CN113868516A

CN113868516A - Object recommendation method and device, electronic equipment and storage medium

Info

Publication number: CN113868516A
Application number: CN202111057304.8A
Authority: CN
Inventors: 蒋敏; 邱明辉; 董浩; 王志华
Original assignee: Beijing Dajia Internet Information Technology Co Ltd
Current assignee: Beijing Dajia Internet Information Technology Co Ltd
Priority date: 2021-09-09
Filing date: 2021-09-09
Publication date: 2021-12-31

Abstract

The method comprises the steps of obtaining an initial generalization attribute, an initial personalized attribute, historical deep behavior information and an object attribute of an object to be recommended of a target user account; performing liveness analysis on the target user account based on the historical deep behavior information to obtain liveness indexes, wherein the liveness indexes comprise a first index and a second index, the first index represents the generalization degree of the target user account, and the second index represents the personalization degree of the target user account; determining a target generalization attribute according to the initial generalization attribute and the first index; determining a target personalized attribute according to the initial personalized attribute and the second index; generating a target interest index based on the target generalization attribute, the target personalized attribute and the object attribute; and recommending the target object in the objects to be recommended to the target user account based on the target interest index. By the aid of the method and the device, recommendation accuracy and effect can be improved.

Description

Object recommendation method and device, electronic equipment and storage medium

Technical Field

The present disclosure relates to the field of internet technologies, and in particular, to an object recommendation method and apparatus, an electronic device, and a storage medium.

Background

With the development of internet technology, a large number of network platforms are continuously upgraded, so that besides some image-text information can be issued, personal users can share daily short videos at any time, and how to accurately capture the interests of the users is a challenge met by a large number of recommendation systems.

In the related technology, in the process of mining user interests, no matter a new user or an old user, the method on the old user is often directly reused for user characterization, that is, all information of the user is directly aggregated, including generalized information such as gender, age, region and mobile phone brand, and personalized information such as user identification and historical behavior. However, any recommendation system is dynamically changed, and the user, the object to be recommended, and the like in the system are also dynamically changed, and the related technology lacks consideration of the dynamic variability, and cannot effectively use the user information to perform user characterization, so that the interest and preference of the user cannot be captured well, and the problem of poor recommendation accuracy and effect in the recommendation system is caused.

Disclosure of Invention

The disclosure provides an object recommendation method, an object recommendation device, an electronic device and a storage medium, which are used for at least solving the problem that the recommendation accuracy and effect in a recommendation system are poor due to the fact that the interest preference of a user cannot be effectively learned in the related art. The technical scheme of the disclosure is as follows:

according to a first aspect of the embodiments of the present disclosure, there is provided an object recommendation method, including:

acquiring an initial generalization attribute, an initial personalized attribute, historical deep behavior information and an object attribute of an object to be recommended of a target user account, wherein the historical deep behavior information represents operation information generated in the process that the target user account executes a preset operation on a historical object;

performing liveness analysis on the target user account based on the historical deep behavior information to obtain liveness indexes, wherein the liveness indexes comprise a first index and a second index, the first index represents the generalization degree of the target user account, and the second index represents the personalization degree of the target user account;

determining a target generalization attribute according to the initial generalization attribute and the first index;

determining a target personalized attribute according to the initial personalized attribute and the second index;

generating a target interest index based on the target generalization attribute, the target personalized attribute and the object attribute;

recommending the target object in the objects to be recommended to the target user account based on the target interest index.

Optionally, the activity analysis of the target user account based on the historical deep behavior information to obtain an activity index includes:

and inputting the historical deep behavior information into an activity analysis network for activity analysis to obtain the activity index.

Optionally, the activity analysis network includes: an active feature extraction layer and a normalization layer;

inputting the historical deep behavior information into an activity analysis network for activity analysis, and obtaining the activity index comprises:

performing active feature extraction on the historical deep behavior information based on the active feature extraction layer to obtain an initial activity index, wherein the initial activity index comprises a first initial index and a second initial index;

and normalizing the first initial index and the second initial index based on the normalization layer to obtain the first index and the second index.

Optionally, the generating a target interest indicator based on the target generalization attribute, the target personalized attribute and the object attribute includes:

and inputting the target generalization attribute, the target personalized attribute and the object attribute into an interest identification network for interest identification to obtain the target interest index.

Optionally, the interest recognition network includes: the system comprises a user characteristic extraction layer, an object characteristic extraction layer and an interest perception layer;

the step of inputting the target generalization attribute, the target personalized attribute and the object attribute into an interest identification network for interest identification to obtain the target interest index comprises:

inputting the target generalization attribute and the target personalized attribute into the user feature extraction layer for user feature extraction to obtain target user feature information;

inputting the object attribute into the object feature extraction layer to extract object features to obtain target object feature information;

and inputting the target object characteristic information and the target user characteristic information into the interest perception layer for interest perception processing to obtain the target interest index.

Optionally, the method further includes:

acquiring an initial sample generalization attribute, an initial sample personalized attribute, historical sample deep behavior information, a sample object attribute of a sample object corresponding to a sample user account and a labeling interest index of the sample user account on the sample object;

inputting the historical sample deep behavior information into an activity analysis network to be trained for activity analysis to obtain a sample activity index, wherein the sample activity index comprises a first sample index and a second sample index, the first sample index represents the generalization degree of the sample user account, and the second sample index represents the personalization degree of the sample user account;

determining a sample generalization attribute according to the initial sample generalization attribute and the first sample index;

determining a sample personalized attribute according to the initial sample personalized attribute and the second sample index;

inputting the sample generalization attribute, the sample personalized attribute and the sample object attribute into an interest recognition network to be trained for interest recognition to obtain a sample interest index;

determining target loss information according to the sample interest indexes and the labeling interest indexes;

and training the activity analysis network to be trained and the interest recognition network to be trained based on the target loss information to obtain the activity analysis network and the interest recognition network.

Optionally, the recommending, based on the target interest indicator, a target object in the objects to be recommended to the target user account includes:

determining the target object from the objects to be recommended according to the target interest index;

and recommending the target object to the target user account.

According to a second aspect of the embodiments of the present disclosure, there is provided an object recommendation apparatus including:

the information acquisition module is configured to execute acquisition of an initial generalization attribute, an initial personalized attribute, historical deep behavior information and an object attribute of an object to be recommended of a target user account;

a first liveness analysis module configured to perform liveness analysis on the target user account based on the historical deep behavior information to obtain liveness indexes, where the liveness indexes include a first index and a second index, the first index represents a generalization degree of the target user account, and the second index represents a personalization degree of the target user account;

a target generalization attribute determination module configured to perform determining a target generalization attribute based on the initial generalization attribute and the first indicator;

a target personalized attribute determination module configured to perform a determination of a target personalized attribute from the initial personalized attribute and the second indicator;

a target interest index generation module configured to perform generation of a target interest index based on the target generalization attribute, the target personalized attribute, and the object attribute;

and the object recommending module is configured to recommend a target object in the objects to be recommended to the target user account based on the target interest index.

Optionally, the first activity analysis module is further configured to perform activity analysis by inputting the historical deep behavior information into an activity analysis network, so as to obtain the activity index.

the first liveness analysis module includes:

an active feature extraction unit configured to perform active feature extraction on the historical deep behavior information based on the active feature extraction layer to obtain an initial activity index, where the initial activity index includes a first initial index and a second initial index;

a normalization processing unit configured to perform normalization processing on the first initial index and the second initial index based on the normalization layer to obtain the first index and the second index.

Optionally, the target interest index generating module is further configured to perform interest recognition by inputting the target generalization attribute, the target personalized attribute and the object attribute into an interest recognition network, so as to obtain the target interest index.

the target interest index generation module comprises:

the user characteristic extraction unit is configured to input the target generalization attribute and the target personalized attribute into the user characteristic extraction layer for user characteristic extraction to obtain target user characteristic information;

an object feature extraction unit configured to perform object feature extraction by inputting the object attribute into the object feature extraction layer, so as to obtain target object feature information;

and the interest perception processing unit is configured to input the target object characteristic information and the target user characteristic information into the interest perception layer for interest perception processing to obtain the target interest index.

Optionally, the apparatus further comprises:

the training data acquisition module is configured to perform acquisition of an initial sample generalization attribute, an initial sample personalized attribute, historical sample deep behavior information, a sample object attribute of a sample object corresponding to the sample user account, and a labeling interest index of the sample user account on the sample object;

the second liveness analysis module is configured to input the historical sample deep behavior information into a liveness analysis network to be trained for liveness analysis to obtain a sample liveness index, wherein the sample liveness index comprises a first sample index and a second sample index, the first sample index represents the generalization degree of the sample user account, and the second sample index represents the personalization degree of the sample user account;

a sample generalization attribute determination module configured to perform determining a sample generalization attribute based on the initial sample generalization attribute and the first sample metric;

a sample personalized attribute determination module configured to perform a determination of a sample personalized attribute from the initial sample personalized attribute and the second sample indicator;

the interest identification module is configured to input the sample generalization attribute, the sample personalized attribute and the sample object attribute into an interest identification network to be trained for interest identification to obtain a sample interest index;

a target loss information determination module configured to perform determining target loss information according to the sample interest indicator and the annotation interest indicator;

a network training module configured to perform training of the activity analysis network to be trained and the interest recognition network to be trained based on the target loss information, so as to obtain the activity analysis network and the interest recognition network.

Optionally, the object recommendation module includes:

a target object determination unit configured to perform determining the target object from the objects to be recommended according to the target interest indicator;

and the object recommending unit is configured to recommend the target object to the target user account.

According to a third aspect of the embodiments of the present disclosure, there is provided an electronic apparatus including: 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 embodiments of the present disclosure, there is provided a computer-readable storage medium, wherein instructions, when executed by a processor of an electronic device, enable the electronic device to perform the method of any one of the first aspects of the embodiments of the present disclosure.

According to a fifth aspect of the embodiments of the present disclosure, there is provided a computer program product containing instructions which, when run on a computer, cause the computer to perform the method of any one of the first aspects of the embodiments of the present disclosure.

The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:

by analyzing the activity of the historical deep behavior information of the target user account, a first index capable of representing the generalization degree of the target user account and a second index capable of representing the personalization degree of the target user account can be rapidly and accurately acquired, and the initial generalization attribute and the initial personalization attribute of the target user account are reasonably and effectively controlled by combining the first index and the second index, so that the proportion of the characterization of the interest and the preference of the user is represented, the precision of the characterization of the users with different user attributes is greatly improved, the precision of object recommendation is effectively improved, and the recommendation effect is improved.

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 present disclosure and, together with the description, serve to explain the principles of the disclosure and are not to be construed as limiting the disclosure.

FIG. 1 is a schematic diagram illustrating an application environment in accordance with an illustrative embodiment;

FIG. 2 is a flow diagram illustrating a method of object recommendation in accordance with an exemplary embodiment;

FIG. 3 is a flowchart illustrating inputting historical deep behavior information into an activity analysis network for activity analysis to obtain an activity indicator, according to an illustrative embodiment;

FIG. 4 is a flowchart illustrating inputting a target generalization attribute, a target personalization attribute, and an object attribute into an interest recognition network for interest recognition to obtain a target interest index in accordance with an illustrative embodiment;

FIG. 5 is a flow diagram illustrating a pre-generation liveness analysis network and interest recognition network in accordance with an exemplary embodiment;

FIG. 6 is a schematic diagram of a determination of a target interest indicator in conjunction with an activity analysis network and an interest recognition network, according to an example embodiment;

FIG. 7 is a block diagram illustrating an object recommendation device in accordance with an exemplary embodiment;

FIG. 8 is a block diagram illustrating an electronic device for object recommendation in accordance with an illustrative embodiment;

FIG. 9 is a block diagram illustrating an electronic device for object recommendation, according to an example embodiment.

Detailed Description

In order to make the technical solutions of the present disclosure better understood by those of ordinary skill in the art, the technical solutions in 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 above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein are capable of operation in sequences other than those illustrated or otherwise described herein. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for presentation, analyzed data, etc.) referred to in the present disclosure are information and data authorized by the user or sufficiently authorized by each party.

Referring to fig. 1, fig. 1 is a schematic diagram illustrating an application environment according to an exemplary embodiment, which may include a server 100 and a terminal 200, as shown in fig. 1.

In an alternative embodiment, the server 100 may be used to train liveness analysis networks and interest recognition networks. Specifically, the server 100 may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as a cloud service, a cloud database, cloud computing, a cloud function, cloud storage, a Network service, cloud communication, a middleware service, a domain name service, a security service, a CDN (Content Delivery Network), a big data and artificial intelligence platform, and the like.

In an alternative embodiment, the terminal 200 may perform the object recommendation process in combination with the liveness analysis network and the interest recognition network trained by the server 100. Specifically, the terminal 200 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. Optionally, 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 shows only one application environment provided by the present disclosure, and in practical applications, other application environments may also be included, for example, object recommendation processing may also be implemented on the server 100.

In the embodiment of the present specification, the server 100 and the terminal 200 may be directly or indirectly connected through wired or wireless communication, and the disclosure is not limited herein.

Fig. 2 is a flowchart illustrating an object recommendation method according to an exemplary embodiment, where the object recommendation method is used in a terminal and a server electronic device, as shown in fig. 2, and includes the following steps.

In step S201, an initial generalization attribute, an initial personalized attribute, historical deep behavior information, and an object attribute of an object to be recommended of a target user account are obtained.

In a specific embodiment, the target user account is any account in the object recommendation system.

In this embodiment, the initial generalization attribute of the target user account may be a user representation (e.g., a vector) obtained based on information of user generalization. In a specific embodiment, the information of the user generalization can be information of gender, age, region, mobile phone brand, and the like. Specifically, feature extraction may be performed on the information of the user generalization in advance to obtain the initial generalization attribute.

In this embodiment of the present specification, the initial personalized attribute of the target user account may be a user representation obtained based on information personalized by the user. In a specific embodiment, the user personalized information may be identification information of a target user account, historical behavior information (information generated during a process in which the target user account performs a preset operation on a historical object), and the like. Optionally, feature extraction may be performed on the personalized information of the user in advance to obtain the initial personalized attribute.

In a specific embodiment, the historical deep behavior information represents operation information generated in the process that the target user account performs a preset operation on the historical object. Specifically, the preset operation may be set in combination with an actual application scenario, and specifically, the preset operation may include, but is not limited to, browsing, clicking, converting (for example, a related product is purchased, or a related application is downloaded), and the like. Accordingly, the operation information may be the number of preset operations performed by the user, the time for the user to perform the preset operations, the operation type of the preset operations performed by the user, and the like. Optionally, feature extraction may be performed on the operation information in advance to obtain the historical deep behavior information.

In a specific embodiment, the object to be recommended may be a recommendable object in an object recommendation platform, and specifically, the object may include, but is not limited to, a short video, an application program, a novel, commodity information, and the like. Specifically, the object attribute may be an object representation obtained by information for describing the object. In a particular embodiment, the information used to describe the object may include, but is not limited to, object identification (e.g., short video identification), object content (e.g., short video title), and the like.

In an optional embodiment, in the process of performing the feature extraction to obtain the initial generalized attribute, the initial personalized attribute, the historical deep behavior information, and the object attribute, the process may include, but is not limited to, combining a one-hot (one-hot) coded network, an N-Gram (chinese language model), and other feature representation networks.

In step S203, liveness analysis is performed on the target user account based on the historical deep behavior information, so as to obtain a liveness index.

In practical application, the historical deep behavior information of the target user account can reflect the activity of the target user account, optionally, the higher the activity is, the greater the influence of the personalized attribute of the user on the favorite interest of the user is, and the smaller the influence of the generalized attribute of the user on the favorite interest of the user is; conversely, the lower the liveness, the smaller the influence of the personalized attributes of the user on the favorite interests of the user, and the larger the influence of the generalized attributes of the user on the favorite interests of the user. In a specific embodiment, the activity index includes a first index and a second index, specifically, the first index may represent a generalization degree of the target user account, and specifically, the first index may be used to control a characterization ratio of an initial generalization attribute of the target user account to a user interest and preference; the second index may represent a degree of personalization of the target user account, and specifically, the second index may be used to control a ratio of an initial personalization attribute of the target user account to a user interest/preference characterization.

In an optional embodiment, the activity analysis performed on the target user account based on the historical deep behavior information to obtain an activity indicator may include:

and inputting the historical deep behavior information into an activity analysis network to carry out activity analysis, thereby obtaining an activity index.

In a particular embodiment, the liveness analysis network may be a deep learning network that is pre-trained for liveness analysis.

In the embodiment, in the process of activity analysis based on historical deep behavior information, the activity analysis network is combined, so that the activity index can be rapidly and accurately acquired, the initial generalization attribute and the initial personalized attribute of the target user account can be conveniently and reasonably controlled subsequently, the user interest and preference characterization proportion is improved, and the characterization accuracy of users with different user attributes is improved.

In an optional embodiment, the activity analysis network may include: an active feature extraction layer and a normalization layer; correspondingly, as shown in fig. 3, the step of inputting the historical deep behavior information into the activity analysis network for activity analysis to obtain the activity index may include the following steps:

in step S301, performing active feature extraction on the historical deep behavior information based on an active feature extraction layer to obtain an initial activity index, where the initial activity index includes a first initial index and a second initial index;

in step S303, normalization processing is performed on the first initial index and the second initial index based on the normalization layer, so as to obtain a first index and a second index.

In an alternative embodiment, the active feature extraction layer may comprise a fully connected layer with an output size of 1 x 2. Specifically, setting the output size of the fully-connected layer to 1 × 2 may ensure that the output is 1 × 2 data including the first index and the second index. In a specific embodiment, the normalization layer may perform normalization processing on the first initial indicator and the second initial indicator by combining with the activation function softmax, so as to obtain the first indicator and the second indicator.

In the embodiment, after the active feature extraction is performed in combination with the historical deep behavior information, the extracted first initial index and the second initial index are normalized, so that the proportion of the initial generalization attribute and the initial personalized attribute of the target user account to the user interest and favorite representation can be controlled more conveniently, and the precision of user representation is improved.

In step S205, a target generalization attribute is determined according to the initial generalization attribute and the first index;

in a specific embodiment, the initial generalization attribute may be multiplied by the first index to obtain the target generalization attribute.

In step S207, determining a target personalized attribute according to the initial personalized attribute and the second index;

in a specific embodiment, the initial personalized attribute and the second index may be multiplied to obtain the target generalized attribute.

In step S209, a target interest index is generated based on the target generalization attribute, the target personalization attribute, and the object attribute;

in a specific embodiment, the target interest indicator may represent an interest condition of the target user account for the object to be recommended; optionally, the target interest indicator may be a numerical value proportional to the degree of interest, or may be a symbolized representation representing the degree of interest of the target user account for the object to be recommended, for example, "medium", and optionally, the word symbol representation may be quantized to a corresponding numerical value in combination with a certain rule.

In an optional embodiment, the generating the target interest indicator based on the target generalization attribute, the target personalized attribute and the object attribute may include:

and inputting the target generalization attribute, the target personalized attribute and the object attribute into an interest identification network for interest identification to obtain a target interest index.

In a specific embodiment, the interest recognition network may be a deep learning network trained in advance for interest recognition.

In the embodiment, in the process of generating the target interest index based on the target generalization attribute, the target personalized attribute and the object attribute, the interest identification of the target user account is performed by combining the interest identification network, so that the target interest index of the target user account can be captured quickly and accurately, and the recommendation accuracy is further improved.

In an alternative embodiment, the interest recognition network comprises: the system comprises a user characteristic extraction layer, an object characteristic extraction layer and an interest perception layer; correspondingly, as shown in fig. 4, the step of inputting the target generalization attribute, the target personalized attribute and the object attribute into the interest recognition network for interest recognition to obtain the target interest indicator may include the following steps:

in step S401, inputting the target generalization attribute and the target personalized attribute into a user feature extraction layer to perform user feature extraction, so as to obtain target user feature information;

in step S403, inputting the object attribute into the object feature extraction layer to perform object feature extraction, so as to obtain target object feature information;

in step S405, the target object feature information and the target user feature information are input to the interest perception layer for interest perception processing, so as to obtain a target interest index.

In an optional embodiment, after the target generalization attribute and the target personalized attribute are spliced, the user feature extraction layer is input to perform user feature extraction, so as to obtain target user feature information.

In an alternative embodiment, the user feature extraction layer and the object feature extraction layer may include, but are not limited to, a feature extraction network in combination with a fully connected layer or the like.

In an optional embodiment, the similarity between the target object feature information and the target user feature information may be calculated in the interest perception layer, and the similarity is normalized to be used as the target interest indicator. Alternatively, the similarity may include, but is not limited to, Euclidean distance, Manhattan distance, cosine distance, and the like. Optionally, the similarity may be normalized by combining with an activation function sigmoid.

In the above embodiment, after the target user characteristic information and the target object characteristic information are respectively extracted by combining the user characteristic extraction layer and the object characteristic extraction layer, the interest perception layer is combined to perform interest perception processing, so that the target interest index of the target user can be rapidly and accurately captured, and the recommendation accuracy is further improved.

In an optional embodiment, in a case that activity analysis needs to be performed in conjunction with the activity analysis network, and interest recognition needs to be performed in conjunction with the interest recognition network, the method may further include: specifically, as shown in fig. 5, the step of generating the activity analysis network and the interest recognition network in advance may include the following steps:

in step S501, an initial sample generalization attribute, an initial sample personalized attribute, historical sample deep behavior information, a sample object attribute of a sample object corresponding to a sample user account, and a labeling interest index of the sample user account for the sample object are obtained;

in step S503, inputting the deep behavior information of the historical sample into an activity analysis network to be trained to perform activity analysis, so as to obtain a sample activity index, where the sample activity index includes a first sample index and a second sample index, the first sample index represents the generalization degree of the sample user account, and the second sample index represents the personalization degree of the sample user account;

in step S505, determining a sample generalization attribute according to the initial sample generalization attribute and the first sample index;

in step S507, determining a sample personalized attribute according to the initial sample personalized attribute and the second sample index;

in step S509, inputting the sample generalization attribute, the sample personalized attribute, and the sample object attribute into the interest recognition network to be trained for interest recognition, so as to obtain a sample interest index;

in step S511, target loss information is determined according to the sample interest index and the annotation interest index;

in step S513, based on the target loss information, the activity analysis network to be trained and the interest recognition network to be trained are trained, and an activity analysis network and an interest recognition network are obtained.

In a specific embodiment, the specific refinements of the initial sample generalization attribute, the initial sample personalized attribute, the historical sample deep behavior information, and the sample object attribute of the sample object corresponding to the sample user account obtained for the sample user account may refer to the specific refinements of the initial generalization attribute, the initial personalized attribute, the historical deep behavior information, and the object attribute of the object to be recommended obtained for the target user account, which is not described herein again.

In an optional embodiment, the labeling interest index of the sample user account for the sample object may represent the probability of the sample user account performing a preset operation on the sample object, specifically, the labeling interest index may be 1 or 0, specifically, the sample user account performs the preset operation on a certain sample object, and the corresponding labeling interest index may be 1; otherwise, the sample user account does not perform a preset operation on a sample object, and the corresponding interest index may be 0.

In a specific embodiment, determining the target loss information according to the sample interest indicator and the annotation interest indicator may include determining the target loss information between the sample interest indicator and the annotation interest indicator based on a preset loss function. Specifically, the target loss information may characterize a difference between the sample interest indicator and the annotation interest indicator.

In a particular embodiment, the predetermined loss function may include, but is not limited to, a cross entropy loss function, a mean square error loss function, a logic loss function, an exponential loss function, and the like.

In a specific embodiment, the training the activity analysis network to be trained and the interest recognition network to be trained based on the target loss information to obtain the activity analysis network and the interest recognition network may include: under the condition that the target loss information does not meet the preset condition, adjusting network parameters of an activity analysis network to be trained and an interest recognition network to be trained based on the target loss information; and repeating the steps S503 and S511 based on the activity analysis network to be trained and the interest recognition network to be trained after the network parameters are adjusted, wherein the corresponding interest recognition network to be trained when the preset condition is met is used as the interest recognition network and the corresponding activity analysis network to be trained when the preset condition is met is used as the activity analysis network under the condition that the target loss information meets the preset condition.

In a specific embodiment, the target loss information meeting the preset condition may be that the target loss information is less than or equal to a specified threshold, or that a difference between corresponding target loss information in two training processes is less than a certain threshold. In the embodiment of the present specification, the specified threshold and a certain threshold may be set in combination with actual training requirements.

In the embodiment, in the activity analysis network training process, activity analysis is performed by combining historical sample deep behavior information of the sample user account, a first sample index capable of representing the generalization degree of the sample user account and a second sample index capable of representing the personalization degree of the sample user account can be rapidly and accurately extracted, the initial sample generalization attribute and the initial sample personalization attribute of the user are reasonably and effectively controlled by combining the first sample index and the second sample index, the proportion of the characterization of the interest and the preference of the user is greatly improved, the characterization accuracy of the user with different user attributes is greatly improved, and the interest identification accuracy of the trained interest identification network can be effectively ensured.

In an alternative embodiment, in a case where activity analysis needs to be performed in combination with an activity analysis network and interest recognition needs to be performed in combination with an interest recognition network, as shown in fig. 6, fig. 6 is a schematic diagram of determining a target interest indicator in combination with an activity analysis network and an interest recognition network according to an exemplary embodiment. Specifically, after historical deep behavior information is input into an activity analysis network for activity analysis, a first index W1 and a second index W2 can be obtained, then, information obtained by multiplying an initial generalized attribute by W1 and information obtained by multiplying an initial personalized attribute by W2 are input into a user feature extraction layer in an interest recognition network for user feature extraction, and target user feature information is obtained; simultaneously, inputting the object attribute into an object feature extraction layer in the interest identification network, and extracting object features to obtain target object feature information; and then, inputting the target user characteristic information and the target object characteristic information into an interest perception layer in the interest recognition network for interest perception processing, so as to obtain a target interest index.

In step S211, a target object of the objects to be recommended is recommended to the target user account based on the target interest indicator.

In an optional embodiment, the recommending, based on the target interest indicator, a target object in the objects to be recommended to the target user account may include:

determining a target object from objects to be recommended according to the target interest index;

and recommending the target object to the account of the target user.

In an optional embodiment, the object to be recommended often includes a large number of objects, specifically, in the process of determining the target object from the object to be recommended according to the target interest indicator, a confidence threshold may be set in advance in combination with a recommendation accuracy requirement (the higher the general confidence threshold is, the more accurate the recommended object is, that is, the more in line with the preference of the user), and accordingly, the object whose value corresponding to the target interest indicator is greater than or equal to the confidence threshold may be used as the target object. Optionally, the target interest indicators may be sorted in a descending order according to the corresponding numerical values, and accordingly, a preset number of objects may be selected as the target objects.

In an optional embodiment, recommending the target object to the target user account may include pushing the target object to a terminal corresponding to the target user account.

In the embodiment, the target object recommended to the target user account is determined by combining the target interest index capable of accurately representing the interest condition of the target user account to the object to be recommended, so that the target object recommended to the target user account can be effectively ensured to accord with the interest preference of the user, and the object recommendation accuracy is further improved.

According to the technical scheme provided by the specification, through activity analysis on the historical deep behavior information of the target user account, the first index capable of representing the generalization degree of the target user account and the second index capable of representing the personalization degree of the target user account can be rapidly and accurately obtained, and the initial generalization attribute and the initial personalization attribute of the target user account are reasonably and effectively controlled in combination with the first index and the second index, so that the proportion of the representation of the interest and the preference of the user is represented, the representation accuracy of the user with different user attributes is greatly improved, the object recommendation accuracy is effectively improved, and the recommendation effect is improved.

FIG. 7 is a block diagram illustrating an object recommendation device according to an example embodiment. Referring to fig. 7, the apparatus includes:

the information acquisition module 710 is configured to perform acquisition of an initial generalization attribute, an initial personalized attribute, historical deep behavior information and an object attribute of an object to be recommended of a target user account;

the first activity analysis module 720 is configured to perform activity analysis on the target user account based on the historical deep behavior information to obtain an activity index, where the activity index includes a first index and a second index, the first index represents the generalization degree of the target user account, and the second index represents the personalization degree of the target user account;

a target generalization attribute determination module 730 configured to perform determining a target generalization attribute based on the initial generalization attribute and the first indicator;

a target personalized attribute determination module 740 configured to perform determining a target personalized attribute from the initial personalized attribute and the second indicator;

a target interest index generation module 750 configured to perform generation of a target interest index based on the target generalization attribute, the target personalization attribute, and the object attribute;

and the object recommending module 760 is configured to recommend a target object in the objects to be recommended to the target user account based on the target interest indicator.

Optionally, the first activity analysis module 720 is further configured to perform activity analysis by inputting the historical deep behavior information into an activity analysis network, so as to obtain an activity index.

the first activity analysis module 720 includes:

the active feature extraction unit is configured to perform active feature extraction on the historical deep behavior information based on an active feature extraction layer to obtain an initial activity index, and the initial activity index comprises a first initial index and a second initial index;

and the normalization processing unit is configured to perform normalization processing on the first initial index and the second initial index based on the normalization layer to obtain a first index and a second index.

Optionally, the target interest index generating module 750 is further configured to perform interest recognition by inputting the target generalization attribute, the target personalized attribute and the object attribute into an interest recognition network, so as to obtain a target interest index.

the target interest index generation module 750 includes:

the user characteristic extraction unit is configured to input the target generalization attribute and the target personalized attribute into a user characteristic extraction layer for user characteristic extraction to obtain target user characteristic information;

the object feature extraction unit is configured to input object attributes into an object feature extraction layer for object feature extraction, and target object feature information is obtained;

and the interest perception processing unit is configured to input the target object characteristic information and the target user characteristic information into the interest perception layer for interest perception processing to obtain a target interest index.

Optionally, the apparatus further comprises:

the training data acquisition module is configured to execute the steps of acquiring the initial sample generalization attribute, the initial sample personalized attribute, the historical sample deep behavior information, the sample object attribute of the sample object corresponding to the sample user account and the labeled interest index of the sample user account on the sample object;

the second liveness analysis module is configured to input the deep behavior information of the historical sample into a liveness analysis network to be trained for liveness analysis to obtain a sample liveness index, the sample liveness index comprises a first sample index and a second sample index, the first sample index represents the generalization degree of the sample user account, and the second sample index represents the personalization degree of the sample user account;

a sample generalization attribute determination module configured to perform determining a sample generalization attribute based on the initial sample generalization attribute and the first sample index;

a target loss information determination module configured to perform determining target loss information according to the sample interest index and the annotation interest index;

and the network training module is configured to train the activity analysis network to be trained and the interest recognition network to be trained based on the target loss information to obtain the activity analysis network and the interest recognition network.

Optionally, the object recommendation module 760 includes:

a target object determination unit configured to perform determining a target object from objects to be recommended according to a target interest index;

With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.

Fig. 8 is a block diagram illustrating an electronic device for object recommendation, which may be a server, according to an exemplary embodiment, and an internal structure thereof may be as shown in fig. 8. 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 equipment comprises 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 an operating system and computer programs in the non-volatile storage medium. The network interface of the electronic device is used for connecting and communicating with an external terminal through a network. The computer program is executed by a processor to implement an object recommendation method.

Fig. 9 is a block diagram illustrating an electronic device for object recommendation, which may be a terminal according to an exemplary embodiment, and an internal structure thereof may be as shown in fig. 9. The electronic device comprises a processor, a memory, a network interface, a display screen and an input device which are connected through a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic equipment comprises 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 an operating system and computer programs in the non-volatile storage medium. The network interface of the electronic device is used for connecting and communicating with an external terminal through a network. The computer program is executed by a processor to implement an object recommendation method. The display screen of the electronic equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the electronic equipment, an external keyboard, a touch pad or a mouse and the like.

It will be understood by those skilled in the art that the configurations shown in fig. 8 or 9 are only block diagrams of some configurations relevant to the present disclosure, and do not constitute a limitation on the electronic device to which the present disclosure is applied, and a particular electronic device may include more or less components than those shown in the figures, or 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, there is also provided a computer-readable storage medium, in which instructions, when executed by a processor of an electronic device, enable the electronic device to perform an object recommendation method in an embodiment of the present disclosure.

In an exemplary embodiment, there is also provided a computer program product containing instructions which, when run on a computer, cause the computer to perform the object recommendation method in the embodiments of the present disclosure.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile 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 DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

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 variations, 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 will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. An object recommendation method, comprising:

2. The object recommendation method of claim 1, wherein performing liveness analysis on the target user account based on the historical deep behavior information to obtain a liveness indicator comprises:

3. The object recommendation method of claim 2, wherein the liveness analysis network comprises: an active feature extraction layer and a normalization layer;

4. The object recommendation method of claim 2, wherein generating a target interest indicator based on the target generalization attribute, the target personalized attribute, and the object attribute comprises:

5. The object recommendation method of claim 4, wherein the interest recognition network comprises: the system comprises a user characteristic extraction layer, an object characteristic extraction layer and an interest perception layer;

6. The object recommendation method according to any one of claims 4 or 5, further comprising:

7. An object recommendation apparatus, comprising:

8. 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 of claims 1 to 6.

9. A computer-readable storage medium, wherein instructions in the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the object recommendation method of any of claims 1-6.

10. A computer program product comprising computer instructions, characterized in that the computer instructions, when executed by a processor, implement the object recommendation method of any one of claims 1 to 6.