CN113515696A - Recommendation method and device, electronic equipment and storage medium - Google Patents
Recommendation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113515696A CN113515696A CN202110537640.6A CN202110537640A CN113515696A CN 113515696 A CN113515696 A CN 113515696A CN 202110537640 A CN202110537640 A CN 202110537640A CN 113515696 A CN113515696 A CN 113515696A
- Authority
- CN
- China
- Prior art keywords
- user
- resource
- target
- portrait
- resources
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 230000006399 behavior Effects 0.000 claims description 58
- 230000015654 memory Effects 0.000 claims description 35
- 230000004927 fusion Effects 0.000 claims description 11
- 238000010586 diagram Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 4
- 230000001360 synchronised effect Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/43—Querying
- G06F16/435—Filtering based on additional data, e.g. user or group profiles
Abstract
The embodiment of the invention relates to a recommendation method, a recommendation device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring historical resources displayed to a target user within a set historical time period and user behavior data of the target user aiming at the historical resources; determining a first user portrait of the target user according to the attribute information of the target user, and determining a second user portrait of different resource types corresponding to the target user according to the historical resource and the user behavior data, wherein the resource characteristics corresponding to the different resource types are not completely the same; and recommending resources to the target user according to the first user portrait and each second user portrait, so that resources of different resource types can be recommended to the user according to the characteristics of the user for the resources of different resource types, the richness of recommended resources is improved, and the user experience is improved.
Description
Technical Field
The embodiment of the invention relates to the field of intelligent recommendation, in particular to a recommendation method, a recommendation device, electronic equipment and a storage medium.
Background
In recent years, with the development of internet technology and multimedia technology, more and more users acquire resources such as videos, pictures and texts, songs and the like through the internet.
In order to meet the personalized requirements of users, recommendation systems widely exist in various websites and are used for recommending resources meeting the requirements of the users or possibly interested in the users through a certain algorithm according to the behavior data of the users.
Disclosure of Invention
In order to meet the personalized requirements of users and simultaneously improve the richness of recommended resources, the embodiment of the invention provides a recommendation method, a recommendation device, electronic equipment and a storage medium.
In a first aspect, an embodiment of the present invention provides a recommendation method, including:
acquiring historical resources displayed to a target user within a set historical time period and user behavior data of the target user aiming at the historical resources;
determining a first user portrait of the target user according to the attribute information of the target user, and determining a second user portrait of different resource types corresponding to the target user according to the historical resource and the user behavior data, wherein the resource characteristics corresponding to the different resource types are not completely the same;
and recommending resources to the target user according to the first user portrait and each second user portrait.
In one possible embodiment, the recommending resources to the target user according to the first user representation and each second user representation includes:
fusing the first user portrait and each second user portrait to obtain a third user portrait of the target user;
inputting the third user portrait and the resource characteristics of each candidate resource into a trained prediction model to obtain the estimated click rate of the target user to each candidate resource;
and recommending resources to the target user according to the estimated click rate of the target user to each candidate resource.
In one possible embodiment, the recommending resources to the target user according to the first user representation and each second user representation includes:
inputting the first user portrait, each second user portrait and resource characteristics of each candidate resource into a trained prediction model, wherein the prediction model comprises a portrait fusion sub-model and a click rate prediction sub-model, the first user portrait and each second user portrait are fused by the fusion sub-model to obtain a third user portrait corresponding to the target user and output to the click rate prediction sub-model, and the click rate prediction sub-model obtains the estimated click rate of the target user to each candidate resource according to the third user portrait and each resource characteristic;
and recommending resources to the target user according to the estimated click rate of the target user to each candidate resource.
In one possible embodiment, the resource types include: at least two of the image-text type, the long video type and the short video type;
wherein, the resource characteristics corresponding to the image-text type comprise: at least one of author information and copyright party information;
the resource characteristics corresponding to the long video type comprise: at least one of director information, actor information, and producer information;
the resource characteristics corresponding to the short video type include: and uploading at least one of short video person information, subject information and partner information.
In one possible embodiment, the method further comprises:
determining a second target resource matched with a first target resource from candidate resources of other resource types, wherein the first target resource refers to a resource recommended to the target user according to the first user portrait and each second user portrait, and the other resource types refer to resource types different from the resource type to which the first target resource belongs;
recommending the second target resource to the target user.
In one possible embodiment, the fusing the first user representation and each of the second user representations to obtain a third user representation of the target user includes:
fusing the first user portrait with each second user portrait respectively to obtain a plurality of fourth user portraits;
and fusing the plurality of fourth user portraits to obtain a third user portrait of the target user.
In one possible embodiment, the candidate resource is determined by:
determining adjacent users similar to the target user interests according to the user behavior data;
selecting resources related to the historical resources from the resources preferred by the adjacent users to determine the resources as the candidate resources.
In a second aspect, an embodiment of the present invention provides a recommendation apparatus, including:
the acquisition module is used for acquiring historical resources displayed to a target user within a set historical time period and user behavior data of the target user aiming at the historical resources;
the portrait module is used for determining a first user portrait of the target user according to the attribute information of the target user and determining a second user portrait of different resource types corresponding to the target user according to the historical resource and the user behavior data, wherein the resource characteristics corresponding to the different resource types are not completely the same;
and the recommending module is used for recommending resources to the target user according to the first user portrait and each second user portrait.
In a third aspect, an embodiment of the present invention provides an electronic device, including: a processor and a memory, wherein the processor is configured to execute the recommendation program stored in the memory to implement the recommendation method in the first aspect.
In a fourth aspect, an embodiment of the present invention provides a storage medium, including: the storage medium stores one or more programs, which are executable by one or more processors to implement the recommendation method described in the first aspect above.
The technical solution provided in the embodiment of the present invention is different from the prior art that different types of resources are normalized to the same feature space when there are multiple types of resources, that is, different from the prior art that the same feature is extracted for different types of resources, that is, by acquiring historical resources displayed to a target user within a set historical time period and user behavior data of the target user for the historical resources, determining a first user portrait of the target user according to attribute information of the target user, determining second user portraits of the target user corresponding to different resource types according to the historical resources and the user behavior data, recommending resources to the target user according to the first user portrait and each second user portrait, and because the second user portrait corresponding to the resource type is respectively drawn for each resource type, the second user portrait can include resource features unique to the resource type, therefore, the characteristics of the user for the resource type resource can be represented, and when the resource recommendation is performed on the target user according to the first user portrait and each second user portrait, the resources of different resource types can be recommended to the user according to the characteristics of the user for the resources of different resource types, the abundance of recommended resources is improved, and the user experience is improved.
Drawings
FIG. 1 is a system architecture diagram according to an embodiment of the present invention;
fig. 2 is a schematic flowchart of a recommendation method according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a prediction model;
FIG. 4 is a flowchart illustrating an image fusion process according to an embodiment of the present invention;
FIG. 5 is a flowchart illustrating another recommendation method according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a recommendation device according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
For the understanding of the embodiments of the present invention, the following description first makes an exemplary description of a system architecture related to the embodiments of the present invention with reference to the accompanying drawings.
Referring to fig. 1, a system architecture diagram provided in the embodiment of the present invention is shown. As shown in fig. 1, the system architecture 100 includes: the system comprises a user terminal 101 and a server 102, wherein the user terminal 101 is in communication connection with the server 102.
The user terminal 101 may be a device that supports various electronic devices having a display screen, including but not limited to: smart phones, desktop computers, tablet computers, laptop portable computers, etc., and fig. 1 only illustrates a smart phone as an example.
In practice, the user terminal 101 may provide a corresponding network service by installing a corresponding client application, for example, the user terminal 101 provides a video service by installing a video APP. Correspondingly, the server 102 provides the corresponding network service by installing the corresponding server application, for example, the server 102 provides the video service by installing the server application corresponding to the video APP.
It will be understood that the number of networks and devices in fig. 1 is merely illustrative. The system architecture 100 may include any number of networks and devices, as desired for an implementation.
Based on the system architecture illustrated in fig. 1, in an exemplary application scenario, the server 102 may be implemented to perform resource recommendation to a user using the user terminal 101 by using the recommendation method provided in the embodiment of the present invention, where resource types of recommended resources include, but are not limited to: long video, short video, graphics, and the like.
The following further explains the recommendation method provided by the embodiment of the present invention with specific embodiments in conjunction with the drawings, and the embodiment does not limit the embodiment of the present invention.
Referring to fig. 2, a schematic flow chart of a recommendation method according to an embodiment of the present invention is shown. As an example, the method may be applied to the server 102 illustrated in fig. 1. As shown in fig. 2, the method may include the steps of:
In the embodiment of the present invention, the historical resource refers to a resource that is presented to the target user within a set historical time period (for example, within the previous 48 hours, within the previous week, within the previous month, and the like), and of course, the resource types of the resource presented herein include, but are not limited to: long video, short video, graphics, and the like.
The user behavior specifically refers to an operation behavior of a user on a historical resource, and by analyzing user behavior data for a certain historical resource, a user's preference degree for the historical resource can be inferred, wherein the user behavior includes but is not limited to: the user behavior data is used for describing user behaviors, such as the number of clicks, the watching duration, the comment content and the like of the user on the historical resources.
As an embodiment, in an application, the terminal may monitor the user behavior at the terminal side by means of embedding the point, and collect and record behavior data of the user and a time point when the behavior data occurs, or when it is monitored that a specific user behavior occurs, collect and record corresponding user behavior data. And then, the terminal sends the collected user behavior data to the server, and the server receives and stores the user behavior data. In this way, the server can acquire the user behavior data of the target user for the historical resource in the set historical time period based on the stored user behavior data.
It can be understood that the resource characteristics corresponding to different resource types are not identical, and the resource of each resource type has its own resource characteristics, for example, the resource characteristics unique to the teletext type include but are not limited to: author information, copyright information, etc., resource characteristics unique to long video types including, but not limited to: director information, actor information, producer information, etc., resource characteristics unique to the short video type include, but are not limited to: up owner (i.e., the person who uploaded the short video), story information, partner information, actor information, etc.
Based on this, in the embodiment of the present invention, according to the historical resources and the user behavior data, a user image (called a second user image for convenience of description) of a target user corresponding to a different resource type is determined, that is, for each resource type, a second user image of the target user corresponding to the resource type is determined according to the historical resources of the resource type and the user behavior data of the user for the historical resources of the resource type.
Specifically, the server may first divide the historical resources obtained in step 201 according to the resource types to obtain a plurality of groups, where the historical resources in the same group belong to the same resource type, and the historical resources in different groups belong to different resource types, that is, each group corresponds to one resource type. Then, for each resource type, based on the historical resource in the corresponding group of the resource type and the user behavior data of the user for the historical resource in the group, performing behavior modeling on the user through a technical means, outputting a series of user labels through the behavior modeling, and forming a set by the user labels, wherein the set is a second user portrait of the user corresponding to the resource type. The technical means can be as follows: the embodiment of the invention does not limit the specific technical means adopted in behavior modeling.
The user behavior data is used for describing the operation behavior of the user on the historical resource (namely the user behavior), and the second user portrait is obtained by modeling the user behavior, so that the second user portrait can represent the characteristics of the operation behavior of the user on the historical resource, further, the operation behavior of the user on the historical resource can truly represent the characteristics of the preference, the preference degree and the like of the user on the historical resource, and therefore, the second user portrait corresponding to a certain resource type can represent the characteristics of the preference, the preference degree and the like of the user on the resource type resource.
For example, assuming that the historical resources displayed to the target user in the historical time period include resources of two resource types, namely, a long video and a short video, in this step 202, a user portrait (i.e., a second user portrait) of the target user for the long video may be determined according to the resource of the long video type and the user behavior data of the target user for the resource, respectively, and a user portrait (i.e., a second user portrait) of the target user for the short video may be determined according to the resource of the short video type and the user behavior data of the target user for the resource. Here, user portrayal for long videos may include, but is not limited to: the user's favorite director, the user's favorite actors, the highest number of offerings in the long video viewed by the user, etc., the user portrayal for short videos may include, but is not limited to: user's favorite up master, user's favorite subject, etc. As can be seen from the above description, in the embodiment of the present invention, by respectively depicting the second user portrait corresponding to the resource type of the target user for each resource type, the second user portrait can include resource features unique to the resource type, so that characteristics of the user for the resource type resource can be represented.
The attribute information of the target user includes, but is not limited to: the age, gender, city, etc. of the target user. As to how to determine the first user representation of the target user according to the attribute information of the target user, the embodiments of the present invention are not described again.
As can be seen from the description in step 203, in the embodiment of the present invention, when resource recommendation is performed on a target user, a second user portrait of the target user for each resource type is considered, which means that characteristics of the target user for each resource type resource are considered, so that resources of different resource types can be recommended to the user according to characteristics of the target user for resources of different resource types.
For example, if the target user likes a director a for the long video type and likes a master up b for the short video type, the target user may recommend the long video shot by the director a, the short video uploaded by the master up b, the long video shot by the master up b, the short video shot by the director a, and the short video clipped from the long video shot by the director a to the target user when recommending the resource to the target user.
In the embodiment of the invention, when resource recommendation is performed on the target user according to the first user portrait and each second user portrait, the first user portrait and each second user portrait are fused to obtain a third user portrait of the target user, and then resource recommendation is performed on the target user according to the third user portrait of the target user. How to fuse the first user representation and the second user representations to obtain a third user representation of the target user is described below by a flow shown in fig. 4, which will not be described in detail.
As an embodiment, after the first user portrait and each second user portrait are fused to obtain a third user portrait of the target user, the third user portrait and resource characteristics of each candidate resource can be input into the trained prediction model to obtain an estimated click rate of the target user on each candidate resource, and then resource recommendation is performed on the target user according to the estimated click rate of the target user on each candidate resource.
Here, the candidate resource refers to a resource that is recalled from a mass resource by the server and is likely to be of interest to the target user, and the resource characteristics of the candidate resource include resource characteristics unique to the candidate resource. The following description will be given as to how the server recalls the candidate resource from the mass resource, and will not be described in detail here.
The prediction model may be a DNN (Deep Neural Network) model, which is input as a third user profile of the user, resource characteristics of the candidate resource, and output as an estimated click rate of the user clicking the candidate resource. As will be understood by those skilled in the art, the prediction model can be obtained by constructing a training set, using the training set, and training the initial model by using a supervised training algorithm, wherein each training sample in the training set includes a third user portrait, a resource feature, and a user click behavior on a resource, and the click behavior includes click and non-click.
As another embodiment, the above-mentioned prediction model may include two parts, as shown in FIG. 3, which is a structural diagram of the prediction model. The prediction model illustrated in fig. 3 includes an image fusion sub-model and a click-through rate prediction sub-model, wherein the image fusion sub-model inputs the first user image and each second user image and outputs a third user image, and the click-through rate prediction sub-model inputs the third user image and resource characteristics of candidate resources and outputs an estimated click-through rate. Based on this, in this embodiment, the server may input the first user portrait, each second user portrait, and the resource features of each candidate resource into the trained prediction model, so as to fuse the first user portrait and each second user portrait by the fusion sub-model, obtain a third user portrait corresponding to the target user, and output the third user portrait to the click rate prediction sub-model, so as to obtain the estimated click rate of the target user for each candidate resource according to the third user portrait and the resource features of each candidate resource by the click rate prediction sub-model.
In the embodiment of the invention, after the server obtains the estimated click rate of the target user on each candidate resource, the candidate resource with the estimated click rate meeting the set conditions can be selected from each candidate resource and determined as the final recommended resource to be recommended to the target user.
Here, the setting conditions may be, for example: according to the order of the estimated click rate from high to low, the candidate resources with the top N (N is a natural number greater than 0 and is 50 for example) are ranked, or the candidate resources with the estimated click rate higher than a set threshold (0.5 for example) are ranked. It should be noted that in the application, the above N and the set threshold may be manually set by a user (e.g., an operator) and may be updated.
In the technical solution provided in the embodiment of the present invention, when there are multiple types of resources, different types of resources are normalized to the same feature space as in the prior art, that is, the same feature is extracted for the different types of resources (for example, an attribute of a director unique to a resource of a long video type, an attribute of an up theme unique to a resource of a short video type, and both have an attribute of a subject, so that when feature extraction is performed for the resource of the long video type, the feature of the director is discarded, and the feature of the subject is extracted, and similarly, when feature extraction is performed for the resource of the short video type, the feature of the up theme is discarded, and the feature of the subject is extracted), the first user portrait of the target user is determined according to the attribute information of the target user by acquiring the historical resource presented to the target user within a set historical time period and the user behavior data of the target user for the historical resource, and according to historical resources and user behavior data, determining second user portraits of different resource types corresponding to a target user, and recommending resources to the target user according to the first user portraits and the second user portraits.
Referring to fig. 4, a schematic flow chart of an image fusion process according to an embodiment of the present invention is shown in fig. 4, where the method may include the following steps:
And step 402, fusing the plurality of fourth user portraits to obtain a third user portrait of the target user.
For example, assuming that the second user representation includes a second user representation of the target user corresponding to the long video asset type (denoted b1) and a second user representation of the target user corresponding to the short video asset type (denoted b2), in an embodiment of the present invention, steps 401 and 402 above may be implemented by the following formulas:
c ═ a × b1+ a × b 2; or c ═ 1-a ═ b1+ a ^ b2
In the above formula, c represents the third user representation, a represents the first user representation, and a × b1, a × b2, (1-a) × b1, a × b2 are all the fourth user representations.
It should be noted that the way of fusing the first user portrait and the second user portraits to obtain the fourth user portraits and then the fourth user portraits to obtain the third user portrait shown in the above formula is only an exemplary one, and in an application, there are other ways of fusing, for example, weighting and fusing the first user portrait and the second user portraits to obtain the fourth user portraits and then the fourth user portraits to obtain the third user portrait, which is not limited in the embodiment of the present invention.
According to the technical scheme provided by the embodiment of the invention, the first user portrait and each second user portrait are fused to obtain the third user portrait, and then the resource recommendation is carried out on the target user according to the third user portrait, so that the characteristic of the target user aiming at different resource types can be introduced when the resource recommendation is carried out on the target user, thereby realizing the purpose of recommending the resources of different resource types to the user, improving the richness of the recommended resources and improving the user experience.
Referring to fig. 5, a schematic flowchart of another recommendation method provided in an embodiment of the present invention is shown, where the flowchart shown in fig. 5 may include the following steps based on the flowchart shown in fig. 2:
The description of step 501 and step 502 may refer to the description of step 201 and step 202 in the embodiment shown in fig. 2, and are not repeated here.
The first target resource in step 503 may refer to the candidate resource whose estimated click rate satisfies the set condition as described in step 203 of the embodiment shown in fig. 2. For how to determine the first target resource according to the first user image and each second user image, refer to the description in step 203, which is not repeated herein.
First, the other resource type is a resource type different from the resource type to which the first target resource belongs.
In the embodiment of the present invention, determining the second target resource matching the first target resource from the candidate resources of other resource types may be implemented by the following processes: and determining the association degree between the candidate resource and the first target resource aiming at the candidate resource of each other resource type, and determining the candidate resource as a second target resource matched with the first target resource when determining that the association degree between the candidate resource and the first target resource meets the set condition. Here, the degree of association may be represented by a cosine distance or a euclidean distance between the feature vector of the candidate resource and the feature vector of the first target resource, and it is understood that a smaller cosine distance or euclidean distance means a larger degree of association. Accordingly, the setting condition may refer to that the cosine distance or the euclidean distance is smaller than a preset distance threshold. Wherein, the distance threshold value can be set by human and can be updated.
And 505, recommending the first target resource and the second target resource to the target user.
According to the technical scheme provided by the embodiment of the invention, on the basis of the flow of fig. 2, the second target resource matched with the first target resource is determined from the candidate resources of other resource types, and the first target resource and the second target resource are recommended to the target user, so that joint recommendation among different resource types can be realized, and the user experience is further improved.
To facilitate understanding of the joint recommendations between different resource types described above, the following examples are shown:
it is assumed that the first target resource obtained by executing the flow shown in fig. 2 or executing the above steps 501 to 503 is a clip in a certain tv play, that is, a short video, and the second target resource obtained by executing the step 504 is an episode of the tv play, that is, a long video. Therefore, when the short video corresponding to a certain television series is recommended to the target user, the complete episode of the television series is recommended to the target user at the same time, and the user experience is further improved.
In the following, a process of the server recalling resources that may be of interest to the target user from the mass resources in the embodiment of the present invention is described:
in the prior art, a server can recall resources that a target user may be interested in from a large number of resources through a collaborative filtering strategy, and specifically, the collaborative filtering strategy simply utilizes the preferences of groups with mutual interests and common experiences to recommend the resources that the target user may be interested in to the user. For example, assuming that the user a and the user b are interested in each other, that is, the user a and the user b are adjacent users with similar interests, the resources that the user a browses (or purchases) and the user b does not browse (or purchase) can be recommended to the user b, and similarly, the resources that the user b browses (or purchases) and the user a does not browse (or purchase) are recommended to the user a.
Based on the above description, the embodiment of the present invention proposes to determine neighboring users with similar interests to the target user according to the user behavior data, and then select resources related to historical resources from the resources preferred by the neighboring users to determine the resources as candidate resources. Therefore, compared with the prior art, in the embodiment of the invention, the resources preferred by the adjacent users of the target user are not simply determined as the candidate resources corresponding to the target user, the correlation between the resources preferred by the adjacent users and the historical resources corresponding to the target user is also considered, and the historical resources can accurately represent the actual interest points of the target user, so that the resources related to the historical resources in the resources preferred by the adjacent users can be matched with the actual interest points of the user, and further, when the resources related to the historical resources are recommended to the target user, the probability of the target user clicking the resources is higher, so that the recall accuracy can be improved.
For example, it is still assumed that the user a and the user B are adjacent users with similar interests, the user a is assumed to browse the brand a car and the brand B watch, and the user B is assumed to browse the brand B car, so when determining the candidate resource corresponding to the user B, the brand a car is determined as the candidate resource corresponding to the user B because the brand a car and the brand B car have strong correlation between furniture.
As one embodiment, deep semantics of the historical resource title and the resource title preferred by the adjacent user can be obtained through a BERT model, and then the correlation between the historical resource title and the adjacent user is determined through the deep semantics between the historical resource title and the adjacent user.
Referring to fig. 6, a schematic structural diagram of a recommendation device according to an embodiment of the present invention is shown in fig. 6, where the recommendation device may include:
the acquiring module 61 is configured to acquire a history resource displayed to a target user within a set history time period, and user behavior data of the target user for the history resource;
the portrait module 62 is configured to determine a first user portrait of the target user according to the attribute information of the target user, and determine a second user portrait of different resource types corresponding to the target user according to the historical resource and the user behavior data, where resource features corresponding to different resource types are not completely the same;
and the recommending module 63 is configured to recommend resources to the target user according to the first user representation and each second user representation.
In a possible embodiment, said recommendation module 63 comprises (not shown in the figures):
the fusion submodule is used for fusing the first user portrait and each second user portrait to obtain a third user portrait of the target user;
the first prediction submodule is used for inputting the third user portrait and the resource characteristics of each candidate resource into a trained prediction model to obtain the estimated click rate of the target user on each candidate resource;
and the first recommending submodule is used for recommending the resources to the target user according to the estimated click rate of the target user to each candidate resource.
In a possible embodiment, said recommendation module 63 comprises (not shown in the figures):
the second prediction submodule is used for inputting the first user portrait, each second user portrait and the resource characteristics of each candidate resource into a trained prediction model, the prediction model comprises a portrait fusion submodel and a click rate prediction submodel, the first user portrait and each second user portrait are fused by the fusion submodel to obtain a third user portrait corresponding to the target user and output the third user portrait to the click rate prediction submodel, and the click rate prediction submodel obtains the estimated click rate of the target user to each candidate resource according to the third user portrait and each resource characteristic;
and the second recommending submodule is used for recommending the resources to the target user according to the estimated click rate of the target user to each candidate resource.
In one possible embodiment, the resource types include: at least two of the image-text type, the long video type and the short video type;
wherein, the resource characteristics corresponding to the image-text type comprise: at least one of author information and copyright party information;
the resource characteristics corresponding to the long video type comprise: at least one of director information, actor information, and producer information;
the resource characteristics corresponding to the short video type include: and uploading at least one of short video person information, subject information and partner information.
In a possible embodiment, the device further comprises (not shown in the figures):
the determining module is used for determining a second target resource matched with a first target resource from candidate resources of other resource types, wherein the first target resource refers to a resource recommended to the target user according to the first user portrait and each second user portrait, and the other resource types refer to resource types different from the resource type to which the first target resource belongs;
the recommending module 63 is further configured to: recommending the second target resource to the target user.
In one possible embodiment, the fusion submodule is specifically configured to:
fusing the first user portrait with each second user portrait respectively to obtain a plurality of fourth user portraits; and fusing the plurality of fourth user portraits to obtain a third user portrait of the target user.
In a possible embodiment, the device further comprises (not shown in the figures):
a recall module to determine the candidate resource by: determining adjacent users similar to the target user interests according to the user behavior data; selecting resources related to the historical resources from the resources preferred by the adjacent users to determine the resources as the candidate resources.
Fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, where the electronic device 700 shown in fig. 7 includes: at least one processor 701, memory 702, at least one network interface 704, and other user interfaces 703. The various components in the electronic device 700 are coupled together by a bus system 705. It is understood that the bus system 705 is used to enable communications among the components. The bus system 705 includes a power bus, a control bus, and a status signal bus in addition to a data bus. But for clarity of illustration the various busses are labeled in figure 7 as the bus system 705.
The user interface 703 may include, among other things, a display, a keyboard, or a pointing device (e.g., a mouse, trackball, touch pad, or touch screen, among others.
It is to be understood that the memory 702 in embodiments of the present invention may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory. The non-volatile memory may be a Read-only memory (ROM), a programmable Read-only memory (PROM), an erasable programmable Read-only memory (erasabprom, EPROM), an electrically erasable programmable Read-only memory (EEPROM), or a flash memory. The volatile memory may be a Random Access Memory (RAM) which functions as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (staticiram, SRAM), dynamic random access memory (dynamic RAM, DRAM), synchronous dynamic random access memory (syncronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM ), Enhanced Synchronous DRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and direct memory bus RAM (DRRAM). The memory 702 described herein is intended to comprise, without being limited to, these and any other suitable types of memory.
In some embodiments, memory 702 stores the following elements, executable units or data structures, or a subset thereof, or an expanded set thereof: an operating system 7021 and application programs 7022.
The operating system 7021 includes various system programs, such as a framework layer, a core library layer, a driver layer, and the like, for implementing various basic services and processing hardware-based tasks. The application 7022 includes various applications, such as a media player (MediaPlayer), a Browser (Browser), and the like, for implementing various application services. Programs that implement methods in accordance with embodiments of the present invention can be included within application program 7022.
In the embodiment of the present invention, the processor 701 is configured to execute the method steps provided by the method embodiments by calling a program or an instruction stored in the memory 702, specifically, a program or an instruction stored in the application 7022, for example, and includes:
acquiring historical resources displayed to a target user within a set historical time period and user behavior data of the target user aiming at the historical resources; determining a first user portrait of the target user according to the attribute information of the target user, and determining a second user portrait of different resource types corresponding to the target user according to the historical resource and the user behavior data, wherein the resource characteristics corresponding to the different resource types are not completely the same; and recommending resources to the target user according to the first user portrait and each second user portrait.
The method disclosed in the above embodiments of the present invention may be applied to the processor 701, or implemented by the processor 701. The processor 701 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be implemented by integrated logic circuits of hardware or instructions in the form of software in the processor 701. The processor 701 may be a general-purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, or discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software elements in the decoding processor. The software elements may be located in ram, flash, rom, prom, or eprom, registers, among other storage media that are well known in the art. The storage medium is located in the memory 702, and the processor 701 reads the information in the memory 702 and performs the steps of the above method in combination with the hardware thereof.
It is to be understood that the embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or any combination thereof. For a hardware implementation, the processing units may be implemented within one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), general purpose processors, controllers, micro-controllers, microprocessors, other electronic units configured to perform the functions described herein, or a combination thereof.
For a software implementation, the techniques described herein may be implemented by means of units performing the functions described herein. The software codes may be stored in a memory and executed by a processor. The memory may be implemented within the processor or external to the processor.
The server provided in this embodiment may be an electronic device as shown in fig. 7, and may execute all the steps of the resource recommendation method shown in fig. 2 and 4-5, so as to achieve the technical effect of the resource recommendation method shown in fig. 2 and 4-5, for which reference is specifically made to the descriptions related to fig. 2 and 4-5, which are not described herein for brevity.
The embodiment of the invention also provides a storage medium (computer readable storage medium). The storage medium herein stores one or more programs. Among others, the storage medium may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, a hard disk, or a solid state disk; the memory may also comprise a combination of memories of the kind described above.
When one or more programs in the storage medium are executable by one or more processors, the recommendation method executed on the electronic device side is realized.
The processor is used for executing the recommendation program stored in the memory to realize the following steps of the recommendation method executed on the electronic equipment side:
acquiring historical resources displayed to a target user within a set historical time period and user behavior data of the target user aiming at the historical resources; determining a first user portrait of the target user according to the attribute information of the target user, and determining a second user portrait of different resource types corresponding to the target user according to the historical resource and the user behavior data, wherein the resource characteristics corresponding to the different resource types are not completely the same; and recommending resources to the target user according to the first user portrait and each second user portrait.
Those of skill would further appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied in hardware, a software module executed by a processor, or a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. A recommendation method, comprising:
acquiring historical resources displayed to a target user within a set historical time period and user behavior data of the target user aiming at the historical resources;
determining a first user portrait of the target user according to the attribute information of the target user, and determining a second user portrait of different resource types corresponding to the target user according to the historical resource and the user behavior data, wherein the resource characteristics corresponding to the different resource types are not completely the same;
and recommending resources to the target user according to the first user portrait and each second user portrait.
2. The method of claim 1, wherein said making resource recommendations to the target user based on the first user representation, each of the second user representations comprises:
fusing the first user portrait and each second user portrait to obtain a third user portrait of the target user;
inputting the third user portrait and the resource characteristics of each candidate resource into a trained prediction model to obtain the estimated click rate of the target user to each candidate resource;
and recommending resources to the target user according to the estimated click rate of the target user to each candidate resource.
3. The method of claim 1, wherein said making resource recommendations to the target user based on the first user representation, each of the second user representations comprises:
inputting the first user portrait, each second user portrait and resource characteristics of each candidate resource into a trained prediction model, wherein the prediction model comprises a portrait fusion sub-model and a click rate prediction sub-model, the first user portrait and each second user portrait are fused by the fusion sub-model to obtain a third user portrait corresponding to the target user and output the third user portrait to the click rate prediction sub-model, and the click rate prediction sub-model obtains the estimated click rate of the target user to each candidate resource according to the third user portrait and each resource characteristic;
and recommending resources to the target user according to the estimated click rate of the target user to each candidate resource.
4. The method of claim 1, wherein the resource types comprise: at least two of the image-text type, the long video type and the short video type;
wherein, the resource characteristics corresponding to the image-text type comprise: at least one of author information and copyright party information;
the resource characteristics corresponding to the long video type comprise: at least one of director information, actor information, and producer information;
the resource characteristics corresponding to the short video type include: and uploading at least one of short video person information, subject information and partner information.
5. The method of claim 1, further comprising:
determining a second target resource matched with a first target resource from candidate resources of other resource types, wherein the first target resource refers to a resource recommended to the target user according to the first user portrait and each second user portrait, and the other resource types refer to resource types different from the resource type to which the first target resource belongs;
recommending the second target resource to the target user.
6. The method of claim 2, wherein said fusing said first user representation and each of said second user representations to obtain a third user representation of said target user comprises:
fusing the first user portrait with each second user portrait respectively to obtain a plurality of fourth user portraits;
and fusing the plurality of fourth user portraits to obtain a third user portrait of the target user.
7. The method according to any one of claims 2, 3, 5 and 6, wherein the candidate resource is determined by:
determining adjacent users similar to the target user interests according to the user behavior data;
selecting resources related to the historical resources from the resources preferred by the adjacent users to determine the resources as the candidate resources.
8. A recommendation device, comprising:
the acquisition module is used for acquiring historical resources displayed to a target user within a set historical time period and user behavior data of the target user aiming at the historical resources;
the portrait module is used for determining a first user portrait of the target user according to the attribute information of the target user and determining a second user portrait of different resource types corresponding to the target user according to the historical resource and the user behavior data, wherein the resource characteristics corresponding to the different resource types are not completely the same;
and the recommending module is used for recommending resources to the target user according to the first user portrait and each second user portrait.
9. An electronic device, comprising: a processor and a memory, the processor being configured to execute the recommendation program stored in the memory to implement the recommendation method of any one of claims 1-7.
10. A storage medium storing one or more programs, the one or more programs being executable by one or more processors to implement the recommendation method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110537640.6A CN113515696A (en) | 2021-05-17 | 2021-05-17 | Recommendation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110537640.6A CN113515696A (en) | 2021-05-17 | 2021-05-17 | Recommendation method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113515696A true CN113515696A (en) | 2021-10-19 |
Family
ID=78064390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110537640.6A Pending CN113515696A (en) | 2021-05-17 | 2021-05-17 | Recommendation method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113515696A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114422584A (en) * | 2021-12-16 | 2022-04-29 | 北京百度网讯科技有限公司 | Resource pushing method, equipment and storage medium |
| CN114840743A (en) * | 2022-03-01 | 2022-08-02 | 深圳市小秤砣科技有限公司 | Model recommendation method and device, electronic equipment and readable storage medium |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107832437A (en) * | 2017-11-16 | 2018-03-23 | 北京小米移动软件有限公司 | Audio/video method for pushing, device, equipment and storage medium |
| CN108024145A (en) * | 2017-12-07 | 2018-05-11 | 北京百度网讯科技有限公司 | Video recommendation method, device, computer equipment and storage medium |
| CN108108998A (en) * | 2017-12-14 | 2018-06-01 | 百度在线网络技术（北京）有限公司 | Showing advertisement method and apparatus, server, storage medium |
| CN110162698A (en) * | 2019-04-18 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of user's representation data processing method, device and storage medium |
| CN110413877A (en) * | 2019-07-02 | 2019-11-05 | 阿里巴巴集团控股有限公司 | A kind of resource recommendation method, device and electronic equipment |
| CN110598011A (en) * | 2019-09-27 | 2019-12-20 | 腾讯科技（深圳）有限公司 | Data processing method, data processing device, computer equipment and readable storage medium |
| CN111722766A (en) * | 2020-06-04 | 2020-09-29 | 北京达佳互联信息技术有限公司 | Multimedia resource display method and device |
| CN111782957A (en) * | 2020-07-13 | 2020-10-16 | 腾讯科技（深圳）有限公司 | Method and device for recommending content, computing equipment and storage medium |
| 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 |
| CN112182460A (en) * | 2019-07-04 | 2021-01-05 | 腾讯科技（北京）有限公司 | Resource pushing method and device, storage medium and electronic device |
| CN112328888A (en) * | 2020-11-20 | 2021-02-05 | Oppo广东移动通信有限公司 | Information recommendation method and device, server and storage medium |
| CN112395496A (en) * | 2020-10-22 | 2021-02-23 | 上海众源网络有限公司 | Information recommendation method and device, electronic equipment and storage medium |
| CN112597395A (en) * | 2020-12-28 | 2021-04-02 | 上海众源网络有限公司 | Object recommendation method, device, equipment and storage medium |
-
2021
- 2021-05-17 CN CN202110537640.6A patent/CN113515696A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107832437A (en) * | 2017-11-16 | 2018-03-23 | 北京小米移动软件有限公司 | Audio/video method for pushing, device, equipment and storage medium |
| CN108024145A (en) * | 2017-12-07 | 2018-05-11 | 北京百度网讯科技有限公司 | Video recommendation method, device, computer equipment and storage medium |
| CN108108998A (en) * | 2017-12-14 | 2018-06-01 | 百度在线网络技术（北京）有限公司 | Showing advertisement method and apparatus, server, storage medium |
| CN110162698A (en) * | 2019-04-18 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of user's representation data processing method, device and storage medium |
| 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 |
| CN110413877A (en) * | 2019-07-02 | 2019-11-05 | 阿里巴巴集团控股有限公司 | A kind of resource recommendation method, device and electronic equipment |
| CN112182460A (en) * | 2019-07-04 | 2021-01-05 | 腾讯科技（北京）有限公司 | Resource pushing method and device, storage medium and electronic device |
| CN110598011A (en) * | 2019-09-27 | 2019-12-20 | 腾讯科技（深圳）有限公司 | Data processing method, data processing device, computer equipment and readable storage medium |
| CN111722766A (en) * | 2020-06-04 | 2020-09-29 | 北京达佳互联信息技术有限公司 | Multimedia resource display method and device |
| CN111782957A (en) * | 2020-07-13 | 2020-10-16 | 腾讯科技（深圳）有限公司 | Method and device for recommending content, computing equipment and storage medium |
| CN112395496A (en) * | 2020-10-22 | 2021-02-23 | 上海众源网络有限公司 | Information recommendation method and device, electronic equipment and storage medium |
| CN112328888A (en) * | 2020-11-20 | 2021-02-05 | Oppo广东移动通信有限公司 | Information recommendation method and device, server and storage medium |
| CN112597395A (en) * | 2020-12-28 | 2021-04-02 | 上海众源网络有限公司 | Object recommendation method, device, equipment and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| 李浩君;张广;王万良;江波;: "基于多维特征差异的个性化学习资源推荐方法", 系统工程理论与实践, no. 11, pages 245 - 255 * |
| 解娜;: "基于用户画像的读者周边好书推荐服务研究", 情报探索, no. 08, pages 109 - 113 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114422584A (en) * | 2021-12-16 | 2022-04-29 | 北京百度网讯科技有限公司 | Resource pushing method, equipment and storage medium |
| CN114840743A (en) * | 2022-03-01 | 2022-08-02 | 深圳市小秤砣科技有限公司 | Model recommendation method and device, electronic equipment and readable storage medium |
| CN114840743B (en) * | 2022-03-01 | 2023-02-07 | 深圳市小秤砣科技有限公司 | Model recommendation method and device, electronic equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11176213B2 (en) | Systems and methods for identifying electronic content using video graphs | |
| CN111078939B (en) | Method, system and recording medium for extracting and providing highlight image in video content | |
| US9684656B2 (en) | Creating personalized and continuous playlists for a content sharing platform based on user history | |
| US9064154B2 (en) | Systems and methods for associating electronic content | |
| CN110139162B (en) | Media content sharing method and device, storage medium and electronic device | |
| CN106326391B (en) | Multimedia resource recommendation method and device | |
| JP2021103543A (en) | Use of machine learning for recommending live-stream content | |
| RU2632131C2 (en) | Method and device for creating recommended list of content | |
| CN110223186B (en) | User similarity determining method and information recommending method | |
| CN111178970B (en) | Advertisement putting method and device, electronic equipment and computer readable storage medium | |
| US10136173B2 (en) | Conditional display of hyperlinks in a video | |
| WO2018011689A1 (en) | Method and system for switching to dynamically assembled video during streaming of live video | |
| CN112100504B (en) | Content recommendation method and device, electronic equipment and storage medium | |
| CN113515696A (en) | Recommendation method and device, electronic equipment and storage medium | |
| CN113254679B (en) | Multimedia resource recommendation method and device, electronic equipment and storage medium | |
| CN107515870B (en) | Searching method and device and searching device | |
| CN112507163A (en) | Duration prediction model training method, recommendation method, device, equipment and medium | |
| US9015607B2 (en) | Virtual space providing apparatus and method | |
| CN114896454A (en) | Short video data recommendation method and system based on label analysis | |
| US10225293B1 (en) | Decorating embedded graphic representations on social shares with metadata | |
| CN110264283B (en) | Popularization resource display method and device | |
| CN115152242A (en) | Machine learning management of videos for selection and display | |
| CN112287799A (en) | Video processing method and device based on artificial intelligence and electronic equipment | |
| CN113836388A (en) | Information recommendation method and device, server and storage medium | |
| US9704178B2 (en) | Customizable sequence of content |
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 |