CN113934941A

CN113934941A - User recommendation system and method based on multi-dimensional information

Info

Publication number: CN113934941A
Application number: CN202111189867.2A
Authority: CN
Inventors: 韩康; 窦铮; 陈正超; 段凌云
Original assignee: Beijing Langma Shulian Technology Co ltd
Current assignee: Beijing Langma Shulian Technology Co ltd
Priority date: 2021-10-12
Filing date: 2021-10-12
Publication date: 2022-01-14

Abstract

The application provides a user recommendation system and method based on multi-dimensional information, user data of a target user is collected and sorted through a first class user library module, UGC text content data in the first class user library module is obtained through a representation information determining module, and representation social information of the target user is determined based on the UGC text content data; determining a user eigen model of the target user based on the representation social information of the target user by using an eigen sampling identification module; and further determining the multi-dimensional social attributes and the social tendency of the target user by using a target user multi-dimensional attribute module based on the user data and the user eigen model of the target user. The first computing module may determine multi-dimensional social attributes and social tendencies of the users to be matched. The matching value calculation module can determine the multi-dimensional matching values of the target user and the user to be matched, and further determine a recommended user list matched with the target user. Therefore, the matching precision and the user experience of the user can be improved.

Description

User recommendation system and method based on multi-dimensional information

Technical Field

The application relates to the technical field of user matching, in particular to a user recommendation system and method based on multi-dimensional information.

Background

In the social network, each user can exchange information with each other, so that the users can communicate with each other on the network. By the end of 2015, the number of Chinese netizens reaches 6.88 hundred million, and the market share of the social network occupies the vast majority. With the increase of users in the social network market, in order to prevent the loss of users, increase the viscosity of users using social products, expand market share, and so on, these social network service platforms or social software will usually introduce a corresponding user recommendation matching function in the social network service platforms or social software, that is, a function of recommending one or part of user groups to users who may be interested in the user recommendation matching function.

The user recommendation matching method of the existing social network service platform or social software is often single, generally, the user recommendation matching method can be carried out according to partial user information filled by the user or the user relationship of the user in the social network service platform or the social software, and the method cannot meet complicated and variable human social requirements.

Firstly, as for user information filled by a user, many inaccurate information exists, the information is too single, and most of the inaccurate information is information such as birthday, blood type, hometown, company, school and the like filled by the user as a resume. For example, user a and user B both have friends of user C, the hometown of user a and user B is the same, and the filled favorite leisure content such as movies and sports are substantially the same, but in reality, user a may have a negative attitude to user B, and this attitude is transparent and uncertain for the social network service platform or social software. Even though both the user a and the user B enjoy movies and sports, from the perspective of deep level multi-dimensionality, the user a likes a horror-type movie, and the user B likes a comedy-type movie. For example, in order to explain that the information filled in by the user cannot be deeply and multi-dimensionally subdivided, the considered social factor is too single, and at this time, if the user a is recommended to the user B by performing recommendation matching using the user relationship and the user information filled in by the user, the social result is often failed.

Therefore, when the user needs are diversified, the single crude user recommendation matching method cannot effectively find true potential friends in a social network service platform or social software for the user.

Disclosure of Invention

An object of the embodiments of the present application is to provide a user recommendation system and method based on multidimensional information, so as to effectively discover real potential friends in a social network service platform or social software for a user, and improve the accuracy of user matching.

In order to achieve the above object, embodiments of the present application are implemented as follows:

in a first aspect, an embodiment of the present application provides a user recommendation system based on multidimensional information, including: the first-class user library module is used for collecting and sorting user data of a target user, wherein the user data of the target user comprises basic information of the target user and UGC text content data determined based on UGC behaviors of the target user; the representation information determining module is used for acquiring UGC text content data in the first-class user library module and determining representation social information of the target user based on the UGC text content data, wherein the representation represents the objective reflection of the psychological activity of the user in the user recommendation system on external things behaviors, the psychological activity of the user can be analyzed by recording the objective reflection, and the representation social information comprises words reflecting the psychological activity of the target user; the intrinsic sampling identification module is used for determining a user intrinsic model of the target user based on the representation social information of the target user, wherein the intrinsic represents the true social tendency of the user, all representation social information of the user in the user recommendation system is analyzed and mapped into the existing intrinsic model, and the user intrinsic model is set based on the personality theory in psychology and can be described and reflected through words; the target user multi-dimensional attribute module is used for determining the multi-dimensional social attributes and social tendencies of the target user based on the user data of the target user and the user eigen model; the first calculation module is used for acquiring user data and a user eigen model of a user to be matched and determining the multi-dimensional social attributes and social tendencies of the user to be matched; and the matching value calculation module is used for determining a multi-dimensional matching value of the target user and the user to be matched based on the multi-dimensional social attribute and the social tendency of the target user and the multi-dimensional social attribute and the social tendency of the user to be matched, and determining a recommended user list matched with the target user based on the multi-dimensional matching value.

In the embodiment of the application, user data (including basic information of a target user and UGC text content data determined based on UGC behavior of the target user) of the target user is collected and sorted through a first-class user library module, UGC text content data in the first-class user library module is obtained through a representation information determination module, and representation social information (including words reflecting psychological activities of the target user) of the target user is determined based on the UGC text content data; determining a user eigen model of the target user by using an eigen sampling recognition module based on the representative social information of the target user (based on the personality theory setting in psychology, the user eigen model can be described and reflected through vocabularies); and further determining the multi-dimensional social attributes and the social tendency of the target user by using a target user multi-dimensional attribute module based on the user data and the user eigen model of the target user. The first calculation module can acquire user data and a user eigen model of the user to be matched, and multi-dimensional social attributes and social tendencies of the user to be matched are determined. Therefore, the matching value calculation module can determine the multidimensional matching values of the target user and the user to be matched based on the multidimensional social attributes and social tendencies of the target user and the multidimensional social attributes and social tendencies of the user to be matched, and determine the recommended user list matched with the target user based on the multidimensional matching values. In such a mode, UGC (user original content) text content data of a user can be used as a basis for determining the representation social information and the user eigen model of the user, and the UGC text content data can usually effectively reflect the psychological activities of the user, so that the eigen model (which personality the user belongs to) is reflected laterally, psychology and internet technology are combined, the personality theory in psychology is applied to a social network and used as a reference when the user is matched, and the matching precision of the user can be effectively improved. The basic information, UGC text content data, representation social information and the user eigen model of the user are used as indexes for determining the multi-dimensional social attributes and social tendency of the target user, information dimensions in the user matching process can be enriched, and the similarity of the user is considered from more dimensions, so that the success rate of user matching is improved, real potential friends in a social network service platform or social software are effectively discovered for the user, and the matching precision and the user experience of the user are improved.

With reference to the first aspect, in a first possible implementation manner of the first aspect, the user data includes basic information and behavior information, and the first class user library module includes: the user basic information main attribute unit is used for receiving and updating basic main attribute data of a target user and classifying and weighting the data, wherein the basic main attribute data comprises one or more of a nickname, a UID, a gender, a sexual orientation, registration time and registration duration; the user basic information secondary attribute unit is used for receiving and updating basic secondary attribute data of a target user and classifying and weighting the data, wherein the basic secondary attribute data comprises one or more of age, birthday, hometown, resident geographic position, academic experience, political tendency and religion; the first calculation unit is used for calculating the weight values of the basic primary attribute data and the basic secondary attribute data of the target user and outputting the weight values to the user basic information module in a classified manner; the user basic information module is used for sorting and sorting the data input by the first computing unit, wherein the basic information comprises the basic primary attribute data and the corresponding weight, and the basic secondary attribute data and the corresponding weight; the user UGC behavior unit is used for receiving and updating UGC text content data of the target user and classifying and empowering the data, wherein the UGC text content data comprises published content text data, published content affiliated fields and published content browsing indexes, and the published content browsing indexes comprise one or more of browsing amount, browsing efficiency, reply amount and collection amount of the published content text data; the user social behavior unit is used for receiving and updating the social behavior data of the target user and classifying and empowering the data, wherein the social behavior data comprises a plurality of items of page access volume, user reply behavior, reply content quality, attention quantity, social attribute dimensionality of the attention user, fan quantity, social attribute dimensionality of fan, social attribute dimensionality of friend application, chat duration with friends, chat times with friends, whether the user actively initiates chat, whether the user actively adds friends and friend quantity; the basic behavior unit is used for receiving and updating basic behavior data of the target user and classifying and empowering the data, wherein the basic behavior data comprises multiple items of LBS geographic tracks, online UGC active time periods, behavior publishing time periods, behavior liking time periods, online social activity time periods, friend chat time periods, behavior replying time periods, friend adding time periods, behavior attention time periods, online reading time periods, content reading time periods, online inertia time periods, first-time-of-day starting time periods and quitting time periods every day; the second calculation unit is used for calculating UGC text content data, social behavior data and basic behavior data of the target user and outputting the UGC text content data, the social behavior data and the basic behavior data to the user behavior module in a classified mode; and the user behavior module is used for sequencing and sorting the data input by the second computing unit, wherein the behavior information comprises the UGC text content data and the corresponding weight thereof, the social behavior data and the corresponding weight thereof, and the basic behavior data and the corresponding weight thereof.

In the implementation mode, the first-class user library module is divided into a user basic information module and a user behavior module, and the user basic information module is subdivided into a user basic information primary attribute unit (nickname, UID, gender, sexual orientation, registration time, registration duration and the like) and a user basic information secondary attribute unit (age, birthday, hometown, resident geographic position, academic experience, political tendency, religion and the like); the user behavior module is subdivided into a user UGC behavior unit (published content text data, published content belonging field and published content browsing index, the published content browsing index comprises browsing amount, browsing efficiency, reply amount, collection amount and the like of the published content text data), a user social behavior unit (webpage access amount, user reply behavior, reply content quality, attention amount, social attribute dimension of the attention user, vermicelli amount, social attribute dimension of vermicelli, social attribute dimension of friend application, chatting duration with friends, chatting times with friends, whether the user actively initiates chatting, whether the user actively adds friends and the like) and a user basic behavior unit (LBS geographical trajectory, online UGC active time period, behavior publishing time period, like behavior time period, online social active time period, friend chatting time period, etc.), a user basic behavior unit (LBS geographical trajectory, online UGC active time period, behavior publishing time period, like behavior time period, online social active time period, friend chatting time period, etc.) A time period for replying to behavior, a time period for adding friends, a time period for paying attention to behavior, an online reading time period, a time period for reading content, an online inertia time period, a daily first start time period, a daily quit time period, etc.). Therefore, the multi-dimensional social attributes of the users can be effectively enriched, and the users which are closer in more dimensions (closer in geographic positions, more consistency in the eigen mode, the representation social information, the subdivision categories and the like of the users) can be matched, so that three essential elements of favorite emotion among people in psychology, namely attraction, proximity and similarity are met. By combining the three elements, the method can be applied to the accurate matching of the users in the social network service or the social software through computer science, so that the success rate of the matching of the users is effectively improved, and the user experience is improved.

With reference to the first aspect, in a second possible implementation manner of the first aspect, the user recommendation system further includes a second-class user library module, where the second-class user library module includes: the interest category unit is used for counting, classifying and collecting interest category data of the target user and clustering and weighting the data, wherein the interest category data are data results selected by the target user based on interest categories built in the user recommendation system or interest categories generated immediately; the vertical field unit is used for counting, classifying and collecting vertical field data of the target user and clustering and weighting the data, wherein the vertical field data are data results selected by the target user based on a preset library provided by the user recommendation system or a vertical field library provided by a third party and are used for representing a single vertical field in user interest; and the third calculating unit is used for calculating the weights of the interest category data and the vertical field data and classifying the user data types.

In the implementation mode, the second-class user library module is divided into an interest category unit and a vertical field unit, the interest category unit can count, classify and collect interest category data of the target user (a data result of the target user selected based on an interest category built in a user recommendation system or an interest category generated immediately), and cluster and weight the data; the vertical field unit can count, classify and collect vertical field data of a target user (a data result selected by the target user based on a preset library provided by a user recommendation system or a vertical field library provided by a third party is used for representing a single vertical field in user interest), and cluster and weight the data. The user database module can not only collect the interest field of the user, but also collect the single vertical interest field of the user, and the second type user database module can effectively reflect the interest of the user. According to the theory of cognitive disorder, people are more suitable for other people with the same idea, the same hobbies and the same situation, people can easily generate sense of identity, satisfaction and dependence under the environment and the condition, and the social network service which is more in line with the requirements of the users can be provided for the users, so that the diversified requirements of the users are met, the illusion of transparency existing in the social network service can be eliminated, and the social network service is more suitable for the requirements of the times and the requirements of the social network users.

With reference to the first aspect, in a third possible implementation manner of the first aspect, the characterization information determining module is specifically configured to: performing preliminary word segmentation on the UGC text content data, wherein the word segmentation method comprises at least one of forward maximum matching, reverse maximum matching and bidirectional maximum matching; carrying out weight distribution on UGC text content data after word segmentation; extracting core words of each sentence in UGC text content data through the core words, marking the core words, determining the emotion expressed by the target user in the UGC text content data based on a semantic probability model, classifying the UGC text content data of the target user, and determining the representative social information of the target user based on all the classified UGC text content data.

In this implementation, the characterization information determination module may perform preliminary word segmentation on the UGC text content data; carrying out weight distribution on UGC text content data after word segmentation; extracting core words of each sentence in UGC text content data through the core words, marking the core words, determining the emotion expressed by the target user in the UGC text content data based on a semantic probability model, classifying the UGC text content data of the target user, and determining the representation social information of the target user based on all the classified UGC text content data. By the method, the emotion of the user can be effectively predicted from the UGC text content data of the target user, so that the UGC text content data of the user can be classified, and the accuracy of mapping the UGC text content data to the eigenmode of the user is facilitated.

With reference to the first aspect, in a fourth possible implementation manner of the first aspect, the eigen-model of the user is set based on the five personality theories, each eigen-model is associated with a plurality of words describing and accommodating the eigen-model, and the eigen-sampling recognition module is specifically configured to: classifying the representative social information of the target user, mapping the representative social information to each eigenmode shape, and determining the contact ratio of the representative social information of the target user in each eigenmode shape; and determining the eigenmode shape with the highest contact ratio and the difference reaching the standard between the contact ratio of the eigenmode shape and the contact ratio of other eigenmode shapes as the user eigenmode of the target user based on the contact ratio corresponding to each eigenmode shape.

In the implementation mode, the user eigen model is set based on the five-personality theory, the representation social information of the target user is classified and mapped to each eigen model, the user eigen model with the highest contact ratio of the representation social information of the target user in the eigen model is determined, the personality type of the user can be effectively and accurately reflected, and therefore accurate matching of the user is facilitated.

With reference to the first aspect, in a fifth possible implementation manner of the first aspect, the matching value calculating module is specifically configured to: for each subdivision index in the multidimensional social attribute and the social tendency of the target user, performing cosine similarity calculation with the multidimensional social attribute and the corresponding subdivision index in the social tendency of the user to be matched to obtain a multidimensional matching value between the target user and the user to be matched; and determining the user with the multi-dimensional matching value reaching the set matching value as a user to be recommended, and sequencing the user to be recommended to obtain a recommended user list matched with the target user.

In the implementation mode, the cosine similarity is utilized to calculate the closeness between each subdivision index in the multidimensional social attribute and the social tendency of the target user and the corresponding subdivision index in the multidimensional social attribute and the social tendency of the user to be matched.

With reference to the first aspect, in a sixth possible implementation manner of the first aspect, the user recommendation system further includes a second calculation module and a verification module, where the second calculation module is configured to perform preliminary screening on the multidimensional social attributes and social tendencies of the user to be matched before the matching value calculation module determines the multidimensional matching values of the target user and the user to be matched, so as to determine the user to be matched that meets a preliminary screening condition; the verification module is used for determining whether a user to be rejected, the social attribute difference of which exceeds the set difference, exists between the preliminarily screened user to be matched and the target user after the user to be matched meeting the preliminary screening condition is determined by the second calculation module, and rejecting the user to be rejected from the user to be matched to obtain the screened user to be matched; correspondingly, the matching value calculation module is used for determining the multidimensional matching value of the target user and the screened user to be matched.

In the implementation mode, the second computing module can perform preliminary screening on the multi-dimensional social attributes and social tendency of the users to be matched to determine the users to be matched which meet the preliminary screening conditions; and then, rejecting the users to be rejected, of which the social attribute difference with the target user exceeds the set difference, from the preliminarily screened users to be matched by using the checking module, so that the users can be effectively screened, users with high matching possibility with the target user are screened from other user groups, and the multidimensional matching value is calculated, thereby effectively reducing the calculated amount and improving the operating efficiency of the user recommendation system.

With reference to the first aspect, in a seventh possible implementation manner of the first aspect, the user recommendation system further includes a third calculation module and a third class user library module, where the third calculation module is configured to, after the matching value calculation module determines a recommended user list matching the target user, collect subsequent matching data between the target user and the recommended users in the recommended user list, and classify and weight matching results based on the subsequent matching data; and the third-class user library module is used for classifying the subsequent matching data and the corresponding classification and weight value collected in the third calculation module into the social tendency of the target user so as to correct the social tendency of the target user.

In this implementation manner, the third calculation module may collect subsequent matching data between the target user and the recommended users in the recommended user list, and classify and weight the matching result based on the subsequent matching data, and the third class user library module may classify the subsequent matching data and the corresponding classification and weight collected in the third calculation module into the social tendency of the target user to correct the social tendency of the target user. And the data in the third-class user library module can also be used as a training set to learn and correct a subsequent matching mechanism, so that the diversified requirements of user matching can be flexibly met.

In a second aspect, an embodiment of the present application provides a method for recommending a user based on multidimensional information, which is applied to the system for recommending a user based on multidimensional information described in any one of the first aspect or possible implementation manners of the first aspect, where the method includes: acquiring user data of a target user, wherein the user data of the target user comprises basic information of the target user and UGC text content data determined based on UGC behaviors of the target user; determining representative social information of the target user based on the UGC text content data, wherein the representative social information represents an objective reflection of the user's psychological activity in the user recommendation system on external things behaviors, the user's psychological activity can be analyzed by recording the objective reflection, and the representative social information comprises words reflecting the target user's psychological activity; determining a user eigen model of the target user based on the representation social information of the target user, wherein the eigen represents the true social tendency of the user, all representation social information of the user in the user recommendation system is analyzed and mapped into the existing eigen model, and the user eigen model is set based on the five-personality theory and can be described and reflected through words; determining multi-dimensional social attributes and social tendencies of the target user based on the user data of the target user and the user eigen model; acquiring user data and a user eigen model of a user to be matched, and determining the multi-dimensional social attributes and social tendency of the user to be matched; and determining a multi-dimensional matching value of the target user and the user to be matched based on the multi-dimensional social attributes and social tendency of the target user and the multi-dimensional social attributes and social tendency of the user to be matched, and determining a recommended user list matched with the target user based on the multi-dimensional matching value.

In a third aspect, an embodiment of the present application provides a method for recommending a user based on multidimensional information, which is applied to the system for recommending a user based on multidimensional information described in any one of the first aspect or possible implementation manners of the first aspect, where the method includes: if the target user uses the user recommendation system for the first time, determining the multi-dimensional social attributes and social tendency of the target user, classifying the target user and storing the target user in a first database; screening users in a second database by utilizing the multidimensional social attributes and social tendency of the target user, screening out users to be matched which are consistent with the multidimensional social attributes and social tendency of the target user, and storing the users to be matched into a third database, wherein the second database stores data information of other users, and the data information comprises the multidimensional social attributes and social tendency of the users; carrying out multidimensional matching on the multidimensional social attributes and social tendencies of the target user in the first database and the multidimensional social attributes and social tendencies of each user to be matched in the third database, determining multidimensional matching values of the target user and the users to be matched, and determining a recommended user list matched with the target user based on the multidimensional matching values; and acquiring subsequent matching data between each recommended user in the recommended user list and the target user, classifying and weighting matching results based on the subsequent matching data, and storing the matching results into a fourth database, wherein the subsequent matching data corresponding to each recommended user, the classification and the weighting thereof can be classified into the social tendency of the target user so as to correct the social tendency of the target user.

In a fourth aspect, an embodiment of the present application provides a storage medium, where the storage medium includes a stored program, and when the program runs, a device on which the storage medium is located is controlled to execute the multidimensional information based user recommendation method of the second aspect or the third aspect.

In a fifth aspect, an embodiment of the present application provides an electronic device, including a memory and a processor, where the memory is configured to store information including program instructions, and the processor is configured to control execution of the program instructions, where the program instructions are loaded and executed by the processor to implement the multi-dimensional information-based user recommendation method according to the second aspect or the third aspect.

In order to make the aforementioned objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments of the present application will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and that those skilled in the art can also obtain other related drawings based on the drawings without inventive efforts.

Fig. 1 is a schematic diagram of a user recommendation system based on multi-dimensional information according to an embodiment of the present application.

Fig. 2 is a schematic diagram of another user recommendation system provided in an embodiment of the present application.

Fig. 3 is a schematic diagram of a social user heterogeneous graph model according to an embodiment of the present disclosure.

Fig. 4 is a schematic diagram of a function of selecting an interest content item according to an embodiment of the present application.

Fig. 5 is a schematic diagram of a function of displaying a user information content item according to an embodiment of the present application.

Fig. 6 is a schematic diagram of a selection function of a vertical field of interest according to an embodiment of the present application.

Fig. 7 is a flowchart of a first user recommendation method based on multi-dimensional information according to an embodiment of the present application.

Fig. 8 is a flowchart of a second user recommendation method based on multi-dimensional information according to an embodiment of the present application.

Fig. 9 is a block diagram of an electronic device according to an embodiment of the present application.

Icon: 1000-user recommendation system; 1100-first class user library module; 1110-user basic information module; 1111-user basic information main attribute unit; 1112-user basic information secondary attribute unit; 1113-first computing unit; 1120-user behavior module; 1121-user UGC behavior unit; 1122-user social behavior unit; 1123-user basic behavior unit; 1124-a second computing unit; 1200-a characterization information determination module; 1300-intrinsic sample identification module; 1400-target user multidimensional attribute module; 1500-a first calculation module; 1600-matching value calculation module; 1700-second class user library module; 1710-interest category unit; 1720-vertical field unit; 1730-user other category behavior units; 1740-a third calculation unit; 1810-a second calculation module; 1820-a check module; 1910-a third calculation module; 1920-third class user library module; 1930-memory module; 1940-cache module; 1950-result output module.

Detailed Description

The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.

Referring to fig. 1, fig. 1 is a schematic diagram of a user recommendation system 1000 based on multi-dimensional information according to an embodiment of the present disclosure.

In this embodiment, the multi-dimensional information based user recommendation system 1000 may include: the system comprises a first user library module 1100, a characterization information determining module 1200, an intrinsic sampling identification module 1300, a target user multi-dimensional attribute module 1400, a first calculating module 1500, a matching value calculating module 1600 and the like.

It should be noted that the user recommendation system 1000 described in fig. 1 is only a simple system structure. In this embodiment, a user recommendation system 1000 with a more complex structure based on the user recommendation system 1000 will be further described (as shown in fig. 2, fig. 2 shows a schematic diagram of another more complex user recommendation system 1000). Therefore, the structure of the user recommendation system 1000 herein should not be construed as limiting the present application.

In this embodiment, the first-class User library module 1100 may be configured to collect and sort User data of a target User, where the User data of the target User includes basic information of the target User and UGC text Content data determined based on a UGC (User Generated Content) behavior of the target User.

For example, to facilitate understanding of information components of user data in the present embodiment, please refer to fig. 3, and fig. 3 is a schematic diagram of a social user heterogeneous graph model according to an embodiment of the present application.

In this embodiment, the first-class user library module 1100 may include a user basic information module 1110 and a user behavior module 1120, which take social users (taking target users as an example) as a core.

Illustratively, the user basic information module 1110 may include a user basic information primary attribute unit 1111 and a user basic information secondary attribute unit 1112.

The user basic information main attribute unit 1111 is configured to receive and update the basic main attribute data of the target user, and perform classification and weighting on the data. The basic main attribute data includes one or more of nickname, UID, gender, sex orientation, registration time, and registration time (in this embodiment, the nickname, UID, gender, sex orientation, registration time, and registration time are included as examples, but not limited thereto). The user basic information main attribute unit 1111 is mainly fixed information filled by the user at the beginning of registration, and is also a main constituent element of the user's own basic information in the social attributes of the user in the scheme, and the information can be used for the user social attribute dimension classification in the first step.

The user basic information secondary attribute unit 1112 is configured to receive and update the basic secondary attribute data of the target user, and to perform classification and weighting on the data. The basic secondary attribute data includes one or more of age, birthday, hometown, resident geographic location, academic experience, political trend, and religion (including age, birthday, hometown, resident geographic location, academic experience, political trend, and religion, but not limited thereto).

The first calculating unit 1113 is configured to calculate weights of basic primary attribute data and basic secondary attribute data of a target user, and output the weights to the user basic information module 1110 in a classified manner, so that the user basic information module 1110 can sort and sort data input by the first calculating unit 1113, where the basic information includes the basic primary attribute data and corresponding weights, the basic secondary attribute data and corresponding weights.

Illustratively, the user behavior module 1120 may include a user UGC behavior unit 1121, a user social behavior unit 1122, and a user basic behavior unit 1123.

The user UGC behavior unit 1121 is configured to receive and update UGC text content data of a target user, and perform classification and weighting on the data, where the UGC text content data includes published content text data, a domain to which published content belongs, and published content browsing indicators, and the published content browsing indicators include one or more of browsing amount, browsing efficiency, reply amount, and collection amount of the published content text data. In addition, statistics can be performed on the classification of the published contents of the user, such as user habits such as user preferences for publishing long articles, short articles or topic articles of interest discussions, and data statistics can be performed on the user preferences for publishing the long articles, the number of text contents in the long articles, the number of content words in the long articles, the number of published contents in all fields, the browsing amount of the published contents of the user, browsing efficiency, the number of replies, the number of collections and the like.

By counting the relevant data such as the text data of the published contents, the domain to which the published contents belong, and the browsing index of the published contents in the user UGC behavior unit 1121, the detailed data of the user in the system about the user UGC behavior content can be clearly counted. For example, if a user inside the system likes to post content in a domain of a certain interest category or browses an article board in a certain interest domain for a time period longer than that of an interest board in other domains, the interest habit of the user can be accurately located, data of the content posted by the user replied or browsed by other users can be analyzed, and weights and relationship chains of various aspects of social attributes can be calculated for the user through the data. Even if the published contents of the user are less in reply amount and much in browsing amount, the later published contents of the user can be recommended to users active in replying, social recognition and group integration of the user can be improved, and the viscosity and the dependency of the user on the platform are increased.

The user social behavior unit 1122 is configured to receive and update social behavior data of a target user, and classify and empower the data, where the social behavior data includes a number of items from a page access amount, a user reply behavior, a reply content quality, a quantity of interest, a social attribute dimension of the user of interest, a fan quantity, a social attribute dimension of the fan, a social attribute dimension of a friend application, a chat duration with a friend, a chat frequency with the friend, whether the user actively initiates a chat, whether the user actively adds a friend, and a friend quantity. By counting the data, the social behavior trend and habit of the user can be accurately analyzed.

A user basic behavior unit 1123, configured to receive and update basic behavior data of a target user, and classify and weight the data, where the basic behavior data includes multiple items of LBS (Location Based Service) geographic trajectory, online UGC active time period, time period for posting behavior, time period for liking behavior, online social active time period, friend chat time period, time period for replying to behavior, time period for adding friends, time period for paying attention to behavior, online reading time period, time period for reading content, online inertia time period, first-time-of-day startup period, and exit-of-day period.

For example, through the analysis of the LBS geographical trajectory of the user, the activity trajectory of the user in reality can be seen, so that the intersection of the LBS geographical trajectories of other users in the system and the user can be calculated, and the proximity between the two users can be estimated by combining the data of various types of behavior time periods in the user basic behavior unit 1123, and the closer the geographical distance position is, the higher the availability of offline activities that the user can perform to the matched user is. Generally speaking, ordinary people have little opportunity to know everyone in the city and can not classify others in detail, and strangers are no more social units for people. According to the scheme, the users in the same city with the user can be selected in a classified mode and displayed, intersection is easy to generate between the users and other users due to the fact that the distances are close, the intersection between the users and other user time periods in the system can be obtained more accurately through multi-dimensional analysis of the user use time periods in the user basic behavior units 1123, and more accurate auxiliary data are made for user matching.

The second calculating unit 1124 is configured to calculate UGC text content data, social behavior data, and basic behavior data of the target user, and output the UGC text content data, the social behavior data, and the basic behavior data to the user behavior module 1120 in a classified manner, so that the user behavior module 1120 can sort and sort the data input by the second calculating unit 1124, where the behavior information includes the UGC text content data and its corresponding weight, the social behavior data and its corresponding weight, and the basic behavior data and its corresponding weight.

Additionally, in this embodiment, social users may also be associated with a second class of user library module 1700, which second class of user library module 1700 may include an interest category element 1710 and a vertical domain element 1720.

Illustratively, the interest category unit 1710 is configured to count, classify, and collect interest category data of the target user, and perform clustering weighting on the data, where the interest category data is a data result selected by the target user based on an interest category built in the user recommendation system 1000 or an immediately generated interest category

An interest category unit 1710, configured to count, classify, and collect interest category data of a target user, and perform clustering weighting on the data. The interest category data is a data result selected by the target user based on the interest category built in the user recommendation system 1000 or the instantly generated interest category.

For example, the initial collection manner of the interest category data of the interest category unit 1710 may be implemented by a selection function (as shown in fig. 4) of an interest content item provided in the database when a new account is established for a new user by logging in, and may be browsed by other users in a display function (shown in fig. 5) of a user information content item, where the interest categories include, but are not limited to, "interest", "game", "movie", "music", "quadratic element", "reading", "sports", "personal traits", and "now", the user selects from the provided selections, and the interest category unit 1710 records the selections.

And the vertical field unit 1720 is used for counting, classifying and collecting vertical field data of a target user, and performing clustering weighting on the data. The vertical domain data is a data result selected by the target user based on a preset library provided by the user recommendation system 1000 or a vertical domain library provided by a third party.

The vertical domain here refers to a single vertical domain in the user's interest, for example, if the user likes to watch movies and books, then the vertical domain unit 1720 can perform statistical sorting for which movie and which book are specified. Specific information of the vertical domain of interest, such as specific sub-level menu options in the second-level menu or the first-level menu options in fig. 4 and fig. 6, may be selected at the beginning of the user registration, and may be sub-options provided to the user by the API, or may be established by the user himself (vertical domain data). If the new sub-option is supported by other users, the system can judge whether to establish the sub-option in the interest category selection module according to a certain threshold value and associate and classify the newly generated sub-option (it can be understood that a user establishes a new sub-option to show the vertical field of the user, and other users subsequently establish the sub-option as well, and when the number of people establishing the sub-option reaches a certain number, the system can establish the sub-option in the interest category selection module and associate and classify the newly generated sub-option, so that the sub-level menu option provided by the system comprises the sub-option).

And the third calculating unit 1740 is configured to calculate weights of the interest category data and the vertical domain data, and classify the user data types.

In this embodiment, in addition to the interest category unit 1710 and the vertical domain unit 1720, the second-class user library module 1700 may additionally include a user other-category behavior unit 1730. The user other category behavior unit 1730 is configured to count, classify, and collect other category behavior data of the target user, and assign a weight to the data (which may be a classification weight or a clustering weight, based on a mode to which data characteristics are applicable, and is not limited herein). And acquiring user behavior statistical data in the information authorized by the target user based on the inside of the system and other platforms by using other types of behavior data. Correspondingly, the third calculating unit 1740 may calculate weights of the interest category data, the vertical domain data and other category behavior data, and classify the user data types thereof.

For example, activities such as offline gathering or topic discussion are performed inside the system, and it is necessary to classify the activities and calculate user behavior statistics (e.g., whether to participate, duration of participation, whether to publish contents, whether to add friends, etc.) of the users in such activities. User behavior data of the user on other platforms can also be included, such as a song library that the user likes to listen to in certain music software, or game records that the user purchases and uses on other game platforms. The data are classified according to the use habits of the user, and finally recorded as other category behavior data in the other category behavior unit 1730 of the user as an index in a multi-dimensional matching mode of the user.

By dividing the first-class user library module 1100 into a user basic information module 1110 and a user behavior module 1120, the user basic information module 1110 is subdivided into a user basic information primary attribute unit 1111 (nickname, UID, gender, sexual orientation, registration time, registration duration, etc.) and a user basic information secondary attribute unit 1112 (age, birthday, hometown, resident geographical location, academic experience, political inclination, religion, etc.); the user behavior module 1120 is subdivided into a user UGC behavior unit 1121 (published content text data, published content belonging field, and published content browsing index including browsing amount of the published content text data, browsing efficiency, reply amount, collection amount, etc.), a user social behavior unit 1122 (page access amount, user reply behavior, reply content quality, attention amount, social attribute dimension of the attention user, fan number, social attribute dimension of fan, social attribute dimension of friend application, chat duration with friends, chat times with friends, whether the user actively initiates chat, whether the user actively adds friends and friend number, etc.), and a user basic behavior unit 1123(LBS geographical trajectory, online UGC active time period, published behavior time period, favorite behavior time period, online social active time period, etc.) A friend chat period, a reply to behavior period, an add friend period, a follow up behavior period, an online reading period, a content reading period, an online inertia period, a daily first start period, a daily quit period, etc.). Therefore, the multi-dimensional social attributes of the users can be effectively enriched, and the users which are closer in more dimensions (closer in geographic positions, more consistency in the eigen mode, the representation social information, the subdivision categories and the like of the users) can be matched, so that three essential elements of favorite emotion among people in psychology, namely attraction, proximity and similarity are met. By combining the three elements, the method can be applied to the accurate matching of the users in the social network service or the social software through computer science, so that the success rate of the matching of the users is effectively improved, and the user experience is improved.

The second-class user library module 1700 is divided into an interest category unit 1710 and a vertical domain unit 1720, and the interest category unit 1710 can count, classify and collect interest category data of a target user (a data result of the target user selecting based on an interest category built in the user recommendation system 1000 or an interest category generated immediately), and cluster and assign weights to the data; the vertical domain unit 1720 may count, classify, and collect vertical domain data of a target user (a data result selected by the target user based on a preset library provided by the user recommendation system 1000 or a vertical domain library provided by a third party, which is used to represent a single vertical domain in the user interest), and cluster and weight the data. The user interest field can be collected, and the single vertical interest field of the user can also be collected, so that the second-class user library module 1700 can effectively reflect the interest of the user. According to the Theory of Cognitive disorder (the Theory of Cognitive disorder is proposed in a book of a Theory of Cognitive disorder of filt, 1957, namely Cognitive disorder Theory), people are more suitable for other people with the same idea, the same hobbies and the same situation, and under the environment and the condition, people can easily generate the sense of identity, the satisfaction and the sense of dependence, and the social network service which is more in line with the requirements of the users can be provided for the users, so that the diversified requirements of the users are met, the ever-existing transparency illusion (namely the transparency of social network service) in the social network service can be eliminated, and the social network service is more in line with the requirements of the times and the requirements of the social network users.

In the above, a social user heterogeneous graph model with a social user as a core is introduced, and a source of user data used in the scheme is introduced in detail. Of course, in other implementations, the social user heterogeneous graph model may also include more types of data, for example, as described below with reference to the third-class user library module 1920, which will not be described herein. The description of other modules in the multi-dimensional information based user recommendation system 1000 will be continued later.

Referring to fig. 1 again, in this embodiment, the representation information determining module 1200 may be configured to obtain UGC text content data in the first-class user library module 1100, and determine the representation social information of the target user based on the UGC text content data. Where the tokens represent an objective reflection of the user's mental activities within the user recommendation system 1000 of external thing behaviors that can be analyzed by recording such objective reflection, the token social information here includes words (which can include other information, such as category labels) that reflect the mental activities of the target user.

Of course, in addition to obtaining UGC text content data in first-class user library module 1100 (e.g., from within user UGC behavior unit 1121), characterization information determination module 1200 may also obtain UGC text content data from elsewhere, e.g., characterization information determination module 1200 may also obtain UGC text content data (e.g., from text content data published by a target user within its vertical domain) from vertical domain unit 1720, user other-class behavior unit 1730, etc. of second-class user library module 1700, without limitation herein.

For example, the specific way for the representation information determining module 1200 to determine the representation social information of the target user based on the UGC text content data may be:

the representation information determining module 1200 may perform preliminary word segmentation on the UGC text content data, for example, may perform preliminary word segmentation on the UGC text content data by using a forward maximum matching method, a reverse maximum matching method, a bidirectional maximum matching method, or the like; then carrying out weight distribution on UGC text content data after word segmentation; and extracting core words of each sentence in the UGC text content data through the core words, marking, determining the emotion expressed by the target user in the UGC text content data based on a semantic probability model, classifying the UGC text content data of the target user, and determining the representation social information of the target user based on all the classified UGC text content data.

For example, UGC text content data may be subjected to preliminary word segmentation by a word segmentation system, wherein the employed word segmentation method includes, but is not limited to, forward maximum matching, reverse maximum matching, bidirectional maximum matching, and the like. It is common practice to predict the tag [ tag ] of each word of a text string, such as B, E, I, S, which identifies beginning, inside, ending, single, respectively, as the words of the beginning, middle, ending, and individual words of a sentence. And then, processing such as part-of-speech tagging, new word discovery and the like is carried out on the text data. The processed UGC text content data is assigned with weight, important words and words can be given higher weight, and text retrieval, relevance and the like in the UGC text content data can be analyzed. And extracting important core words in a sentence and giving marks through the core words, and extracting the importance labels of the implied users to the words and the characters from numerous data by utilizing the data automatically mined from the library by the system. A semantic probability represented by a sentence itself can then be calculated using a language model (a probabilistic model) and a mood in the UGC text content data can be predicted by the language model to classify the user's UGC text content data. Thus, representative social information of the target user may be determined based on all of the categorized UGC textual content data.

In this way, the representation information determining module 1200 can effectively predict the emotion of the user from the UGC text content data of the target user, so as to classify the UGC text content data of the user, which is beneficial to accurately mapping the UGC text content data to the eigen-mode of the user, thereby ensuring the accuracy and effectiveness of the eigen-mode of the user.

In this embodiment, the eigen-sampling identification module 1300 may be configured to determine a user eigen-model of the target user based on the representative social information of the target user. Here, the eigen symbolizes the true social tendency of the user's mind, and all the representative social information of the user in the user recommendation system 1000 is analyzed and mapped into the existing eigen model, and the eigen model of the user is set based on the personality theory (such as the quintessence theory, the personality trait theory, etc.) in the psychology, and can be described and reflected through vocabularies.

In the scheme, the intrinsic characteristic represents the true social tendency of the user, and the data acquired by analyzing all characteristic behaviors and social attributes of the user in the platform are mapped into the existing user intrinsic model; and the representation refers to the objective reflection of the external object behaviors by one psychological activity of all social attributes of the user in the system, and the internal psychological activity of the user can be analyzed in detail by recording the objective reflection.

For example, taking the user eigenmode as an example based on the five-personality theory, each eigenmode may be associated with a plurality of words describing and outlining the eigenmode. That is, by adopting the principle of the five-personality theory in the user eigenmodel, each face in human life is endowed with words describing the nature of the face, and not only, if something is really important and ubiquitous, but it is endowed with more words in all languages to describe it, so that the theory that the personality traits are important for the Lexical Hypothesis (Lexical hyperthesis) is discovered from the words. The personality traits formed by five robust factors of the user, namely the five personalities, are preset in the intrinsic sampling model of the user. Therefore, each user can at least correspond to five eigenmodes set based on the five-personality theory in the system.

The eigen-sampling recognition module 1300 may classify the representative social information of the target user, map the representative social information to each eigen-mode shape, and determine the contact ratio of the representative social information of the target user in each eigen-mode shape. Then, the eigen-sampling recognition module 1300 may determine, based on the degree of coincidence corresponding to each eigen-mode shape, an eigen-mode shape with the highest degree of coincidence and with which the difference between the degree of coincidence and the other eigen-mode shapes meets the standard (i.e., the representative social information of the user obviously meets a certain eigen-mode shape) as the user eigen-model of the target user.

In addition, in each eigen model provided in the present solution, words for describing and accommodating personality differences mapped to the eigen model are collected, and the eigen attributes of the user can be improved by mapping the characterizing social information determined by the characterizing information determining module 1200 to the eigen model. Assuming that the user cannot match the eigen model preset by the system, the eigen sampling recognition module 1300 may further perform manual collection and verification according to the representation social information determined by the representation information determination module 1200, and combine the user data to generate other richer user eigen models, which is not limited herein.

The user eigen model is set based on the five-personality theory, the representation social information of the target user is classified and mapped to each eigen model, the fact that the highest coincidence degree of the representation social information of the target user in the eigen model is the user eigen model is determined, the personality type of the user can be effectively and accurately reflected, and therefore accurate matching of the user is facilitated.

In this embodiment, the target user multidimensional attribute module 1400 is configured to determine multidimensional social attributes and social tendencies of the target user based on the user data and the user eigen model of the target user. The main function of the target user multidimensional attribute module 1400 can be understood as statistics (e.g. extraction of descriptive words) and summarization of user data and user eigen models (as multidimensional social attributes of the target user and social trends of the target user), and obtaining characterization attributes, eigen attributes, other social attributes and social trends of the target user.

In this embodiment, the first computing module 1500 is configured to obtain user data and a user eigen model of a user to be matched, and determine a multi-dimensional social attribute and a social tendency of the user to be matched. Here, the function of the first calculation module 1500 may be aligned with the function of the target user multidimensional attribute module 1400, but the difference is that the specific object of the data source obtained by the two modules is different, and the function of the first calculation module may refer to the target user multidimensional attribute module 1400, which is not described herein again.

It should be noted that, in order to improve the operation efficiency of the user recommendation system 1000, before the matching value calculation module 1600 determines the multidimensional matching value between the target user and the user to be matched, the user to be matched may be screened first. Based on this, the user recommendation system 1000 further provides a second calculation module 1810 and a verification module 1820.

Referring to fig. 2 again, for example, the second calculating module 1810 may perform preliminary screening on the multi-dimensional social attributes and social tendencies of the users to be matched, and determine the users to be matched that satisfy the preliminary screening conditions. For example, the second calculation module 1810 may perform screening on other users meeting a preset threshold rule (such a screening threshold may flexibly select an index and set a value according to actual needs) according to each data weight between the target user and the other users in the first-class user library module 1100 and the second-class user library module 1700, generate a to-be-checked data sample (a user meeting a condition, that is, a to-be-matched user meeting a preliminary screening condition), and then perform checking by the checking module 1820.

For example, the checking module 1820 may determine, after the second computing module 1810 determines the to-be-matched users meeting the preliminary screening condition, whether to-be-rejected users with social attribute differences exceeding set differences exist between the preliminarily screened to-be-matched users and the target users, and reject the to-be-rejected users from the to-be-matched users to obtain the screened to-be-matched users. For example, the verification module 1820 may be configured to calculate the difference of social attributes between the target user and the other users, and if the difference is too large and exceeds the set difference of the system, the multidimensional matching value calculation between the other users and the target user cannot be triggered (which is equivalent to removing the user from the preliminarily screened group of users to be matched).

Correspondingly, subsequently, when the matching value calculating module 1600 runs, the matching value calculating module 1600 may determine the multidimensional matching values of the target user and the screened user to be matched (i.e., only the multidimensional matching values between the screened user to be matched and the target user are calculated).

By the method, the users can be effectively screened, the users with high matching possibility with the target user are screened from other user groups, the multi-dimensional matching value is calculated, the calculation amount can be effectively reduced, and the operation efficiency of the user recommendation system 1000 is improved.

Referring to fig. 1 again, in this embodiment, the matching value calculation module 1600 may be configured to determine a multidimensional matching value between the target user and the user to be matched based on the multidimensional social attribute and the social tendency of the target user and the multidimensional social attribute and the social tendency of the user to be matched, and determine a recommended user list matched with the target user based on the calculated multidimensional matching value (a plurality of multidimensional matching values corresponding to a plurality of users to be matched).

Illustratively, the match value calculation module 1600 may be specifically configured to: and aiming at each subdivision index in the multidimensional social attribute and the social tendency of the target user, performing cosine similarity calculation on the multidimensional social attribute and the corresponding subdivision index in the social tendency of the user to be matched to obtain a multidimensional matching value between the target user and the user to be matched. And then, determining that the users with the multidimensional matching value reaching the set matching value (for example, more than 60%) are to-be-recommended users, and performing sorting processing on the to-be-recommended users to obtain a recommended user list matched with the target user.

And calculating the closeness between each subdivision index in the multidimensional social attribute and the social tendency of the target user and the corresponding subdivision index in the multidimensional social attribute and the social tendency of the user to be matched by utilizing the cosine similarity, wherein the method is accurate and efficient, and can well compare the multidimensional social attribute and the social tendency to determine a better result.

Referring again to fig. 2, in the present embodiment, the user recommendation system 1000 may further include a third calculation module 1910 and a third-class user library module 1920.

A third calculating module 1910, configured to, after the matching value calculating module 1600 determines the recommended user list matching the target user, collect subsequent matching data between the target user and the recommended users in the recommended user list, and classify and weight the matching result based on the subsequent matching data.

And the third-class user library module 1920 is configured to classify the subsequent matching data and the corresponding classification and weight value collected by the third calculation module 1910 into the social tendency of the target user, so as to correct the social tendency of the target user.

For example, if the target user matches user a, the target user and user a no longer generate interactive behavior for a certain period of time, or the target user has a low subsequent matching degree to the matched user a, and these data with similar characteristics (indicating that the recommended user matching the target user is not suitable) may be recorded into the third type user library module 1920 by the third calculation module 1910, and the data of the target user and user a in the first type user library module 1100, the second type user library module 1700, and the user book model may be analyzed in detail, and the social tendencies of the two users (the target user and user a) are recorded, so as to avoid the similar situation still occurring in the subsequent matching process (this similar situation refers to the negative result of no interactive behavior between the matched users).

The third calculation module 1910 may collect subsequent matching data between the target user and the recommended users in the recommended user list and classify and weight the matching result based on the subsequent matching data, and the third-class user library module 1920 may classify the subsequent matching data and the corresponding classification and weight collected in the third calculation module 1910 into the social tendency of the target user to correct the social tendency of the target user. In addition, the data in the third-class user library module 1920 can also be used as a training set to learn and correct a subsequent matching mechanism, so that the diversified requirements of user matching can be flexibly met.

In addition, the user recommendation system 1000 may further include a storage module 1930 for completely storing all data generated between the system and the user according to the scheme.

And, the user recommendation system 1000 may further include a cache module 1940, configured to transmit the data to the cache module 1940 when there is a problem with the data verified by the verification module 1820, so that the user recommendation system 1000 may quickly recalculate the multidimensional matching value of the user. The cache module 1940 may also perform data caching on the generated multidimensional matching data from the storage module 1930, so as to reduce the time for system retrieval and calculation and improve the hardware performance.

Of course, the user recommendation system 1000 may further include a result output module 1950 for displaying a recommended user list for the target user, so that the target user may select a desirable recommended user for a friend-making attempt.

Based on the multi-dimensional information-based user recommendation system 1000 provided by the scheme, the embodiment of the application further provides two multi-dimensional information-based user recommendation methods applied to the user recommendation system 1000.

First, referring to fig. 7, fig. 7 is a flowchart of a first user recommendation method based on multidimensional information according to an embodiment of the present application. In the present embodiment, this user recommendation method may include step S11, step S12, step S13, step S14, step S15, and step S16.

Step S11: acquiring user data of a target user, wherein the user data of the target user comprises basic information of the target user and UGC text content data determined based on UGC behaviors of the target user.

Step S12: and determining the representation social information of the target user based on the UGC text content data, wherein the representation represents the objective reflection of the psychological activities of the user in the user recommendation system to the external things behaviors, the psychological activities of the user can be analyzed by recording the objective reflection, and the representation social information comprises words reflecting the psychological activities of the target user.

Step S13: and determining a user eigen model of the target user based on the representative social information of the target user, wherein the eigen represents the true social tendency of the user, all the representative social information of the user in the user recommendation system is analyzed and mapped into the existing eigen model, and the user eigen model is set based on the five-personality theory and can be described and reflected through words.

Step S14: and determining the multi-dimensional social attributes and social tendencies of the target user based on the user data of the target user and the user eigen model.

Step S15: the method comprises the steps of obtaining user data and a user eigen model of a user to be matched, and determining the multi-dimensional social attributes and social tendencies of the user to be matched.

Step S16: and determining a multi-dimensional matching value of the target user and the user to be matched based on the multi-dimensional social attributes and social tendency of the target user and the multi-dimensional social attributes and social tendency of the user to be matched, and determining a recommended user list matched with the target user based on the multi-dimensional matching value.

Since the flow of the user recommendation method corresponds to a manner in which the functions of the user recommendation system 1000 based on multidimensional information described in the foregoing are implemented in cooperation with the modules of the user recommendation system 1000 based on multidimensional information, reference may be made to the description of the user recommendation system 1000 based on multidimensional information, and details are not described here.

Next, referring to fig. 8, fig. 8 is a flowchart of a second user recommendation method based on multi-dimensional information according to an embodiment of the present application. In the present embodiment, the user recommendation method may include steps S21, S22, S23, and S24.

Step S21: if the target user uses the user recommendation system for the first time, determining the multi-dimensional social attributes and the social tendency of the target user, classifying the target user and storing the classified target user in a first database.

In this embodiment, if the target user uses the user recommendation system 1000 for the first time, the multidimensional social attributes and social tendencies of the target user may be determined, and the target user is categorized and stored in the first database. Here, a statistical analysis of the target user may be performed using basic information in the user data of the target user and user behavior data (e.g., UGC text content data), and then the target user may be categorized and stored in the first database.

Then, step S22 may be performed.

Step S22: and screening the users in the second database by utilizing the multi-dimensional social attributes and the social tendency of the target user, screening out the users to be matched which are consistent with the multi-dimensional social attributes and the social tendency of the target user, and storing the users to be matched into a third database, wherein the second database stores data information of other users, and the data information comprises the multi-dimensional social attributes and the social tendency of the users.

In this embodiment, the data information of other users (including the multidimensional social attributes and social tendencies of the users) is stored in the second database, the users in the second database can be screened by using the multidimensional social attributes and social tendencies of the target user, the users to be matched which are consistent with the multidimensional social attributes and social tendencies of the target user are screened, and the users to be matched are stored in the third database.

Step S23: and carrying out multi-dimensional matching on the multi-dimensional social attributes and social tendency of the target user in the first database and the multi-dimensional social attributes and social tendency of each user to be matched in the third database, determining multi-dimensional matching values of the target user and the users to be matched, and determining a recommended user list matched with the target user based on the multi-dimensional matching values.

In this embodiment, multidimensional matching may be performed on the multidimensional social attributes and social tendencies of the target user in the first database and the multidimensional social attributes and social tendencies of each user to be matched in the third database, multidimensional matching values between the target user and the users to be matched are determined, and a recommended user list matched with the target user is determined based on the multidimensional matching values. Since the process of determining the multidimensional matching value can be referred to the above description, it is not described here.

After the recommended user list is determined, step S24 may be performed.

Step S24: and acquiring subsequent matching data between each recommended user in the recommended user list and the target user, classifying and weighting matching results based on the subsequent matching data, and storing the matching results into a fourth database, wherein the subsequent matching data corresponding to each recommended user, the classification and the weighting thereof can be classified into the social tendency of the target user so as to correct the social tendency of the target user.

In this embodiment, subsequent matching data between each recommended user in the recommended user list and the target user may be obtained, and the matching result is classified and weighted based on the subsequent matching data and stored in the fourth database. Here, the subsequent matching data corresponding to each recommended user and the classification and weight thereof may be classified into the social tendency of the target user to correct the social tendency of the target user. The specific functions may be referred to the description of the third-class user library module 1920 in the foregoing, and are not described herein again.

In addition, for the case where the target user does not use the user recommendation system 1000 for the first time, the user recommendation method may further include step S25 and step S26. In this case, step S21, step S22, and step S23 may be skipped, and the process may be started from step S25.

Step S25: and acquiring the information of the target user in the first database, the third database and the fourth database.

In this embodiment, information of the target user in the first database (for example, the multidimensional social attributes and social tendencies of the target user), the third database (for example, the recommended user list matching the target user), and the fourth database (for example, the recommended user list matching the target user, the user list with poor subsequent matching effect, and corresponding data) may be obtained.

After the information of the target user in the first database, the third database and the fourth database is obtained, step S26 may be executed.

Step S26: and screening the users to be matched, matching the target user with the screened users to be matched according to a multi-dimensional matching mode, and outputting a matching result, wherein the matching result is a recommended user list.

In this embodiment, based on the information of the target user in the first database, the third database, and the fourth database, the target user is screened (please refer to the description of the second calculation module 1810 and the verification module 1820, which is not described herein again), the target user is matched with the screened user to be matched according to the multidimensional matching method (i.e., the multidimensional matching value between the target user and the screened user to be matched is calculated, which may also refer to the description herein, which is not described herein again), and the matching result (the recommended user list) is output.

Correspondingly, after the matching result is output, step S24 may be further executed to obtain subsequent matching data between each recommended user and the target user in the recommended user list, classify and weight the matching result based on the subsequent matching data, and store the result in the fourth database, thereby further updating the fourth database.

In this embodiment, the method can record the subsequent matching data of the target user again, further follow-up and record the subsequent matching of the target user and the recommended user, and is beneficial to obtaining instant feedback data as a training set to improve the effectiveness of multi-dimensional matching.

The embodiment of the present application further provides a storage medium, where the storage medium includes a stored program, and when the program runs, the device where the storage medium is located is controlled to execute any one of the multi-dimensional information-based user recommendation methods described in the present embodiment.

Referring to fig. 9, fig. 9 is a block diagram of an electronic device 20 according to an embodiment of the present disclosure.

In this embodiment, the electronic device 20 is a server, such as a cloud server, a web server, a server cluster, and the like, in which the multi-dimensional information-based user recommendation system 1000 is built, and is not limited herein.

Illustratively, the electronic device 20 may include: a communication module 22 connected to the outside world via a network, one or more processors 24 for executing program instructions, a bus 23, and a different form of memory 21, such as a disk, ROM, or RAM, or any combination thereof. The memory 21, the communication module 22, and the processor 24 may be connected by a bus 23.

Illustratively, the memory 21 has stored therein a program. Processor 24 may invoke and run these programs from memory 21 so that any of the multi-dimensional information based user recommendation methods may be implemented by running the programs.

In summary, the embodiment of the present application provides a user recommendation system and method based on multi-dimensional information, which includes collecting and sorting user data (including basic information of a target user and UGC text content data determined based on UGC behavior of the target user) of a target user through a first-class user library module 1100, acquiring the UGC text content data in the first-class user library module 1100 by using a representation information determining module 1200, and determining representation social information (including words reflecting psychological activities of the target user) of the target user based on the UGC text content data; determining a user eigen model of the target user (based on the personality theory setting in psychology, and describing and reflecting the user eigen model through vocabularies) by using the eigen sampling recognition module 1300 based on the representative social information of the target user; further, the multidimensional social attributes and social tendencies of the target user are determined by using the target user multidimensional attribute module 1400 based on the user data and the user eigen model of the target user. The first computing module 1500 may obtain user data and a user eigen model of the user to be matched, and determine a multi-dimensional social attribute and a social tendency of the user to be matched. Therefore, the matching value calculation module 1600 may determine the multidimensional matching values of the target user and the user to be matched based on the multidimensional social attributes and social tendencies of the target user and the multidimensional social attributes and social tendencies of the user to be matched, and determine the recommended user list matched with the target user based on the multidimensional matching values. In such a mode, UGC (user original content) text content data of a user can be used as a basis for determining the representation social information and the user eigen model of the user, and the UGC text content data can usually effectively reflect the psychological activities of the user, so that the eigen model (which personality the user belongs to) is reflected laterally, psychology and internet technology are combined, the personality theory in psychology is applied to a social network and used as a reference when the user is matched, and the matching precision of the user can be effectively improved. The basic information, UGC text content data, representation social information and the user eigen model of the user are used as indexes for determining the multi-dimensional social attributes and social tendency of the target user, information dimensions in the user matching process can be enriched, and the similarity of the user is considered from more dimensions, so that the success rate of user matching is improved, real potential friends in a social network service platform or social software are effectively discovered for the user, and the matching precision and the user experience of the user are improved.

In the embodiments provided in the present application, it should be understood that the disclosed system and method may be implemented in other ways. The above-described system embodiments are merely illustrative, and for example, the division of the units/modules is only one logical division, and other divisions may be realized in practice, and for example, a plurality of units or modules may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of modules or units through some communication interfaces, and may be in an electrical, mechanical or other form.

In this document, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions.

The above description is only an example of the present application and is not intended to limit the scope of the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A user recommendation system based on multi-dimensional information is characterized by comprising:

the first-class user library module is used for collecting and sorting user data of a target user, wherein the user data of the target user comprises basic information of the target user and UGC text content data determined based on UGC behaviors of the target user;

the representation information determining module is used for acquiring UGC text content data in the first-class user library module and determining representation social information of the target user based on the UGC text content data, wherein the representation represents the objective reflection of the psychological activity of the user in the user recommendation system on external things behaviors, the psychological activity of the user can be analyzed by recording the objective reflection, and the representation social information comprises words reflecting the psychological activity of the target user;

the intrinsic sampling identification module is used for determining a user intrinsic model of the target user based on the representation social information of the target user, wherein the intrinsic represents the true social tendency of the user, all representation social information of the user in the user recommendation system is analyzed and mapped into the existing intrinsic model, and the user intrinsic model is set based on the personality theory in psychology and can be described and reflected through words;

the target user multi-dimensional attribute module is used for determining the multi-dimensional social attributes and social tendencies of the target user based on the user data of the target user and the user eigen model;

the first calculation module is used for acquiring user data and a user eigen model of a user to be matched and determining the multi-dimensional social attributes and social tendencies of the user to be matched;

and the matching value calculation module is used for determining a multi-dimensional matching value of the target user and the user to be matched based on the multi-dimensional social attribute and the social tendency of the target user and the multi-dimensional social attribute and the social tendency of the user to be matched, and determining a recommended user list matched with the target user based on the multi-dimensional matching value.

2. The multi-dimensional information-based user recommendation system according to claim 1, wherein said user data comprises basic information and behavior information, and said first-class user library module comprises:

the user basic information main attribute unit is used for receiving and updating basic main attribute data of a target user and classifying and weighting the data, wherein the basic main attribute data comprises one or more of a nickname, a UID, a gender, a sexual orientation, registration time and registration duration;

the user basic information secondary attribute unit is used for receiving and updating basic secondary attribute data of a target user and classifying and weighting the data, wherein the basic secondary attribute data comprises one or more of age, birthday, hometown, resident geographic position, academic experience, political tendency and religion;

the first calculation unit is used for calculating the weight values of the basic primary attribute data and the basic secondary attribute data of the target user and outputting the weight values to the user basic information module in a classified manner;

the user basic information module is used for sorting and sorting the data input by the first computing unit, wherein the basic information comprises the basic primary attribute data and the corresponding weight, and the basic secondary attribute data and the corresponding weight;

the user UGC behavior unit is used for receiving and updating UGC text content data of the target user and classifying and empowering the data, wherein the UGC text content data comprises published content text data, published content affiliated fields and published content browsing indexes, and the published content browsing indexes comprise one or more of browsing amount, browsing efficiency, reply amount and collection amount of the published content text data;

the user social behavior unit is used for receiving and updating the social behavior data of the target user and classifying and empowering the data, wherein the social behavior data comprises a plurality of items of page access volume, user reply behavior, reply content quality, attention quantity, social attribute dimensionality of the attention user, fan quantity, social attribute dimensionality of fan, social attribute dimensionality of friend application, chat duration with friends, chat times with friends, whether the user actively initiates chat, whether the user actively adds friends and friend quantity;

the basic behavior unit is used for receiving and updating basic behavior data of the target user and classifying and empowering the data, wherein the basic behavior data comprises multiple items of LBS geographic tracks, online UGC active time periods, behavior publishing time periods, behavior liking time periods, online social activity time periods, friend chat time periods, behavior replying time periods, friend adding time periods, behavior attention time periods, online reading time periods, content reading time periods, online inertia time periods, first-time-of-day starting time periods and quitting time periods every day;

the second calculation unit is used for calculating UGC text content data, social behavior data and basic behavior data of the target user and outputting the UGC text content data, the social behavior data and the basic behavior data to the user behavior module in a classified mode;

and the user behavior module is used for sequencing and sorting the data input by the second computing unit, wherein the behavior information comprises the UGC text content data and the corresponding weight thereof, the social behavior data and the corresponding weight thereof, and the basic behavior data and the corresponding weight thereof.

3. The multi-dimensional information based user recommendation system according to claim 1, further comprising a second class user library module, said second class user library module comprising:

the interest category unit is used for counting, classifying and collecting interest category data of the target user and clustering and weighting the data, wherein the interest category data are data results selected by the target user based on interest categories built in the user recommendation system or interest categories generated immediately;

the vertical field unit is used for counting, classifying and collecting vertical field data of the target user and clustering and weighting the data, wherein the vertical field data are data results selected by the target user based on a preset library provided by the user recommendation system or a vertical field library provided by a third party and are used for representing a single vertical field in user interest;

and the third calculating unit is used for calculating the weights of the interest category data and the vertical field data and classifying the user data types.

4. The multi-dimensional information-based user recommendation system according to claim 1, wherein the characterization information determining module is specifically configured to:

performing preliminary word segmentation on the UGC text content data, wherein the word segmentation method comprises at least one of forward maximum matching, reverse maximum matching and bidirectional maximum matching;

carrying out weight distribution on UGC text content data after word segmentation;

extracting core words of each sentence in UGC text content data through the core words, marking the core words, determining the emotion expressed by the target user in the UGC text content data based on a semantic probability model, classifying the UGC text content data of the target user, and determining the representative social information of the target user based on all the classified UGC text content data.

5. The multidimensional information based user recommendation system of claim 1, wherein the user eigen model is set based on the five-personality theory, each eigen mode is associated with a plurality of words describing and accommodating the eigen mode, and the eigen sampling recognition module is specifically configured to:

classifying the representative social information of the target user, mapping the representative social information to each eigenmode shape, and determining the contact ratio of the representative social information of the target user in each eigenmode shape;

and determining the eigenmode shape with the highest contact ratio and the difference reaching the standard between the contact ratio of the eigenmode shape and the contact ratio of other eigenmode shapes as the user eigenmode of the target user based on the contact ratio corresponding to each eigenmode shape.

6. The multi-dimensional information-based user recommendation system according to claim 1, wherein the matching value calculation module is specifically configured to:

for each subdivision index in the multidimensional social attribute and the social tendency of the target user, performing cosine similarity calculation with the multidimensional social attribute and the corresponding subdivision index in the social tendency of the user to be matched to obtain a multidimensional matching value between the target user and the user to be matched;

and determining the user with the multi-dimensional matching value reaching the set matching value as a user to be recommended, and sequencing the user to be recommended to obtain a recommended user list matched with the target user.

7. The multi-dimensional information-based user recommendation system according to claim 1, further comprising a second calculation module and a verification module,

the second calculation module is used for preliminarily screening the multidimensional social attributes and social tendencies of the users to be matched before the matching value calculation module determines the multidimensional matching values of the target user and the users to be matched, and determining the users to be matched which meet preliminary screening conditions;

the verification module is used for determining whether a user to be rejected, the social attribute difference of which exceeds the set difference, exists between the preliminarily screened user to be matched and the target user after the user to be matched meeting the preliminary screening condition is determined by the second calculation module, and rejecting the user to be rejected from the user to be matched to obtain the screened user to be matched;

correspondingly, the matching value calculation module is used for determining the multidimensional matching value of the target user and the screened user to be matched.

8. The multi-dimensional information-based user recommendation system according to claim 1, further comprising a third computing module and a third class user library module,

the third calculation module is used for collecting subsequent matching data between the target user and the recommended users in the recommended user list after the matching value calculation module determines the recommended user list matched with the target user, and classifying and weighting the matching result based on the subsequent matching data;

and the third-class user library module is used for classifying the subsequent matching data and the corresponding classification and weight value collected in the third calculation module into the social tendency of the target user so as to correct the social tendency of the target user.

9. A multi-dimensional information-based user recommendation method is applied to the multi-dimensional information-based user recommendation system of any one of claims 1-8, and the method comprises the following steps:

acquiring user data of a target user, wherein the user data of the target user comprises basic information of the target user and UGC text content data determined based on UGC behaviors of the target user;

determining representative social information of the target user based on the UGC text content data, wherein the representative social information represents an objective reflection of the user's psychological activity in the user recommendation system on external things behaviors, the user's psychological activity can be analyzed by recording the objective reflection, and the representative social information comprises words reflecting the target user's psychological activity;

determining a user eigen model of the target user based on the representation social information of the target user, wherein the eigen represents the true social tendency of the user, all representation social information of the user in the user recommendation system is analyzed and mapped into the existing eigen model, and the user eigen model is set based on the five-personality theory and can be described and reflected through words;

determining multi-dimensional social attributes and social tendencies of the target user based on the user data of the target user and the user eigen model;

acquiring user data and a user eigen model of a user to be matched, and determining the multi-dimensional social attributes and social tendency of the user to be matched;

and determining a multi-dimensional matching value of the target user and the user to be matched based on the multi-dimensional social attributes and social tendency of the target user and the multi-dimensional social attributes and social tendency of the user to be matched, and determining a recommended user list matched with the target user based on the multi-dimensional matching value.

10. A multi-dimensional information-based user recommendation method is applied to the multi-dimensional information-based user recommendation system of any one of claims 1-8, and the method comprises the following steps:

if the target user uses the user recommendation system for the first time, determining the multi-dimensional social attributes and social tendency of the target user, classifying the target user and storing the target user in a first database;

screening users in a second database by utilizing the multidimensional social attributes and social tendency of the target user, screening out users to be matched which are consistent with the multidimensional social attributes and social tendency of the target user, and storing the users to be matched into a third database, wherein the second database stores data information of other users, and the data information comprises the multidimensional social attributes and social tendency of the users;

carrying out multidimensional matching on the multidimensional social attributes and social tendencies of the target user in the first database and the multidimensional social attributes and social tendencies of each user to be matched in the third database, determining multidimensional matching values of the target user and the users to be matched, and determining a recommended user list matched with the target user based on the multidimensional matching values;

and acquiring subsequent matching data between each recommended user in the recommended user list and the target user, classifying and weighting matching results based on the subsequent matching data, and storing the matching results into a fourth database, wherein the subsequent matching data corresponding to each recommended user, the classification and the weighting thereof can be classified into the social tendency of the target user so as to correct the social tendency of the target user.