined time period belong, and use it as the target category. It can also obtain the user's satisfaction with each target category, and then obtain the average value of each satisfaction. Furthermore, the target category corresponding to the satisfaction level that is higher than the average value can be used as the user's confidence point of interest.
[0084] Thus, for any target resource, the processing module 203 can determine whether its category is a user's trusted interest point. If so, the first fusion score of the target resource can be calculated according to formula (3). If not, the estimated value of the non-basic target of the target resource can be corrected first, and the corrected estimated value can replace the estimated value in formula (3). Then, the first fusion score of the target resource can be calculated according to formula (3).
[0085] The initial correction may include: correcting the estimated value of the non-basic target of the target resource based on constraint rules. Preferably, the processing module 203 may perform the following processing for any non-basic target of any target resource: obtaining the average value of the estimated value of the non-basic target of each target resource; and correcting the estimated value of the non-basic target of the target resource to the average value in response to the fact that the estimated value of the non-basic target of the target resource is greater than the average value.
[0086] Figure 2 The specific workflow of the device embodiment shown can be found in the relevant descriptions in the foregoing method embodiments, and will not be repeated here.
[0087] In summary, the solution described in this disclosure can reduce implementation costs, improve the accuracy of processing results, thereby increasing user retention rate, and has wide applicability.
[0088] The solutions described in this disclosure can be applied to the field of artificial intelligence, particularly in areas such as intelligent recommendation, big data processing, and cloud computing. Artificial intelligence is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It involves both hardware and software technologies. Artificial intelligence hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. Artificial intelligence software technologies mainly include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.
[0089] Furthermore, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.
[0090] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.
[0091] Figure 3A schematic block diagram of an electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.
[0092] like Figure 3 As shown, device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 302 or a computer program loaded from storage unit 308 into random access memory (RAM) 303. The RAM 303 may also store various programs and data required for the operation of device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via bus 304. Input / output (I / O) interface 305 is also connected to bus 304.
[0093] Multiple components in device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of monitors, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0094] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as those described in this disclosure. For example, in some embodiments, the methods described in this disclosure can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program can be loaded and / or installed on device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of the methods described in this disclosure can be performed. Alternatively, in other embodiments, the computing unit 301 can be configured to perform the methods described in this disclosure by any other suitable means (e.g., by means of firmware).
[0095] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0096] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0097] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0098] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0099] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.
[0100] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.
[0101] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.
[0102] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.
Claims
1. A multi-target fusion processing method, comprising: Obtain N estimated values for each candidate resource to be recommended to the user. Different estimated values correspond to different fusion targets. The fusion targets include: basic targets and non-basic targets. N is a positive integer greater than one. The basic targets are targets related to the estimated user's browsing time and are used to reflect the user's preference for candidate resources. Based on the estimated values of the basic objectives of each candidate resource, a portion of the candidate resources are selected as target resources, including: for any candidate resource, determining the second fusion score of the candidate resource based on the estimated values of each basic objective of the candidate resource; designating the categories to which each candidate resource belongs as categories to be processed; and selecting at least one candidate resource with the highest second fusion score from the candidate resources under each category to be processed. For any target resource, the following processing is performed: It is determined whether the category to which the target resource belongs is a user's confidence point of interest. If it is, a first fusion score for the target resource is determined based on the N estimated values of the target resource and the bias values corresponding to different estimated values. If it is not, the estimated values of the non-basic targets of the target resource are first corrected, including: for any non-basic target of the target resource, the average of the estimated values of the non-basic targets of each target resource is obtained; in response to the estimated value of the non-basic target of the target resource being greater than the average, the estimated value of the non-basic target of the target resource is corrected to the average; and the first fusion score for the target resource is determined based on the first corrected estimated values of the non-basic targets, the estimated values of the basic targets, and the bias values corresponding to different estimated values.
2. The method according to claim 1, wherein, The step of determining the first fusion score of the target resource based on the N estimated values of the target resource and the bias values corresponding to the different estimated values includes: Based on the bias values corresponding to the different estimated values, the N estimated values of the target resource are corrected respectively, and the first fusion score of the target resource is determined based on the corrected N estimated values.
3. The method according to claim 1, further comprising: The category to which the resources viewed by the user within the most recent scheduled time period belong is determined and designated as the target category; Obtain the user's satisfaction with each target category, and obtain the average of each satisfaction level; The target category corresponding to the satisfaction level that is higher than the mean among all satisfaction levels is taken as the confidence point of interest.
4. The method according to claim 1, wherein, The step of determining the first fusion score of the target resource based on the estimated value of the non-basic target after the first correction, the estimated value of the basic target, and the bias value corresponding to each of the different estimated values includes: performing a second correction on the estimated value of the basic target and the estimated value of the non-basic target of the target resource based on the bias value corresponding to each of the different estimated values, and determining the first fusion score of the target resource based on the N estimated values after the second correction.
5. A multi-object fusion processing apparatus comprising: The module includes an acquisition module, a filtering module, and a processing module. The acquisition module is used to acquire N estimated values corresponding to each candidate resource to be recommended to the user. Different estimated values correspond to different fusion targets. The fusion targets include: basic targets and non-basic targets. N is a positive integer greater than one. The basic targets are targets related to the estimated user's browsing time and are used to reflect the user's preference for candidate resources. The filtering module is used to select a portion of candidate resources as target resources based on the estimated values of the basic objectives of each candidate resource. This includes: for any candidate resource, determining the second fusion score of the candidate resource based on the estimated values of each basic objective of the candidate resource; designating the categories to which each candidate resource belongs as categories to be processed; and selecting at least one candidate resource with the highest second fusion score from the candidate resources under each category to be processed. The processing module is configured to perform the following processing for any target resource: determine whether the category to which the target resource belongs is a user's confidence interest point; if it is a confidence interest point, determine a first fusion score for the target resource based on the N estimated values of the target resource and the bias values corresponding to the different estimated values; if it is not a confidence interest point, perform an initial correction on the estimated values of the non-basic targets of the target resource, including: for any non-basic target of the target resource, obtain the average value of the estimated values of the non-basic targets of each target resource; in response to the estimated value of the non-basic target of the target resource being greater than the average value, correct the estimated value of the non-basic target of the target resource to the average value; and determine the first fusion score for the target resource based on the estimated value of the non-basic target after the initial correction, the estimated value of the basic target, and the bias values corresponding to the different estimated values.
6. The apparatus according to claim 5, wherein, For any target resource, the processing module performs the following processing: based on the bias values corresponding to different estimated values, it corrects the N estimated values of the target resource respectively, and determines the first fusion score of the target resource based on the corrected N estimated values.
7. The apparatus according to claim 5, wherein, The processing module is further configured to determine the category to which the resources browsed by the user within the most recent predetermined time period belong, as the target category, obtain the user's satisfaction with each target category, obtain the average value of each satisfaction, and take the target category corresponding to the satisfaction level that is higher than the average value as the confidence point of interest.
8. The apparatus according to claim 5, wherein, The processing module performs secondary corrections on the estimated values of the basic target and the estimated values of the non-basic target of the target resource based on the bias values corresponding to different estimated values, and determines the first fusion score of the target resource based on the N estimated values after secondary correction.
9. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-4.
10. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-4.
11. A computer program product comprising a computer program / instructions that, when executed by a processor, implement the method of any one of claims 1-4.