WO2022228075A1 - 确定推广方的方法、装置、设备、存储介质和程序产品 - Google Patents

确定推广方的方法、装置、设备、存储介质和程序产品 Download PDF

Info

Publication number
WO2022228075A1
WO2022228075A1 PCT/CN2022/085597 CN2022085597W WO2022228075A1 WO 2022228075 A1 WO2022228075 A1 WO 2022228075A1 CN 2022085597 W CN2022085597 W CN 2022085597W WO 2022228075 A1 WO2022228075 A1 WO 2022228075A1
Authority
WO
WIPO (PCT)
Prior art keywords
promoters
candidate
promoter
target
priority level
Prior art date
Application number
PCT/CN2022/085597
Other languages
English (en)
French (fr)
Inventor
李松松
赵旭
冯思源
林恩禄
张军
Original Assignee
北京有竹居网络技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京有竹居网络技术有限公司 filed Critical 北京有竹居网络技术有限公司
Publication of WO2022228075A1 publication Critical patent/WO2022228075A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0242—Determining effectiveness of advertisements
    • G06Q30/0245—Surveys
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data

Definitions

  • Various implementations of the present disclosure relate to the field of computers, and more particularly, to methods, apparatuses, devices, and computer storage media for entity clustering.
  • Some providers eg, stores or service providers, etc.
  • cooperate with promoters to promote objects to be better known by users, or guide more users to acquire these objects.
  • a method for determining a promoter includes: recalling a plurality of candidate promoters for a target provider from a set of promoters, the target provider can provide at least one object that can be obtained by a user, and the plurality of candidate promoters can publish a message for guiding the user to acquire the corresponding object. guiding content; determining a priority level of the plurality of candidate promoters based on the first characteristic of the target provider and the second characteristic of the plurality of candidate promoters; and determining, based on the priority level, the target provider from among the plurality of candidate promoters party's target promotion party.
  • an apparatus for determining a promoter includes: a recall module configured to recall a plurality of candidate promoters for a target provider from a set of promoters, the target provider can provide at least one object that can be obtained by a user, and the plurality of candidate promoters can be published for guiding the user to obtain the guiding content of the corresponding object; the sorting module is configured to determine the priority level of the plurality of candidate promoters based on the first characteristic of the target provider and the second characteristic of the plurality of candidate promoters; and the determining module is configured by Configured to determine a target promoter for a target provider from a plurality of candidate promoters based on a priority level.
  • an electronic device comprising: a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement according to the present disclosure method of the first aspect.
  • a computer-readable storage medium having stored thereon one or more computer instructions, wherein the one or more computer instructions are executed by a processor to implement the method according to the first aspect of the present disclosure .
  • a computer program product comprising one or more computer instructions, wherein the one or more computer instructions are executed by a processor to implement the method according to the first aspect of the present disclosure.
  • Figure 1 shows a schematic diagram of an example environment in which various embodiments of the present disclosure can be implemented
  • FIG. 2 illustrates a flowchart of an example process for determining a promoter according to some embodiments of the present disclosure
  • FIG. 3 illustrates a flowchart of an example process for recalling candidate promoters in accordance with some embodiments of the present disclosure
  • FIG. 4 shows a schematic structural block diagram of an apparatus for determining a promoter according to some embodiments of the present disclosure.
  • FIG. 5 illustrates a block diagram of a computing device capable of implementing various embodiments of the present disclosure.
  • the term “comprising” and the like should be understood as open-ended inclusion, ie, “including but not limited to”.
  • the term “based on” should be understood as “based at least in part on”.
  • the terms “one embodiment” or “the embodiment” should be understood to mean “at least one embodiment”.
  • the terms “first”, “second”, etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.
  • example embodiments of the present disclosure propose a scheme for determining a promoter.
  • targeted providers eg, brick and mortar stores, virtual stores, service providers, etc.
  • candidate promoters e.g., anchors, video creators, text content creators, etc.
  • target providers e.g., fans of anchors, students or parents attending training, viewers of videos, readers of articles readers, etc.
  • object for example, tangible goods, digital content or specific services
  • multiple candidate promoters can publish guiding content (for example, online live content, video documents, online articles, etc.).
  • priority levels of the plurality of candidate promoters may be determined based on the first characteristic of the target provider and the second characteristic of the plurality of candidate promoters, and based on the priority levels, the target provider may be determined from the plurality of candidate promoters. party's target promotion party. Based on the facts of the present disclosure, promoters that better meet the needs of the target provider can be determined more efficiently.
  • FIG. 1 shows a schematic diagram of an example environment 100 in which various embodiments of the present disclosure can be implemented.
  • computing device 130 recalls a plurality of candidate promoters 150-1, 150-2 through 150-M (individually or collectively referred to as candidate promoters 150) for target provider 110 from promoter set 145 .
  • the target provider 110 can provide at least one object 170 (eg, tangible goods, digital content, or a particular service) that is available to the user 180 .
  • object 170 eg, tangible goods, digital content, or a particular service
  • Such target providers 110 may include any individual or organization capable of providing objects 170 .
  • examples of target providers 110 may include, but are not limited to, physical or virtual stores that sell merchandise, news service providers that provide news subscription services, restaurants that provide catering services, and music service providers that provide music services.
  • Promoter set 145 may include a plurality of promoters capable of publishing introductory content.
  • a promoter is any individual or organization capable of providing guiding content 190 for guiding users 180 to acquire corresponding objects.
  • examples of promoters may include, but are not limited to, hosts who deliver goods in live broadcasts, authors who write restaurant reviews, radio hosts who provide music sharing, creators who publish video works, and so on.
  • the promoter set 145 may include all the anchors who can undertake the delivery service. The process of recalling multiple candidate promoters 150 from the promoter set 145 will be described in detail below, and will not be described in detail here.
  • the computing device 130 may determine multiple candidate promoters based on the first feature 120 of the target provider 110 and the second feature 140 of the multiple candidate promoters 150 priority level.
  • the first feature 120 may, for example, be used to characterize the target provider 110 or at least one object 170 provided by the first provider.
  • the second feature 140 may be used, for example, to characterize the candidate promoter 150 or the guiding content 190 published by the candidate promoter 150 .
  • the process of determining the priority level based on the first feature 120 and the second feature 140 will be described in detail below, and will not be described in detail here.
  • Computing device 130 may further the determined priority level to determine target promoters, such as promoters 150-1 and 150-2, for target provider 110 from the recalled plurality of candidate promoters 150. For example, the computing device 130 may select a candidate promoter whose priority level is greater than a threshold level from the plurality of candidate promoters 150 as a target promoter. Alternatively, the computing device 130 may also select a predetermined number of target promoters with higher priority levels according to the order of priority levels.
  • target promoters such as promoters 150-1 and 150-2
  • FIG. 2 shows a flowchart of an example process 200 of determining a promoter according to some embodiments of the present disclosure.
  • the process 200 may be implemented, for example, at the computing device 130 of FIG. 1 .
  • the process 200 will be described below in conjunction with the scenario of the merchant and the host, and it should be understood that the process 200 can also be applied to other suitable providers or promoters.
  • computing device 130 recalls a plurality of candidate promoters 150 from promoter set 145 for target provider 110 capable of providing at least one object 170 available to user 180,
  • the plurality of candidate promoters 150 can publish the guiding content 190 for guiding the user to acquire the corresponding object.
  • "recall" refers to the process of selecting a plurality of candidate promoters 150 from a plurality of promoter sets 145 .
  • the computing device 130 may recall multiple candidate anchors from the anchor set in response to the target merchant's request. For example, computing device 130 may initiate recall of candidate anchors in response to the target merchant logging into the anchor acquisition page. Alternatively, the computing device 130 may also determine that the target merchant has a need to cooperate with the anchor in response to the target merchant listing new goods, and automatically initiates the recall of the candidate anchor.
  • computing device 130 may recall a plurality of candidate promoters 150 from promoter set 145 using a predetermined recall strategy or strategies. In some embodiments, computing device 130 may recall multiple candidate promoters 150 using a combination of multiple recall strategies in order to enrich the results. The detailed process of recalling the target promoter will be described below with reference to FIG. 3 .
  • FIG. 3 shows a flowchart of an example process 300 of recalling a promoter according to some embodiments of the present disclosure.
  • computing device 130 may determine sets of candidate promoters from promoter set 145 corresponding to various recall strategies.
  • computing device 130 may utilize a field-aware factorization machine FFM recall strategy for recall. Specifically, the computing device 130 may utilize the trained FFM model to determine the first domain vector of the target provider 110 and the second domain vector of the plurality of candidate promoters 150, and based on the distance of the first domain vector and the second domain vector , to determine a set of candidate promoters.
  • FFM recall strategy for recall.
  • the computing device 130 may utilize the trained FFM model to determine the first domain vector of the target provider 110 and the second domain vector of the plurality of candidate promoters 150, and based on the distance of the first domain vector and the second domain vector , to determine a set of candidate promoters.
  • the FFM model can be trained, for example, based on the cooperation information of the provider and the promoter, so that the partnered provider and the promoter have a close distance in the vector space. Conversely, the providers and promoters that have not cooperated have a large distance in the vector space. Based on this manner, the computing device 130 may search for promoters that are closer to the first domain vector of the target provider 110 in the vector space as candidate promoters.
  • computing device 130 may utilize a collaborative recall strategy for recall. Specifically, the computing device 130 may determine the historical promoters with which the target provider has cooperated based on the historical cooperation information of the target provider 110 . Subsequently, computing device 130 may obtain a set of candidate promoters that differ from historical promoters by less than a predetermined threshold.
  • the computing device 130 may determine the historical anchors that the target merchant has cooperated with, and recall anchors similar to the historical anchors as candidate anchors. For example, if the target merchant once cooperated with the first anchor who carried goods in the beauty field, the computing device 130 can recall, for example, the second anchor who also carried goods in the beauty field and had a similar size of fans.
  • the difference between promoters may be determined in an appropriate manner. Taking the anchor as an example, the difference can be determined based on the attribute information of the anchor, such as the category of goods brought, the live broadcast duration, and the like. Alternatively, the differences between the anchors, such as the number of fans or the distribution of fan attributes, etc., may also be determined based on attribute information of the fans of the anchors.
  • computing device 130 may utilize a joint recall strategy for recall. Specifically, the computing device 130 may determine, based on the historical contact information of the target provider 110, the promoters previously contacted by the target provider 110 as a set of candidate promoters.
  • the computing device 130 may obtain the anchors that the target merchant has contacted in the platform, and recall such anchors as candidate anchors. It should be understood that such anchors may be anchors who have cooperated before, or may be anchors who have only contacted without cooperation.
  • computing device 130 may utilize an extended lookalike recall strategy for recall. Specifically, computing device 130 may determine a set of extended users based on a set of seed users associated with the target provider, wherein the set of seed users acquired objects provided by the target provider within a predetermined period of time. Subsequently, computing device 130 may obtain the promoters associated with the extended set of users as a set of candidate promoters.
  • the computing device 130 can use users who have purchased goods from the target merchant as positive samples, that is, seed users, and users who click but have not purchased as negative samples, so as to construct a seed population for the merchant. Learning models. Accordingly, the computing device 130 may utilize the crowd learning model to determine an extended set of users from the users of the live streaming platform. Such extended users, for example, also have the potential to purchase goods from the target merchant.
  • the computing device 130 may obtain a group of candidate anchors based on the anchors that the expanded user follows, the anchors who have watched the live broadcast, and the anchors who have purchased goods through the live broadcast with goods.
  • computing device 130 may utilize a hot recall strategy for recall. Specifically, computing device 130 may determine a set of candidate promoters whose popularity exceeds a threshold, where popularity indicates the degree to which the promoter is followed by the user.
  • the computing device 130 may recall, for example, a predetermined number of the most popular current anchors on the platform as candidate anchors. It should be understood that such popularity can be based on, for example, the number of fans, the number of users who watch the live broadcast, the number of users who have successfully brought goods, and the user's support for the anchor (for example, the number of likes, reposts, comments, and gifts) wait to be determined.
  • computing device 130 may utilize a similar object recall strategy for recall. Specifically, the computing device 130 may determine a group of similar objects that differ from at least one object 170 by less than a predetermined threshold; and obtain promoters associated with the group of similar objects as a group of candidate promoters.
  • the computing device 130 may determine a group of commodities currently sold by the target merchant, and determine a group of similar commodities based on the similarity between the commodities. For example, the target merchant may currently sell beauty products of brand A, and the computing device 130 may determine beauty products of brand B that are close to the price of the beauty product. The computing device 130 may then recall the anchors who carried the similar set of merchandise as candidate anchors. For example, the computing device 130 may recall anchors who have carried more than a threshold number of beauty products of brand B in the past predetermined period of time as candidate anchors.
  • recall strategies may be employed to recall multiple groups of candidate promoters.
  • computing device 130 may select a plurality of candidate promoters 150 from the plurality of sets of candidate promoters.
  • the computing device 130 may, for example, select multiple candidate promoters 150 from multiple groups of candidate promoters.
  • the computing device 130 may select a plurality of candidate promoters 150 from a plurality of groups of candidate promoters not exceeding a predetermined number by means of serpentine merging.
  • computing device 130 may select one candidate promoter as the recalled plurality of candidates 150 in accordance with the 4 groups of candidate promoters.
  • the computing device 130 may further constrain individual candidate promoters such that at most a threshold number of candidate promoters in each group of candidate promoters are included in the among the selected multiple candidate promoters.
  • computing device 130 may also filter self-broadcasting promoters, considering that some promoters may have a weaker willingness to cooperate with other providers. Specifically, the computing device 130 may exclude the self-broadcasting promoter from the multiple groups of candidate promoters to obtain a plurality of candidate promoters, wherein the object that the self-broadcasting promoter guides the user to acquire within a predetermined time period is the self-broadcasting promoter or Provided by affiliates of the self-broadcasting promoter.
  • such a self-broadcasting promoter may, for example, refer to a self-broadcast anchor, and the self-broadcast anchor has sold the goods in the live broadcast in the past month, or is sold by the self-broadcast anchor. Sold by affiliates (eg, affiliated companies) of self-broadcast hosts.
  • Such self-broadcast anchors usually have weak cooperation willingness, and by filtering such self-broadcast anchors, it is possible to avoid providing anchors with low cooperation willingness to target merchants.
  • the computing device 130 is based on the first feature 120 of the target provider 110 and the multiple candidate promoters 150.
  • the second feature 140 is to determine the priority level of the plurality of candidate promoters.
  • the first characteristic may characterize the user attributes of a first set of associated users associated with the target provider 110
  • the second characteristic may characterize user attributes of a second set of associated users associated with the candidate promoter
  • the first feature may represent, for example, relevant attributes of purchasing users who have purchased goods sold by the target merchant.
  • the second feature may, for example, represent the relevant attributes of a group of fans who have paid attention to the candidate anchor.
  • the first characteristic may characterize first statistical information associated with the target provider
  • the second characteristic may characterize second statistical information associated with the candidate promoter, wherein the first statistical information and the second statistical information At least one of is updated in real-time or periodically in response to user operations.
  • the first statistical information associated with the target merchant may include, for example, some data that is updated in real time, such as the target merchant's sales, the number of positive reviews from users, the number of negative reviews from users, the number of product views, and the number of product reviews. Add to cart times and more.
  • the computing device 130 may, for example, utilize a pre-buried point so that a specific operation of the user can trigger real-time updating of the first statistical information.
  • the first statistical information can also be updated periodically, for example.
  • the first statistical information may indicate the sales of the target merchant in the past 30 days, the total number of user reviews in the past 30 days, and the like. Such statistical information may, for example, be regularly updated by the platform on a daily basis.
  • the second statistical information may also include, for example, some data updated in real time, such as the total amount of goods brought by the candidate anchors, the number of fans, and the like.
  • the computing device 130 may, for example, utilize a pre-buried point so that a specific operation of the user can trigger a real-time update of the second statistical information.
  • the second statistical information can also be updated periodically, for example.
  • the second statistical information may indicate the amount of goods brought by the candidate anchor in the past 30 days, the number of new fans in the past 30 days, and the like.
  • Such statistical information may, for example, be regularly updated by the platform on a daily basis.
  • the first characteristic may represent a first attribute of a history promoter that has worked with the target provider
  • the second characteristic may represent a second attribute of a history provider that has worked with the candidate promoter.
  • Such a first attribute and a second attribute are intended to describe the characteristics of the promoters with which the target provider has cooperated, and the characteristics of the providers with which the candidate promoter has cooperated.
  • computing device 130 may generate input features based on the first feature representation of first feature 120 and the second feature representation of second feature 140 in order to determine the priority level.
  • computing device 130 may, for example, concatenate the first feature representation and the second feature representation to obtain input features.
  • feature representations may include, for example, feature portions corresponding to user attributes of associated users, feature portions corresponding to statistical information, and/or feature portions corresponding to attributes of historical partners. Based on this method, the target provider 110 and the multiple candidate promoters 150 can be more comprehensively characterized.
  • computing device 130 may process the input features using a priority model that is trained based on historical collaboration information for a set of training providers and a set of training promoters to determine priority levels.
  • computing device 130 may obtain a trained priority model.
  • Such prioritization models can be implemented by appropriate machine learning models (eg, deep neural networks). It should be understood that the priority model may be trained by the same or different training device as computing device 130 .
  • the training device can obtain a set of training providers and a set of training promoters, and construct multiple provider-promoter sample pairs. For each sample pair, the training device may determine the input features of the input-valued model based on the manner discussed above, and use whether the provider-promoter has ever cooperated as the ground truth for model training (eg, 1 may represent cooperated, 0 means not cooperated) to train the priority model.
  • the training device may determine the input features of the input-valued model based on the manner discussed above, and use whether the provider-promoter has ever cooperated as the ground truth for model training (eg, 1 may represent cooperated, 0 means not cooperated) to train the priority model.
  • the trained priority model can receive input features and input probabilities of 0-1 to characterize the probability of cooperation between the target provider 110 and the candidate promoter 150, such probability may be determined as, for example, The priority level of the candidate promoter.
  • the computing device 130 may determine a target promoter for the target provider from the plurality of candidate promoters based on the priority level.
  • the computing device 130 may select a candidate promoter with a priority level greater than a threshold level from the plurality of candidate promoters 150 as a target promoter.
  • the computing device 130 may also select a predetermined number of target promoters with higher priority levels according to the order of priority levels.
  • computing device 130 may also adjust the priority level of at least one candidate promoter before sorting to obtain the final target promoter, and determine the target promoter based on the adjusted priority level.
  • computing device 130 may lower the priority level of at least one candidate promoter that has worked with the targeting provider, given that targeting provider 110 may be more expected to collaborate with the new promoter.
  • computing device 130 may determine the degree to which the priority level is reduced based on first lead information of the at least one candidate sponsor, the first lead information indicating leads issued via the at least one candidate sponsor within a predetermined period of time The amount of objects fetched for the content. Further, computing device 130 may lower the priority level based on the degree.
  • the first guidance information may, for example, indicate that the host who has worked with the host is the quantity of goods delivered by the target merchant.
  • the larger the amount of goods brought by the anchor it means that the target merchant may be in close enough contact with the anchor, and the computing device 130 may not provide the anchor additionally.
  • the computing device 130 may, for example, calculate the proportion of the amount of goods brought by the anchor to the amount of goods sold by the target merchant through the anchor, and determine the degree to which the priority level should be lowered based on the proportion. .
  • the priority level of anchors with a larger proportion may be reduced to a greater extent, and conversely, the priority levels of anchors with a smaller proportion may be reduced to a lesser extent, for example.
  • the computing device 130 may also determine differences between the first evaluation information of the plurality of candidate promoters and the second evaluation information of the historical promoters, including the promoters who have previously cooperated with the target provider . Further, computing device 130 may adjust the priority level based on the difference such that the priority level of candidate promoters with a difference greater than a threshold is lowered.
  • the first evaluation information may include, for example, the level of the candidate anchor on the platform
  • the second evaluation information may include, for example, the level of the historical anchors that the target merchant has cooperated with.
  • the level of anchors that merchants are willing to cooperate with is relatively stable. For example, larger-scale stores are generally reluctant to cooperate with anchors with lower levels, and smaller stores are generally difficult to pay for anchors with higher levels. cooperation costs. Based on this method, the provided results can be more in line with the expectations of the merchants.
  • computing device 130 may also determine at least one candidate promoter based on second lead information of the plurality of candidate promoters, wherein the second lead information indicates a lead published via the plurality of candidate promoters within a predetermined period of time content and the amount associated with at least one candidate promoter is below a threshold amount. Further, computing device 130 may lower the priority level of at least one candidate promoter.
  • the second guide information may, for example, represent the amount or quantity (eg, also referred to as the amount of merchandise) that the host guides the user to purchase commodities within a predetermined period of time.
  • Computing device 130 may, for example, lower the priority level of anchors carrying less than a threshold.
  • the computing device 130 may filter target promoters based on the adjusted priority level.
  • computing device 130 may also present information associated with the target promoter to the target provider.
  • computing device 130 may send information associated with the target promoter to the target promoter, such information may include, for example, a description about the target promoter.
  • the computing device 130 may, for example, provide the target merchant with the determined information of the target anchor, such as the anchor's level, number of fans, category of merchandise, recent merchandise volume, contact information, and the like. Such information can help the target merchant to more easily understand the characteristics of the target anchor, thereby promoting cooperation between the two parties.
  • the target promoters include a first promoter and a second promoter, wherein the first promoter has a higher priority level than the second promoter. Accordingly, the first information associated with the first promoter may have a higher presentation priority than the second information associated with the second promoter.
  • computing device 130 may present information for a limited number of target promoters, such as through a list, such that information for target promoters with a higher priority level may be presented higher up in the list. It should be understood that the first information may also be presented more prominently in other suitable manners.
  • embodiments of the present disclosure can utilize feature engineering to determine the priority level from among a plurality of candidate promoters that are initially recalled, and thereby more accurately determine the target promoter suitable for the target provider , thereby increasing the possibility of further cooperation between the two parties.
  • FIG. 4 shows a schematic structural block diagram of an apparatus 400 for entity clustering according to some embodiments of the present disclosure.
  • the apparatus 400 may include a recall module 410 configured to recall a plurality of candidate promoters for the target provider from the promoter set, and the target provider can provide at least one object available to the user, a plurality of Candidate promoters can publish guiding content for guiding users to acquire corresponding objects.
  • the apparatus 400 also includes a ranking module 420 configured to determine the priority level of the plurality of candidate promoters based on the first characteristic of the target provider and the second characteristic of the plurality of candidate promoters.
  • the apparatus 400 also includes a determination module 430 configured to determine a target promoter for the target provider from the plurality of candidate promoters based on the priority level.
  • recall module 410 is further configured to: determine sets of candidate promoters corresponding to the plurality of recall strategies from the set of promoters; and select a plurality of candidate promoters from the sets of candidate promoters.
  • the plurality of recall strategies include a field-aware factorization machine FFM recall strategy
  • the recall module 410 is further configured as an FFM module configured to utilize the FFM model to determine the first domain vector of the target provider and the plurality of a second domain vector of promoters; and determining a set of candidate promoters based on the distance of the first domain vector and the second domain vector.
  • the plurality of recall strategies include a collaborative recall strategy
  • the recall module 410 is further configured to: determine, based on historical cooperation information of the target provider, historical promoters with which the target provider has cooperated; and obtain and historical promoters A set of candidate promoters whose difference is less than a predetermined threshold.
  • the plurality of recall strategies include a joint recall strategy
  • the recall module 410 is further configured to: based on historical contact information of the target provider, determine the promoters previously contacted by the target provider as a set of candidate promotions square.
  • the plurality of recall strategies include an extended lookalike recall strategy
  • the recall module 410 is further configured to: determine a set of extended users based on a set of seed users associated with the target provider, a set of seed users in acquiring objects provided by the target provider within a predetermined time period; and acquiring promoters associated with a set of extended users as a set of candidate promoters.
  • the plurality of recall strategies includes a popular recall strategy
  • the recall module 410 is further configured to: determine a set of candidate promoters whose popularity exceeds a threshold, the popularity indicating the degree of attention of the promoter by the user.
  • the plurality of recall strategies includes a similar object recall strategy
  • the recall module 410 is further configured to: determine a set of similar objects that differ from at least one object by less than a predetermined threshold; and obtain information related to the set of similar objects Associated promoters as a group of candidate promoters.
  • At most a threshold number of candidate promoters in each set of candidate promoters are included in the selected plurality of candidate promoters.
  • the recall module 410 is further configured to: exclude self-broadcasting promoters from multiple groups of candidate promoters to obtain multiple candidate promoters, and the objects that the self-broadcasting promoters guide the user to acquire within a predetermined period of time are: Provided by the self-broadcasting promoter or an affiliate of the self-broadcasting promoter.
  • the first feature characterizes user attributes of a first set of associated users associated with the target provider
  • the second feature characterizes user attributes of a second set of associated users associated with the candidate promoter
  • the first characteristic represents first statistical information associated with the target provider
  • the second characteristic represents second statistical information associated with the candidate promoter
  • An item is updated in real time or periodically in response to user actions.
  • the first feature represents a first attribute of a history promoter that has worked with the target provider
  • the second feature represents a second attribute of a history provider that has worked with the candidate promoter
  • ranking module 420 is further configured to: generate input features based on the first feature representation of the first feature and the second feature representation of the second feature; and process the input features using a priority model to determine priorities At the priority level, the priority model is trained based on the historical cooperation information of a set of training providers and a set of training promoters.
  • determination module 430 is further configured to: adjust a priority level of at least one candidate promoter of the plurality of candidate promoters; and determine a target promoter based on the adjusted priority level.
  • the determination module 430 is further configured to reduce the priority level of at least one candidate promoter that has worked with the target provider.
  • the determining module 430 is further configured to: determine a degree to which the priority level is reduced based on first guidance information of the at least one candidate promoter, the first guidance information indicating promotion via the at least one candidate within a predetermined period of time The amount of objects that are fetched to guide content published by a party; and the lowering of the priority level based on the degree.
  • the determining module 430 is further configured to: determine the difference between the first evaluation information of the plurality of candidate promoters and the second evaluation information of the historical promoters, including the historical promoters who have previously cooperated with the target provider and adjusting the priority level based on the difference such that the priority level of candidate promoters with a difference greater than a threshold is lowered.
  • the determining module 430 is further configured to: determine at least one candidate promoter based on second guidance information of the plurality of candidate promoters, wherein the second guidance information indicates that the plurality of candidate promoters are routed through the plurality of candidate promoters within a predetermined period of time. an amount of objects that were acquired for the published directed content, and an amount associated with the at least one candidate promoter is below a threshold amount; and reducing the priority level of the at least one candidate promoter.
  • the apparatus 400 further includes a providing module configured to present information associated with the target promoter to the target provider.
  • the target promoter includes a first promoter and a second promoter, the first promoter has a higher priority level than the second promoter, and the first information associated with the first promoter has a higher priority than the second promoter The second information associated with the second promoter has a higher presentation priority.
  • the units included in the apparatus 400 may be implemented in various manners, including software, hardware, firmware, or any combination thereof.
  • one or more units may be implemented using software and/or firmware, such as machine-executable instructions stored on a storage medium.
  • some or all of the units in apparatus 400 may be implemented, at least in part, by one or more hardware logic components.
  • exemplary types of hardware logic components include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standards (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLD), etc.
  • FIG. 5 illustrates a block diagram of a computing device/server 500 in which one or more embodiments of the present disclosure may be implemented. It should be understood that the computing device/server 500 shown in FIG. 5 is merely exemplary and should not constitute any limitation on the functionality and scope of the embodiments described herein.
  • computing device/server 500 is in the form of a general-purpose computing device.
  • Components of computing device/server 500 may include, but are not limited to, one or more processors or processing units 510, memory 520, storage devices 530, one or more communication units 540, one or more input devices 550, and one or more Output device 560.
  • the processing unit 510 may be an actual or virtual processor and can perform various processes according to programs stored in the memory 520 . In a multi-processor system, multiple processing units execute computer-executable instructions in parallel to increase the parallel processing capabilities of the computing device/server 500 .
  • Computing device/server 500 typically includes multiple computer storage media. Such media can be any available media that can be accessed by computing device/server 500, including but not limited to volatile and nonvolatile media, removable and non-removable media.
  • Memory 520 may be volatile memory (eg, registers, cache, random access memory (RAM)), non-volatile memory (eg, read only memory (ROM), electrically erasable programmable read only memory (EEPROM) , Flash) or some combination of them.
  • Storage device 530 may be removable or non-removable media, and may include machine-readable media, such as flash drives, magnetic disks, or any other media that may be capable of storing information and/or data (eg, training data for training). ) and can be accessed within the computing device/server 500.
  • Computing device/server 500 may further include additional removable/non-removable, volatile/non-volatile storage media.
  • disk drives for reading or writing from removable, non-volatile magnetic disks eg, "floppy disks" and for reading or writing from removable, non-volatile optical disks
  • CD-ROM drive for reading or writing.
  • each drive may be connected to a bus (not shown) by one or more data media interfaces.
  • Memory 520 may include a computer program product 525 having one or more program modules configured to perform various methods or actions of various embodiments of the present disclosure.
  • the communication unit 540 enables communication with other computing devices through a communication medium. Additionally, the functions of the components of computing device/server 500 may be implemented in a single computing cluster or multiple computing machines capable of communicating through a communication connection. Accordingly, computing device/server 500 may operate in a networked environment using logical connections to one or more other servers, network personal computers (PCs), or another network node.
  • PCs network personal computers
  • Input device 550 may be one or more input devices, such as a mouse, keyboard, trackball, and the like.
  • Output device 560 may be one or more output devices, such as a display, speakers, printer, and the like.
  • the computing device/server 500 may also communicate with one or more external devices (not shown), such as storage devices, display devices, etc., through the communication unit 540, as needed, with one or more external devices that enable the user to communicate with the computing device/server. 500 interacts with any device (eg, network card, modem, etc.) that enables computing device/server 500 to communicate with one or more other computing devices. Such communication may be performed via an input/output (I/O) interface (not shown).
  • I/O input/output
  • a computer-readable storage medium having stored thereon one or more computer instructions, wherein the one or more computer instructions are executed by a processor to implement the method described above.
  • These computer readable program instructions may be provided to the processing unit of a general purpose computer, special purpose computer or other programmable data processing apparatus to produce a machine that causes the instructions when executed by the processing unit of the computer or other programmable data processing apparatus , resulting in means for implementing the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • These computer readable program instructions can also be stored in a computer readable storage medium, these instructions cause a computer, programmable data processing apparatus and/or other equipment to operate in a specific manner, so that the computer readable medium storing the instructions includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • Computer-readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other equipment to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other equipment to produce a computer-implemented process , thereby causing instructions executing on a computer, other programmable data processing apparatus, or other device to implement the functions/acts specified in one or more blocks of the flowcharts and/or block diagrams.
  • each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executables for implementing the specified logical function(s) instruction.
  • the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations can be implemented in dedicated hardware-based systems that perform the specified functions or actions , or can be implemented in a combination of dedicated hardware and computer instructions.

Landscapes

  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Finance (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

根据本公开的实施例，提供了一种用于确定推广方的方法、装置、设备、存储介质和程序产品。该方法包括：从推广方集合中召回针对目标提供方的多个候选推广方，目标提供方能够提供用户可获取的至少一项对象，多个候选推广方能够发布用于引导用户获取相应对象的引导内容；基于目标提供方的第一特征和多个候选推广方的第二特征，确定多个候选推广方的优先级水平；以及基于优先级水平，从多个候选推广方中确定针对目标提供方的目标推广方。根据本公开的事实，可以更为高效地确定更符合目标提供方需求的推广方。

Description

确定推广方的方法、装置、设备、存储介质和程序产品
本申请要求于2021年4月30日提交中国专利局、申请号为202110484423.5、发明名称为“确定推广方的方法、装置、设备、存储介质和程序产品”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本公开的各实现方式涉及计算机领域，更具体地，涉及实体聚类的方法、装置、设备和计算机存储介质。
背景技术
随着信息技术的发展，人们在日常生活中能够接触到各种各样的引导内容，例如，文字或视频广告或带货视频直播等。这些引导内容能够引导人们去获取相应的对象，这样的对象例如可以包括有形的商品、数字内容或者是特定的服务等。
一些提供方(例如，店铺或服务提供商等)通过与推广方合作来促进对象更好地被用户了解，或引导更多用户获取这些对象。然而，对于提供方而言，需要耗费大量的时间成本和人力成本，才能够从海量的推广方中筛选出符合自己需要的推广方。因此，如何有效地为提供方提供符合需要的推广方已经成为关注的焦点。
发明内容
在本公开的第一方面，提供了一种用于确定推广方的方法。该方法包括：从推广方集合中召回针对目标提供方的多个候选推广方，目标提供方能够提供用户可获取的至少一项对象，多个候选推广方能够发布用于引导用户获取相应对象的引导内容；基于目标提供方的第一特征和多个候选推广方的第二特征，确定多个候选推广方的优先级水平；以及基于优先级水平，从多个候选推广方中确定针对 目标提供方的目标推广方。
在本公开的第二方面中，提供了一种用于实确定推广方的装置。该装置包括：召回模块，被配置为从推广方集合中召回针对目标提供方的多个候选推广方，目标提供方能够提供用户可获取的至少一项对象，多个候选推广方能够发布用于引导用户获取相应对象的引导内容；排序模块，被配置为基于目标提供方的第一特征和多个候选推广方的第二特征，确定多个候选推广方的优先级水平；以及确定模块，被配置为基于优先级水平，从多个候选推广方中确定针对目标提供方的目标推广方。
在本公开的第三方面，提供了一种电子设备，包括：存储器和处理器；其中存储器用于存储一条或多条计算机指令，其中一条或多条计算机指令被处理器执行以实现根据本公开的第一方面的方法。
在本公开的第四方面，提供了一种计算机可读存储介质，其上存储有一条或多条计算机指令，其中一条或多条计算机指令被处理器执行实现根据本公开的第一方面的方法。
在本公开的第五方面，提供了一种计算机程序产品，其包括一条或多条计算机指令，其中一条或多条计算机指令被处理器执行实现根据本公开的第一方面的方法。
根据本公开的各种实施例，能够有效地基于目标提供方合作过的历史推广方以及候选推广方合作过的历史提供方的信息，来高效地从召回的多个候选推广方中筛选出符合目标提供方需要的推广方。
附图说明
结合附图并参考以下详细说明，本公开各实施例的上述和其他特征、优点及方面将变得更加明显。在附图中，相同或相似的附图标注表示相同或相似的元素，其中：
图1示出了本公开的多个实施例能够在其中实现的示例环境的 示意图；
图2示出了根据本公开的一些实施例的确定推广方的示例过程的流程图；
图3示出了根据本公开的一些实施例的召回候选推广方的示例过程的流程图；
图4示出了根据本公开的一些实施例的确定推广方的装置的示意性结构框图；以及
图5示出了能够实施本公开的多个实施例的计算设备的框图。
具体实施方式
下面将参照附图更详细地描述本公开的实施例。虽然附图中显示了本公开的某些实施例，然而应当理解的是，本公开可以通过各种形式来实现，而且不应该被解释为限于这里阐述的实施例，相反提供这些实施例是为了更加透彻和完整地理解本公开。应当理解的是，本公开的附图及实施例仅用于示例性作用，并非用于限制本公开的保护范围。
在本公开的实施例的描述中，术语“包括”及其类似用语应当理解为开放性包含，即“包括但不限于”。术语“基于”应当理解为“至少部分地基于”。术语“一个实施例”或“该实施例”应当理解为“至少一个实施例”。术语“第一”、“第二”等等可以指代不同的或相同的对象。下文还可能包括其他明确的和隐含的定义。
如上文所讨论的，越来越多的提供方期望通过与推广方合作来引导用户获取提供方所提供的对象。例如，一些商家可以与一些带货主播合作，并通过主播的直播内容来引导用户购买商家所销售的商品。
然而，随着直播行业的飞速发展，为了进行商品的推广，商家通常需要耗费大量的时间去从海量的主播中筛选出合适的主播。这将耗费商家较多的时间成本和人力成本。
由此可见，目前的方案难以为提供方有效地确定符合其需要的 推广方。
为了至少部分地解决上述问题以及其他潜在问题中的一个或者多个问题，本公开的示例实施例提出了确定推广方的方案。总体而言，根据在此描述的实施例，可以从推广方集合中召回针对目标提供方(例如，实体店铺、虚拟店铺、服务提供方等能够提供用户可获取的物理对象或者虚拟对象的实体)的多个候选推广方(例如，主播、视频创作者、文字内容创造者等)，其中目标提供方能够提供用户(例如，主播的粉丝、参加培训的学生或家长、视频的观看者、文章的读者等)可获取的至少一项对象(例如，有形的商品、数字内容或者特定的服务)，多个候选推广方能够发布用于引导用户获取相应对象的引导内容(例如，在线直播内容、视频文件、在线文章等)。
随后，可以基于目标提供方的第一特征和多个候选推广方的第二特征来确定多个候选推广方的优先级水平，并基于优先级水平来从多个候选推广方中确定针对目标提供方的目标推广方。根据本公开的事实，可以更为高效地确定更符合目标提供方需求的推广方。
以下将参照附图来具体描述本公开的实施例。
图1示出了本公开的多个实施例能够在其中实现的示例环境100的示意图。在该示例环境100中，计算设备130从推广方集合145中召回针对目标提供方110的多个候选推广方150-1、150-2至150-M(单独或统一称为候选推广方150)。如上文所讨论的，目标提供方110能够提供用户180可获取的至少一项对象170(例如，有形的商品、数字内容或者特定的服务)。这样的目标提供方110可以包括能够提供对象170的任何个体或组织。例如，目标提供方110的示例可以包括但不限于：销售商品的实体店铺或者虚拟店铺、提供新闻订阅服务的新闻服务提供方、提供餐饮服务的饭店和提供音乐服务的音乐服务提供方等。
推广方集合145可以包括能够发布引导内容的多个推广方。如上文所讨论的，推广方是能够提供用于引导用户180获取对应的对 象的引导内容190的任何个体或者组织。例如，推广方的示例可以包括但不限于：直播带货的主播、撰写餐厅评价的作者、提供音乐分享的电台节目主持人、发布视频作品的创作者等。以直播带货平台为例，推广方集合145可以包括能够承接带货服务的全部主播。关于从推广方集合145中召回得到多个候选推广方150的过程将在下文详细描述，在此暂不详叙。
如图1所示，在召回得到多个候选推广方150后，计算设备130可以基于目标提供方110的第一特征120和多个候选推广方150的第二特征140来确定多个候选推广方的优先级水平。第一特征120例如可以用于表征目标提供方110或者第一提供方所提供的至少一项对象170。第二特征140例如可以用于表征候选推广方150或者候选推广方150所发布的引导内容190。关于基于第一特征120和第二特征140来确定优先级水平的过程将在下文详细介绍，在此暂不详叙。
计算设备130可以进一步所确定的优先级水平来从所召回的多个候选推广方150中确定针对目标提供方110的目标推广方，例如推广方150-1和150-2。例如，计算设备130可以从多个候选推广方150中选择优先级水平大于阈值水平的候选推广方，以作为目标推广方。或者，计算设备130也可以根据优先级水平的排序，选择优先级水平较高的预定数目的目标推广方。
应当理解，图1中所示的候选推广方和目标推广方的数目仅是示意性地，不旨在构成对本公开的限制。
以下将结合图2来详细描述计算设备130确定目标推广方的过程。图2示出了根据本公开的一些实施例的确定推广方的示例过程200的流程图。该过程200例如可以在图1的计算设备130处实施。仅是出于方便描述的目的，以下将结合商家与主播的场景来描述过程200，应当理解，过程200同样可以被应用于其他适当的提供方或推广方。
如图2所示，在框202，计算设备130从推广方集合145中召回 针对目标提供方110的多个候选推广方150，目标提供方110能够提供用户180可获取的至少一项对象170，多个候选推广方150能够发布用于引导用户获取相应对象的引导内容190。在本公开中，“召回”表示从多个推广方集合145中筛选得到多个候选推广方150的过程。
以直播带货为例，计算设备130可以响应于目标商家的请求而从主播集合中召回得到多个候选主播。例如，计算设备130可以响应于目标商家登录到主播获取页面来启动候选主播的召回。或者，计算设备130也可以响应于目标商家上架了新的货物而判断目标商家存在与主播合作的需求，而自动地启动候选主播的召回。
在一些实施例中，计算设备130可以利用预定的一种或多种召回策略来从推广方集合145中召回多个候选推广方150。在一些实施例中，为了丰富结果，计算设备130可以利用多中召回策略的组合来召回多个候选推广方150。以下将结合图3来描述召回目标推广方的详细过程。图3示出了根据本公开的一些实施例的召回推广方的示例过程300的流程图。
如图3所示，在框302，计算设备130可以从推广方集合145中确定与多种召回策略相对应的多组候选推广方。
在一些实施例中，计算设备130可以利用场感知因子分解机FFM召回策略来进行召回。具体地，计算设备130可以利用经训练的FFM模型来确定目标提供方110的第一域向量和多个候选推广方150的第二域向量，并基于第一域向量和第二域向量的距离，来确定一组候选推广方。
在一些实施例中，FFM模型例如可以基于提供方和推广方的合作信息而被训练，以使得合作过的提供方和推广方在向量空间中具有较近的距离。相反，没有合作过提供方和推广方在向量空间中具有较远的距离。基于这样的方式，计算设备130可以寻找在向量空间中与目标提供方110的第一域向量距离较近的推广方，以作为候选推广方。
在另一些实施例中，计算设备130可以利用协同召回策略进行 召回。具体地，计算设备130可以基于目标提供方110的历史合作信息来确定目标提供方合作过的历史推广方。随后，计算设备130可以获取与历史推广方的差异小于预定阈值的一组候选推广方。
以直播带货作为示例，计算设备130可以确定目标商家曾经合作过的历史主播，并召回与历史主播相似的主播，以作为候选主播。例如，目标商家曾经与进行美妆领域带货的第一主播合作，则计算设备130例如可以召回同样进行美妆领域带货，且粉丝规模接近的第二主播。
应当理解，可以利用适当的方式来确定推广方之间的差异。以主播作为示例，可以基于主播的属性信息，例如带货类别、直播时长等来确定差异。或者，还可以基于主播的粉丝的属性信息来确定主播之间的差异，例如粉丝数量或粉丝属性分布等等。
基于这样的方式，可以召回与目标提供方110曾经合作的历史推广方类似的其他推广方。
在一些实施例中，计算设备130可以利用建联召回策略进行召回。具体地，计算设备130可以基于目标提供方110的历史联系信息，确定目标提供方110先前联系的推广方，以作为一组候选推广方。
以直播带货作为示例，计算设备130可以获取目标商家在平台中曾经联系过的主播，并将这样的主播召回以作为候选主播。应当理解，这样的主播可能是曾经合作过的主播，也可能是仅仅联系而没有合作的主播。
在一些实施例中，计算设备130可以利用相似扩展lookalike召回策略进行召回。具体地，计算设备130可以基于与目标提供方相关联的一组种子用户来确定一组扩展用户，其中一组种子用户在预定时间段内获取了目标提供方提供的对象。随后，计算设备130可以获取与一组扩展用户相关联的推广方，以作为一组候选推广方。
以直播带货作为示例，计算设备130可以将曾经购买了目标商家的货物的用户作为正样本，也即种子用户，并将点击但未购买的 用户作为负样本，从而构建针对该商家的种子人群学习模型。相应地，计算设备130可以利用该人群学习模型来从直播带货平台的用户中确定一组扩展用户。这样的扩展用户例如同样具有购买该目标商家的货物的潜力。
进一步地，计算设备130可以基于这样的扩展用户所关注的主播、曾经观看直播的主播、曾经经由带货直播而购买过商品的主播来获取一组候选主播。
在一些实施例中，计算设备130可以利用热门召回策略来进行召回。具体地，计算设备130可以确定热度超过阈值的一组候选推广方，其中热度指示推广方受用户关注的程度。
以直播带货作为示例，计算设备130例如可以召回平台中当前最为热门的预定数量的主播，以作为候选主播。应当理解，这样的热度例如可以基于粉丝的数目、观看直播的用户的数目、带货成功的用户的数目、用户对主播的支持情况(例如，点赞数、转发数、评论数、礼物数目)等来被确定。
在一些实施例中，计算设备130可以利用相似对象召回策略来进行召回。具体地，计算设备130可以确定与至少一项对象170的差异小于预定阈值的一组相似对象；以及获取与一组相似对象相关联的推广方，以作为一组候选推广方。
以直播带货作为示例，计算设备130可以确定目标商家当前出售的一组商品，并基于商品之间的相似性来确定一组类似的商品。例如，目标商家当前可能出售的是A品牌的美妆商品，计算设备130可以确定与该美妆商品价格接近的B品牌的美妆商品。随后，计算设备130可以召回带货了该组类似的商品的主播，以作为候选主播。例如，计算设备130可以召回在过去预定时间段内带货了B品牌的美妆商品的数量超过阈值的主播，以作为候选主播。
应当理解，可以采用其上示例召回策略中的一项或多项来召回得到多组候选推广方。
框304，计算设备130可以从多组候选推广方中选择多个候选推 广方150。
在一些实施例中，由于最终提供给目标提供方110的目标推广方的数量通常是有限的，为了降低计算负担，计算设备130例如可以从多组候选推广方中选择多个候选推广方150。
在一些实施例中，为了保证召回结果的丰富性，计算设备130可以利用蛇形合并的方式来从多组候选推广方中选择不超过预定数目的多个候选推广方150。
示例性地，当例如利用4种召回策略获得4组候选推广方时，计算设备130可以依从4组候选推广方选择一个候选推广方以作为经召回的多个候选方150。
在一些实施例中，考虑到不同召回策略的结果可能存在重复，计算设备130例如还可以对单独候选推广方进行约束，以使得每组候选推广方中至多有阈值数目的候选推广方被包括在经选择的多个候选推广方中。
在一些实施例中，考虑到某些推广方可能具有较弱的与其他提供方合作的意愿，计算设备130还可以过滤自播推广方。具体地，计算设备130可以从多组候选推广方中排除自播推广方，以获得多个候选推广方，其中自播推广方在预定时间段内引导用户获取的对象是由自播推广方或自播推广方的关联方所提供的。
以直播带货作为示例，这样的自播推广方例如可以是指自播主播，该自播主播在过去一个月中直播带货的货物均是由该自播主播所销售的，或者是由该自播主播的关联方(例如，所属公司)所销售的。这样的自播主播通常具有较弱的合作意愿，通过过滤这样的自播主播可以避免将具有较低合作意愿的主播提供给目标商家。
以上介绍了从推广方集合145中召回得到多个候选推广方150的过程，继续参考图2，在框204，计算设备130基于目标提供方110的第一特征120和多个候选推广方150的第二特征140来确定多个候选推广方的优先级水平。
在一些实施例中，第一特征可以表征与目标提供方110相关联 的第一组关联用户的用户属性，第二特征可以表征与候选推广方相关联的第二组关联用户的用户属性。
以直播带货作为示例，第一特征例如可以表征包括购买了目标商家所销售的货物的购买用户的相关属性。相应地，第二特征例如可以表征关注了候选主播的一组粉丝的相关属性。
在一些实施例中，第一特征可以表征与目标提供方相关联的第一统计信息，第二特征可以表征与候选推广方相关联的第二统计信息，其中第一统计信息和第二统计信息中的至少一项响应于用户操作而被实时地更新或周期性地更新。
以直播带货作为示例，与目标商家相关联的第一统计信息例如可以包括一些被实时更新的数据，例如目标商家的销售额、用户好评数目、用户差评数目、商品的浏览次数、商品的添加购物车次数等等。计算设备130例如可以利用预先的埋点，以使得用户的特定操作能够触发第一统计信息的实时更新。另一方面，第一统计信息例如还可以是被周期性地更新。例如，第一统计信息可以指示该目标商家在过去30天的销售额、过去30天用户评价的总数目等。这样的统计信息例如可以是由平台每天定期地更新。
相应地，第二统计信息例如也可以包括一些被实时更新的数据，例如候选主播的带货总量、粉丝的数目等等。计算设备130例如可以利用预先的埋点，以使得用户的特定操作能够触发第二统计信息的实时更新。
另一方面，第二统计信息例如还可以是被周期性地更新。例如，第二统计信息可以指示该候选主播在过去30天的带货量、过去30天粉丝新增数目等。这样的统计信息例如可以是由平台每天定期地更新。
在一些实施例中，第一特征可以表征与目标提供方合作过的历史推广方的第一属性，第二特征可以表征与候选推广方合作过的历史提供方的第二属性。这样的第一属性和第二属性旨在描述目标提供方曾经合作过的推广方的特性，以及候选推广方曾经合作过的提 供方的特性。
以上讨论了第一特征120和第二特征140可能表征的具体信息。在一些实施例中，为了确定优先级水平，计算设备130可以基于第一特征120的第一特征表示和第二特征140的第二特征表示来生成输入特征。
在一些实施例中，计算设备130例如可以将第一特征表示和第二特征表示进行级联，以获得输入特征。这样的特征表示例如可以包括与关联用户的用户属性对应的特征部分、与统计信息对应的特征部分、和/或与历史合作方的属性所对应的特征部分。基于这样的方式，可以更为全面的表征目标提供方110和多个候选推广方150。
进一步地，计算设备130可以利用优先级模型来处理输入特征，以确定优先级水平，其中优先级模型基于一组训练提供方与一组训练推广方的历史合作信息而被训练。
在一些实施例中，计算设备130可以获取经训练的优先级模型。这样的优先级模型可以通过适当的机器学习模型(例如，深度神经网络)来实现。应当理解，可以由与计算设备130相同或不同的训练设备来训练该优先级模型。
在训练过程，训练设备可以获取一组训练提供方和一组训练推广方，并构建多个提供方-推广方样本对。针对每个样本对，训练设备可以基于上文所讨论的方式来确定被输入值模型的输入特征，并且以该提供方-推广方是否曾经合作过作为模型训练的真值(例如，1可以表示合作过，0表示未合作过)来训练该优先级模型。
经过该训练过程，经训练的优先级模型能够接收输入特征，并输入0-1的概率，以表征目标提供方110与候选推广方150之间进行合作的概率，这样的概率例如可以被确定作为候选推广方的优先级水平。
继续参考图2，在框206，计算设备130可以基于优先级水平，从多个候选推广方中确定针对目标提供方的目标推广方。
在一些实施例中，计算设备130可以从多个候选推广方150中 选择优先级水平大于阈值水平的候选推广方，以作为目标推广方。或者，计算设备130也可以根据优先级水平的排序，选择优先级水平较高的预定数目的目标推广方。
在一些实施例中，在排序以获得最终的目标推广方之前，计算设备130还可以调整至少一个候选推广方的优先级水平，并基于经调整的优先级水平来确定目标推广方。
在一些实施例中，考虑到目标提供方110可以更加期望与新的推广方进行合作，计算设备130可以降低曾经与目标提供方合作的至少一个候选推广方的优先级水平。
在一些实施例中，计算设备130可以基于至少一个候选推广方的第一引导信息来确定优先级水平被降低的程度，第一引导信息指示在预定时间段内经由至少一个候选推广方发布的引导内容而被获取的对象的量。进一步地，计算设备130可以基于程度来降低优先级水平。
以直播带货作为示例，第一引导信息例如可以指示该曾经合作过的主播为该目标商家的带货量。相应地，该主播的带货量越大，则表示该目标商家可能与该主播之间的联系已经足够近，计算设备130可以不再额外地提供该主播。
在一些实施例中，计算设备130例如还可以计算该主播的带货量占该目标商家通过主播而出售的货物量的占比，并基于该占比来去确定优先级水平应当被下降的程度。例如，占比越大的主播的优先级水平例如可以被更大程度地降低，相反，占比较小的主播的优先级水平例如可以被更小程度地降低。
在一些实施例中，计算设备130还可以确定多个候选推广方的第一评价信息与历史推广方的第二评价信息之间的差异，历史推广方包括先前与目标提供方合作过的推广方。进一步地，计算设备130可以基于差异来调整优先级水平，以使得差异大于阈值的候选推广方的优先级水平被降低。
以直播带货作为示例，第一评价信息例如可以包括候选主播在 平台的等级，第二评价信息例如可以包括目标商家曾经合作过的历史主播的等级。通常而言，商家愿意合作的主播的等级是较为稳定的，例如，规模较大的店家一般不愿意与等级较低的主播合作，而规模较小的店家一般也难以支付等级较高主播可能产生的合作费用。基于这样的方式，可以使得所提供的结果更加符合商家的预期。
在一些实施例中，计算设备130还可以基于多个候选推广方的第二引导信息，确定至少一个候选推广方，其中第二引导信息指示在预定时间段内经由多个候选推广方发布的引导内容而被获取的对象的量，并且与至少一个候选推广方相关联的量低于阈值量。进一步地，计算设备130可以降低至少一个候选推广方的优先级水平。
以直播带货作为示例，第二引导信息例如可以表征主播在预定时间段内引导用户购买商品的金额或者数量(例如，也称为带货量)。计算设备130例如可以降低带货量低于阈值的主播的优先级水平。
进一步地，计算设备130可以基于经调整的优先级水平来筛选得到目标推广方。
在一些实施例中，计算设备130还可以向目标提供方呈现与目标推广方相关联的信息。示例性地，计算设备130可以向目标推广方发送与目标推广方相关联的信息，这样的信息例如可以包括关于目标推广方的描述。
以直播带货作为示例，计算设备130例如可以向目标商家提供所确定的目标主播的信息，例如，主播的等级、粉丝数目、带货类别、最近带货量、联系方式等。这样的信息能够帮助目标商家更加便捷地了解该目标主播的特征，进而促进双方开展合作。
在一些实施例中，其中目标推广方包括第一推广方和第二推广方，其中第一推广方的优先级水平高于第二推广方。相应地，与第一推广方相关联的第一信息可以具有比与第二推广方相关联的第二信息更高的呈现优先级。
示例性地，计算设备130例如可以通过列表来呈现不多个目标推广方的信息，并使得具有更高优先级水平的目标推广方的信息可 以被呈现在列表的更上端。应当理解，还可以通过其他适当的方式来使得第一信息被更加突出的呈现。
应当理解，上文所提及的涉及主播、商家、用户或粉丝的任何属性或特征均应当是在获取相应的主体许可的情况下所获取的。
基于上文所讨论的过程，本公开的实施例能够从初步召回的多个候选推广方中利用特征工程来进行优先级水平的确定，并从而更为准确地确定适合目标提供方的目标推广方，进而提高了双方开展进一步合作的可能性。
本公开的实施例还提供了用于实现上述方法或过程的相应装置。图4示出了根据本公开的一些实施例的实体聚类的装置400的示意性结构框图。
如图4所示，装置400可以包括召回模块410，被配置为从推广方集合中召回针对目标提供方的多个候选推广方，目标提供方能够提供用户可获取的至少一项对象，多个候选推广方能够发布用于引导用户获取相应对象的引导内容。装置400还包括排序模块420，被配置为基于目标提供方的第一特征和多个候选推广方的第二特征，确定多个候选推广方的优先级水平。装置400还包括确定模块430，被配置为基于优先级水平，从多个候选推广方中确定针对目标提供方的目标推广方。
在一些实施例中，召回模块410还被配置为：从推广方集合中确定与多种召回策略相对应的多组候选推广方；以及从多组候选推广方中选择多个候选推广方。
在一些实施例中，多个召回策略包括场感知因子分解机FFM召回策略，并且召回模块410还被配置为：FFM模块，被配置为利用FFM模型确定目标提供方的第一域向量和多个推广方的第二域向量；以及基于第一域向量和第二域向量的距离，来确定一组候选推广方。
在一些实施例中，多个召回策略包括协同召回策略，并且召回模块410还被配置为：基于目标提供方的历史合作信息，确定目标 提供方合作过的历史推广方；以及获取与历史推广方的差异小于预定阈值的一组候选推广方。
在一些实施例中，多个召回策略包括建联召回策略，并且召回模块410还被配置为：基于目标提供方的历史联系信息，确定目标提供方先前联系的推广方，以作为一组候选推广方。
在一些实施例中，多个召回策略包括相似扩展lookalike召回策略，并且召回模块410还被配置为：基于与目标提供方相关联的一组种子用户，确定一组扩展用户，一组种子用户在预定时间段内获取了目标提供方提供的对象；以及获取与一组扩展用户相关联的推广方，以作为一组候选推广方。
在一些实施例中，多个召回策略包括热门召回策略，并且召回模块410还被配置为：确定热度超过阈值的一组候选推广方，热度指示推广方受用户关注的程度。
在一些实施例中，多个召回策略包括相似对象召回策略，并且召回模块410还被配置为：确定与至少一项对象的差异小于预定阈值的一组相似对象；以及获取与一组相似对象相关联的推广方，以作为一组候选推广方。
在一些实施例中，每组候选推广方中至多有阈值数目的候选推广方被包括在经选择的多个候选推广方中。
在一些实施例中，召回模块410还被配置为：从多组候选推广方中排除自播推广方，以获得多个候选推广方，自播推广方在预定时间段内引导用户获取的对象是由自播推广方或自播推广方的关联方所提供的。
在一些实施例中，第一特征表征与目标提供方相关联的第一组关联用户的用户属性，第二特征表征与候选推广方相关联的第二组关联用户的用户属性。
在一些实施例中，第一特征表征与目标提供方相关联的第一统计信息，第二特征表征与候选推广方相关联的第二统计信息，第一统计信息和第二统计信息中的至少一项响应于用户操作而被实时地 更新或周期性地更新。
在一些实施例中，第一特征表征与目标提供方合作过的历史推广方的第一属性，第二特征表征与候选推广方合作过的历史提供方的第二属性。
在一些实施例中，排序模块420还被配置为：基于第一特征的第一特征表示和第二特征的第二特征表示，生成输入特征；以及利用优先级模型来处理输入特征，以确定优先级水平，优先级模型基于一组训练提供方与一组训练推广方的历史合作信息而被训练。
在一些实施例中，确定模块430还被配置为：调整多个候选推广方中的至少一个候选推广方的优先级水平；以及基于经调整的优先级水平，确定目标推广方。
在一些实施例中，确定模块430还被配置为：降低曾经与目标提供方合作的至少一个候选推广方的优先级水平。
在一些实施例中，确定模块430还被配置为：基于至少一个候选推广方的第一引导信息，确定优先级水平被降低的程度，第一引导信息指示在预定时间段内经由至少一个候选推广方发布的引导内容而被获取的对象的量；以及基于程度来降低优先级水平。
在一些实施例中，确定模块430还被配置为：确定多个候选推广方的第一评价信息与历史推广方的第二评价信息之间的差异，历史推广方包括先前与目标提供方合作过的推广方；以及基于差异来调整优先级水平，以使得差异大于阈值的候选推广方的优先级水平被降低。
在一些实施例中，确定模块430还被配置为：基于多个候选推广方的第二引导信息，确定至少一个候选推广方，其中第二引导信息指示在预定时间段内经由多个候选推广方发布的引导内容而被获取的对象的量，并且与至少一个候选推广方相关联的量低于阈值量；以及降低至少一个候选推广方的优先级水平。
在一些实施例中，装置400还包括提供模块，被配置为向目标提供方呈现与目标推广方相关联的信息。
在一些实施例中，目标推广方包括第一推广方和第二推广方，第一推广方的优先级水平高于第二推广方，并且与第一推广方相关联的第一信息具有比与第二推广方相关联的第二信息更高的呈现优先级。
装置400中所包括的单元可以利用各种方式来实现，包括软件、硬件、固件或其任意组合。在一些实施例中，一个或多个单元可以使用软件和/或固件来实现，例如存储在存储介质上的机器可执行指令。除了机器可执行指令之外或者作为替代，装置400中的部分或者全部单元可以至少部分地由一个或多个硬件逻辑组件来实现。作为示例而非限制，可以使用的示范类型的硬件逻辑组件包括现场可编程门阵列(FPGA)、专用集成电路(ASIC)、专用标准品(ASSP)、片上系统(SOC)、复杂可编程逻辑器件(CPLD)，等等。
图5示出了其中可以实施本公开的一个或多个实施例的计算设备/服务器500的框图。应当理解，图5所示出的计算设备/服务器500仅仅是示例性的，而不应当构成对本文所描述的实施例的功能和范围的任何限制。
如图5所示，计算设备/服务器500是通用计算设备的形式。计算设备/服务器500的组件可以包括但不限于一个或多个处理器或处理单元510、存储器520、存储设备530、一个或多个通信单元540、一个或多个输入设备550以及一个或多个输出设备560。处理单元510可以是实际或虚拟处理器并且能够根据存储器520中存储的程序来执行各种处理。在多处理器系统中，多个处理单元并行执行计算机可执行指令，以提高计算设备/服务器500的并行处理能力。
计算设备/服务器500通常包括多个计算机存储介质。这样的介质可以是计算设备/服务器500可访问的任何可以获得的介质，包括但不限于易失性和非易失性介质、可拆卸和不可拆卸介质。存储器520可以是易失性存储器(例如寄存器、高速缓存、随机访问存储器(RAM))、非易失性存储器(例如，只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、闪存)或它们的某种组合。 存储设备530可以是可拆卸或不可拆卸的介质，并且可以包括机器可读介质，诸如闪存驱动、磁盘或者任何其他介质，其可以能够用于存储信息和/或数据(例如用于训练的训练数据)并且可以在计算设备/服务器500内被访问。
计算设备/服务器500可以进一步包括另外的可拆卸/不可拆卸、易失性/非易失性存储介质。尽管未在图5中示出，可以提供用于从可拆卸、非易失性磁盘(例如“软盘”)进行读取或写入的磁盘驱动和用于从可拆卸、非易失性光盘进行读取或写入的光盘驱动。在这些情况中，每个驱动可以由一个或多个数据介质接口被连接至总线(未示出)。存储器520可以包括计算机程序产品525，其具有一个或多个程序模块，这些程序模块被配置为执行本公开的各种实施例的各种方法或动作。
通信单元540实现通过通信介质与其他计算设备进行通信。附加地，计算设备/服务器500的组件的功能可以以单个计算集群或多个计算机器来实现，这些计算机器能够通过通信连接进行通信。因此，计算设备/服务器500可以使用与一个或多个其他服务器、网络个人计算机(PC)或者另一个网络节点的逻辑连接来在联网环境中进行操作。
输入设备550可以是一个或多个输入设备，例如鼠标、键盘、追踪球等。输出设备560可以是一个或多个输出设备，例如显示器、扬声器、打印机等。计算设备/服务器500还可以根据需要通过通信单元540与一个或多个外部设备(未示出)进行通信，外部设备诸如存储设备、显示设备等，与一个或多个使得用户与计算设备/服务器500交互的设备进行通信，或者与使得计算设备/服务器500与一个或多个其他计算设备通信的任何设备(例如，网卡、调制解调器等)进行通信。这样的通信可以经由输入/输出(I/O)接口(未示出)来执行。
根据本公开的示例性实现方式，提供了一种计算机可读存储介质，其上存储有一条或多条计算机指令，其中一条或多条计算机指 令被处理器执行以实现上文描述的方法。
这里参照根据本公开实现的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其他可编程数据处理装置的处理单元，从而生产出一种机器，使得这些指令在通过计算机或其他可编程数据处理装置的处理单元执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其他可编程数据处理装置、或其他设备上，使得在计算机、其他可编程数据处理装置或其他设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其他可编程数据处理装置、或其他设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本公开的多个实现的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以 用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
以上已经描述了本公开的各实现，上述说明是示例性的，并非穷尽性的，并且也不限于所公开的各实现。在不偏离所说明的各实现的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实现的原理、实际应用或对市场中的技术的改进，或者使本技术领域的其他普通技术人员能理解本文公开的各实现。

Claims (20)

  1. 一种用于确定推广方的方法，包括：
    从推广方集合中召回针对目标提供方的多个候选推广方，所述目标提供方能够提供用户可获取的至少一项对象，所述多个候选推广方能够发布用于引导用户获取相应对象的引导内容；
    基于所述目标提供方的第一特征和所述多个候选推广方的第二特征，确定所述多个候选推广方的优先级水平；以及
    基于所述优先级水平，从所述多个候选推广方中确定针对所述目标提供方的目标推广方。
  2. 根据权利要求1所述的方法，其中从推广方集合中召回针对目标提供方的多个候选推广方包括：
    从所述推广方集合中确定与多种召回策略相对应的多组候选推广方；以及
    从所述多组候选推广方中选择所述多个候选推广方。
  3. 根据权利要求2所述的方法，其中所述多个召回策略包括协同召回策略，并且获取与多种召回策略相对应的多组候选推广方包括：
    基于所述目标提供方的历史合作信息，确定所述目标提供方合作过的所述历史推广方；以及
    获取与所述历史推广方的差异小于预定阈值的一组候选推广方。
  4. 根据权利要求2所述的方法，其中所述多个召回策略包括相似扩展lookalike召回策略，并且获取与多种召回策略相对应的多组候选推广方包括：
    基于与所述目标提供方相关联的一组种子用户，确定一组扩展用户，所述一组种子用户在预定时间段内获取了所述目标提供方提供的对象；以及
    获取与所述一组扩展用户相关联的推广方，以作为一组候选推 广方。
  5. 根据权利要求2所述的方法，其中所述多个召回策略包括相似对象召回策略，并且获取与多种召回策略相对应的多组候选推广方包括：
    确定与所述至少一项对象的差异小于预定阈值的一组相似对象；以及
    获取与所述一组相似对象相关联的推广方，以作为一组候选推广方。
  6. 根据权利要求2所述的方法，其中每组候选推广方中至多有阈值数目的候选推广方被包括在经选择的所述多个候选推广方中。
  7. 根据权利要求2所述的方法，其中从所述多组候选推广方中选择所述多个候选推广方包括：
    从所述多组候选推广方中排除自播推广方，以获得所述多个候选推广方，所述自播推广方在预定时间段内引导用户获取的对象是由所述自播推广方或所述自播推广方的关联方所提供的。
  8. 根据权利要求1所述的方法，其中所述第一特征表征与所述目标提供方相关联的第一组关联用户的用户属性，所述第二特征表征与所述候选推广方相关联的第二组关联用户的用户属性。
  9. 根据权利要求1所述的方法，其中所述第一特征表征与所述目标提供方相关联的第一统计信息，所述第二特征表征与所述候选推广方相关联的第二统计信息，所述第一统计信息和所述第二统计信息中的至少一项响应于用户操作而被实时地更新或周期性地更新。
  10. 根据权利要求1所述的方法，其中所述第一特征表征与所述目标提供方合作过的历史推广方的第一属性，所述第二特征表征与所述候选推广方合作过的历史提供方的第二属性。
  11. 根据权利要求1所述的方法，其中从所述多个候选推广方中确定针对所述目标提供方的目标推广方包括：
    调整所述多个候选推广方中的至少一个候选推广方的所述优先 级水平；以及
    基于经调整的所述优先级水平，确定所述目标推广方。
  12. 根据权利要求11所述的方法，其中调整所述多个候选推广方中的至少一个候选推广方的所述优先级水平包括：
    降低曾经与所述目标提供方合作的所述至少一个候选推广方的所述优先级水平。
  13. 根据权利要求12所述的方法，其中降低所述至少一个候选推广方的所述优先级水平包括：
    基于所述至少一个候选推广方的第一引导信息，确定所述优先级水平被降低的程度，所述第一引导信息指示在预定时间段内经由所述至少一个候选推广方发布的引导内容而被获取的对象的量；以及
    基于所述程度来降低所述优先级水平。
  14. 根据权利要求11所述的方法，其中调整所述多个候选推广方中的至少一个候选推广方的所述优先级水平包括：
    确定所述多个候选推广方的第一评价信息与历史推广方的第二评价信息之间的差异，所述历史推广方包括先前与所述目标提供方合作过的推广方；以及
    基于所述差异来调整所述优先级水平，以使得差异大于阈值的候选推广方的优先级水平被降低。
  15. 根据权利要求11所述的方法，其中调整所述多个候选推广方中的至少一个候选推广方的所述优先级水平包括：
    基于所述多个候选推广方的第二引导信息，确定所述至少一个候选推广方，其中所述第二引导信息指示在预定时间段内经由所述多个候选推广方发布的引导内容而被获取的对象的量，并且与所述至少一个候选推广方相关联的所述量低于阈值量；以及
    降低所述至少一个候选推广方的所述优先级水平。
  16. 根据权利要求1所述的方法，还包括：
    向所述目标提供方呈现与所述目标推广方相关联的信息；
    其中所述目标推广方包括第一推广方和第二推广方，所述第一推广方的所述优先级水平高于所述第二推广方，并且与所述第一推广方相关联的第一信息具有比与所述第二推广方相关联的第二信息更高的呈现优先级。
  17. 一种用于确定推广方的装置，包括：
    召回模块，被配置为从推广方集合中召回针对目标提供方的多个候选推广方，所述目标提供方能够提供用户可获取的至少一项对象，所述多个候选推广方能够发布用于引导用户获取相应对象的引导内容；
    排序模块，被配置为基于所述目标提供方的第一特征和所述多个候选推广方的第二特征，确定所述多个候选推广方的优先级水平；以及
    确定模块，被配置为基于所述优先级水平，从所述多个候选推广方中确定针对所述目标提供方的目标推广方。
  18. 一种电子设备，包括：
    存储器和处理器；
    其中所述存储器用于存储一条或多条计算机指令，其中所述一条或多条计算机指令被所述处理器执行以实现根据权利要求1至16中任一项所述的方法。
  19. 一种计算机可读存储介质，其上存储有一条或多条计算机指令，其中所述一条或多条计算机指令被处理器执行以实现根据权利要求1至16中任一项所述的方法。
  20. 一种计算机程序产品，包括一条或多条计算机指令，其中所述一条或多条计算机指令被处理器执行以实现根据权利要求1至16中任一项所述的方法。
PCT/CN2022/085597 2021-04-30 2022-04-07 确定推广方的方法、装置、设备、存储介质和程序产品 WO2022228075A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110484423.5 2021-04-30
CN202110484423.5A CN113205362A (zh) 2021-04-30 2021-04-30 确定推广方的方法、装置、设备、存储介质和程序产品

Publications (1)

Publication Number Publication Date
WO2022228075A1 true WO2022228075A1 (zh) 2022-11-03

Family

ID=77028200

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/085597 WO2022228075A1 (zh) 2021-04-30 2022-04-07 确定推广方的方法、装置、设备、存储介质和程序产品

Country Status (2)

Country Link
CN (1) CN113205362A (zh)
WO (1) WO2022228075A1 (zh)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113205362A (zh) * 2021-04-30 2021-08-03 北京有竹居网络技术有限公司 确定推广方的方法、装置、设备、存储介质和程序产品
CN113205382B (zh) * 2021-04-30 2024-02-20 北京有竹居网络技术有限公司 确定对象的方法、装置、设备、存储介质和程序产品
CN113837792A (zh) * 2021-08-23 2021-12-24 武汉卓尔数字传媒科技有限公司 推广合作对象推荐方法、装置、设备和存储介质
CN113850416A (zh) * 2021-08-26 2021-12-28 武汉卓尔数字传媒科技有限公司 广告推广合作对象确定方法和装置
CN116701770B (zh) * 2023-08-01 2023-10-27 北京创智汇聚科技有限公司 基于决策场景的请求响应优化方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140330647A1 (en) * 2013-05-03 2014-11-06 International Business Machines Corporation Application and service selection for optimized promotion
CN111178970A (zh) * 2019-12-30 2020-05-19 微梦创科网络科技（中国）有限公司 广告投放的方法及装置、电子设备和计算机可读存储介质
CN112100558A (zh) * 2020-09-02 2020-12-18 北京字节跳动网络技术有限公司 用于对象推荐的方法、装置、设备和存储介质
CN112118489A (zh) * 2020-09-07 2020-12-22 北京字节跳动网络技术有限公司 群组管理的方法、装置、设备和介质
CN113205362A (zh) * 2021-04-30 2021-08-03 北京有竹居网络技术有限公司 确定推广方的方法、装置、设备、存储介质和程序产品

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108829771A (zh) * 2018-05-29 2018-11-16 广州虎牙信息科技有限公司 主播推荐方法、装置、计算机存储介质及服务器
CN112672188B (zh) * 2019-10-15 2023-05-26 阿里巴巴集团控股有限公司 一种主播推荐方法、设备及存储介质
CN111241388A (zh) * 2019-12-13 2020-06-05 北京三快在线科技有限公司 多策略召回方法、装置、电子设备及可读存储介质
CN111988636B (zh) * 2020-08-21 2022-03-01 广州方硅信息技术有限公司 主播推荐方法和装置、服务器、计算机可读存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140330647A1 (en) * 2013-05-03 2014-11-06 International Business Machines Corporation Application and service selection for optimized promotion
CN111178970A (zh) * 2019-12-30 2020-05-19 微梦创科网络科技（中国）有限公司 广告投放的方法及装置、电子设备和计算机可读存储介质
CN112100558A (zh) * 2020-09-02 2020-12-18 北京字节跳动网络技术有限公司 用于对象推荐的方法、装置、设备和存储介质
CN112118489A (zh) * 2020-09-07 2020-12-22 北京字节跳动网络技术有限公司 群组管理的方法、装置、设备和介质
CN113205362A (zh) * 2021-04-30 2021-08-03 北京有竹居网络技术有限公司 确定推广方的方法、装置、设备、存储介质和程序产品

Also Published As

Publication number Publication date
CN113205362A (zh) 2021-08-03

Similar Documents

Publication Publication Date Title
WO2022228075A1 (zh) 确定推广方的方法、装置、设备、存储介质和程序产品
JP6732001B2 (ja) クロスプラットフォーム広告ターゲティング
TWI409712B (zh) 以消費者推薦之社交網路行銷方法與裝置
US11544737B2 (en) Attention application user classification privacy
US10380501B2 (en) Lookalike evaluation
US8768774B2 (en) Advertisements with multiple targeting criteria bids
US20130226711A1 (en) Monetizing images in publishing networks
US11853983B1 (en) Video revenue sharing program
WO2022228074A1 (zh) 确定对象的方法、装置、设备、存储介质和程序产品
US20210192460A1 (en) Using content-based embedding activity features for content item recommendations
CN111095330B (zh) 用于预测在线用户交互的机器学习方法和系统
US20210350419A1 (en) Method and system for facilitating a provisioning of advertisements to customers
US11367110B2 (en) Personalized mobile application re-engagement
WO2020172443A1 (en) Method and apparatus for providing web advertisements to users
TW201335874A (zh) 用於搜尋結果之即時廣告
WO2023091733A1 (en) Techniques for transactions associated with non-fungible tokens (nft) using artificial intelligence (ai) and machine learning (ml)
JP7463286B2 (ja) 高速かつ安全なコンテンツ提供のためのシステム
US11019379B2 (en) Stage-based content item selection and transmission
WO2023091736A1 (en) Techniques for transactions associated with non-fungible tokens (nft) using artificial intelligence (ai) and machine learning (ml)
US20200410588A1 (en) Determining value of source of data
US20130339154A1 (en) System and method for assembling personalized offers
US20150213467A1 (en) Metadata rich tag for survey re-targeting
US11341541B2 (en) Pruning field weights for content selection
US11182390B2 (en) Content item selection
JP2023138029A (ja) 情報処理装置、情報処理方法、及び情報処理プログラム

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22794546

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18555100

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22794546

Country of ref document: EP

Kind code of ref document: A1