CN112054949A

CN112054949A - User information processing method, information pushing method and device and electronic equipment

Info

Publication number: CN112054949A
Application number: CN201910497310.1A
Authority: CN
Inventors: 丁建栋
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2019-06-06
Filing date: 2019-06-06
Publication date: 2020-12-08

Abstract

The embodiment of the invention provides a user information processing method, an information pushing device and electronic equipment. The method for processing the organization information comprises the following steps: acquiring user information of each user in a first user set and a link relation between the users; extracting the characteristics of the user information and the link relation to generate a user view characteristic set and a link relation view characteristic set of the first user set; and inputting the user view feature set and the link relation view feature set into a classifier model corresponding to the first user set for processing so as to determine users belonging to the specified role in the first user set. In the embodiment of the invention, by adopting a multi-view feature extraction mode, the link relation between the user information and the users in the user set is analyzed, the targeted analysis and prediction of various user sets on a network platform are realized, the users with the designated roles are found out, and the targeted service processing is realized.

Description

User information processing method, information pushing method and device and electronic equipment

Technical Field

The application relates to a user information processing method, an information pushing device and electronic equipment, and belongs to the technical field of computers.

Background

With the development of the internet and computer technologies, there are a large number of enterprises or institutions that use services provided by internet platforms, which may include office systems built based on platforms, business systems based on even communication applications, cloud-based architectures, etc., and these enterprises or institutions also form their own organizational architectures in virtual environments. For example, many companies use an OA (Office Automation System) Office platform provided by an internet service provider, an instant messaging application, and the like, and these companies also register as an organization in the platform or application to realize Office Automation.

In some cases, the platform needs to push some information to a specific user in the organization, which requires the platform to be able to identify which users in the organization are the objects to be pushed, for example, the platform needs to push some important notification information or alarm information to the decision maker of the organization.

However, because there are a large number of organizations and users in the organizations do not provide complete organization architecture information to the platform, the platform cannot know the roles of the users in the organizations, thereby causing a certain difficulty in accurately pushing the information. In this case, the interface personnel in the organization can only be pushed first, and then the personnel in the organization can carry out targeted input through the background, which causes great waste of labor and time.

Disclosure of Invention

The embodiment of the invention provides a user information processing method, an information pushing device and electronic equipment, and aims to effectively reach users with designated roles in an organization by information.

In order to achieve the above object, an embodiment of the present invention provides a method for processing user information, including:

acquiring user information of each user in a first user set and a link relation between the users;

extracting the characteristics of the user information and the link relation to generate a user view characteristic set and a link relation view characteristic set of the first user set;

and inputting the user view feature set and the link relation view feature set into a classifier model corresponding to the first user set for processing so as to determine users belonging to the specified role in the first user set.

The embodiment of the invention provides an information pushing method, which comprises the following steps:

acquiring information to be pushed and determining role information of a pushed object;

acquiring user set characteristic data of a target user set, and determining a classifier model according to the user set characteristic data and the role information;

and inputting the characteristic data of the user set into the classifier model for processing so as to determine the user meeting the role information in the target, and pushing the information to be pushed to the user.

An embodiment of the present invention provides a device for processing user information, including:

the first acquisition module is used for acquiring the user information of each user in the first user set and the link relation between the users;

the characteristic extraction module is used for extracting characteristics of the user information and the link relation to generate a user view characteristic set and a link relation view characteristic set of the first user set;

and the role identification module is used for inputting the user view feature set and the link relation view feature set into a classifier model corresponding to the first user set for processing so as to determine users belonging to the specified role in the first user set.

An embodiment of the present invention provides an information pushing apparatus, including:

the second acquisition module is used for acquiring information to be pushed and determining role information of a pushed object;

the third acquisition module is used for acquiring the user set characteristic data of the target user set and determining a classifier model according to the user set characteristic data and the role information;

and the information pushing module is used for inputting the organization characteristic data into the classifier model for processing so as to determine the users in the target user set, which accord with the role information, and pushing the information to be pushed to the users.

The embodiment of the invention provides an information push method in instant messaging, which comprises the following steps:

acquiring user information of each user in an instant address book set of an enterprise;

acquiring the use data of each user according to the user information of each user;

and determining the user corresponding to the information to be pushed based on the user information and the use data of each user.

An embodiment of the present invention provides an electronic device, including:

a memory for storing a program;

and the processor is used for operating the program stored in the memory so as to execute the processing method of the user information.

a memory for storing a program;

and the processor is used for operating the program stored in the memory so as to execute the information pushing method.

a memory for storing a program;

and the processor is used for operating the program stored in the memory so as to execute the information push method in the instant messaging.

The method, the device and the electronic equipment for processing the user information adopt a multi-view characteristic extraction mode to analyze the link relation between the user information and the users in the user set, realize targeted analysis and prediction of various user sets on a network platform, find out the users with the designated roles in the user sets, and further realize targeted service processing. In addition, the information pushing method, the information pushing device and the electronic equipment in the embodiment of the invention determine the pushed role based on the information content, and further carry out targeted analysis and identification on the user set based on the classifier model to find out the user corresponding to the role to be pushed, thereby realizing timely and effective delivery of the information.

The foregoing description is only an overview of the technical solutions of the present invention, and the embodiments of the present invention are described below in order to make the technical means of the present invention more clearly understood and to make the above and other objects, features, and advantages of the present invention more clearly understandable.

Drawings

FIG. 1 is a schematic structural diagram of a multi-view bifurcation learning model according to an embodiment of the present invention;

FIG. 2 is a flowchart illustrating a method for processing user information according to an embodiment of the present invention;

FIG. 3 is a second flowchart illustrating a method for processing user information according to an embodiment of the present invention;

FIG. 4 is a flowchart illustrating an information pushing method according to an embodiment of the present invention;

FIG. 5 is a schematic structural diagram of a user information processing apparatus according to an embodiment of the present invention;

FIG. 6 is a second schematic structural diagram of a user information processing apparatus 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

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

The technical solution of the present invention is further illustrated by some specific examples.

The user set referred to in the embodiment of the present invention may be an organization formed by a plurality of users and established on an internet platform, where the internet platform provides services to an enterprise or an organization, and the services may include an office system constructed based on the platform, an instant messaging application, a business system based on a cloud architecture, and the like, and the organization corresponds to an entity enterprise or an organization in the real world. For example, an organization formed by all members of a company based on an instant messaging application, or an organization formed by an organization on an office service platform provided by the internet, and the organization and the members in the organization use services provided by the office service platform or the instant messaging application.

In some cases, the platform on the network side needs to push some information to a specific user in the organization, which requires the platform to identify which users in the organization are the objects to be pushed, for example, the platform needs to push some important notification information or alarm information to the decision maker of the organization, where the decision maker of the organization may be the users with decision status in the organization, for example, the manager or the core member of the organization.

On some office service platforms, instant messaging applications are also provided to users. On the platforms, an intelligent robot can be operated, and the intelligent robot can realize a collaborative office scene, instantly and accurately push messages or related organization data to users and groups, so as to assist organizations or individuals in making decisions or processing or arousing attention and the like depending on applications. In order to make the intelligent robot reach the user more efficiently and improve the effect of data-driven services, the intelligent robot hopes to push some key data to a specific role in the organization in a targeted manner, and as mentioned above, some information hopes to be pushed directly to a decision maker.

In view of real organizations in the real world, when the office service platform is used, detailed organization architecture information may not be registered on the platform, and therefore, the intelligent robot cannot directly find a required role and push information.

Therefore, the embodiment of the invention provides a technical scheme for intelligently identifying the designated role based on the organization and the related data of the users in the organization, and the technical scheme is applied to information push of a platform.

According to the technical scheme of the embodiment of the invention, based on the application data of organizations and users, a multi-view divergence learning algorithm is adopted to identify each organization, and the designated role is found from the organization.

In the aspect of model architecture, the embodiment of the invention introduces a multi-view bifurcation learning mechanism, adopts the architecture of multi-view and multi-classifier, and predicts the classification result by independently processing the characteristics of the corresponding view by each classifier. In the aspect of feature extraction, user information (static information) of each user in an organization and a link relationship (dynamic information) between the users are emphasized. The user information can be extracted from the registration information disclosed by the user, and the link relationship can be a friend relationship, a message interaction relationship and the like. The embodiment of the invention adopts a Divide-and-conquer (Divide-and-conquer) framework, and the model matched with each organization is used for carrying out independent analysis aiming at each organization so as to find out the designated role in the organization, rather than directly finding out the designated role from massive users, thereby being capable of more accurately identifying the designated role in the organization.

In addition, in view of the fact that the number of tissues on the platform is possibly huge and marked training data is limited, a transfer learning mode can be adopted for the situation, training is performed on the basis of the marked training data to form some existing models, then transfer learning is performed on model parameters, and the model parameters are applied to the models of the tissues with higher similarity, so that the limited training data are effectively utilized, and the predicted cold start problem is optimized by transferring the hyper-parameters of the similar tissues.

Aiming at the similarity between the tissues, the embodiment of the invention adopts a mode of clustering the tissues to judge the similarity between the tissues. In the aspect of feature selection of clustering, the features of the organization can be selected from the following aspects: organization personnel data, group data in the organization, message data in the organization, and data accessed to applications in the organization. By reasonably clustering the similarity between the tissues, the parameter migration between the models can obtain good technical effect.

An exemplary model structure employed in the embodiment of the present invention is described below. Fig. 1 is a schematic structural diagram of a multi-view bifurcation learning model according to an embodiment of the present invention, and as shown in the diagram, the model includes the following parts: the system comprises a plurality of view generators, a feature extraction module, a plurality of classifiers and a result output module. This model is illustrated in this example by taking a decision maker identifying an organization that is registered on a platform of an office-like instant messaging application, where the organization includes a plurality of users using the instant messaging application, and where there are numerous organizations of different sizes and scales. In this example, two view generators and two classifiers corresponding to the two views may be employed, with the various portions of the model being described separately below.

View generator

Two view generators 1 and 2 in the model extract information of a view one and a view two from the acquired organization data, and in this example, the information of the view one may include: the output form of the link relation between users in the instant messaging application can be a bipartite graph, and the left side and the right side of the bipartite graph are members in an organization. For the recognition goal of this example, i.e., finding a decision maker within an organization, the link relationship within the bipartite graph may include two types of information: and the communication link and the approval link of the user in the instant messaging application. The communication link means that a communication interaction relationship based on the instant messaging application exists between users, that is, if a friend relationship is established between the user a and the user B, a communication link exists between the users AB. In order to determine the communication link more accurately, the chat information existing in the instant messaging application may also be used as a standard for determining whether the communication link information exists between the user a and the user B, that is, if the user a and the user B have one-to-one chat records, the communication link exists between the user a and the user B. The examination and approval link means that an administrative examination and approval relationship exists between users, that is, if the process of the user A needs the user B to carry out examination and approval, examination and approval link information from the user A to the user B exists, and examination and approval link information does not exist from the user B to the user A. The information of the view two is mainly user information in an organization, and can comprise personal information which is publicly collected in an instant messaging application, and the output form can be a feature vector. Such information may include public user data such as name, cell phone number, work related information, and the like.

The information contained in the two views is almost completely different, and when the tissue is identified by using the model, the tissue can be comprehensively analyzed from the two different views, so that a better identification effect is realized. In addition, in the training process, different classifiers are cooperatively trained by using the information of the two types of views, so that the trained classifier can obtain a better classification result.

Feature extraction module

In the embodiment of the invention, a model architecture of a plurality of classifiers can be adopted, and the plurality of classifiers are cooperatively trained in a cooperative training mode. Therefore, the wrapping (wrapper) feature method can be used, which is suitable for the co-training model of any number of classifiers, and can provide a feature subset more suitable for the model than the filter feature extraction method. Taking the model structure shown in fig. 1 as an example, the feature extraction module performs feature extraction from the information obtained by the two view generators, respectively, for the following classifiers.

Classifier and collaborative training

In the face of massive organization users, if the model is constructed and trained only based on the labeled data, it is difficult to obtain a satisfactory generalization result on a huge unlabeled data set. In order to use the user data to the maximum extent and improve the generalization performance of the model, the embodiment of the present invention may use a Semi-supervised Learning (Semi-supervised Learning) mode to train the model. Specifically, the collaborative training of the classifier can be performed by a method of the bifurcation learning of the semi-supervised learning. The specific process is as follows:

first, a plurality of classifiers are respectively trained based on labeled data, and taking two classifiers shown in fig. 1 as an example, two classifiers M1 and M2 can be respectively trained by using labeled sample data. The unlabeled sample data is then used for processing by classifiers M1 and M2 to pseudo label the unlabeled samples, respectively. The pseudo label is a prediction label given by a trained model, is not an actual label, and is updated according to a certain constraint to be incorporated into a label sample set. For any classifier, the previous round of newly completing the pseudo-labeled samples will be added as the training samples of the next round to the training sets of other classifiers, such as the samples labeled by the classifier M1 of the previous round, and will be added to the training set of the classifier M2 of the next round for training of the classifier M2 in the new round, and the process is continued until the classifiers M1 and M2 are no longer lifted or reach the agreed maximum training times or no unlabeled samples exist.

In the example shown in fig. 1, the classifiers M1 and M2 respectively learn the feature vectors corresponding to the two views, and the feature extraction module in fig. 1 extracts the feature data corresponding to the two views from the information generated by the two view generators. Considering that the data feature types of the two views are different, the classifier M1 may adopt a Bi-Rank classifier to correspondingly process features in the link relationship between users in the organization, the classifier M2 may adopt a RankSVM classifier to correspondingly process user information of each user in the organization, the two classifiers take the length of each user, finally, when the termination condition is satisfied, a pseudo tag is output in the training process, for the marked organization, the pseudo tag may be verified, and specifically, NDCG (Normalized discrete calibrated cumulative gain) may be selected as a specific verification index.

Through the cooperative training, under the conditions of sufficiency and relative independence, the cooperative training can generate a model for identifying specific roles in an organization with high accuracy.

Result output module

The above-mentioned multiple classifiers will output the classification result made for each user whether belonging to the designated role, and the following strategy can be adopted for the classification results of the multiple classifiers to determine the final output result. In view of recognition efficiency, the output result of any one classifier can be used as the final output result, that is, for a certain user, as long as the classification result output by one classifier is that the user belongs to the designated role, the user is determined to be the designated role. In view of improving the recognition accuracy, the user may be determined as the designated role only when all the classifiers satisfying the preset number output a positive classification result.

It should be noted that, in fig. 1, only two views and two classifiers are taken as examples to illustrate the model architecture of the embodiment of the present invention, and in practical applications, the model architecture can be extended to a plurality of views and a plurality of classifiers based on specific requirements, and the basic principle is the same.

Based on the model architecture and the analysis mode provided by the embodiment of the invention, the targeted analysis and prediction of various organizations on the network platform can be realized, and the users with the designated roles in the organizations can be found, so that the targeted service processing is realized.

Example one

As shown in fig. 2, which is a schematic flow chart of a user information processing method according to an embodiment of the present invention, the method includes:

s101: and acquiring user information of each user in the first user set and the link relation between the users. The first set of users referred to herein may be an organization registered on the platform described above, and may correspond to a business, organization, or group in reality. The first set of users includes a plurality of users using the platform, which may provide services such as instant messaging applications, office automation systems, and e-commerce platforms to the users. Based on the service provided by the platform, the user can make user information on the platform and interact with other users in the organization, such as information of message interaction, transaction processing and the like, the information can be acquired from a database or a system log of the platform, and the interaction relationship forms a link relationship between the users.

S102: and performing feature extraction on the user information and the link relation to generate a user view feature set and a link relation view feature set of the first user set. As described above, the link relationship between the user information in the organization and the users in the organization forms the features of two different views of the first user set, and since the organization includes a plurality of users, the finally extracted features are also in units of users, so that two feature sets are formed.

S103: and inputting the user view feature set and the link relation view feature set into a classifier model corresponding to the first user set for processing so as to determine users belonging to the specified role in the first user set. In the embodiment of the present invention, the main purpose of the classifier model is to identify which user or users in the organization belong to a designated role, where the designated role may be: such as an organization's decision maker, an organization's transaction coordinator, or a critical member of an organization, etc. According to different designated roles, the classifier model can be trained in a targeted manner so as to realize the purpose of identifying different roles.

Further, the classifier model described above may include a first sub-classifier and a second sub-classifier. Wherein, the first sub-classifier and the second sub-classifier can form a divergence learning model, the first sub-classifier performs classification processing based on communication characteristics among users aiming at the link relation view characteristic set, and the second sub-classifier performs classification processing aiming at personal characteristics and/or behavior characteristics of users aiming at the user view characteristic set.

Accordingly, inputting the user view feature set and the link relation view feature set into a classifier for processing, so as to determine users belonging to the designated role in the first user set, may include:

and respectively inputting the user information characteristics and the link relation characteristics into the first sub-classifier and the second sub-classifier for processing, and determining the user of the designated role according to the classification result output by the first sub-classifier and/or the second sub-classifier. The first and second sub-classifiers will output for each user whether the classification result belongs to the designated role, and the following strategy can be adopted for the classification results of the two sub-classifiers to determine the final output result:

for a certain user, when any one of the first sub-classifier and the second sub-classifier identifies the user as the designated role, the user belonging to the designated role to be found can be output. Alternatively, when both the first sub-classifier and the second sub-classifier output that a certain user belongs to a designated role, the user may be determined as the user of the designated role to be found.

Based on the model architecture and the analysis mode provided by the invention, the targeted analysis and prediction can be carried out on various organizations based on a network platform or application, and the users with the designated roles in the organizations can be found out, thereby realizing the targeted service processing. In practical application, some important notification information or alarm information is often pushed to an organization decision maker, and the designated role may be the organization decision maker, that is, the organization decision maker is bound to be reached by the method of the present embodiment and the information is efficiently delivered, and is not indirectly delivered to the decision maker by an interface person in the organization.

Based on the organization processing method provided by the embodiment of the invention, the multi-view feature extraction and classifier processing are respectively carried out on each organization on the platform, so that a user with a designated role in the organization is found out, and the designated role is not directly found out from users in the sea.

Furthermore, labeled training data is limited in view of the large number of tissues on the platform. The embodiment of the invention adopts a transfer learning mode to transfer the parameters of the existing classifier model and applies the parameters to the classifier model of the user set with higher similarity, thereby effectively utilizing limited training data and optimizing the cold start problem predicted by the classifier model by transferring the parameters of the similar user set as the initial parameters of the classifier model. It should be noted that, in the embodiment of the present invention, different classifier models are used for different types of user sets, so that the identification of a designated role in a user set is more targeted. For example, the user set may be divided into user sets of multiple scale levels according to scale, or may be classified according to industries in which the user sets are located, and the like. Different classifier models may be employed for different sized sets of users, respectively. The different classifier models are different in parameters in the models, the structures of the classifier models adopted for tissue recognition can be the same, and the classifier models can be trained independently for different types of user sets, so that the parameters of the classifier models are improved, and the analysis and the processing of the classifier model tissues have pertinence.

For the aspect of judging the similarity of the user set, the judgment can be performed according to the characteristics of the user set personnel data, the group data in the user set, the message data in the user set, the data accessed to the application in the user set and the like, and a specific judgment mode can adopt a mode of clustering the user set. Based on the determination of the similarity of the user set, model parameters of a classifier at the training position based on the marked sample data can be used for reference. For example, a small-scale enterprise corresponds to a small-scale organization, and because fewer users are used, the roles of all members can be marked artificially, so that marked sample data is formed, and a more accurate classifier model is trained. Because the medium-scale organization and the small-scale organization corresponding to the medium-scale enterprise have certain similarity, parameters for the classifier model of the small-scale organization can be used for reference, the parameters are transferred to the classifier model for the medium-scale organization and serve as initial parameters to carry out training or executing role recognition processing, and the parameters of the classifier model are continuously optimized in the subsequent training or role recognition process.

Based on the above technical ideas, as shown in fig. 3, which is a second flowchart of the method for processing organization information according to the embodiment of the present invention, the embodiment of the present invention may further include the following processing of model parameter migration:

s201: a second set of users is determined that satisfies a similarity threshold condition with the first set of users. Wherein, the step can be specifically as follows: performing feature extraction on each tissue in a tissue group where the first user set and the second user set are located to generate tissue feature data, wherein the feature extraction is performed from one or more of the following items of data: organization personnel data, group data in the organization, message data in the organization and data accessed to application in the organization; and clustering each organization in the organization group according to the organization characteristic data, and determining a second user set meeting the similarity threshold condition according to the clustering result.

S202: and acquiring parameters of the classifier corresponding to the second user set, and applying the parameters of the classifier corresponding to the second user set to the classifier model corresponding to the first user set. As described above, the migrated parameters may serve as initial parameters of the classifier model used in step S103 described above.

The method for processing the organization information of the embodiment of the invention adopts a multi-view characteristic extraction mode to analyze the link relation between the user information and the users in the user set, thereby realizing targeted analysis and prediction of various user sets on a network platform and finding out the users with specified roles, thereby realizing targeted service processing.

Example two

Fig. 4 is a schematic flow chart of an information pushing method according to an embodiment of the present invention. In some scenarios, a platform on the network side needs to push some information, referred to herein as information to be pushed, to a user in a specific role in a user set, and what information needs to be pushed to a user in what role may depend on the information itself. For example, the platform may wish to push some important notification or alert information to the decision maker of the user set, and may wish to push some campaign information to the responsible transaction coordinator of the user set, etc. Based on such a requirement, and in combination with the above-described identification technology for a designated role in a user set, the following information pushing method is proposed, which includes:

s301: and acquiring information to be pushed, and determining role information of a pushed object. Some platforms have an intelligent robot system, and the intelligent robot can analyze the role information of the push object according to the content of the information to be pushed, and can also determine the role information of the push object based on a preset rule. For example, when the information to be pushed is alarm information, the push object is determined to be a persona of the user set.

S302: and acquiring user set characteristic data organized by a user set, and determining a classifier model according to the user set characteristic data and the role information. As described in the foregoing, the classifier models constructed in the embodiments of the present invention use a divide-and-conquer approach, and different classifier models are used for different types of user sets and requirements of character recognition, instead of performing direct recognition on massive users. The user set feature data can perform feature extraction on the user set in a multi-view mode, and accordingly, a multi-classifier structure corresponding to multiple views can be adopted in the classifier model. Further, the step may specifically include:

acquiring feature information of a plurality of views in a target user set, wherein the step can be further embodied as acquiring user information of each user in the target user set and link relation information among the users, performing feature extraction, and generating a user view feature set and a link relation view feature set of the target user set;

and determining a plurality of corresponding sub-classifiers in the model according to the characteristic information of the views, wherein the plurality of sub-classifiers are independent from each other and respectively correspond to the views.

In practical application, a platform can pre-train a plurality of classifier models to form a model library, then select a proper classifier model for role recognition according to different push objects and target user sets, and determine a specific push object.

S303: and inputting the characteristic data of the user set into a classifier model for processing so as to determine the user which accords with the role information in the target, and pushing information to be pushed to the user. The specific processing of this step may adopt the processing mechanism of the classifier model described in the foregoing embodiment, and after determining the users in the user set that meet the role information, the automatic pushing may be performed based on the platform.

The information pushing method of the embodiment of the invention determines the pushed role based on the information content, and further carries out targeted analysis and identification on the user set based on the classifier model to find out the user corresponding to the role to be pushed, thereby realizing timely and effective delivery of the information.

EXAMPLE III

As shown in fig. 5, which is a schematic structural diagram of a user information processing apparatus according to an embodiment of the present invention, the apparatus may be disposed on the foregoing internet platform for providing services to enterprises or institutions, and the apparatus includes:

the first obtaining module 11 is configured to obtain user information of each user in the first user set and a link relationship between the users.

And the feature extraction module 12 is configured to perform feature extraction on the user information and the link relationship, and generate a user view feature set and a link relationship view feature set of the first user set. As described above, the link relationship between the user information in the user set and the users in the organization forms the features of two different views of the first user set, and since the user set includes a plurality of users, the finally extracted features are also in units of users, so that two feature sets are formed.

And the role identification module 13 is configured to input the user view feature set and the link relation view feature set into a classifier model corresponding to the first user set, and process the user view feature set and the link relation view feature set to determine users belonging to a specified role in the first user set. The user in the designated role here may be an organizational decision maker.

The classifier model may include a first sub-classifier and a second sub-classifier, and accordingly, inputting the user view feature set and the link relation view feature set into the classifier for processing to determine users belonging to a designated role in the first user set may include: respectively inputting the user information characteristics and the link relation characteristics into a first sub-classifier and a second sub-classifier for processing; and determining the user of the designated role according to the classification result output by the first sub-classifier and/or the second sub-classifier respectively. Further, a first sub-classifier and a second sub-classifier form a bifurcation learning model, the first sub-classifier performs classification processing based on communication characteristics among users aiming at the link relation view characteristic set, and the second sub-classifier performs classification processing aiming at personal characteristics and/or behavior characteristics of the users aiming at the user view characteristic set.

In addition, in order to effectively utilize limited labeled training data and trained classifier models, parameter migration is also performed between the classifier models based on the similarity of the user set. Specifically, the apparatus may further include:

and the organization similarity judging module is used for determining a second user set meeting the similarity threshold condition with the first user set. Specifically, feature extraction may be performed on each organization in a user set group in which the first user set and the second user set are located, so as to generate user set feature data, where the feature extraction is performed from one or any plurality of data in the following aspects: the method comprises the steps of collecting personnel data of a user set, group data in the user set, message data in the user set and application access data in the user set, clustering each user set in a user set group according to user set characteristic data, and determining a second user set meeting a similarity threshold condition according to a clustering result.

And the parameter migration module is used for acquiring the parameters of the classifier corresponding to the second user set and applying the parameters of the classifier corresponding to the second user set to the classifier model corresponding to the first user set. The parameters of the migration may be used as initial parameters of the classifier model.

The detailed description of the above processing procedure, the detailed description of the technical principle, and the detailed analysis of the technical effect are described in the foregoing embodiments, and are not repeated herein.

The processing device of the user information of the embodiment of the invention adopts a multi-view characteristic extraction mode to analyze the link relation between the user information and the users in the user set, thereby realizing targeted analysis and prediction of various user sets on a network platform, finding out the users with specified roles in the user sets, and further realizing targeted service processing.

Example four

As shown in fig. 6, which is a second schematic structural diagram of a user information processing apparatus according to an embodiment of the present invention, the apparatus may be installed on the foregoing internet platform for providing services to enterprises or institutions to perform automatic information push, and the apparatus includes:

the second obtaining module 21 is configured to obtain information to be pushed, and determine role information of a pushed object. Some platforms have an intelligent robot system, and the intelligent robot can analyze the role information of the push object according to the content of the information to be pushed, and can also determine the role information of the push object based on a preset rule. For example, when the information to be pushed is alarm information, the push object is determined as a persona of an organization.

And a third obtaining module 22, configured to obtain organization characteristic data of the target user set, and determine a classifier model according to the user set characteristic data and the role information. As described in the foregoing, the classifier models constructed in the embodiments of the present invention use a divide-and-conquer approach, and different classifier models are used for different types of user sets and requirements of character recognition, instead of performing direct recognition on massive users. The above-mentioned feature data of the user set may perform feature extraction on the tissue in a multi-view manner, and accordingly, a multi-classifier structure corresponding to multiple views may be adopted in the classifier model. Further, the processing of the module may specifically include: acquiring the characteristic information of a plurality of views in the target user set, and then determining a plurality of corresponding sub-classifiers in the model according to the characteristic information of the plurality of views, wherein the plurality of sub-classifiers are independent from each other and respectively correspond to the plurality of views. The obtaining of the feature information of the multiple views in the target user set may further include: and acquiring user information of each user in the target user set and link relation information among the users, extracting features, and generating a user view feature set and a link relation view feature set of the target user set.

And the information pushing module 23 is configured to input the feature data of the user set into the classifier model for processing, so as to determine a user in the target user set, which meets the role information, and push information to be pushed to the user. In this module, the processing mechanism of the classifier model described in the foregoing embodiment may be adopted, and after determining users in the organization that meet the role information, the automatic pushing may be performed based on the platform.

The information pushing device of the embodiment of the invention determines the pushed role based on the information content, and further performs targeted analysis and identification on the user set based on the classifier model to find out the user corresponding to the role to be pushed, thereby realizing timely and effective delivery of the information.

EXAMPLE five

The embodiment provides an information push method in instant messaging, aiming at the instant messaging of enterprises, some enterprises can be registered as group users or organization users on instant messaging application, and the instant messaging is used as an office communication tool, an OA office system and the like. Based on such a scenario, the method provided by the present embodiment can find out a specific user as a message pushing object based on the user information and daily usage data of the enterprise user, thereby achieving accurate delivery of information. The information pushing method in the instant messaging comprises the following steps:

s401: and acquiring user information of each user in the instant address book set of the enterprise. The user information referred to herein may be registration information of the user, such as user name, post, phone call, address, and the like.

S402: and acquiring the use data of each user according to the user information of each user. The usage data herein may include: chat records of users, approval relationships between users, project authoring relationships between users, and the like.

S403: and determining the user corresponding to the information to be pushed based on the user information and the use data of each user. Wherein, the step may specifically include:

s4031: and acquiring user view characteristic data and link relation view characteristic data according to the use data and the user information of each user. The user view feature data may include a feature set formed by a plurality of user view features for each extracted user view feature as the user view feature data. The link relation view characteristic data may also include link relations between a plurality of users, and a set of these link relations constitutes the link relation view characteristic data.

S4032: and determining a user corresponding to the information to be pushed according to the user view characteristic data, the link relation view characteristic data and the content of the push message. In the process of determining a specific user, the classifier model mentioned in the foregoing embodiment may be used, and based on the classifier model, the role of the user is determined, so that the user corresponding to the information to be pushed is selected.

EXAMPLE six

The foregoing embodiment describes the foregoing flow processing and device structure, and the functions of the foregoing method and device can be implemented by an electronic device, as shown in fig. 7, which is a schematic structural diagram of the electronic device according to an embodiment of the present invention, and specifically includes: a memory 110 and a processor 120.

And a memory 110 for storing a program.

In addition to the programs described above, the memory 110 may also be configured to store other various data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on the electronic device, contact data, phonebook data, messages, pictures, videos, and so forth.

The memory 110 may be implemented by any type or combination of volatile or non-volatile memory devices, such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disks.

The processor 120, coupled to the memory 110, is configured to execute the program in the memory 110, so as to perform the operation steps of the method for processing organization information and the method for pushing information described in the foregoing embodiments and the method for pushing information in instant messaging.

Further, the processor 120 may also include various modules described in the foregoing embodiments to perform parameter determination operations, and the memory 110 may be used, for example, to store data required by the modules to perform the operations and/or output data.

Further, as shown, the electronic device may further include: communication components 130, power components 140, audio components 150, display 160, and other components. Only some of the components are schematically shown in the figure and it is not meant that the electronic device comprises only the components shown in the figure.

The communication component 130 is configured to facilitate wired or wireless communication between the electronic device and other devices. The electronic device may access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 130 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 130 further includes a Near Field Communication (NFC) module to facilitate short-range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

The power supply component 140 provides power to the various components of the electronic device. The power components 140 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for an electronic device.

The audio component 150 is configured to output and/or input audio signals. For example, the audio component 150 includes a Microphone (MIC) configured to receive external audio signals when the electronic device is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal may further be stored in the memory 110 or transmitted via the communication component 130. In some embodiments, audio assembly 150 also includes a speaker for outputting audio signals.

The display 160 includes a screen, which may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

Those of ordinary skill in the art will understand that: all or a portion of the steps of implementing the above-described method embodiments may be performed by hardware associated with program instructions. The program may be stored in a computer-readable storage medium. When executed, the program performs steps comprising the method embodiments described above; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A user information processing method comprises the following steps:

2. The method of claim 1, wherein the classifier model includes a first sub-classifier and a second sub-classifier,

inputting the user view feature set and the link relation view feature set into a classifier for processing, so as to determine users belonging to a specified role in the first user set, wherein the step of:

inputting the user information characteristics and the link relation characteristics into a first sub-classifier and a second sub-classifier respectively for processing;

and determining the user of the designated role according to the classification result output by the first sub-classifier and/or the second sub-classifier respectively.

3. The method according to claim 2, wherein the first and second sub-classifiers form a bifurcation learning model, the first sub-classifier performs a classification process based on inter-user communication features on the set of link relationship view features, and the second sub-classifier performs a classification process for user personal and/or behavioral features on the set of user view features.

4. The method of claim 1, further comprising:

determining a second set of users satisfying a similarity threshold condition with the first set of users;

and acquiring parameters of a classifier corresponding to the second user set, and applying the parameters of the classifier corresponding to the second user set to a classifier model corresponding to the first user set.

5. The method of claim 4, wherein determining a second set of users that satisfies a similarity threshold condition with the first set of users comprises:

performing feature extraction on each user set in a user set group where the first user set and the second user set are located to generate user set feature data, wherein the feature extraction is performed from one or more of the following data: the method comprises the following steps that personnel data in a user set, group data in the user set, message data in the user set and data of access application in the user set are collected;

and clustering each user set in the user set group according to the user set characteristic data, and determining whether the first user set and the second user set meet the similarity threshold condition according to the clustering result.

6. The method of claim 1, wherein the determining users of the first set of users belonging to a specified role comprises determining a decision maker of the first set of users.

7. An information push method, comprising:

8. The method of claim 7, wherein obtaining user set feature data for a target user set and determining a classifier model based on the user set feature data and the role information comprises:

acquiring characteristic information of a plurality of views in a target user set;

and determining a plurality of corresponding sub-classifiers in the model according to the characteristic information of the plurality of views, wherein the plurality of sub-classifiers are independent from each other and respectively correspond to the plurality of views.

9. The method of claim 8, wherein obtaining feature information for a plurality of views in a set of target users comprises: and acquiring user information of each user in the target user set and link relation information among the users, extracting features, and generating a user view feature set and a link relation view feature set of the target user set.

10. The method of claim 7, wherein the role information is a decision maker.

11. An apparatus for processing user information, comprising:

12. The apparatus of claim 11, wherein the classifier model includes a first sub-classifier and a second sub-classifier,

13. The apparatus of claim 11, further comprising:

the user set similarity judging module is used for determining a second user set meeting the similarity threshold condition with the first user set;

and the parameter migration module is used for acquiring parameters of the classifier corresponding to the second user set and applying the parameters of the classifier corresponding to the second user set to the classifier model corresponding to the first user set.

14. An information pushing apparatus comprising:

and the information pushing module is used for inputting the characteristic data of the information pushing module into the classifier model for processing so as to determine the users in the target user set, which accord with the role information, and pushing the information to be pushed to the users.

15. An information push method in instant messaging comprises the following steps:

16. The method of claim 15, wherein determining the user corresponding to the information to be pushed based on the usage data of each user comprises:

acquiring user view characteristic data and link relation view characteristic data according to the use data and the user information of each user;

and determining a user corresponding to the information to be pushed according to the user view characteristic data, the link relation view characteristic data and the content of the push message.

17. An electronic device, comprising:

a memory for storing a program;

a processor for executing the program stored in the memory to perform the user information processing method of any one of claims 1 to 6.

18. An electronic device, comprising:

a memory for storing a program;

a processor for executing the program stored in the memory to execute the information pushing method according to any one of claims 7 to 10.

19. An electronic device, comprising:

a memory for storing a program;

a processor configured to execute the program stored in the memory to execute the information pushing method according to claim 15 or 16.