CN112818249B - A method and system for constructing a multi-dimensional portrait of a specific tendency group - Google Patents

Abstract

The present invention relates to a method for constructing a multi-dimensional portrait of a specific tendency group, which specifically includes the following steps: S1: taking the input method data source as the target group data source; S2: constructing at least one image with at least one specific tendency according to the input method data source and actual needs S3: Match the input method data source of the population to be analyzed with the feature database, and filter out the target population with specific tendencies; S4: Extract the input original text data of the target population, and construct a single Channel target group portrait library; S5: Export individual identification information, and associate the individual's online accounts on all Internet platforms; S6: Cross-Internet platform network data fusion, fusion of heterogeneous network data of each Internet platform, according to The analysis results adjust the specific tendency of the individual and form a multi-dimensional portrait library of the specific tendency target population, and adjust and improve the related tendency classification feature library in S2.

Description

A method and system for constructing multi-dimensional portraits of a specific tendency group

technical field

The invention relates to a method for constructing a multi-dimensional portrait of a specific tendency group.

The invention also relates to a multi-dimensional portrait construction system of a specific tendency group.

Background technique

With the popularization of mobile Internet and social media, the dissemination of various illegal information has intensified, the means of dissemination have become increasingly concealed, and the insufficiency of existing monitoring methods has become increasingly prominent. The development of the Internet has brought unprecedented challenges for relevant departments to carry out special work such as anti-illegal activities. The popularization of various network applications, including cloud disks, instant messaging, forums and blogs, and online finance, makes people with specific illegal tendencies make extensive use of private and group-based network tools to incite, recruit, organize, plan and The implementation of illegal activities has brought great convenience. The traditional method of collecting web page data based on web crawlers has many problems such as difficulty in discovery, difficulty in tracking, insufficient deterrence, and high cost. It is necessary to find another way to innovate a new mode of special information management and control, and effectively improve the ability to perceive all kinds of illegal information. Big data technology has become one of the important technical means to transform social governance.

Due to the intensified dissemination of various types of illegal and illegal network public nuisance information, the means of dissemination have become increasingly concealed, and the insufficiency of traditional monitoring methods has become increasingly prominent. Therefore, it is necessary to change thinking, find new ways, innovate online public nuisance information management and control modes such as illegal activities on the Internet, and effectively improve the perception and prevention and control of the production and dissemination of network public nuisance information. In recent years, the total amount of global Internet data has maintained a relatively high growth rate. On the one hand, the technical monitoring area is far less than the growth rate of the entire network data, and it is difficult to achieve results simply by adding human or technical resources; Monitoring blind spots mainly include closed communication platforms such as various social media, online disks, email groups, and instant messaging groups, as well as semi-closed communication platforms such as circle of friends, content sharing communities, and live broadcast bullet screens. As a result, the management and control of special harmful information is "fortified and difficult to prevent".

In addition, the dissemination of various types of illegal information is more organized and purposeful than conventional harmful information, and has obvious characteristics: content production is more concealed, dissemination circles are grouped and private, outbreaks are more random, and domestic and foreign exchanges are frequent. . These characteristics make it very difficult to identify and monitor such harmful information, and it is difficult to find the source and predict in advance. Only by thoroughly studying the law of production and dissemination of such illegal information can we avoid the dilemma of "finding a needle in a haystack" and achieve "targeted" and multiplier effects.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a multi-dimensional portrait construction method of a specific tendency group, which can quickly and accurately perceive the specific tendency group, and conduct multi-platform tracking and comprehensive research and judgment according to the individual identification information, and form an individual tendency group. The multi-dimensional image library is convenient for the research and tracking analysis of groups and individuals of tendentious groups, and it is also conducive to timely control of the production and dissemination sources of such illegal and harmful information, which is convenient for users to follow up on such illegal and harmful information and Its production and dissemination populations are managed in a timely manner in accordance with laws and regulations.

Another object of the present invention is to provide a multi-dimensional portrait construction system of a specific tendency group, which can realize the above method.

A method for constructing a multi-dimensional portrait of a specific tendency group of the present invention specifically includes the following steps:

S1: Use the input method data source as the data source of the population to be analyzed;

S2: Build at least one feature library according to the input method data source and actual requirements, and the feature library has a set of at least one specific tendency feature;

S3: Match the input method data source of the population to be analyzed with the feature database, and screen out the target population with at least one specific tendency;

S4: Extract the input original text data of the target group, and construct a single-channel target group portrait library based on the input method;

S5: Export the individual identification information in the single-channel target group portrait database, push it to each Internet platform, and associate the individual's online accounts on all Internet platforms;

S6: Cross-Internet platform network data fusion, fuse the heterogeneous network data of each Internet platform in S5; adjust the specific tendency of the individual according to the analysis results and form a multi-dimensional portrait library of the target population with specific tendency, and analyze the data in S2. The related propensity grading feature library can be adjusted and improved.

Using the above method, people with specific tendencies can be quickly and accurately sensed, and multi-platform tracking and comprehensive research and judgment can be carried out according to individual identification information, and a multi-dimensional portrait library of individual tendent groups can be formed, which is convenient for the analysis of tendent groups. The research and tracking analysis of groups and individuals is also conducive to timely control of the source of production and dissemination of such illegal and harmful information, and it is convenient for users to manage such illegal and harmful information and its production and dissemination groups in a timely manner in accordance with laws and regulations.

Further, step S5 includes:

S5.1 Export the individual identification information in the single-channel target group portrait database, including but not limited to: the device identification code used by the user when using the input method, the registered mobile phone number of the input method, and the registered mailbox of the input method;

S5.2 Each Internet platform checks the identification information of each individual, and provides the account and related data information of the individual on its platform, obtains the possible online accounts of the target group on each Internet platform and associates them with each other to form the target group Network account library.

The above method is used to associate individual Internet accounts with individual identification information such as device identification code, input method registered mobile phone number, and input method registered email address, and conduct multi-platform tracking and analysis of the target group of individuals, so as to more comprehensively analyze the individual Track and manage relevant information.

Step S6 includes:

S6.1 Data fusion is performed on the input method data source of each individual in the target group's network account database and the data information of each Internet platform;

S6.2 conducts comprehensive research and judgment according to the fusion data of S6.1, and adjusts the specific tendencies and/or levels of the individual target groups to form a hierarchical multi-dimensional portrait library of the target groups with specific tendencies. The feature library is adjusted and improved.

The above method is used to conduct comprehensive research and judgment through the fusion of data, so as to form an accurate hierarchical multi-dimensional portrait database of specific tendentious target groups based on relatively comprehensive Internet data resources, and adjust and improve the corresponding feature database accordingly to provide screening. precision.

It also includes step S7: user business support, based on a hierarchical multi-dimensional portrait database of specific tendentious target groups, and developing different usage models according to the actual situation, including but not limited to: entity discovery, target activity trajectory restoration and tracking, accompanying relationship analysis, information traceability and Diffusion analysis, social relationship network restoration and social relationship network mining or other standardized basic data analysis models.

Using the above method, use the fused data to set up different usage models in subsequent research and development to meet the complex and diverse user needs

Step S1 includes:

S1.1 Multi-source data collection: collect data based on different input methods, including but not limited to: input text, input time, input platform, device identification code, registered account;

S1.2 Multi-source heterogeneous data processing: preprocess the collected data, and remove noise information or blank information according to the cleaning mechanism or screening mechanism;

S1.3 Establishment of input method data source base library: build input method data source base library based on the preprocessed data source, perform storage management and establish a query and retrieval mechanism.

Using the above method, on the one hand, different input data sources are collected by collecting different types of input method tools, and the data includes but not limited to input text, input time, input platform, device identification code, registered account, etc. The data thus collected has The multi-source and richness makes the data relatively complete and comprehensive, and also facilitates the subsequent analysis and management of the target population; effectiveness.

Step S2 includes:

S2.1 Initial feature library construction: build an initial feature library with at least one specific tendency according to actual needs;

S2.2 Grading identification: attribute classification and identification of a specific tendency in the initial feature library according to the grading standard, and can distinguish the degree of tendency when characterizing a specific tendency feature;

S2.3 Feature Library Supplement: Analyze and study the text information in S6, and add the newly discovered text information used to characterize specific tendency features to the relevant specific tendency feature library;

S2.4 Feature library adjustment: According to the text information input by the input method of the target group in S6, check the text information representing the specific tendency feature, and adjust the content and graded identification of the corresponding specific tendency feature library according to the check result. .

Using the above method, an initial feature database with at least one set of specific tendency features is initially constructed, and the target population is graded and identified according to the frequency of occurrence of the relevant feature words or the feature words of the target population in the feature database, so that In the follow-up, different management can be carried out according to its identification level; in addition, the feature library and classification standard are adjusted and improved according to the final matching result, so as to continuously improve the judgment and analysis accuracy of this method. At the same time, intelligently analyze the input method raw data of some target groups with specific inclinations, and extract new words that reflect specific inclinations, including specific terms, code words, slang, jargon, slogans, abbreviations, and abbreviations with a certain degree of distinction. , aliases and other text information and their combinations, or text information with specific tendent meanings such as names, place names, organization names and their combinations to construct corresponding extraction algorithms, discover new words and cryptic terms in a timely manner, and continuously update and improve the feature database. Improve the judgment analysis accuracy and screening speed of this method.

The present invention also provides a multi-dimensional portrait construction system for a specific tendency group, including:

The input method data source subsystem collects and stores the input method data information of the population to be analyzed;

A propensity feature library subsystem, having at least one specific propensity feature and a collection of feature ranking information; and

The propensity matching subsystem compares the data information in the input method data source subsystem with the characteristic information in the propensity feature library subsystem, marks the people with a specific tendency, and filters out at least one specific tendencies of the target population;

The target group's network account subsystem, which associates the individual's online accounts on all Internet platforms according to the target group's individual identification information;

The cross-Internet platform network data fusion subsystem integrates the heterogeneous network data of each Internet platform, adjusts the specific tendencies of the individual according to the analysis results, and forms a multi-dimensional portrait database of the target population with specific tendencies, and classifies the characteristics of related tendencies. library to adjust and improve.

The target group network account subsystem includes:

Target group individual identification information database, from the input method-based single-channel target group portrait library of the propensity matching subsystem, output the device identification code, registered mobile phone number, registered email address or other individual identification types involved when the individual uses the input method information, and collect it to form the database; and based on this, Internet companies including video websites, cloud disks, social networks and other Internet companies are required to conduct their own investigations on these individual identification categories;

The target group network account database, each Internet company conducts investigations based on the target group’s individual identification information database, and collects the network accounts that may be registered by the target group individuals with one or more specific tendencies on the Internet platform operated by them to form Target population network account database.

The cross-platform network data fusion subsystem includes:

Multi-source heterogeneous network data fusion module, according to the target population network account database, each Internet platform requires the network data of the relevant account, and analyzes the heterogeneous network data of individuals with specific tendencies on multiple Internet platforms and their input method data sources. Data Fusion;

The comprehensive tendency judgment and classification module conducts comprehensive research and judgment on the cross-platform fusion data of individuals, conducts secondary confirmation of its specific tendency and its classification, and adjusts and adjusts the relevant tendency classification feature library in the tendency feature library subsystem. Complete;

The hierarchical multi-dimensional portrait library of specific tendentious target groups is based on relatively comprehensive Internet data resources, and is formed according to the analysis and confirmation results of the comprehensive tendentiousness judgment and the grading module.

Also included is a propensity feature discovery and verification subsystem, which includes:

The original input text library of the target group, extracting the original text data of the target group input in the single-channel target group portrait library based on the input method in the propensity matching subsystem, as the basic data source of this subsystem;

The tendency feature discovery module extracts the text information that is used for representation in the original input text library of the target group and is not entered in the tendency feature library subsystem. After verification, the text information is added to the corresponding feature library as a new feature. middle;

The propensity feature verification module conducts secondary comprehensive research and judgment on the input original text data of the target population of the single-channel target population portrait library based on the input method in the propensity matching subsystem, and verifies the propensity and its characteristics of the target population. Grading. And according to the verification conclusion, adjust the content and classification mark of the tendency feature library subsystem.

It also includes the user service support subsystem, which includes:

Standardized basic data analysis model library, based on the hierarchical multi-dimensional portrait library of specific tendentious target groups as data support, develop data analysis modules with small granularity, low coupling degree and unified interface, each analysis module operates independently for data analysis and standardized input and output interfaces , forming the standardized basic data analysis model library;

The user business support module library is based on the standardized basic data analysis model library, combined with the user's own business needs, and forms the required user business support modules by assembling each analysis module.

In this embodiment, the core database formed by the system is based on the hierarchical multi-dimensional portrait library of specific tendent target groups, which supports users to carry out diversified business work.

The invention takes the relevant data obtained by the input method tool as the basic data source, can quickly and accurately perceive the crowd with a specific tendency, and conduct multi-platform tracking and comprehensive research and judgment according to the individual identification information, and form a multi-dimensional portrait of the individual tendency group It is convenient for the research and tracking analysis of groups and individuals of tendentious populations, and it is also conducive to timely control of relevant population information, which is convenient for subsequent analysis and management; Conduct comprehensive research and judgment on tendencies and their grading, so as to further confirm and adjust the specific tendencies of target groups and their tendencies grading; input method data sources can be used to perceive people as various tendencies, especially Categorizing the dissemination groups of dangerous information is conducive to controlling the source of dissemination of dangerous information in a timely manner, and facilitates subsequent timely processing of dangerous information and dissemination groups of dangerous information.

Description of drawings

FIG. 1 is a schematic flowchart of a method for constructing a multi-dimensional portrait of a specific tendency group according to the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. The indicated orientation or positional relationship is based on the orientation or positional relationship shown in the accompanying drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the indicated device or element must have a specific orientation or a specific orientation. construction and operation, and therefore should not be construed as limiting the invention; the terms "first", "second", "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance; furthermore, unless otherwise Clearly stipulated and defined, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or a Electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and it can be the internal connection of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood in specific situations.

Example 1

As shown in FIG. 1 , this embodiment provides a method for constructing a multi-dimensional portrait of a specific tendency group, which specifically includes the following steps:

S1: Use the input method data source as the data source of the population to be analyzed;

S2: constructing at least one feature library according to the input method data source and actual requirements, the feature library having at least one set of specific tendency features;

S3: Match the input method data source of the population to be analyzed with the feature database, and screen out the target population with at least one specific tendency;

S4: Extract the input original text data of the target group, and construct a single-channel target group portrait library based on the input method;

S5: Export the individual identification information in the single-channel target group portrait database, push it to each Internet platform, and associate the individual's online accounts on all Internet platforms;

S6: Cross-Internet platform network data fusion, integrate the heterogeneous network data of each Internet platform in S5, adjust the specific tendency of the individual according to the analysis results, and form a multi-dimensional portrait library of the target population with specific tendency, and analyze the data in S2. The related propensity grading feature library can be adjusted and improved.

For example, in a real situation, a target individual may seldom input text information related to a specific tendency when using the input method, but he may share relevant resources with a specific tendency through programs such as cloud disks, etc. The tracked data information is adjusted to the specific tendency and attribute level of the individual to further improve the accuracy or precision of the screened target population.

In this embodiment, the relevant data obtained by the input method tool is used as the basic data source to screen and classify groups of people with a specific tendency, which is conducive to timely control of relevant group information and facilitates subsequent analysis and management; it is relatively For other analysis methods, the screening accuracy is high, the target population is highly identifiable, and the follow-up management and tracking are convenient. Conduct comprehensive research and judgment, so as to further confirm and adjust the specific tendencies of the target population and their tendencies grading; at the same time, conduct multi-platform tracking and comprehensive research and judgment based on individual identification information, and form a multi-dimensional portrait library of individual tendencies. , so as to facilitate the research and tracking analysis of groups and individuals of tendentious populations; this embodiment classifies and divides the population of dangerous information dissemination through multi-dimensional perception of the crowd, which is conducive to timely control of the source of dangerous information dissemination and facilitates follow-up to dangerous information. Information and dangerous information dissemination groups should be dealt with in a timely manner.

In this embodiment, the input method data source is obtained under the condition of legal compliance, and the basic data source obtained by the input method tool has the following characteristics compared with the traditional data acquisition method based on web page collection (such as web crawler, etc.) Advantage:

1. The data sources are comprehensive and stable. The input method is a necessary software for netizens to input text and publish information online. Whether they use forums, Weibo, QQ, or WeChat, they are inseparable from the support of the input method. It has wide user coverage, high user stickiness, low replacement rate, and permissions. High level features. In order to improve the user experience, each input method has a built-in data collection function, that is, when the user uses the Internet device, the input method will automatically collect the user's input content, device identification code, the network platform used and other information and upload it to the background database. What needs to be explained here is: when the input method company installs the input method tool, the user needs to confirm its lengthy "user agreement" before further installation, and the "user agreement" has clear collection of relevant information, that is, the input method is Collect data with the user's permission.

Second, the input method data supply side is relatively concentrated. At present, the input method software with the highest market share monopolizes the vast majority of the market share, which provides convenience for coordinating data sources.

Third, the data types are more comprehensive. In order to improve the analysis ability and user experience, each input method has built-in data collection function, that is, when the user uses the software, the software will automatically collect the user's input time, input text content, device identification code, input network platform and other information. And upload it to the background database for storage, which provides important support for users to carry out data analysis.

In this embodiment, the method is used for the screening and tracking of network crowds with illegal activities such as telecommunications, and the relevant responsible departments can adopt different management methods for key individuals according to the analysis results, thereby improving the supervision efficiency of the above-mentioned dangerous tendency groups. and track processing speed. That is to say, this embodiment intelligently classifies and divides relevant groups of people with specific tendencies, which is beneficial to timely control the source of production and dissemination of such illegal and harmful information, and facilitates users to follow up on such illegal and harmful information and its production and dissemination groups in a timely manner. Carry out management in accordance with laws and regulations.

Further, step S5 includes:

S5.1 Export the individual identification information in the single-channel target group portrait database, including but not limited to: the device identification code used by the user when using the input method, the registered mobile phone number of the input method, and the registered mailbox of the input method;

S5.2 Each Internet platform checks the identification information of each individual, and provides the account and related data information of the individual on its platform, obtains the possible online accounts of the target group on each Internet platform and associates them with each other to form the target group Network account library.

In this embodiment, individual identification information such as device identification code, input method registered mobile phone number, and input method registered mailbox is used to associate individual Internet accounts, and to perform multi-platform tracking and analysis on individuals of the target population, so as to more comprehensively analyze the correlation between individuals. Information is tracked and managed.

Step S6 includes:

S6.1 Data fusion is performed on the input method data source of each individual in the target group's network account database and the data information of each Internet platform;

S6.2 conducts comprehensive research and judgment according to the fusion data of S6.1, and adjusts the specific tendencies and/or levels of the individual target groups to form a hierarchical multi-dimensional portrait library of the target groups with specific tendencies. The feature library is adjusted and improved.

In this embodiment, comprehensive research and judgment is carried out by fusing data, so as to form an accurate hierarchical multi-dimensional portrait library of specific tendentious target groups based on relatively comprehensive Internet data resources, and adjust and improve the corresponding feature library accordingly, so as to provide accurate screening. Spend.

It also includes step S7: user business support, using the data fused in S5 to develop different usage models according to the actual situation, including but not limited to: entity discovery, target activity trajectory restoration and tracking, accompanying relationship analysis, information traceability and diffusion analysis, social relationship Network restoration and social relationship network mining or other standardized basic data analysis models. These models can also be assembled and assembled by custom, including but not limited to: ①sensitive topic discovery and tracking module; ②gang discovery and control module; ③gang organization structure and personnel relationship analysis module; ④gang regional change monitoring module ; ⑤ group behavior early warning module; ⑥ domestic and foreign specific tendentious personnel collusion monitoring module; ⑦ information visualization and monitoring module to meet their complex and diverse business needs. This embodiment uses the fused data to set up different usage models in subsequent research and development to meet complex and diverse user usage requirements.

Step S1 of this embodiment includes:

S1.1 Multi-source data collection: collect data based on different input methods, including but not limited to: input text, input time, input platform, device identification code, registered account;

S1.2 Multi-source heterogeneous data processing: preprocess the collected data, and remove noise information or blank information according to the cleaning mechanism or screening mechanism;

S1.3 Establishment of input method data source base library: build input method data source base library based on the preprocessed data source, carry out storage management and establish a query and retrieval mechanism.

Using the above method, on the one hand, different input data sources are collected by collecting different types of input method tools, the data includes but not limited to input text, input time, input platform, device identification code, registered account, etc. The data thus collected has The multi-source and richness makes the data more complete and comprehensive, and it is also convenient for the subsequent analysis and management of the target population; effectiveness.

Step S2 includes:

S2.1 Initial feature library construction: build an initial feature library with at least one specific tendency according to actual needs;

S2.2 Grading identification: attribute classification and identification of a specific tendency in the initial feature database according to the grading standard, and can distinguish the degree of tendency when characterizing a specific tendency feature;

S2.3 Feature Library Supplement: Analyze and study the text information in S6, and add the newly discovered text information used to characterize specific tendency features to the relevant specific tendency feature library;

S2.4 Feature library adjustment: According to the text information input by the input method of the target group in S6, check the text information representing the specific tendency feature, and adjust the content and graded identification of the corresponding specific tendency feature library according to the check result. .

In layman's terms, the specific scheme of this embodiment is: S1: use the input method data source as the data source of the population to be analyzed;

S2: Build at least one feature library according to the input method data source and actual requirements, the feature library has a set of at least one specific tendency feature, each specific tendency feature has a graded identification, and each tendency feature library can be continuously updated and iterated ;

The definition of the so-called specific tendency characteristics: including specific terms, code words, slang, jargon, slogans, abbreviations, abbreviations, aliases and other text information and their combinations with a certain degree of distinction, or names of people, places, and organizations with specific tendency meanings textual information such as names and combinations thereof;

The so-called graded identification of a specific tendency feature refers to the distinction of the "propensity degree" embodied by the feature when it represents a specific tendency.

S3: Match the input method data source of the population to be analyzed with the feature database, perform matching of various specific tendencies of the population and their classification, and screen out the target population with at least one specific tendencies. The grading mark grades the target population.

In this embodiment, an initial feature database with at least one set of specific tendency features is initially constructed, and the target population is classified and identified according to the frequency of occurrence of related feature words or feature words of the target population in the feature database. Differential management can be carried out according to its identification level; in addition, the feature library and grading standard are adjusted and improved according to the final matching result, so as to continuously improve the judgment and analysis accuracy of this method. At the same time, intelligently analyze the input method raw data of some target groups with specific inclinations, and extract new words that reflect specific inclinations, including specific terms, code words, slang, jargon, slogans, abbreviations, and abbreviations with a certain degree of distinction. , aliases and other text information and their combinations, or text information with specific tendent meanings such as names, place names, organization names and their combinations to construct corresponding extraction algorithms, discover new words and cryptic terms in a timely manner, and continuously update and improve the feature database. Improve the judgment analysis accuracy and screening speed of this method.

To sum up, in this embodiment, multi-platform tracking analysis is carried out on the individuals of the target group, and their specific tendencies and their grading are comprehensively judged, so as to realize the specific tendencies of the target groups and their tendencies grading. Make further confirmation and adjustments; at the same time, you can also use the fused data to set up different usage models for subsequent research and development to meet complex and diverse user needs.

Example 2

This embodiment provides a multi-dimensional portrait construction system for a specific tendency group, including:

The input method data source subsystem collects and stores the input method data information of the population to be analyzed;

A propensity feature library subsystem, having a set of feature information for at least one specific propensity; and

The propensity matching subsystem compares the data information in the input method data source subsystem with the characteristic information in the propensity feature library subsystem, marks the people with a specific tendency, and filters out at least one specific tendencies of the target population;

The target group's network account subsystem, which associates the individual's online accounts on all Internet platforms according to the target group's individual identification information;

The cross-Internet platform network data fusion subsystem integrates the heterogeneous network data of each Internet platform, adjusts the specific tendencies of the individual according to the analysis results, and forms a multi-dimensional portrait library of the target population with specific tendencies, and classifies the related tendencies. library to adjust and improve.

Wherein, the input method data subsystem includes

The multi-source data collection module collects various input method data of the population to be analyzed. The input method data includes but is not limited to: input text, input time, input platform, device identification code, registered account;

The multi-source data preprocessing module preprocesses the collected data and removes noise information or blank information according to the cleaning mechanism or screening mechanism; and the input method database module, which builds the input method data source foundation based on the preprocessed data source database, carry out storage management and establish a query and retrieval mechanism, and provide a variety of data interfaces to support the mining of specific tendentious groups;

The tendency feature library subsystem includes

A feature library initialization module, which builds an initial feature library with at least one specific tendency feature or imports an original initial feature library according to actual needs; and

The feature grading and identification module is used to classify and identify a specific tendency of the initial feature library according to the grading standard;

The tendency feature library subsystem updates or perfects the feature library initialization module and the feature classification identification module according to the matching structure of the tendency matching subsystem;

The propensity matching subsystem includes:

The crowd-specific tendency matching module performs correlation analysis between the data information of the input method database module in the input method data source subsystem and the characteristic information of the feature grading identification module in the tendency feature library subsystem, and analyzes the characteristics of one or more Marking groups of people with at least one specific tendency, screening out target groups with at least one specific tendency, and grading and identifying their specific tendency;

The single-channel target group portrait library based on the input method, the text information input by the input method of the target group individual obtained by the group-specific tendency matching module is combined with the input time, input platform, device identification code, registered account and other information. Data fusion and analysis form a single-channel target population portrait library.

In this embodiment, the input method data subsystem, on the one hand, collects different input data sources by collecting different types of input method tools, and the data includes but is not limited to input text, input time, input platform, device identification code , registered accounts, etc. The data collected in this way has multiple sources and richness, making the data relatively complete and comprehensive, and also facilitating the subsequent analysis and management of the target population; on the other hand, the data is cleaned or screened before processing to remove noise. information or blank information to improve the validity of the population data to be analyzed. The tendency feature library subsystem initially constructs an initial feature library with at least one specific tendency, and classifies and identifies the target population according to the frequency of occurrence of the relevant feature words or the feature words of the target population in the feature database, so that Subsequent management can be differentiated according to its identification level; in addition, according to the final matching result, the feature library and classification standard are adjusted and improved, so as to continuously improve the judgment and analysis accuracy of this method.

That is to say, in this embodiment, the relevant data obtained by the input method tool is used as the basic data source, and the crowds with a certain tendency are screened and classified, which is conducive to timely control of the relevant crowd information and facilitates subsequent analysis and management; Compared with other analysis methods, the screening accuracy is high, the target population is highly recognizable, and the follow-up management and tracking are convenient; Comprehensively research and judge the situation, so as to further confirm and adjust the specific tendencies of the target population and their tendencies classification; this embodiment classifies and divides the population of dangerous information dissemination through the multi-dimensional perception of the population, which is conducive to timely control The source of the dissemination of dangerous information is convenient for subsequent timely processing of dangerous information and the dissemination of dangerous information.

The target group network account subsystem includes:

Target group individual identification information database, from the input method-based single-channel target group portrait library of the propensity matching subsystem, output the device identification code, registered mobile phone number, registered email address or other individual identification types involved when the individual uses the input method information, and aggregated to form the database;

The target group network account database, each Internet company conducts investigations based on the target group’s individual identification information database, and collects the network accounts that may be registered by the target group individuals with one or more specific tendencies on the Internet platform operated by them to form Target population network account database.

In this embodiment, the single-channel target group portrait library based on the input method of the propensity matching subsystem outputs the device identification code, registered mobile phone number, registered mailbox and other individual identification information that each character uses when using the input method. The database is assembled and based on this, Internet companies including video websites, cloud disks, and social networks are required to conduct their own investigations on these individual identification categories. After checking the individual identification information database of the target group, each Internet enterprise collects the online accounts that may be registered on the Internet platform operated by the target group with one or more specific tendencies, so as to form a comprehensive network account of the target group. The database can be used to track and manage individual-related information more comprehensively in the future.

The cross-platform network data fusion subsystem includes:

Multi-source heterogeneous network data fusion module, according to the target population network account database, each Internet platform requires the network data of the relevant account, and analyzes the heterogeneous network data of individuals with specific tendencies on multiple Internet platforms and their input method data sources. Data Fusion;

The comprehensive tendency judgment and classification module conducts comprehensive research and judgment on the cross-platform fusion data of individuals, conducts secondary confirmation of its specific tendency and its classification, and adjusts and adjusts the relevant tendency classification feature library in the tendency feature library subsystem. Complete;

The hierarchical multi-dimensional portrait library of the specific tendency target group is formed according to the analysis and confirmation results of the comprehensive tendency judgment and classification module.

In this embodiment, the formed target population network account database is handed over to each Internet platform to provide the network data of the relevant accounts, and the heterogeneous network data of multiple Internet platforms is fused, and finally generated after comprehensive tendency judgment and classification A high-accuracy multi-dimensional portrait library of specific tendent target groups.

Further, this embodiment also includes a tendency feature discovery and verification subsystem, which includes:

The original input text library of the target group, extracting the original text data of the target group input in the single-channel target group portrait library based on the input method in the propensity matching subsystem, as the basic data source of this subsystem;

The tendency feature discovery module extracts the text information that is used for representation in the original input text library of the target group and is not entered in the tendency feature library subsystem. After verification, the text information is added to the corresponding feature library as a new feature. middle;

The propensity feature verification module conducts secondary comprehensive research and judgment on the input original text data of the target population of the single-channel target population portrait library based on the input method in the propensity matching subsystem, and verifies the propensity and its characteristics of the target population. grading, and adjust the content and grading identification of the tendency feature library subsystem according to the verification conclusion.

In this embodiment, a specific tendency feature that is not stored is extracted through the tendency feature discovery module, and is added to the corresponding feature library as a new feature after verification. This method is used to intelligently analyze the input method raw data of target groups with specific tendencies, and extract the newly discovered specific tendencies features, so as to discover new words, obscure terms and their combinations in time, and constantly update and improve specific tendencies. The feature library is used to improve the accuracy and adaptability of the judgment and analysis of the method.

Through the propensity feature verification module, a secondary comprehensive judgment is made on the input original text data of the target population of S3, and the propensity and classification of these target populations are further verified. And based on these verification conclusions, the content and classification identification of the relevant tendency feature library in S2 are further adjusted, so as to continuously improve the tendency feature library of S2, so as to improve the accuracy and adaptability of the judgment and analysis of this method.

Further, this embodiment also includes a user service support subsystem, which includes:

Standardized basic data analysis model library, based on the hierarchical multi-dimensional portrait library of specific tendentious target groups as data support, develop data analysis modules with small granularity, low coupling degree and unified interface, each analysis module operates independently for data analysis and standardized input and output interfaces , to form the user business support module library of the standardized basic data analysis model library. Based on the standardized basic data analysis model library, combined with the user's own business needs, the required user business support modules are formed by assembling each analysis module.

In this embodiment, on the one hand, a data analysis component with small granularity, low coupling degree and unified interface is developed based on the data support of the hierarchical multi-dimensional portrait database of specific tendentious target groups. Each analysis model can run independently for data analysis and provide important Intelligence clues can also use standardized input and output interfaces to form a standardized basic data analysis model library, which can be developed including but not limited to "entity discovery, target activity trajectory restoration and tracking, accompanying relationship analysis, information traceability and diffusion analysis, social relations. Standardized basic data analysis models such as network restoration and social relationship network mining”; and write scripts according to predetermined standards to quickly assemble specific data analysis models to form a monitoring, control, and perception chain.

On the other hand, based on the standardized basic data analysis model library, combined with the user's own business needs, through the assembly of the above analysis modules, it can be formed including but not limited to sensitive topic discovery and tracking module, gang discovery and control module, gang organizational structure and personnel. User business support modules such as relationship analysis module, gang regional change monitoring module, group behavior early warning module, domestic and foreign specific tendentious personnel association monitoring module, information visualization and monitoring monitoring module.

To sum up, in this embodiment, the core database formed by the system is based on the hierarchical multi-dimensional portrait database of specific tendent target groups, which supports users to carry out diversified business work.

Example 3

The present embodiment provides a method for constructing a multi-dimensional portrait of a group of people with a tendency to, which specifically includes the following steps:

S1: Use the input method data source as the data source of the population to be analyzed;

S1.1 Multi-source data collection: collect data based on different input methods, including but not limited to: input text, input time, input platform, device identification code, registered account; the input methods include Tencent input method, Baidu input method, Sogou input method Input methods, etc.; the data collected in this way has multiple sources and richness, which makes the data relatively complete and comprehensive, and also facilitates the subsequent analysis and management of the target population;

S1.2 Multi-source heterogeneous data processing: Preprocess the collected data, and remove noise information or blank information according to the cleaning mechanism or screening mechanism; Mechanisms or screening mechanisms, such as removing stop words, too short or too long or even blank data, or specifying data source types to achieve cleaning or screening effects;

S1.3 Establishment of input method data source basic library: build input method data source basic library based on the preprocessed data source, carry out storage management and establish a query retrieval mechanism; realize efficient management, storage management and query retrieval of massive text information, so that for subsequent use of this data source and efficient use of multiple different analysis directions;

S2: construct a feature library according to the input method data source, and the feature library has a certain tendency characteristic;

S2.1 Initial feature library construction: build an initial feature library with at least a certain tendency characteristic according to regulatory requirements;

S2.2 Hierarchical identification: Attribute grading and identification according to the tendencies of the feature database of the regulatory department, so that the regulatory department can adopt different tracking or management methods for personnel at different levels, so as to deploy the supervision work in a focused manner;

S2.3 Feature Library Supplement: Analyze and study the text information in S6, and add the newly discovered text information used to characterize specific tendency features to the relevant specific tendency feature library;

S2.4 Feature library adjustment: According to the text information input by the input method of the target group in S6, check the text information representing the specific tendency feature, and adjust the content and graded identification of the corresponding specific tendency feature library according to the check result. ; Adjust and improve the feature library and grading standards according to the final matching results, so as to continuously improve the screening accuracy of the target population under supervision;

S3: Match the input method data source of the population to be analyzed with the feature database, and screen out the target population with a certain tendency; in order to clarify its risk level or danger, the regulatory department will take corresponding measures as needed;

S4: Extract the input original text data of the target group, and construct a single-channel target group portrait library based on the input method;

S5: Export the individual identification information in the single-channel target group portrait database, push it to each Internet platform, and associate the individual's online accounts on all Internet platforms;

Specifically, step S5 includes:

S5.1 Export the individual identification information in the single-channel target group portrait database, including but not limited to: the device identification code used by the user when using the input method, the registered mobile phone number of the input method, and the registered mailbox of the input method;

S5.2 Each Internet platform checks the identification information of each individual, and provides the account and related data information of the individual on its platform, obtains the possible online accounts of the target group on each Internet platform and associates them with each other to form the target group network account library;

S6: Cross-Internet platform network data fusion, integrate the heterogeneous network data of each Internet platform in S5, adjust the specific tendencies of the individual according to the analysis results and form a hierarchical multi-dimensional portrait library of the target population, and analyze the relevant information in S2. Adjust and improve the tendency classification feature library;

Specifically, step S6 includes:

S6.1 Data fusion is performed on the input method data source of each individual in the target group's network account database and the data information of each Internet platform;

S6.2 conducts comprehensive research and judgment based on the fusion data of S6.1, and classifies the inclinations of the individual target groups, forms a multi-dimensional portrait library for the classification of target groups, and adjusts and improves the predisposition feature library in S2.

In this embodiment, the relevant data obtained by the input method tool is used as the basic data source to screen and classify the people with tendencies, which is conducive to timely control of the relevant population information and facilitates subsequent analysis and management; compared with other analysis In terms of methods, the screening accuracy is high, the target population is highly identifiable, and the follow-up management and tracking are convenient, etc.; and the multi-platform tracking and analysis of the individuals of the target population is carried out, and their specific tendencies and their classification are comprehensively judged, so as to realize the Focus on the specific tendencies of character groups and further confirm and adjust their tendencies grading; classify and classify the dissemination of dangerous information, which is conducive to timely control of the source of dangerous information dissemination, and facilitates follow-up of dangerous information and dangerous information dissemination groups. deal with in a timely manner.

This embodiment intelligently analyzes the input method raw data of some target groups with a tendency to extract new words that reflect a specific tendency, including specific terms with a certain degree of discrimination, code words, slang words, jargon, slogans, abbreviations, and abbreviations. , aliases and other text information and their combinations, or text information with specific tendent meanings such as names, place names, organization names and their combinations to construct corresponding extraction algorithms, discover new words and cryptic terms in a timely manner, and continuously update and improve the feature database. Improve the accuracy of judgment and analysis of tendencies and the speed of screening.

The above content is a further detailed description of the present invention in combination with specific preferred embodiments, and it cannot be considered that the specific implementation of the present invention is limited to these descriptions. For those skilled in the technical field of the present invention, without departing from the concept of the present invention, several equivalent substitutions or obvious modifications can be made, and the performance or use is the same, which should be regarded as belonging to the protection scope of the present invention. Inside.

Claims (6)

1. A multi-dimensional image construction method for crowds with specific tendencies is characterized by comprising the following steps:

s1: taking an input method data source as a data source of a crowd to be analyzed;

s2: constructing at least one feature library according to an input method data source and actual requirements, wherein the feature library has at least one set of specific tendency features;

s3: matching an input method data source of a crowd to be analyzed with a feature library, and screening out a target crowd with at least one specific tendency;

s4: extracting input original text data of a target crowd, and constructing a single-channel target crowd image library based on an input method;

s5: deriving individual identification information in a single-channel target crowd image library, pushing the individual identification information to each internet platform, and associating network account numbers of the individual on all internet platforms;

s6: performing network data fusion across the Internet platforms, fusing heterogeneous network data of each Internet platform in S5, adjusting the specific tendency of the individual according to an analysis result, forming a multi-dimensional image library of a target crowd with the specific tendency, and adjusting and perfecting a related tendency grading characteristic library in S2;

step S5 includes:

s5.1, deriving individual identification class information in a single-channel target crowd image library, wherein the individual identification class information comprises but is not limited to: the method comprises the steps that when a user uses an input method, the user uses an equipment identification code, the input method registration mobile phone number and the input method registration mailbox;

s5.2, each Internet platform checks the identification information of each individual, provides the account number and related data information of the individual on the platform, and acquires the network account numbers possibly existing on each Internet platform of the target population and correlates the network account numbers to form a network account number library of the target population;

step S6 includes:

s6.1, performing data fusion on the data source of the input method of each individual in the target crowd network account database and the data information of each Internet platform;

s6.2, comprehensive research and judgment are carried out according to the fusion data of S6.1, the specific tendentiousness and/or the grade of the target crowd individual are/is adjusted, a specific tendentiousness target crowd grading multidimensional image library is formed, and the characteristic library of the relevant tendentiousness in S2 is adjusted and perfected.

2. The method for constructing a multi-dimensional image of a specific tendency crowd as claimed in claim 1, further comprising the step of S7: and (4) the user business support develops different use models according to actual situations by using the data fused in the S5, including but not limited to: entity discovery, target activity track restoration and tracking, companion relationship analysis, information tracing and diffusion analysis, social relationship network restoration and social relationship network mining or other standardized basic data analysis models.

3. The method of claim 1, wherein the multi-dimensional image of the crowd with a specific tendency is constructed,

step S1 includes:

s1.1, multi-source data acquisition: data is collected based on different input methods including, but not limited to: inputting a text, inputting time, inputting a platform where the text is located, an equipment identification code and a registered account;

s1.2, multi-source heterogeneous data processing: preprocessing the acquired data, and removing noise information or blank information according to a cleaning mechanism or a screening mechanism;

s1.3, establishing a data source base library of an input method: constructing an input method data source base based on the preprocessed data source, performing storage management and establishing a query retrieval mechanism;

step S2 includes:

s2.1, constructing an initial feature library: constructing an initial feature library with at least one specific tendency according to actual requirements;

s2.2 hierarchical identification: performing attribute classification and identification on a certain specific tendency of the initial feature library according to a classification standard, and distinguishing the tendency degree when a certain specific tendency feature is represented;

s2.3 feature library supplementation: analyzing and researching the text information in S6, and adding the newly found text information for representing the specific tendency characteristics into the related specific tendency characteristic library;

s2.4, adjusting a feature library: and according to the text information input by the input method of the target population in the S6, checking the text information for representing the specific tendency characteristics, and adjusting the content and the grading identification of the corresponding specific tendency characteristic library according to the checking result.

4. A system for constructing a multi-dimensional image of a population with a specific tendency, comprising:

the input method data source subsystem is used for collecting and storing the input method data information of the crowd to be analyzed;

a tendency feature library subsystem having a set of at least one specific tendency feature information; and

the tendency matching subsystem is used for comparing the data information in the input method data source subsystem with the characteristic information in the tendency characteristic library subsystem, marking the crowd with a certain specific tendency and screening out the target crowd with at least one specific tendency;

the target population network account subsystem associates the network accounts of the individual on all the Internet platforms according to the individual identification information of the target population;

the cross-internet platform network data fusion subsystem fuses heterogeneous network data of each internet platform, adjusts the specific tendency of the individual according to the analysis result, forms a specific tendency target crowd multi-dimensional image library, and adjusts and perfects a related tendency grading feature library;

the target crowd network account subsystem comprises:

the target population individual identification information database outputs equipment identification codes, registered mobile phone numbers, registered mailboxes or other individual identification information related to individual use of the input method from a single-channel target population image library based on the input method of the tendency matching subsystem, and integrates the equipment identification codes, the registered mobile phone numbers, the registered mailboxes or other individual identification information to form the database;

the system comprises a target group network account library, a target group individual identification information database, a target group network account library and a target group network account library, wherein each internet enterprise checks based on the target group individual identification information database, and collects network accounts which are possibly registered on an internet platform operated by a target group individual with a certain/a plurality of specific tendencies to form the target group network account library;

the cross-platform network data fusion subsystem comprises:

the multi-source heterogeneous network data fusion module is used for performing data fusion on the individual heterogeneous network data with specific tendencies on the plurality of internet platforms and the input method data sources thereof according to the network account database of the target crowd and the network data of the related accounts required by each internet platform;

the comprehensive tendency studying and judging and grading module is used for comprehensively studying and judging the cross-platform fusion data of the individuals, secondarily confirming the specific tendency and the grading thereof and adjusting and perfecting a related tendency grading feature library in the tendency feature library subsystem;

and forming the grading multidimensional image library of the target crowd with the specific tendency according to the analysis and confirmation results of the comprehensive tendency studying and judging and grading module.

5. The system of claim 4, further comprising a tendency feature discovery and verification subsystem, comprising:

extracting input original text data of a target crowd of a single-channel target crowd image library based on an input method in a tendency matching subsystem as a basic data source of the subsystem;

the tendency characteristic finding module is used for extracting text information which is used for representation and is not recorded in the tendency characteristic library subsystem from the original input text library of the target crowd, and supplementing the text information serving as a new characteristic into a corresponding characteristic library after verification;

and the tendency characteristic verification module is used for carrying out secondary comprehensive study and judgment on the input original text data of the target population of the single-channel target population image library based on the input method in the tendency matching subsystem, verifying the tendency and the grading of the target population, and adjusting the content and the grading identification of the tendency characteristic library subsystem according to the verification conclusion.

6. The system of claim 4, further comprising a user services support subsystem comprising:

the standardized basic data analysis model base is formed by taking a classified multidimensional image base of a specific tendency target crowd as a data support, developing data analysis modules with small granularity, low coupling degree and uniform interfaces, and independently operating each analysis module to perform data analysis and standardized input and output interfaces;

and the user service support module library is based on the standardized basic data analysis model library, combines the service requirements of the user and assembles each analysis module to form the required user service support module.

Images