CN114547434B - Object recommendation method and device, electronic equipment and storage medium - Google Patents
Object recommendation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN114547434B CN114547434B CN202111652817.3A CN202111652817A CN114547434B CN 114547434 B CN114547434 B CN 114547434B CN 202111652817 A CN202111652817 A CN 202111652817A CN 114547434 B CN114547434 B CN 114547434B
- Authority
- CN
- China
- Prior art keywords
- objects
- recall
- information
- target
- sequence
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 230000004927 fusion Effects 0.000 claims abstract description 68
- 238000012545 processing Methods 0.000 claims description 33
- 230000008707 rearrangement Effects 0.000 claims description 14
- 238000012216 screening Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 7
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 238000003780 insertion Methods 0.000 claims description 3
- 230000037431 insertion Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 12
- 230000003993 interaction Effects 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 7
- 238000013473 artificial intelligence Methods 0.000 description 5
- 230000006399 behavior Effects 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 230000001186 cumulative effect Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000001960 triggered effect Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 239000000843 powder Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02W—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO WASTEWATER TREATMENT OR WASTE MANAGEMENT
- Y02W90/00—Enabling technologies or technologies with a potential or indirect contribution to greenhouse gas [GHG] emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The disclosure relates to an object recommendation method, an object recommendation device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a plurality of first recall objects and a plurality of second recall objects matched with a target account; sorting the plurality of first recall objects and the plurality of second recall objects to obtain an initial object sequence; based on the matching degree information of each of the plurality of first recall objects and the target account and the index fusion information of each of the plurality of first recall objects under the multi-service index, adjusting the arrangement positions of the plurality of first recall objects in the initial object sequence to obtain a target object sequence; and recommending the target object sequence to the target account. According to the technical scheme provided by the disclosure, the display quantity of the first recall object is ensured, and the correlation between the target object sequence and the user can be ensured.
Description
Technical Field
The disclosure relates to the technical field of internet application, and in particular relates to an object recommendation method, an object recommendation device, electronic equipment and a storage medium.
Background
In order to realize the recommendation of the preset object, the current recommendation system selects a strong interpolation strategy in the related technology, and directly and strong inserts the preset object into an object sequence recommended to the user, so that the correlation between the object recommended to the user and the user is low, and the recommendation effect is poor.
Disclosure of Invention
The disclosure provides an object recommendation method, an object recommendation device, electronic equipment and a storage medium, and the technical scheme of the disclosure is as follows:
according to a first aspect of an embodiment of the present disclosure, there is provided an object recommendation method, including:
acquiring a plurality of first recall objects and a plurality of second recall objects matched with a target account, wherein the plurality of first recall objects are screened from a plurality of candidate objects;
sorting the plurality of first recall objects and the plurality of second recall objects to obtain an initial object sequence;
Based on the matching degree information of the plurality of first recall objects and the target account and the index fusion information of the plurality of first recall objects under the multi-service index, adjusting the arrangement positions of the plurality of first recall objects in the initial object sequence to obtain a target object sequence; the arrangement positions of the plurality of second recall objects in the initial object sequence are the same as the arrangement positions of the plurality of second recall objects in the target object sequence;
recommending the target object sequence to the target account.
In a possible implementation manner, the adjusting the arrangement positions of the plurality of first recall objects in the initial object sequence to obtain a target object sequence based on the matching degree information of the plurality of first recall objects and the target account and the index fusion information of the plurality of first recall objects under the multi-service index respectively includes:
Grouping the plurality of first recall objects to obtain at least two object groups;
Determining intra-group ordering information of the first recall object in each object group based on the matching degree information and the index fusion information;
and based on the intra-group ordering information, adjusting the arrangement positions of the plurality of first recall objects in the initial object sequence to obtain the target object sequence.
In a possible implementation manner, the adjusting, based on the intra-group ordering information, arrangement positions of the plurality of first recall objects in the initial object sequence to obtain a target object sequence includes:
determining a target arrangement position of a first recall object in the initial object sequence in each object group;
and inserting the first recall object in each object group into the target arrangement position based on the intra-group ordering information of each object group to obtain the target object sequence.
In one possible implementation manner, the grouping processing is performed on the plurality of first recall objects to obtain at least two object groups, including:
acquiring a plurality of preset object category information;
And based on the preset object category information, grouping the plurality of first recall objects to obtain at least two object groups, wherein each object group corresponds to one preset object category information.
In one possible implementation manner, the multi-service indicator corresponds to a target service type, and the method further includes:
acquiring the plurality of candidate objects;
Determining a plurality of index information of each candidate object under the multi-service index;
Determining index fusion information of each candidate object according to the plurality of index information of each candidate object;
And carrying out grading processing on the plurality of candidate objects based on the index fusion information to obtain a plurality of object recall sets, wherein the grade information of the candidate objects in each object recall set is the same.
In one possible implementation manner, the obtaining a plurality of first recall objects includes:
Determining at least one target object recall set from the plurality of object recall sets;
the plurality of first recall objects are obtained from the at least one target object recall set.
In one possible implementation, the obtaining the plurality of second recall objects that match the target account includes:
Acquiring a plurality of objects, associated information of the plurality of objects and account attribute information of the target account;
inputting the association information and the account attribute information into a matching prediction model, and performing matching processing to obtain matching prediction information;
and screening the plurality of second recall objects from the plurality of objects based on the matching prediction information.
In one possible implementation manner, the sorting processing is performed on the first recall objects and the second recall objects to obtain an initial object sequence, including:
Acquiring index fusion information of each of the plurality of first recall objects and matching prediction information of the plurality of second recall objects;
And based on the index fusion information and the matching prediction information, sequencing the plurality of first recall objects and the plurality of second recall objects to obtain the initial object sequence.
In one possible implementation, the method further includes:
acquiring a resource consumption threshold value, unit resource consumption information corresponding to the plurality of candidate objects and accumulated recommended times;
determining current resource consumption information according to the unit resource consumption information and the accumulated recommended times;
the obtaining a plurality of first recall objects and a plurality of second recall objects matched with the target account includes:
And under the condition that the current resource consumption information is smaller than the resource consumption threshold, acquiring the plurality of first recall objects and a plurality of second recall objects matched with the target account.
According to a second aspect of the embodiments of the present disclosure, there is provided an object recommendation apparatus, including:
The first acquisition module is configured to execute acquisition of a plurality of first recall objects and a plurality of second recall objects matched with the target account, wherein the plurality of first recall objects are screened from a plurality of candidate objects;
The initial ordering module is configured to perform ordering processing on the plurality of first recall objects and the plurality of second recall objects to obtain an initial object sequence;
The rearrangement module is configured to execute index fusion information based on the matching degree information of the plurality of first recall objects and the target account and the index fusion information of the plurality of first recall objects under the multi-service index, and adjust the arrangement positions of the plurality of first recall objects in the initial object sequence to obtain a target object sequence; the arrangement positions of the plurality of second recall objects in the initial object sequence are the same as the arrangement positions of the plurality of second recall objects in the target object sequence;
and the recommending module is configured to execute recommending the target object sequence to the target account.
In one possible implementation, the rearrangement module includes:
A grouping unit configured to perform grouping processing on the plurality of first recall objects to obtain at least two object groups;
An intra-group ranking unit configured to perform determining intra-group ranking information of the first recall object in each object group based on the matching degree information and the index fusion information;
And the rearrangement unit is configured to execute adjustment of arrangement positions of the plurality of first recall objects in the initial object sequence based on the intra-group ordering information to obtain the target object sequence.
In one possible implementation, the rearrangement unit includes:
A target position determination subunit configured to perform determining a target arrangement position of the first recall object in the initial sequence of objects in each object group;
And the interpolation subunit is configured to perform the insertion of the first recall object in each object group into the target arrangement position based on the intra-group ordering information of each object group to obtain the target object sequence.
In one possible implementation, the grouping unit includes:
an object category acquisition subunit configured to perform acquisition of a plurality of preset object category information;
And the grouping subunit is configured to perform grouping processing on the plurality of first recall objects based on the preset object category information to obtain at least two object groups, and each object group corresponds to one preset object category information.
In one possible implementation manner, the multi-service indicator corresponds to a target service type, and the apparatus further includes:
A second acquisition module configured to perform acquisition of the plurality of candidate objects;
an index information determining module configured to perform determining a plurality of index information of each candidate object under the multi-service index;
an index fusion information determining module configured to perform determining index fusion information of each candidate object according to a plurality of index information of each candidate object;
And the object recall set acquisition module is configured to execute grading processing on the plurality of candidate objects based on the index fusion information to obtain a plurality of object recall sets, wherein the grade information of the candidate objects in each object recall set is the same.
In one possible implementation manner, the first obtaining module includes:
a first acquisition unit configured to perform determining at least one target object recall set from the plurality of object recall sets;
a second acquisition unit configured to perform acquisition of the plurality of first recall objects from the at least one target object recall set.
In one possible implementation manner, the first obtaining module includes:
A third acquisition unit configured to perform acquisition of a plurality of objects, association information of the plurality of objects, and account attribute information of the target account;
The matching unit is configured to input the association information and the account attribute information into a matching prediction model for matching processing to obtain matching prediction information;
And a screening unit configured to perform screening of the plurality of second recall objects from the plurality of objects based on the matching prediction information.
In one possible implementation, the initial ordering module includes:
a fourth acquisition unit configured to perform acquisition of index fusion information of each of the plurality of first recall objects and matching prediction information of the plurality of second recall objects;
and the initial sorting unit is configured to perform sorting processing on the plurality of first recall objects and the plurality of second recall objects based on the index fusion information and the matching prediction information, so as to obtain the initial object sequence.
In one possible implementation, the method further includes:
The third acquisition module is configured to acquire a resource consumption threshold, unit resource consumption information corresponding to the plurality of candidate objects and accumulated recommended times;
A resource consumption determination module configured to perform determination of current resource consumption information according to the unit resource consumption information and the accumulated recommended number of times;
The first obtaining module is further configured to obtain the plurality of first recall objects and a plurality of second recall objects matched with the target account if the current resource consumption information is less than the resource consumption threshold.
According to a third aspect of embodiments of the present disclosure, there is provided an electronic device, comprising: a processor; a memory for storing the processor-executable instructions; wherein the processor is configured to execute the instructions to implement the method of any of the first aspects above.
According to a fourth aspect of the disclosed embodiments, there is provided a computer readable storage medium, which when executed by a processor of an electronic device, causes the electronic device to perform the method of any of the first aspects of the disclosed embodiments.
According to a fifth aspect of embodiments of the present disclosure, there is provided a computer program product comprising computer instructions which, when executed by a processor, cause the computer to perform the method of any one of the first aspects of embodiments of the present disclosure.
The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:
the first recall objects and the second recall objects matched with the target account number are taken as recall objects, and subsequent sequencing is carried out to determine a target object sequence, so that the probability that the first recall objects are recommended is ensured; and the arrangement positions of the plurality of first recall objects in the initial object sequence are adjusted based on the matching degree information and the index fusion information of the plurality of first recall objects and the target account, so that the target object sequence can effectively represent the correlation between the first recall objects and the target account, the recommended probability of the first recall objects is ensured, and the object recommendation accuracy is also ensured.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure and do not constitute an undue limitation on the disclosure.
FIG. 1 is a schematic diagram of an application environment, shown in accordance with an exemplary embodiment.
FIG. 2 is a flowchart illustrating an object recommendation method according to an exemplary embodiment.
FIG. 3 is a schematic diagram illustrating a recommendation flow, according to an example embodiment.
FIG. 4 is a flowchart illustrating a method for adjusting the arrangement positions of a plurality of first recall objects in an initial object sequence to obtain a target object sequence based on matching degree information of each of the plurality of first recall objects with a target account and index fusion information of each of the plurality of first recall objects under a multi-business index, according to an exemplary embodiment.
Fig. 5 is a schematic diagram illustrating an intra-group rearrangement, according to an exemplary embodiment.
FIG. 6 is a flowchart illustrating a method of obtaining a plurality of object recall sets, according to an example embodiment.
FIG. 7 is a schematic diagram illustrating ranking of a candidate object according to an example embodiment.
FIG. 8 is a flow diagram illustrating an object recommendation according to an example embodiment.
FIG. 9 is a block diagram of an object recommendation device, according to an example embodiment.
FIG. 10 is a block diagram of an electronic device for object recommendation, according to an example embodiment.
Detailed Description
In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.
It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the foregoing figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.
Artificial intelligence (ARTIFICIAL INTELLIGENCE, AI) is the theory, method, technique, and application system that simulates, extends, and extends human intelligence using a digital computer or a machine controlled by a digital computer, perceives the environment, obtains knowledge, and uses the knowledge to obtain optimal results. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.
In recent years, with research and progress of artificial intelligence technology, the artificial intelligence technology is widely applied in a plurality of fields, and the scheme provided by the embodiment of the application relates to technologies such as machine learning/deep learning, and the like, and is specifically described by the following embodiments:
referring to fig. 1, fig. 1 is a schematic diagram illustrating an application environment according to an exemplary embodiment, and as shown in fig. 1, the application environment may include a server 01 and a terminal 02.
In an alternative embodiment, server 01 may be used for processing of object recommendations. Specifically, the server 01 may be an independent physical server, or may be a server cluster or a distributed system formed by a plurality of physical servers, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs (Content Delivery Network, content delivery networks), and basic cloud computing services such as big data and artificial intelligence platforms.
In an alternative embodiment, the terminal 02 may send a recommendation request and present the target object in the target object sequence. Specifically, the terminal 02 may include, but is not limited to, a smart phone, a desktop computer, a tablet computer, a notebook computer, a smart speaker, a digital assistant, an Augmented Reality (AR)/Virtual Reality (VR) device, a smart wearable device, and other types of electronic devices. Alternatively, the operating system running on the electronic device may include, but is not limited to, an android system, an IOS system, linux, windows, and the like.
In addition, it should be noted that, fig. 1 is only one application environment of the object recommendation method provided in the present disclosure.
In the embodiment of the present disclosure, the server 01 and the terminal 02 may be directly or indirectly connected through a wired or wireless communication method, which is not limited herein.
It should be noted that, a possible sequence of steps is shown in the following figures, and is not limited to the strict order of the sequence. Some steps may be performed in parallel without mutual dependency. Account information (including but not limited to device information, attribute information, behavior information, etc. corresponding to an account) and data (including but not limited to data for presentation, training, etc.) related to the present disclosure are both information and data authorized by a user or sufficiently authorized by each party.
FIG. 2 is a flowchart illustrating an object recommendation method according to an exemplary embodiment. As shown in fig. 2, the following steps may be included.
In step S201, a plurality of first recall objects and a plurality of second recall objects matched with the target account are acquired, the plurality of first recall objects being screened from a plurality of candidate objects. The plurality of candidate objects are preset recommended objects, where the preset recommended objects may refer to preset objects to be recommended. It can be appreciated that the preset recommended object is applicable to all accounts, can be directly used as a recall object, and does not need to be matched with the accounts in a personalized way. As an example, the preset recommended object may be a preset object that needs to exhibit compensation.
In the embodiment of the present specification, the target account may be any user account in the object recommendation platform. The object recommendation method of the present disclosure may be triggered by the object recommendation platform or may be triggered by the terminal side, which is not limited by the present disclosure. Taking the triggering of the terminal side as an example, the recommendation request can be triggered on the terminal side, so that the recommendation request can be sent to the object recommendation platform. Accordingly, the object recommendation platform receives the recommendation request, and can extract the target account from the recommendation request, namely, extract the account triggering the recommendation request, so that a subsequent recommendation process can be performed to realize recommendation to the target account, as shown in fig. 3. The recommendation procedure may include recall, sort, reorder, and send the target object sequence to the terminal, as will be described in detail below.
In practical applications, a plurality of second recall objects matched with the target account may be acquired, for example, portrait information (such as preference information, gender information, age information, etc.) of the target account and a plurality of objects in the object recommendation platform may be acquired, so that matching prediction information of the target account and each object in the plurality of objects may be determined based on the portrait information. Further, a plurality of second recall objects may be screened from a plurality of objects in the object recommendation platform based on the matching prediction information. The portrait information can be obtained based on account attribute information and/or historical behavior information of the target account, and the historical behavior information can comprise interaction behaviors of the target account in the object recommendation platform, such as praise, attention, forwarding and the like; the account attribute information may include account identification information of the target account, etc., none of which are limited by the present disclosure.
In one possible implementation, obtaining a plurality of second recall objects that match the target account may include: acquiring a plurality of objects, association information of the plurality of objects and account attribute information of a target account; the associated information and account attribute information can be input into a matching prediction model for matching processing to obtain matching prediction information; and screening a plurality of second recall objects from the plurality of objects based on the matching prediction information. And acquiring a plurality of second recall objects based on the matching prediction model, so that the acquisition efficiency and accuracy of the second recall objects can be improved. The association information of the plurality of objects may include category information (sports class, game class, etc.), interaction information, etc. of each of the plurality of objects.
In one possible implementation manner, the plurality of first recall objects may be screened from a plurality of candidate objects based on index fusion information under a multi-service index. For example, an object of an author (object producer) having a low attention and/or an object whose release time period is less than a preset time period threshold may be acquired as a plurality of candidate objects. Further, index fusion information of the plurality of candidate objects under the multi-service index can be determined, so that the plurality of candidate objects can be ranked based on the index fusion information, and a ranking result is obtained; and a preset number of candidate objects can be screened from the plurality of candidate objects based on the sorting result to serve as a plurality of first recall objects. The multi-business index may refer to a plurality of business indexes, which may represent the interaction degree of the object, and may include, for example, click rate, praise rate, forwarding rate, etc., which is not limited in this disclosure. Based on the above, a plurality of pieces of business index information corresponding to each candidate object can be counted, so that the plurality of pieces of business index information can be weighted to obtain index fusion information of each candidate object.
The objects in the object recommendation platform may be any objects that can be used for recommendation, and may include advertisements, multimedia, images, text, objects, and the like, for example, which are not limited in this disclosure.
Alternatively, a PID (proportional-integral-DERIVATIVE CONTROL) measure may be used in the initial sorting in step S203 below to control the number of recall objects participating in the initial sorting to a predetermined number. Based on this, at the time of screening of the plurality of first recall objects, the number of the plurality of first recall objects may be dynamically adjusted to satisfy the predetermined number. The number of the candidate objects can be guaranteed, and the display number of the candidate objects can be further guaranteed.
In step S203, the plurality of first recall objects and the plurality of second recall objects are sorted to obtain an initial object sequence.
In the embodiment of the present disclosure, the first recall objects and the second recall objects may be sequenced to obtain an initial object sequence. The initial sequence of objects may be n objects arranged in a sequence as shown to the left in fig. 5. The objects may be represented in the initial sequence of objects using object identification information, which is not limited by this disclosure. The n objects may include a plurality of first recall objects and a plurality of second recall objects, each object having a corresponding arrangement position in the initial sequence of objects.
Optionally, index fusion information of each of the plurality of first recall objects and matching prediction information of the plurality of second recall objects may be obtained; and the first recall objects and the second recall objects can be sequenced based on the index fusion information and the matching prediction information, so that an initial object sequence is obtained. The index fusion information and the matching prediction information may be quantized information, for example, the index fusion information may be a numerical value representing the index degree, and the matching prediction information may be a matching probability. Based on this, the first recall objects and the second recall objects may be directly sorted according to the index fusion information and the matching prediction information according to a preset sequence, where the preset sequence may be from high to low or from low to high, which is not limited in the disclosure. The initial sorting process can be compatible with index fusion information of the first recall object and matching prediction information of the second recall object, so that the initial object sequence can effectively reflect the priority orders of the recall objects corresponding to different recall modes.
In step S205, based on the matching degree information of each of the plurality of first recall objects and the target account and the index fusion information of each of the plurality of first recall objects under the multi-service index, the arrangement positions of the plurality of first recall objects in the initial object sequence are adjusted to obtain the target object sequence.
In the embodiment of the specification, the matching degree information of each of the plurality of first recall objects and the target account can be determined, and the higher the matching degree information is, the more the first recall object is matched with the target account. In one example, portrait information for the target account may be obtained such that, based on the portrait information, degree of matching information, such as a probability of matching, for each of the plurality of first recall objects with the target account may be determined. In another example, account attribute information of the target account, such as gender information, age information, historical interaction behavior information, etc., of the target account may be obtained. And the associated information of each first recall object, such as category information of the first recall object, historical interaction information of the first recall object and the like, can be obtained. Therefore, the associated information and the account attribute information can be input into a matching degree model to be matched, and the matching degree information is obtained. The matching degree model may be obtained by training a preset machine learning model in advance based on training sample data. The training sample data comprises a plurality of sample objects and corresponding label information, wherein the label information is a matching probability; the preset machine learning model may include a deep learning network, etc., which is not limited by the present disclosure.
In step S207, the target object sequence is recommended to the target account.
In the embodiment of the specification, the target object sequence may be recommended to the target account. For example, the target object sequence may be sent to a terminal corresponding to the target account.
The first recall objects and the second recall objects matched with the target account number are taken as recall objects, and subsequent sequencing is carried out to determine a target object sequence, so that the probability that the first recall objects are recommended is ensured; and the arrangement positions of the plurality of first recall objects in the initial object sequence are adjusted based on the matching degree information and the index fusion information of the plurality of first recall objects and the target account, so that the target object sequence can effectively represent the correlation between the first recall objects and the target account, the recommended probability of the first recall objects is ensured, and the object recommendation accuracy is also ensured.
FIG. 4 is a flowchart illustrating a method for adjusting the arrangement positions of a plurality of first recall objects in an initial object sequence to obtain a target object sequence based on matching degree information of each of the plurality of first recall objects with a target account and index fusion information of each of the plurality of first recall objects under a multi-business index, according to an exemplary embodiment. As shown in fig. 4, in one possible implementation, the step S205 may include:
in step S401, a plurality of first recall objects are subjected to grouping processing, resulting in at least two object groups.
In one possible implementation manner, the plurality of first recall objects may be grouped based on the index fusion information of each of the plurality of first recall objects to obtain at least two object groups. For example, the plurality of first recall objects may be sorted based on the index fusion information, the first recall objects with the sorting greater than the sorting threshold may be divided into one object group, and the first recall objects with the sorting lower than or equal to the sorting threshold may be divided into one object group, so that two object groups may be obtained. Or multiple sorting thresholds may be set so that multiple object groups are available.
In another possible implementation manner, this step S401 may be further implemented based on the following steps:
acquiring a plurality of preset object category information;
And based on the preset object category information, grouping the plurality of first recall objects to obtain at least two object groups, wherein each object group corresponds to one preset object category information.
In this embodiment of the present disclosure, the preset object type information may refer to a content type of an object, and taking the object as a multimedia as an example, the preset object type information may include a game, a movie, a make-up, and the like, which is not limited in this disclosure. As one example, the target object class information of the plurality of first recall objects may be determined based on preset object class information, which may be one of the preset object class information. Further, a plurality of first recall objects can be clustered according to the target object category information to obtain at least two object groups, and the first recall objects with the same target object category information are divided into the same object group, so that the target object category information of the first recall objects in the same object group is the same. The grouping processing is performed based on the preset object category information, so that the intra-group ordering information is performed based on the same preset object category information, the ordering deviation among different types of objects is effectively avoided, the accuracy of the intra-group ordering information is improved, the accuracy of the target object sequence can be improved, the arrangement position of the second recall object in the initial object sequence cannot be influenced, and the ordering among different object groups cannot be influenced.
In step S403, the intra-group ranking information of the first recall object in each object group is determined based on the matching degree information and the index fusion information.
In the embodiment of the present disclosure, the intra-group ranking information of the first recall object in each object group may be determined based on the matching degree information and the index fusion information. For example, the matching degree information of one first recall object H is represented by f (xtr), the index fusion information of the one first recall object H is represented by manual_score, and the intra-group recommendation parameter information global_score of the one first recall object H can be determined using the following formula (1) as follows:
global_score=f(xtr)*g(manual_score) (1)
Where g () may be a transform function, and g () may be in various forms, such as g (x) =x, x t, logx, etc., which is not limited by the present disclosure.
Further, the first recall objects in each object group may be ranked based on the intra-group recommendation parameter information, so that intra-group ranking information may be obtained, where the intra-group ranking information may include a correspondence between an intra-group ranking number and the intra-group first recall objects.
In step S405, based on the intra-group ordering information, the arrangement positions of the plurality of first recall objects in the initial object sequence are adjusted, so as to obtain the target object sequence.
In the embodiment of the present disclosure, as shown in fig. 5, in one possible implementation manner, the step S405 may include:
Determining a target arrangement position of a first recall object in the initial object sequence in each object group;
the first recall object in each object group is inserted into the target permutation location based on the intra-group ordering information for each object group.
As shown in fig. 5, the left side is an initial object sequence of objects 1 to n, where the objects 1 to n may be object identification information of the objects. The plurality of first recall objects may include object 1, object 2, and objects 4 to 7, and the objects other than object 1, object 2, and objects 4 to 7 in object 1 to object n are the plurality of second recall objects. The plurality of first recall objects may be divided into two object groups, an object group corresponding to a movie class, and an object group corresponding to a game class. In the initial object sequence, the target arrangement positions of the video class objects (object 1, object 4 and object 7) are 1,4 and 7; the target arrangement positions of the game class objects (object 2, object 5, and object 6) are 2,5, and 6. The rearrangement of the video in the group can be realized based on the ordering information in the group, and then the target arrangement position corresponding to the object group is inserted back, so that the target object sequence can be obtained, as shown on the right side of fig. 5. Wherein, the video class object and the game class object are respectively changed from the target arrangement position in the initial object sequence to the arrangement position in the target object sequence. As shown in fig. 5, the arrangement position of the movie class object changes: 1,4,7→1,7,4; the arrangement position of the game class objects changes: 2,5,6→ 6,2,5. The first recall objects in each object group are inserted into the target arrangement positions through the intra-group ordering information, so that the target arrangement positions corresponding to each object group are not changed, the order adjustment of the first recall objects in the group can be realized, and the ordering of the first recall objects in the group is realized under the condition that the order of the second recall objects is not influenced.
The method comprises the steps of carrying out grouping processing on a plurality of first recall objects to obtain at least two object groups, so that sequencing adjustment of the first recall objects in the object groups is realized, local adjustment in the groups is realized, the object groups are not mutually influenced, the second recall objects are not influenced, the correlation between a target object sequence and a target account can be ensured, the influence of a preset recommended first recall object on the second recall object is further avoided, and the recommendation effect is ensured.
FIG. 6 is a flowchart illustrating a method of obtaining a plurality of object recall sets, according to an example embodiment. As shown in fig. 6, in one possible implementation, the method may further include:
In step S601, a plurality of candidates are acquired.
In the embodiment of the present disclosure, a plurality of preset recommended objects corresponding to a target service type may be obtained as a plurality of candidate objects, and the plurality of candidate objects may be formed into an object candidate set. Different business requirements correspond to different business targets, and based on the business targets, different business types can be obtained by dividing the business, and the target business type can be one of the different business types. As one example, the different traffic types may include: the present disclosure is not limited in terms of the type of traffic targeted to maximize the powder expansion efficiency, the type of traffic targeted to maximize the display volume, the type of traffic targeted to operation and product statistics, etc. For example, in the case that the target service type is a service type counted by operation and product, a plurality of preset recommended objects given by the operation and the product may be directly acquired, where the operation and the product may be obtained based on interaction statistics of a plurality of objects in the object recommendation platform, or the operation and the product may be given based on actual requirements, which is not limited in this disclosure. Or in the case that the target service type is the service type targeted to maximize the exposure, the object related to the current trending topic may be acquired as a plurality of preset recommended objects. Or under the condition that the target service type is the service type with the maximum powder expansion efficiency as the target, the object issued by the cold start account and/or the cold start object can be obtained and used as a plurality of preset recommended objects, so that the object with higher service potential can be mined for recommendation through the screening of the subsequent index fusion information, and the interactive effective growth of the preset recommended objects is realized.
Alternatively, the object candidate set may be acquired periodically, such as once per hour, to ensure timeliness. The obtained plurality of preset recommended objects may be used as a plurality of candidate objects in the object candidate set, and the plurality of candidate objects may be put into a candidate object pool, as shown in fig. 7.
In step S603, a plurality of index information of each candidate object under the multi-service index is determined.
In practical application, index information of each candidate object under each service index can be determined. For example, a statistical manner may be used to determine index information of each candidate object under each service index, for example, the service index is a click rate index, and the click rate of each candidate object in a preset period of time may be counted as the service index information. Or the neural network model can be used for predicting the index information of each service, such as a click rate prediction model, the object attribute information and the historical interaction operation information of each candidate object can be input into the click rate prediction model to perform click rate prediction processing, so that the click rate prediction information is obtained, and the click rate prediction information can be used as index information under the click rate index. The click rate prediction model may be obtained by training a preset neural network model in advance based on a sample object data set, where the sample object data set may include sample object attribute information, sample interaction operation information, and corresponding sample labels of a plurality of sample objects, and the sample labels may include clicking and non-clicking.
As one example, the business metrics may include an object publisher metric, an object metric, and a feedback metric. Wherein the object publisher index may include at least one of: number of publishers, number of publishers' works, number of last N days of works, average attention rate of publishers, total display amount of publishers, etc. The object index may include at least one of: average click rate, average praise rate, average attention number, average forwarding number, etc. The feedback indicator may comprise a content quality indicator of the object.
In practical application, the multi-service index (multiple service indexes) can be dynamically determined based on the target service type, so that the flexibility of the multi-service index can be improved. In the case where the target traffic type is a traffic type targeting the maximum exposure amount, in order to promote the exposure amount (exposure amount), a traffic index of an object growth rate may be set, and traffic index information at the object growth rate may be determined in the following manner. Multiple candidate objects can be classified into buckets according to two dimensions of [ release time, release duration ], such as [ release at 3 pm for 5 hours ], wherein the bucket intervals can be customized and divided. For all objects in a certain sub-bucket, the objects can be equally divided into preset numbers of bits, such as 100, from large to small based on the display amount of the objects. The quantiles show_100, show_99, show_1 may be calculated. If the display amount show_real of a certain candidate object in the sub-bucket is satisfied between [ show_ { k+1}, show_k), the index information of the certain candidate object at the object growth rate can be determined to be k.
In step S605, index fusion information of each candidate is determined from a plurality of index information of each candidate.
In practical application, after obtaining the multiple index information of each candidate object, the index fusion information of each candidate object may be determined based on the following disclosure (2):
Wherein Score i may be index fusion information of candidate i; w j may be the weight of different traffic indexes, and rank i,j may be the arrangement position of candidate object i in traffic index j, such as the sequence number; as one example, f (x) =x, which the present disclosure does not limit. The ranking number is inversely related to the index information, i.e. the ranking number obtained by ranking the index information from high to low.
In step S607, the plurality of candidate objects are ranked based on the index fusion information, so as to obtain a plurality of object recall sets, where the ranking information of the candidate objects in each object recall set is the same.
In the embodiment of the present disclosure, the plurality of candidate objects may be ranked based on the index fusion information, for example, the plurality of candidate objects may be ranked based on the index fusion information, to obtain a ranking result. Therefore, the multiple candidate objects can be classified based on the sorting result, and multiple object recall sets are obtained. Taking three levels of information as an example, the candidate objects ranked in the previous N1 can be taken as an object recall set 1 (gear 1) according to the sequence of the plurality of candidate objects in the ranking result; taking the candidate objects ordered between N1 and N2 as an object recall set 2 (gear 2); and candidate objects ordered after N2 may be referred to as object recall set 3 (gear 3).
Through the setting of a plurality of service indexes, a plurality of candidate objects can be screened based on a plurality of service index information, and recommendation compensation of a horse racing preferred mode is realized, so that the accuracy of object recommendation can be ensured.
Accordingly, the step S201 may include: determining at least one target object recall set from a plurality of object recall sets; and a plurality of first recall objects may be obtained from at least one target object recall set. Therefore, a plurality of first recall objects with better index fusion information can be obtained according to the grade information, the flexibility of the plurality of first recall objects is improved, and the display quantity of the first recall objects can be effectively ensured; in addition, since the plurality of object recall sets are prefabricated, the acquisition efficiency of the plurality of first recall objects can be improved, and therefore the recommendation efficiency can be improved.
Optionally, in consideration of the ordering effect of the preset recommended objects, to avoid excessive recommended intervention, control may be performed, for example, by the following steps, and the method may further include:
Acquiring unit resource consumption information corresponding to a preset recommended object, accumulated recommended times of the preset recommended object and a resource consumption threshold; the cumulative recommendation number may be obtained from a counter, which may be incremented by 1 and may be initialized to 0 when performing the object recommendation operation of the present disclosure on a recommendation request. Alternatively, if the first recall object is not acquired, the counter may not be incremented by 1, i.e., not counted.
Determining current resource consumption information according to the unit resource consumption information and the accumulated recommended times; for example, the product of the unit resource consumption information and the cumulative recommended number of times may be determined as the current resource consumption information.
Accordingly, the step S201 may include: and under the condition that the current resource consumption information is smaller than a resource consumption threshold value, acquiring a plurality of first recall objects and a plurality of second recall objects matched with the target account.
Optionally, in a case that the current resource consumption information is greater than or equal to the resource consumption threshold, only the plurality of second recall objects may be acquired, so that the plurality of second recall objects may be ordered based on the matching prediction information, and a target object sequence is obtained. That is, in the case where the current resource consumption information is greater than or equal to the resource consumption threshold, the recommendation operation of the preset recommendation object is not performed.
The unit resource consumption information may be determined by a prior experimental estimation method or a post-hoc algorithm. The business index is taken as the click rate ctr to describe, the one-click value is p_per_click, and the value definition modes of other business indexes are similar in the case of a plurality of business indexes.
A. pre-experimental estimation method:
The average click rate of the objects in the target object sequence is measured to be ctr_ nature and the average click rate of the first recall objects in the target object sequence is measured to be ctr_manual through small flow experimental measurement; based on this, the unit resource consumption information may be (ctr_ nature-ctr_manual) p_per_click. By such a real measurement, the unit resource consumption information is made more accurate.
B. Post hoc algorithm:
Every time an object in the preset recommended objects is recommended, a first recall object replaces a second recall object, and the first recall object causes the sequence of the second recall object to move downwards. Thus, the unit resource consumption information can be determined by the predicted click rate pctr instead of the actual ctr. For example, the unit resource consumption information may be (pctr _ nature-pctr _manual) p_per_click, where pxtr _manual may be the average predicted click rate of the first recall object in the target object sequence and pxtr _ nature is the average predicted click rate of the second recall object that was replaced the time (current recommendation request). The click rate can be predicted by a machine learning model, and the accuracy is higher.
In the case of a plurality of business indexes, the sum or weighted sum of the unit resource consumption information of each business index may be used as the target unit resource consumption information, so that the current resource consumption information may be determined based on the target unit resource consumption information and the cumulative recommended number of times. For example, the product of the target unit resource consumption information and the cumulative recommended number of times may be used as the current resource consumption information.
FIG. 8 is a flow diagram illustrating an object recommendation according to an example embodiment. As shown in fig. 8, in one possible implementation, it may include:
in step S801, at least one target object recall set is determined from a plurality of object recall sets;
In step S803, a plurality of first recall objects are obtained from at least one recall set of target objects;
in step S805, a plurality of second recall objects matched with the target account and a plurality of preset object class information are acquired;
In step S807, sorting the plurality of first recall objects and the plurality of second recall objects to obtain an initial object sequence;
In step S809, based on the preset object class information, grouping the plurality of first recall objects to obtain at least two object groups, each object group corresponding to one preset object class information;
In step S811, intra-group ranking information of the first recall object in each object group is determined based on the matching degree information and the index fusion information;
in step S813, the target arrangement position of the first recall object in the initial object sequence in each object group is determined;
in step S815, the first recall object in each object group is inserted into the target arrangement position corresponding to each object group based on the intra-group ordering information of each object group, to obtain a target object sequence.
In practical application, at least one target object recall set can be determined from a plurality of object recall sets recommended in advance, so that a plurality of first recall objects can be obtained from the at least one target object recall set. And may obtain natural recall objects, i.e., a plurality of second recall objects that match the target account. And the first recall objects and the second recall objects may be ranked to obtain an initial sequence of objects, as shown in fig. 5.
Further, the arrangement positions of the first recall objects in the initial object sequence may be adjusted, for example, based on preset object category information, grouping processing may be performed on a plurality of first recall objects to obtain at least two object groups, where each object group corresponds to one preset object category information; and intra-group ordering information of the first recall object in each object group can be determined based on the matching degree information and the index fusion information; so that intra-group reordering can be achieved. As an example, a target arrangement position of the first recall object in each object group in the initial object sequence may be determined, so that the first recall object in each object group may be inserted into the target arrangement position corresponding to each object group based on the intra-group ordering information of each object group, to obtain the target object sequence. For example, object group H: object 2, object 5, and object 6, which may be ordered within the group, object 6, object 2, and object 5. Therefore, the objects 2, 5 and 6 in the initial object sequence on the left side can be extracted, and then the ordered objects 6, 2 and 5 in the group are inserted back to the arrangement position occupied by the object group H in the initial sequence object. Based on the same back-inserting mode, the intra-group sequencing and back-inserting processing of all object groups can be realized, and a target object sequence is obtained.
FIG. 9 is a block diagram of an object recommendation device, according to an example embodiment. Referring to fig. 9, the apparatus may include:
A receiving module 901, configured to perform obtaining a plurality of first recall objects and a plurality of second recall objects matched with a target account, where the plurality of first recall objects are screened from a plurality of candidate objects;
an initial ordering module 903 configured to perform ordering processing on the plurality of first recall objects and the plurality of second recall objects, to obtain an initial object sequence;
A rearrangement module 905 configured to execute adjustment of arrangement positions of the plurality of first recall objects in the initial object sequence based on matching degree information of the plurality of first recall objects and the target account, and index fusion information of the plurality of first recall objects under multi-service indexes, so as to obtain a target object sequence; the arrangement positions of the plurality of second recall objects in the initial object sequence are the same as the arrangement positions of the plurality of second recall objects in the target object sequence;
A recommending module 907 configured to execute a recommendation of the target object sequence to the target account.
In one possible implementation, the rearrangement module 905 may include:
A grouping unit configured to perform grouping processing on the plurality of first recall objects to obtain at least two object groups;
An intra-group ranking unit configured to perform determining intra-group ranking information of the first recall object in each object group based on the matching degree information and the index fusion information;
And the rearrangement unit is configured to execute adjustment of arrangement positions of the plurality of first recall objects in the initial object sequence based on the intra-group ordering information to obtain the target object sequence.
In one possible implementation, the rearrangement unit includes:
A target position determination subunit configured to perform determining a target arrangement position of the first recall object in the initial sequence of objects in each object group;
And the interpolation subunit is configured to perform the insertion of the first recall object in each object group into the target arrangement position based on the intra-group ordering information of each object group to obtain the target object sequence.
In one possible implementation, the grouping unit includes:
an object category acquisition subunit configured to perform acquisition of a plurality of preset object category information;
And the grouping subunit is configured to perform grouping processing on the plurality of first recall objects based on the preset object category information to obtain at least two object groups, and each object group corresponds to one preset object category information.
In one possible implementation manner, the multi-service indicator corresponds to a target service type, and the apparatus further includes:
A second acquisition module configured to perform acquisition of the plurality of candidate objects;
an index information determining module configured to perform determining a plurality of index information of each candidate object under the multi-service index;
an index fusion information determining module configured to perform determining index fusion information of each candidate object according to a plurality of index information of each candidate object;
And the object recall set acquisition module is configured to execute grading processing on the plurality of candidate objects based on the index fusion information to obtain a plurality of object recall sets, wherein the grade information of the candidate objects in each object recall set is the same.
In one possible implementation manner, the first obtaining module 901 may include:
a first acquisition unit configured to perform determining at least one target object recall set from the plurality of object recall sets;
a second acquisition unit configured to perform acquisition of the plurality of first recall objects from the at least one target object recall set.
In one possible implementation manner, the first obtaining module 901 may include:
A third acquisition unit configured to perform acquisition of a plurality of objects, association information of the plurality of objects, and account attribute information of the target account;
The matching unit is configured to input the association information and the account attribute information into a matching prediction model for matching processing to obtain matching prediction information;
And a screening unit configured to perform screening of the plurality of second recall objects from the plurality of objects based on the matching prediction information.
In one possible implementation, the initial ordering module 903 may include:
a fourth acquisition unit configured to perform acquisition of index fusion information of each of the plurality of first recall objects and matching prediction information of the plurality of second recall objects;
and the initial sorting unit is configured to perform sorting processing on the plurality of first recall objects and the plurality of second recall objects based on the index fusion information and the matching prediction information, so as to obtain the initial object sequence.
In one possible implementation manner, the method may further include:
The third acquisition module is configured to acquire a resource consumption threshold, unit resource consumption information corresponding to the plurality of candidate objects and accumulated recommended times;
A resource consumption determination module configured to perform determination of current resource consumption information according to the unit resource consumption information and the accumulated recommended number of times;
The first obtaining module is further configured to obtain the plurality of first recall objects and a plurality of second recall objects matched with the target account if the current resource consumption information is less than the resource consumption threshold.
The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.
FIG. 10 is a block diagram illustrating an electronic device for object recommendation, which may be a server, whose internal structure may be as shown in FIG. 10, according to an exemplary embodiment. The electronic device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic device includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the electronic device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of object recommendation.
It will be appreciated by those skilled in the art that the structure shown in fig. 10 is merely a block diagram of a portion of the structure associated with the disclosed aspects and is not limiting of the electronic device to which the disclosed aspects apply, and that a particular electronic device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In an exemplary embodiment, there is also provided an electronic device including: a processor; a memory for storing the processor-executable instructions; wherein the processor is configured to execute the instructions to implement the object recommendation method as in the embodiments of the present disclosure.
In an exemplary embodiment, a computer readable storage medium is also provided, which when executed by a processor of an electronic device, causes the electronic device to perform the object recommendation method in the embodiments of the present disclosure. The computer readable storage medium may be ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.
In an exemplary embodiment, a computer program product containing instructions that, when run on a computer, cause the computer to perform the method of object recommendation in the embodiments of the present disclosure is also provided.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (SYNCHLINK) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (21)
1. An object recommendation method, comprising:
acquiring a plurality of first recall objects and a plurality of second recall objects matched with a target account, wherein the plurality of first recall objects are screened from a plurality of candidate objects;
sorting the plurality of first recall objects and the plurality of second recall objects to obtain an initial object sequence;
Based on the matching degree information of the plurality of first recall objects and the target account and the index fusion information of the plurality of first recall objects under the multi-service index, adjusting the arrangement positions of the plurality of first recall objects in the initial object sequence to obtain a target object sequence; the arrangement positions of the plurality of second recall objects in the initial object sequence are the same as the arrangement positions of the plurality of second recall objects in the target object sequence;
recommending the target object sequence to the target account.
2. The method of claim 1, wherein adjusting the arrangement positions of the plurality of first recall objects in the initial object sequence based on the matching degree information of the plurality of first recall objects and the target account, and the index fusion information of the plurality of first recall objects under the multi-business index, comprises:
Grouping the plurality of first recall objects to obtain at least two object groups;
Determining intra-group ordering information of the first recall object in each object group based on the matching degree information and the index fusion information;
and based on the intra-group ordering information, adjusting the arrangement positions of the plurality of first recall objects in the initial object sequence to obtain the target object sequence.
3. The method of claim 2, wherein adjusting the arrangement positions of the plurality of first recall objects in the initial sequence of objects based on the intra-group ordering information to obtain a target sequence of objects comprises:
determining a target arrangement position of a first recall object in the initial object sequence in each object group;
and inserting the first recall object in each object group into the target arrangement position based on the intra-group ordering information of each object group to obtain the target object sequence.
4. A method according to claim 2 or 3, wherein said grouping said plurality of first recall objects to obtain at least two object groups comprises:
acquiring a plurality of preset object category information;
And based on the preset object category information, grouping the plurality of first recall objects to obtain at least two object groups, wherein each object group corresponds to one preset object category information.
5. The method of claim 1, wherein the multi-service indicator corresponds to a target service type, the method further comprising:
acquiring the plurality of candidate objects;
Determining a plurality of index information of each candidate object under the multi-service index;
Determining index fusion information of each candidate object according to the plurality of index information of each candidate object;
And carrying out grading processing on the plurality of candidate objects based on the index fusion information to obtain a plurality of object recall sets, wherein the grade information of the candidate objects in each object recall set is the same.
6. The method of claim 5, wherein the obtaining the plurality of first recall objects comprises:
Determining at least one target object recall set from the plurality of object recall sets;
the plurality of first recall objects are obtained from the at least one target object recall set.
7. The method of claim 1, wherein the obtaining the plurality of second recall objects that match the target account comprises:
Acquiring a plurality of objects, associated information of the plurality of objects and account attribute information of the target account;
inputting the association information and the account attribute information into a matching prediction model, and performing matching processing to obtain matching prediction information;
and screening the plurality of second recall objects from the plurality of objects based on the matching prediction information.
8. The method of claim 7, wherein the sorting the plurality of first recall objects and the plurality of second recall objects to obtain an initial sequence of objects comprises:
Acquiring index fusion information of each of the plurality of first recall objects and matching prediction information of the plurality of second recall objects;
And based on the index fusion information and the matching prediction information, sequencing the plurality of first recall objects and the plurality of second recall objects to obtain the initial object sequence.
9. The method according to claim 1, wherein the method further comprises:
acquiring a resource consumption threshold value, unit resource consumption information corresponding to the plurality of candidate objects and accumulated recommended times;
determining current resource consumption information according to the unit resource consumption information and the accumulated recommended times;
the obtaining a plurality of first recall objects and a plurality of second recall objects matched with the target account includes:
And under the condition that the current resource consumption information is smaller than the resource consumption threshold, acquiring the plurality of first recall objects and a plurality of second recall objects matched with the target account.
10. An object recommendation device, characterized by comprising:
The first acquisition module is configured to execute acquisition of a plurality of first recall objects and a plurality of second recall objects matched with the target account, wherein the plurality of first recall objects are screened from a plurality of candidate objects;
The initial ordering module is configured to perform ordering processing on the plurality of first recall objects and the plurality of second recall objects to obtain an initial object sequence;
The rearrangement module is configured to execute index fusion information based on the matching degree information of the plurality of first recall objects and the target account and the index fusion information of the plurality of first recall objects under the multi-service index, and adjust the arrangement positions of the plurality of first recall objects in the initial object sequence to obtain a target object sequence; the arrangement positions of the plurality of second recall objects in the initial object sequence are the same as the arrangement positions of the plurality of second recall objects in the target object sequence;
and the recommending module is configured to execute recommending the target object sequence to the target account.
11. The apparatus of claim 10, wherein the rearrangement module comprises:
A grouping unit configured to perform grouping processing on the plurality of first recall objects to obtain at least two object groups;
An intra-group ranking unit configured to perform determining intra-group ranking information of the first recall object in each object group based on the matching degree information and the index fusion information;
And the rearrangement unit is configured to execute adjustment of arrangement positions of the plurality of first recall objects in the initial object sequence based on the intra-group ordering information to obtain the target object sequence.
12. The apparatus of claim 11, wherein the rearrangement unit comprises:
A target position determination subunit configured to perform determining a target arrangement position of the first recall object in the initial sequence of objects in each object group;
And the interpolation subunit is configured to perform the insertion of the first recall object in each object group into the target arrangement position based on the intra-group ordering information of each object group to obtain the target object sequence.
13. The apparatus according to claim 11 or 12, wherein the grouping unit comprises:
an object category acquisition subunit configured to perform acquisition of a plurality of preset object category information;
And the grouping subunit is configured to perform grouping processing on the plurality of first recall objects based on the preset object category information to obtain at least two object groups, and each object group corresponds to one preset object category information.
14. The apparatus of claim 10, wherein the multi-service indicator corresponds to a target service type, the apparatus further comprising:
A second acquisition module configured to perform acquisition of the plurality of candidate objects;
an index information determining module configured to perform determining a plurality of index information of each candidate object under the multi-service index;
an index fusion information determining module configured to perform determining index fusion information of each candidate object according to a plurality of index information of each candidate object;
And the object recall set acquisition module is configured to execute grading processing on the plurality of candidate objects based on the index fusion information to obtain a plurality of object recall sets, wherein the grade information of the candidate objects in each object recall set is the same.
15. The apparatus of claim 14, wherein the first acquisition module comprises:
a first acquisition unit configured to perform determining at least one target object recall set from the plurality of object recall sets;
a second acquisition unit configured to perform acquisition of the plurality of first recall objects from the at least one target object recall set.
16. The apparatus of claim 10, wherein the first acquisition module comprises:
A third acquisition unit configured to perform acquisition of a plurality of objects, association information of the plurality of objects, and account attribute information of the target account;
The matching unit is configured to input the association information and the account attribute information into a matching prediction model for matching processing to obtain matching prediction information;
And a screening unit configured to perform screening of the plurality of second recall objects from the plurality of objects based on the matching prediction information.
17. The apparatus of claim 16, wherein the initial ordering module comprises:
a fourth acquisition unit configured to perform acquisition of index fusion information of each of the plurality of first recall objects and matching prediction information of the plurality of second recall objects;
and the initial sorting unit is configured to perform sorting processing on the plurality of first recall objects and the plurality of second recall objects based on the index fusion information and the matching prediction information, so as to obtain the initial object sequence.
18. The apparatus as recited in claim 10, further comprising:
The third acquisition module is configured to acquire a resource consumption threshold, unit resource consumption information corresponding to the plurality of candidate objects and accumulated recommended times;
A resource consumption determination module configured to perform determination of current resource consumption information according to the unit resource consumption information and the accumulated recommended number of times;
The first obtaining module is further configured to obtain the plurality of first recall objects and a plurality of second recall objects matched with the target account if the current resource consumption information is less than the resource consumption threshold.
19. An electronic device, comprising:
A processor;
a memory for storing the processor-executable instructions;
wherein the processor is configured to execute the instructions to implement the object recommendation method of any one of claims 1 to 9.
20. A computer readable storage medium, characterized in that instructions in the computer readable storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the object recommendation method of any one of claims 1 to 9.
21. A computer program product comprising computer instructions which, when executed by a processor, implement the object recommendation method of any one of claims 1 to 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111652817.3A CN114547434B (en) | 2021-12-30 | 2021-12-30 | Object recommendation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111652817.3A CN114547434B (en) | 2021-12-30 | 2021-12-30 | Object recommendation method and device, electronic equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114547434A CN114547434A (en) | 2022-05-27 |
CN114547434B true CN114547434B (en) | 2024-07-09 |
Family
ID=81669884
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111652817.3A Active CN114547434B (en) | 2021-12-30 | 2021-12-30 | Object recommendation method and device, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114547434B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103440259A (en) * | 2013-07-31 | 2013-12-11 | 亿赞普(北京)科技有限公司 | Network advertisement push method and device |
CN112131456A (en) * | 2019-06-24 | 2020-12-25 | 腾讯科技(北京)有限公司 | Information pushing method, device, equipment and storage medium |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113407816A (en) * | 2020-03-17 | 2021-09-17 | 北京达佳互联信息技术有限公司 | Object recommendation method and device, electronic equipment and storage medium |
CN111753204B (en) * | 2020-06-28 | 2022-04-26 | 北京字节跳动网络技术有限公司 | Information pushing method and device, electronic equipment and storage medium |
CN111859126B (en) * | 2020-07-09 | 2024-05-14 | 有半岛(北京)信息科技有限公司 | Recommended item determining method, device, equipment and storage medium |
CN113536104A (en) * | 2020-10-10 | 2021-10-22 | 腾讯科技(深圳)有限公司 | Information recommendation method, device, equipment and storage medium |
CN112328882B (en) * | 2020-11-05 | 2023-07-21 | 中国平安人寿保险股份有限公司 | Information recommendation method, device, terminal equipment and storage medium |
CN112507196A (en) * | 2020-12-18 | 2021-03-16 | 北京百度网讯科技有限公司 | Training method, search ordering method, device and equipment of fusion ordering model |
CN112785397B (en) * | 2021-03-09 | 2024-03-29 | 中国工商银行股份有限公司 | Product recommendation method, device and storage medium |
CN113641835B (en) * | 2021-08-27 | 2022-10-11 | 北京达佳互联信息技术有限公司 | Multimedia resource recommendation method and device, electronic equipment and medium |
-
2021
- 2021-12-30 CN CN202111652817.3A patent/CN114547434B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103440259A (en) * | 2013-07-31 | 2013-12-11 | 亿赞普(北京)科技有限公司 | Network advertisement push method and device |
CN112131456A (en) * | 2019-06-24 | 2020-12-25 | 腾讯科技(北京)有限公司 | Information pushing method, device, equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN114547434A (en) | 2022-05-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110929206B (en) | Click rate estimation method and device, computer readable storage medium and equipment | |
CN110941740B (en) | Video recommendation method and computer-readable storage medium | |
EP3819821B1 (en) | User feature generating method, device, and apparatus, and computer-readable storage medium | |
CN111538912B (en) | Content recommendation method, device, equipment and readable storage medium | |
CN110941764A (en) | Object recommendation method and device, computer equipment and storage medium | |
CN106407364B (en) | Information recommendation method and device based on artificial intelligence | |
CN113641835B (en) | Multimedia resource recommendation method and device, electronic equipment and medium | |
CN110413867B (en) | Method and system for content recommendation | |
CN114461871B (en) | Recommendation model training method, object recommendation device and storage medium | |
CN111400546B (en) | Video recall method and video recommendation method and device | |
CN111858969B (en) | Multimedia data recommendation method, device, computer equipment and storage medium | |
CN113536105A (en) | Recommendation model training method and device | |
CN112883257A (en) | Behavior sequence data processing method and device, electronic equipment and storage medium | |
CN114329028A (en) | Data processing method, data processing equipment and computer readable storage medium | |
CN114201626B (en) | Multimedia recommendation method and device, electronic equipment and storage medium | |
CN113343024B (en) | Object recommendation method and device, electronic equipment and storage medium | |
CN113704511B (en) | Multimedia resource recommendation method and device, electronic equipment and storage medium | |
CN113742580B (en) | Recall method and device for target type data, electronic equipment and storage medium | |
CN113704509A (en) | Multimedia recommendation method and device, electronic equipment and storage medium | |
CN114547434B (en) | Object recommendation method and device, electronic equipment and storage medium | |
CN114222146B (en) | Data delivery method and device, electronic equipment and storage medium | |
CN114139045A (en) | Object recommendation method and device, electronic equipment and storage medium | |
CN115687691A (en) | Video recommendation method and device | |
CN114491093A (en) | Multimedia resource recommendation and object representation network generation method and device | |
CN113868522B (en) | Object recommendation method and device, electronic equipment and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |