CN108255933A

CN108255933A - A kind of social media dynamic event develops visual analysis method and system

Info

Publication number: CN108255933A
Application number: CN201711284129.XA
Authority: CN
Inventors: 袁晓如; 陈思明
Original assignee: Peking University
Current assignee: Peking University
Priority date: 2017-12-07
Filing date: 2017-12-07
Publication date: 2018-07-06

Abstract

The invention discloses a kind of social media dynamic events to develop visual analysis method and system, and for certain event, keyword is extracted from social media, and then according to keyword, social media data are obtained from social media；According to based on map metaphor processing rule, by the way of map maps, the social media data of input are converted into event and develop map E Map；It is described to be included based on map metaphor processing rule：City, cities and towns and river, formed structuring and include semantic projector space, and user can go to explore the projector space by interaction；It is described to be sent in the social media data of processing rule at least data below：Keyword, releasing news comprising keyword and forwarding information, user information dissemination.This method and system can support that, to the event visual analysis in social media, the visualization is capable of providing the high-level summary developed to event of user one and possesses the ability of further investigation data different aspect.

Description

A kind of social media dynamic event develops visual analysis method and system

Technical field

The present invention relates to the propagation analysis technical fields of event in social media, and in particular to a kind of social media dynamic thing Part develops visual analysis method and system.Social media dynamic event espespecially based on map metaphor develop visual analysis method and System.The event refers to social event.It is described based on map metaphor be the visualization metaphor based on map.

Background technology

The social event (referred to as event) issued in social media, propagated, with the propagation (diffusion of information) of event, It forms and is developed, such as incipient stage, intermediate stage, ending phase etc. by the event that several stages are formed, in order to fully manage It solves and talks the matter over the behavior of the social media user (referred to as user) developed and in event differentiation (such as information passes Broadcast behavior), it would be desirable to there is the ability for analyzing dynamic diffusion of information.

In the research work of the existing dynamic diffusion of information of analysis, present case is such：

The propagation path of some researchers from a microblogging or theme (being also known as topic) start with and visualized [1, 2].They can explore the propagation path of some specifying information well, which derives from a microblogging Or a theme, but be difficult the different phase of the general view and the event differentiation that obtain the overall situation developed to event.

Researcher also is had studied about keyword in social media as time change [3], theme are as the time drills Become the emotion of [4,5] and user with the problem of time-evolution [6], and propose corresponding visual analysis method.They Method has used the not same order for based on river metaphor (the visualization metaphor i.e. based on river), explicitly defining event differentiation Section, visualization space main projection is on the two dimensional surface of screen space.However, since two dimensional surface is there are two dimension direction, It is needed based on river metaphor with a dimension direction creation time axis, other information can only be compressed to another dimension direction On, such as：

With lateral dimension direction creation time axis, then other information can only on the dimension direction of longitudinal direction, it is described its His information includes：In social media between personage (referring to key person) relationship, the forwarding relationship of information, keyword and theme Correlativity etc. so, is difficult to find out forwarding relationship and semantic change of the social media user in different time sections Change.

The present invention relates to following technical terms：

Dynamic diffusion of information, refers to：In social media, after a user issues certain information for event, the information It can be forwarded by other users, be forwarded with this and further by more other users.This forwarding row of the other users To generate forwarding chain, leading to dynamic diffusion of information, also can be regarded as the dynamic diffusion of information.

In event differentiation, the meaning of dynamic diffusion of information and effect are：Constantly allow more users in short time Know the situation and progress of event, and each user can serve as participating in main body, is commented on and is retouched in forwarding information It states, more generations for commenting on views and comments content can influence the trend of event differentiation with the time in social media.

Global general view, refers to：It is timely to the trend for the user group of forwarding information being participated in social media, event develops Between the progress totality such as trend assurance.

The different phase that event develops, refers to：Starting, development, climax and the extinction of event.

Based on river metaphor, refer to：Using the shape in similar river, the temporal behavior in information is visually given expression to.

Below with reference to document involved in above-mentioned：

[1]N.Cao,Y.Lin,X.Sun,D.Lazer,S.Liu and H.Qu.“Whisper:Tracing the Spat iotemporal Process of Informat ion Diffusion in Real Time”.IEEE Transactions on Visualization and Computer Graphics,2012,18(12):2649–2658.http:// doi.ieeecomputersociety.org/10.1109/TVCG.2012.291.

[2]D.Ren,X.Zhang,Z.Wang,J.Li and X.Yuan.“WeiboEvents:A Crowd Sourcing Weibo Visual Analytic System”.In:Proc.of IEEE PacificVis Notes,2014:330–334.

[3]D.Archambault,D.Greene,P.Cunningham and N.Hurley.“ThemeCrowds: Multiresolution Sum-maries of Twitter Usage”.In:Proceedings of the 3rd International Workshop on Search and Mining User-generated Contents.Glasgow, Scotland,UK:ACM,2011:77–84.http://doi.acm.org/10.1145/2065023.2065041.

[4]W.Dou,X.Wang,D.Skau,W.Ribarsky and M.X.Zhou.“LeadLine:Interactive visual analysis of text data through event identification and exploration” .In:Proc.of IEEE Visual Analytics Science and Technology,2012:93–102.http:// dx.doi.org/10.1109/VAST.2012.6400485.

[5]Y.Wu,S.Liu,K.Yan,M.Liu and F.Wu.“OpinionFlow:Visual Analysis of Opinion Diffusion on Social Media”.IEEE Transactions on Visualization and Computer Graphics,2014,20(12):1763–1772.http://dx.doi.org/10.1109/ TVCG.2014.2346920.

[6]J.Zhao,L.Gou,F.Wang and M.Zhou.“PEARL:An interactive visual analytic tool for understanding personal emotion style derived from social media”.In:Proc.of IEEE Visual Analytics Sci ence and Technology,2014:203–212.

In general, great event can be in social media by a large number of users community review and propagation (forwarding).Understand user It releases news, the information dissemination of forwarding information, better understood when that event develops.However, publication and forwarding information Dynamic and the diversity of the theme of user group's comment bring very big difficulty to event evolution analysis.

Event differentiation includes two parts, has information source first, that is, releases news, next has forwarding information.The two exists Whole event all always exists in developing.

In event, theme is important part, and the polymerization that user group comments on content is the theme, therefore, an event There may be multiple and different themes.Since information (word) from different users, includes different comments in social media Views and comments content, this diversity being the theme.

Invention content

For defect in the prior art, the purpose of the present invention is to provide a kind of differentiation of social media dynamic event Visual analysis method and system can be supported by this method to the event visual analysis (visualization) in social media, described Visualization is capable of providing the high-level summary developed to event of user one and possesses further investigation data different aspect Ability, the data refer to social media data.

To achieve the above object, the technical solution adopted by the present invention is as follows：

A kind of social media dynamic event develops visual analysis method, includes the following steps：

For certain event, keyword is extracted from social media, then according to keyword, is obtained from social media social Media data；

Rule is handled according to based on map metaphor, by the way of map maps, by the social media data conversion of input Map E-Map is developed for event；

It is described to be included based on map metaphor processing rule：City, cities and towns and river, formed structuring and comprising Semantic projector space, user can go to explore the projector space by interaction；

It is described to be sent in the social media data of processing rule at least data below：

Keyword,

Releasing news comprising keyword and forwarding information,

User information dissemination.

Further, a kind of social media dynamic event as described above develops visual analysis method, based on map metaphor In processing rule, semantic map and vision mapping are preset in the following manner, is specifically included：

The semanteme map includes：

Keyword is mapped as city,

Releasing news comprising keyword and forwarding information are mapped as cities and towns, and cities and towns are distributed around city,

User information dissemination, is mapped as river, corresponding to the contact between the track in space and city or region,

It releases news and the corresponding information subsection of forwarding information, is mapped as continent or island；

The Connected degree in region that continent and island are formed by city with cities and towns is determined.

Further, a kind of social media dynamic event as described above develops visual analysis method, and semantic map is using system One color mapping is mapped, and the time-varying attribute of affair character is characterized using color, and the unified color mapping includes But it is not limited to " inferno " color mapping of perceptibility equilibrium.

Further, a kind of social media dynamic event as described above develops visual analysis method, described to be converted to event Map is developed, including：Multiple entity figure constructs and event develops map generation.

Further, a kind of social media dynamic event as described above develops visual analysis method, and multiple entity figure is i.e. from society The graph structure extracted in media information is handed over, is described in social media about the keyword of some event, information and the phase of time Mutual relation；

In multiple entity figure construction, input is social media information M=(m1, m2 ..., mi ..., mnm), and output is one Multiple entity figure G=(N, E), wherein：

N=(K, M, T) includes keyword K, information M and virtual time point T,

E=(KK, KT, KM) includes the company side KK of keyword and keyword, keyword and the company side KT at time point and closes The company side KM of keyword and information.

Further, a kind of social media dynamic event as described above develops visual analysis method, during multiple entity figure constructs, Including two-wheeled layout processing, respectively initial layout and second stage layout.

Further, a kind of social media dynamic event as described above develops visual analysis method, and event develops map life Cheng Zhong, based on multiple entity figure G as input, ultimately generate according to the following steps with city, cities and towns, region, river, continent and The event on island develops map：

Node is placed, node includes：Seed point, city and cities and towns；

Construct city, cities and towns and region；

Construct river；

Merging/separation land constructs continent and island；

Map tuning.

A kind of social media dynamic event develops visual analysis system, including：

Keyword extracting module for being directed to certain event, extracts keyword from social media；

Social media data acquisition module, for according to keyword, social media data to be obtained from social media；

Event develops map E-Map modular converters, for handling rule according to based on map metaphor, using map maps The social media data of input are converted to event and develop map E-Map by mode.

Further, a kind of social media dynamic event as described above develops visual analysis system, and the event develops ground Scheme E-Map modular converters, including：

Multiple entity figure constructs submodule, for constructing multiple entity figure；Multiple entity figure is what is extracted from social media information Graph structure is described in social media about the keyword of some event, information and the correlation of time；

Event develops map generation submodule, and city, cities and towns, region, river, continent and island are carried for ultimately generating The event of small island develops map E-Map.

Further, a kind of social media dynamic event as described above develops visual analysis system, multiple entity figure constructor Module specifically includes：

Initial layout unit, for generating forwarding relationship and moving the keyword occurred in same time period to similar Position；

Second stage layout units for adjusting the position of node, make node loose as possible while state is kept building up It scatters and, further layout calculation is carried out for node M and side KiMj.

Further, a kind of social media dynamic event as described above develops visual analysis system, and event develops map life It is specifically included into submodule：

Node unit is placed, node includes：Seed point, city and cities and towns, seed point for so that map edge more Clearly so that the Voronoi grids of generation are more uniform, and city is directly projected on cities and towns position in Voronoi grids；

City, cities and towns and territory element are constructed, the position in city is placed for the position based on figure, for traversing They, are then assigned to nearest city by all polygonal mesh in Voronoi, to obtain the region model in each city Enclose, for finding the node that adjacent area belongs to different cities, connect into the boundary in region, for by cities and towns according to distance from The information issued earliest is placed from inside to outside, to the information issued the latest；

River unit is constructed, for source-purpose keyword to be carried out even side, by the Voronoi nets where corresponding city Regional location in lattice is connected, and for all region connections to be ranked up according to quantity, river is filtered out according to a parameter The number of stream only shows and connects mostly concerned river, for the river with same paths to be merged；

Continent and island unit are constructed, for being based on Density Distribution, region of the density value more than 0 is attached, bulk Join domain for continent, independent join domain is ocean, for extracting the edge in all continents and island；

Map tuning unit, it is corrosion simulated for being carried out to the edge in continent and island, for what river was each put A little and subsequent point is into row interpolation, and entire river is become relative smooth.

The beneficial effects of the present invention are：Realize following demand (Requirement, hereinafter, abbreviated as letter r)：

·R1:The semantic summary developed to event in time trend is provided.In this view, different themes are commented By should be detected and visualize presentation.This view is capable of providing the in-depth analysis of the different phase of event differentiation.

·R2:It is polymerize according to the forwarding relationship of theme and information.The distribution of different themes should visualize presentation, People is allowed to understand the correlation between the pouplarity and different themes of theme, theme.The forwarding relationship of information similarly needs It is summarised and visualizes presentation.

·R3:Theme in different time periods is presented with visualization in detection.The distribution of theme in different time periods should be strong It adjusts, and visually arranges, people to be allowed intuitively to understand the dynamic change of theme, so as to which people further be allowed to understand in difference The dynamic event of period develops.

·R4:Extract the information dissemination pattern of user.Understand the information dissemination of user (espespecially key person) It is particularly significant how event differentiation is influenced.Visualization, which is presented, needs that user is supported to explore problems with：Who is key person, crucial Personage has commented on any theme, and when key person in which way and participates in commenting on, the behavior of key person in terms of comprehensive How the trend of event differentiation is affected.

In the present invention, the semantic map for meeting more than demand R1~R3 is constructed, and provides a series of interaction sides Formula allows the exploration of User support demand R4.

Description of the drawings

Fig. 1 is that a kind of social media dynamic event provided in the specific embodiment of the invention develops visual analysis method Flow chart.

Fig. 2 is that a kind of social media dynamic event provided in the specific embodiment of the invention develops visual analysis system Structure diagram.

Fig. 3 is event evolving stage schematic diagram.

Fig. 4 is the design and main feature of semantic map.

Fig. 5 develops map flow diagram for tectonic event.

Fig. 6 is visual analysis system interface.

Fig. 7 is visual analysis flow diagram.

Specific embodiment

The present invention is described in further detail with specific embodiment with reference to the accompanying drawings of the specification.

The focus of the present invention is and the relevant social media data of event.

The present invention is defined as follows event.

Event is referred in social media, (true with the real event of certain hashtag (label) and key definition Social event in the world).I.e. event can regard the abbreviation of social event as.

The present invention is defined as follows social media.

Social media (Social Media) refers to contents production and switching plane based on customer relationship on internet, Such as Twitter and Sina weibo etc..

Most of social medias all allow user with hashtag to emphasize some theme or viewpoint (comment viewpoint).Usually One hashtag can be associated with event.For example, when user comments on what this year was assigned together by Trump in Twitter When White House's administrative decree (limit certain national people and enter the U.S.), many users have used #travelban (travellings Ban) this word.Certainly, although this is popular, indicate #travelban can not be represented as hashtag it is all about this The comment of a event, therefore when carrying out event related data collection, the search based on keyword still needs with reference to progress. For example, more, new information about this event can be provided by simultaneously scanning for travelban this word, in this citing, We can see that：Event refers to White House's administrative decree, hashtag #travelban, keyword travelban.

We are its variation with the time to the concern of event.It is understood that event differentiation usually has several stages, The different phase that event develops has identical or different user and plays crucial effect, and the user, that is, key person is (referred to as For personage), it may be possible to key person issues certain information to make a strong impact or certain information is largely forwarded by other users , evoked new comment, it is also possible to which event has new discovery and progress, is published to social matchmaker in time by key person In body, these factors can all influence event differentiation.During event develops, the user of participation, the theme of comment are all therewith Variation.

Based on this, it is believed that event can be defined by a series of feature, and the feature includes<Time, personage, Forwarding information, it is semantic>, these are the important features for describing an event.

As shown in figure 3, the example is event evolving stage schematic diagram：

From time shaft, event, which develops, can be divided into event stage 1~N, N=1,2, and 3 ..., it is clear that multiple event ranks In Duan Yi event, the new event stage may be from new information dissemination or new information.On the other hand, may be used It can be new round contention, amplification or the comment evoked by the new comment views and comments content of user in the new event stage.

From personage's axis, each event stage corresponds to a key person, and key person may be theme Guide or the people with a large amount of beans vermicelli, the trend that may change or guide event to develop.In addition, in event by Evil person or sounder may also can be the key persons of this event, the information that their statement, behavior even provide, all Affect the trend of event differentiation.

From forwarding information axis, some specifying information of key person's publication is forwarded by several other users, with letter Breath is propagated, and user can allow an event duration constantly by issuing or forward the information with keyword or hashtag Ferment in social media.Forwarding information causes more users to be seen, and then more ideas and contention occurs, this is directly Result in dynamic diffusion of information.

From semantic axis, the theme 1 of key person's publication can form theme 2~M, M=2,3 after diffusion of information, 4 ..., a certain theme before some is individually recited between these themes, multiple themes before some references, the master also having It is cited mutually between topic, different themes include different semantemes, and theme merges such as the time, division, withers away or in thing The different phase that part develops occurs.The theme of comment different users and in the event stage it is also very different.

It can be seen that：One event contains multiple event stages in developing, the releasing news of user, the letter of forwarding information Breath dissemination causes event differentiation and the comment to event.The theme of comment may be divided, merge and be fused into newly Theme.

The present invention is defined as follows social media data.

To the definition of social media data with the data instance of microblogging (such as Sina weibo), it is apparent that it should be realized that should Example, which is not configured to restriction social media data, can only derive from microblogging, and the method for the invention and system are to be common to a variety of societies Hand over media.

In microblogging, we use a keyword or hashtag into line search first, each microblogging searched As seed microblogging, the obvious more than one of quantity of seed microblogging, then we further collect all any one of forwardings The microblogging of seed microblogging.We can cover a large amount of microblogging in this way, it will be appreciated that theme is developed to be propagated with information.Wherein：Kind Sub- microblogging, that is, seed data, the microblogging of forwarding seed microblogging forward data.Obviously, seed microblogging is the starting point searched for, often It is comprising the information largely forwarded, we are crawled with this along the forwarding chain of the seed microblogging is forwarded to carry out data.

Every microblogging includes timestamp, and user id, information id forward user id, forwarding information id, content of text etc., I Can be according to user id and the forwarding tree of forwarding user id construction microblogging.

There are these seed datas and all forwarding data, the time that we can further analyze event is special Sign, personage's behavior pattern, forwarding information rule and semantic information.Therefore, social media data include：Seed data and all Forwarding data.Social media data belong to multi-modal information propagation data.

Social media dynamic event of the present invention develops visual analysis method, includes the following steps：

By event develop map allow user's fast understanding event various aspects and deeply understand event develop；It is logical It crosses event and develops map, space-time visual analysis that can be multi-level is explored, and user is allowed to find that event is drilled in event develops map Become how rule, regularity of information dissemination, the information dissemination pattern of key person and key person influence event differentiation Trend；

Keyword,

Releasing news comprising keyword and forwarding information,

User information dissemination.

Using based on the considerations of map metaphor have following 2 points it is main.

First point, map can give a structuring with carrying out organizational information with semantic space.Event and event Feature be all based on social media information extraction, social media information includes releasing news and forwarding information, usually these Information is all non-structured, and is difficult to other features that people is directly allowed to understand theme, event stage and event.One ground Figure can be with the different aspect of display data, such as the shape of object, color and landform.By relevant information MAP a to packet People can be allowed to more fully understand event in 2D spaces containing semanteme.

Second point, map are a kind of familiar modes.User is familiar with the object on map, including continent, city, river Stream, island and land, the relationship between them are also people to be allowed to understand naturally.If a map rationally designs, use Family can natively be familiar with the visual pattern of similar map.

Based on the above technical solution, as shown in figure 4, in based on map metaphor processing rule, in the following manner Default semanteme map and vision mapping, specifically include：

The semanteme map includes：

Keyword is mapped as city,

The Connected degree in region that continent and island are formed by city with cities and towns is determined；

To ensure that visualization is presented, semantic map is mapped using unified color mapping, and the unified color is reflected Penetrate " inferno " color mapping of including but not limited to perceptibility equilibrium.

Specifically：

E-Map use semantic map, formed based on social media data configuration, as raw information release news and All for constructing semantic map, the feature of wherein event has been mapped to semanteme forwarding information as the information extracted Geographical feature, it is described to include with semantic geographical feature：

City, a city represent a keyword, and the size in city represents the information content comprising the keyword How much, depending on intercity distance is by the degree of correlation of keyword, specifically includes forwarding relationship and while relationship occur.

These keywords are combined in most high-frequency key words from different time sections.One city represents geographical sky Between a panel region center, the Semantic center in information space is represented just as most high-frequency key words.In this way, user can be on ground The time trend of keyword is observed on figure.

Cities and towns, cities and towns represent the information for belonging to some city, and each cities and towns are pertaining only to a city, and cities and towns are scattered in it Affiliated peri-urban, the intercity distance belonging to cities and towns to it have mapped the relative time of information.

It is earliest to release news in down town in each cities and towns for belonging to a city, the latest release news from Down town is farthest.In this way, user can be with the Annual distribution and theme distribution of perception information.

Region, city and belongs to the cities and towns in the city and their region, an area have been constructed in manor among them Domain is comprising there are one city (keyword) and multiple cities and towns (information).

The size and shape in region is determined that the boundary in region is represented by dashed line in we by city with the cities and towns for belonging to the city Limit.Different area connectivities represents the correlation of forwarding and temporal correlation.

River, a river on map signify different cities or interregional forwarding relationship, the starting point in a river Illustrate that the two cities or two interregional information have a large amount of forwarding relationship with terminal.

River characterize information propagate and theme develop prompting, this be based on user behavior and sum up come feature.

Continent and island, the region of sheet connection form continent, and small and isolated region forms island.

To sum up, the connectivity between region reflects the characteristics in spreading information based on keyword, and overall includes city, city The continent in town and the layout on island, show the distribution of theme and n-tuple relation.

Other than the vision mapping of object on map, color is also one of most important design consideration.We are to dynamic Event differentiation is very interested, therefore the time-varying attribute of affair character is characterized using color.Specifically, it is contemplated that following three A design requirement：

First, color mapping needs the linearly for the variation and affair character that people is allowed to understand event.

Second, color mapping should be linear and equally distributed perceptually.

Third, all objects using color all should be consistent on map.

Based on this, we have finally selected " inferno " color mapping of perceptibility equilibrium.We are color mapping to city In city, cities and towns and region.Other color mapping modes for meeting condition can also use on our map.

Based on the above technical solution, the event that is converted to develops map, including：Multiple entity figure constructs and thing Part develops map generation, wherein：

The first step, multiple entity figure construction.

Multiple entity figure is the graph structure extracted from social media information, is described in social media about some event Keyword, information (social media information) and the correlation of time (virtual time point, also known as virtual time node), referring to Fig. 5；

N=(K, M, T) includes keyword K, information M and virtual time point T,

Keyword is extracted with time change in different time sections, and different themes is also changing.We are first from all Keyword is extracted in information.In event analysis, we are not only concerned about high frequency words generally, we are also concerned about only at certain The high frequency words of one period burst.Based on this, timeline is divided into multiple periods by us, for the letter in each period It ceases us and calculates the TF-IDF values (word frequency/inverted file word frequency) of keyword.We extract has height within each period The keyword of word frequency.After merging same keyword, we obtain a keyword set K=(k1, k2 ..., Ki ..., knk).In addition, we for segmentation period, also construct virtual time point set T=(t1, t2 ..., Ti ..., tnt), each virtual time point represents a period.We are by the keyword Ki of each period high frequency and its phase Increase a line KiTj on the virtual time point Tj answered.

For each keyword, we can obtain all information for including the keyword.We are turned by information Hair relationship, to obtain the forwarding relationship between keyword.I.e. when an information contains keyword Ki, it forwarded another and include The information of keyword Kj, we just add a line KiKj between two keywords.

Multiple entity figure construction is carried out based on above operation, we obtain an initial graph G '=(N ', E '), wherein saving Point is N '=(K, T), while being E '=(KK, KT).Initial graph G ' in order to obtain, we have used first is oriented to based on power The algorithm of (Force-directed Layout) is laid out, and which use stable simulation processes, it disclosure satisfy that same One data is repeatedly laid out the result to keep relative stability.Specific steps are as shown in figure 5, wherein：

Initial layout shown in step (a) in Fig. 5, its object is to reflect forwarding relationship (KK) and will be in same time period The keyword of appearance moves similar position (KT) to.It is too intensive that the result of initial layout may result in some nodes, need into One step reduces intensive visual occlusion.

The layout of second stage shown in step (b) (layout tuning), has used Lloyd algorithms to adjust the position of node in Fig. 5 It puts, node is allowed to be loosed off as possible while state is kept building up.

In addition, second stage is laid out, also adds node M and side KiMj carries out further layout calculation.Wherein side KiMj contains information node Mj and keyword node Ki.The rule for increasing side KM is that only increase information node is most representative Keyword can so reduce structural complexity, and two can make result become to be appreciated that, can be explained.Otherwise it is a plurality of People can not be allowed to understand while pullling the information position of a final map of node.In this adjustment, more importantly keyword (wraps Contain more information) more spaces can be obtained, then other nodes are pushed open.Meanwhile second stage layout is kept as possible G ' layout results before, we are made the power on the node newly added and side relatively small, are closed with the association in time before highlighting and forwarding Connection.

After two-wheeled layout (referring to initial layout and second stage layout), city (keyword) and the position in cities and towns (information) Putting can determine, this is also the input that our second step events develop map generating algorithm.

Second step, event develop map generation.

Based on multiple entity figure G (the multiple entity figure G obtained after layout tuning) as input, map generation portion will most throughout one's life Map E-Map is developed into the event with city, cities and towns, region, river, continent and island.The basic number that we use It is Voronoi grids according to structure.Specific steps are as shown in figure 5, wherein：

Node is placed in Fig. 5 shown in step (c), node includes：Seed point, city and cities and towns, based on Voronoi grids For Splatting initially according to screen size, we will randomly place seed point in 2D spaces.More seed points can cause The edge of map is more clear.In our realization, it is contemplated that computational efficiency and the balance of aesthetic measure, we select 4096 points are as seed point.We use Lloyd algorithms and are more uniformly distributed so that randomly placing seed point, generate in this way Voronoi grids are more uniform.We have at 2 points using the reasons why Voronoi, and the polygon in Voronoi first is irregular , this, than more consistent, Voronoi grids is also usually used in map generation with landform.On the other hand, from calculate angle and It says, the calculating of the adjacent polygons in Voronoi grids can be by calculating Delaunay triangularizations and very flexibly.Therefore We use Voronoi grids.The city calculated before and cities and towns position are directly projected in Voronoi grids by we, right In each subpoint, the number that we include according to it places a Gaussian kernel and carries out Splatting (Splatting).I The radius of Gaussian kernel is limited, setting maximum radius be width r ratio values, here we r=is rule of thumb set 1/10.After all nodes are thrown, we can obtain the density value in all Voronoi grids.

Structure realm shown in step (d) in Fig. 5, when constructing city, cities and towns and region, we place city in the position based on figure The position in city's (keyword).Size is used for characterizing the information content included.For zoning, we have been traversed in Voronoi Then they are assigned to nearest city by all polygonal mesh, we are obtained with the region in each city in this way Range.Based on this, we traverse all grids, find the node that adjacent area belongs to different cities, connect into the side in region Boundary.In each area, the information issued earliest is placed in cities and towns by we from inside to outside according to distance, to the letter issued the latest Breath.

Construction river shown in step (e) in Fig. 5, river illustrates the forwarding relationship between different zones.We are by before Source-purpose keyword in the G of calculating carries out even side (KK), we are by the area in the Voronoi grids where corresponding city Domain position is connected.Then all regions are connected and are ranked up according to quantity by we, we filter out river according to a parameter The number of stream only shows and connects mostly concerned river.However even if in this way, still there is very big intersection, visual effect is influenced. Therefore, we further merge the river with same paths.About the decision of parameter, our Primary References are final Display effect, in test we have found that 5~10 river effects are relatively good, can not only show main forward-path, but also It can keep certain aesthetics.We providing a series of interactions later allows user to explore further forwarding details.

Merging shown in step (f)/separation land in Fig. 5, for constructing continent and island, based on city before, cities and towns and The Splatting in river, we can obtain the Density Distribution in entire Voronoi grids, based on this Density Distribution, I Density value is attached more than 0 region.Density be 0 it is considered that be sea level, then the join domain of bulk is just Become continent, independent join domain becomes ocean.Have calculating as a result, we will extract all continents and island Edge, i.e. adjacent area are the region on sea level and non-sea level as borderline region.

Final map tuning shown in step (g) in Fig. 5, after this step, we have been obtained for all of map Feature.Tuning step includes following two aspects.First, we carry out the edge in continent and island corrosion simulated, because very The continent island edge in figure is all by long-term seawater scouring and corrosion on the spot.We used common in geography Planchon-Darboux algorithms carry out this part of tuning so that the map structure of generation is truer.Its principle is logical The corrosion for crossing simulation ocean is iterated fringe region the change in location for calculating edge.Secondly, original river might have Entire river is become opposite by the pattern of zig-zag, our upper and subsequent points by the way that river is each put into row interpolation Smoothly.Finally we obtain the geographical maps similar with true map.

Algorithm shown in Fig. 5 is known as Algorithm1 Map building algorithms by us, below the example for the algorithm.

Input:

Keyword node set Ki includes its position Ki.pos and size Ki.size, i=1,2...nk；

Information node set Mi includes its position Mi.pos and size Mi.size, i=1,2...nm；

Key word information connects side KiMj；I=1,2...nk, j=1,2...nm；

The forwarding relationship Li of keyword, comprising source keyword Li.source and purpose keyword Li.target, i=1, 2...nl；

Output:

Voronoi node arrays Vi, i=1,2...nv；

Urban node array Ci, i=1,2...nk；

Cities and towns node array Ti, i=1,2...nt；

Area array Bi, i=1,2...nb；

River array Ri, i=1,2...nr；

Continent and the boundary array Pi of ocean, i=1,2...np.

Algorithm1 algorithms mainly include the following steps (including partial code)：

1://Stepc:It constructs Voronoi grids and carries out Splatting

2:GridpointsV=Lloyd (newPoints (v))；

3:Fori=0；i<nk；i++do

4:The corresponding position point 5 of index=FindMeshGrid (V, Ki.pos) // find on Voronoi nodes: Vindex.size=Ki.size

6:Forj=0；j<Vindex.neighbors.length；j++do

7:Vindex.density=Gaussian (Vindex.pos, Vindex.neighbors [j] .pos, Ki.size)

8:endfor

9:endfor

10:Information node M is also repeated above operation

11://Stepd:Construct city, region and cities and towns

12:Initialize urban node C, borderline region B and downtown areas T

13:Fori=0；i<nk；i++do

14:Index=FindMeshGrid (V, Ki.pos)

15:Ci.pos=Vindex.pos//construction city

16:endfor

17:Fori=0；i<nv；i++do

18:CityIndex=FindNearestCityPointIndex (C, Vi.pos)

19:CcityIndex.area.push (Vi) // setting urban area

20:endfor

21:Fori=0；i<nv；i++do

22:if！isAllEqual(Vi.neighbors.cityIndex)then

23:B.push (Vi) // find the different node of periphery neighbours as boundary node

24:endif

25:endfor

26:Adjacent boundary node is connected to ultimate bound B

27:Based on connection KM constructions T.cityIndex

28:Fori=0；i<nk；i++do

29:Ci.timeMapping={ domain:Ci.timeRange, range:[0, Ci.area] }

30:endfor

31:Fori=0；i<nm；i++do

32:Ti.pos=CTi.cityIndex.timeMapping (Ti.timeRange) // construction cities and towns

33:endfor

34://Stepe:Construct river

35:Initialize river node R

36:Fori=0；i<nl；i++do

37:(source, target)=FindMeshGrid (V, Li.source, Li.target)

38:Ri.curve=BuildCurves (source, target)

39:endfor

40:River node R is ranked up, according to the node that they are accessed, then merges R according to identical access

41://Stepf:Construct continent and island

42:P=Contour (gridpointsV, seaLevel=0) // region in the P of region forms continent 43:// Stepg:Final tuning

44:P=Erosion (P) // marine corrosion effect simulation is carried out to boundary

45:Fori=0；i<nr；i++do

46:Forj=1；j<Ri.nodes.length-1；j++do 21

47:Ri.nodesj.pos=Interpolate (Ri.nodesj-1.pos, Ri.nodesj+1.pos)

48:endfor

49:endfor

In Figure 5 after step (g) processing, the event that we have obtained in Fig. 6 shown in a views develops map E-Map, in order to Further support to carry out it space-time visual analysis with exploring, we provide a set of visual analysis flow, support pair accordingly Event and its exploration of differentiation, it is desired the result is that the deep understanding that acquisition develops event.

The visual analysis flow specifically includes three big steps, first there is provided the general overview of space-time, then It provides multi-level space-time to explore, and the analysis that the exploration of details is supported to develop with event.

In order to realize this process, supported as shown in fig. 6, providing several views in visual analysis system interface and giving, Including：

Time view in Fig. 6 shown in b views, user can explore the time of keyword and information with usage time view Rule,

Information list view in Fig. 6 shown in c views, user can with use information List View come browse, search for row The information of sequence selection,

Key person's List View in Fig. 6 shown in d views, user can assist looking for using key person's List View Whom is key person to and what they have said,

Details keyword relational view in Fig. 6 shown in e views, user can be with details of use keyword relational views simultaneously Auxiliary developed to event map E-Map is explored.

F views show event cutting and animated function in Fig. 6, and it is more preferable to provide the different phase that user develops event Exploration.In figure 6 in Event Example shown in f views, there are 5 event stage S1~S5 to be detected, it may be seen that crucial Personage and analyze they are how to influence event to develop.

The visual analysis flow is as shown in fig. 7, first, when a time and space general view allows the user to carry out multi-level Sky is explored, and including guide to visitors, circle choosing and temporal filtering, rule is found to update the feature on map.We support the time to cut simultaneously Piece and animation effect allow user to be expressly understood map is how dynamically to be constructed out in this way, and which reflects events to be How to develop.In order to further support the exploration of event, user can with the track of the social media user on analytical map with And their propagation association, to judge what key person at different times and behavior play the role of to event.More than being based on Exploration, user to event and its can develop and obtain deep understanding.

Space-time general view shown in Fig. 7

A space-time general view is we provided user to be helped intuitively, fast to understand whole event.Wherein, space view Map view displaying keyword overall distribution.User can be allowed to understand their important journey the perception of the size of keyword Degree.Adjacent city might have more relevant contact, and this contact can also be in time phase in forwarding relationship It shuts.The distribution on continent and island embodies the contiguity of different themes.We provide multiple buttons by user's control The switch of figure can show that be turned on and off density expression, city, cities and towns, river, zone boundary etc. (regards referring to a in Fig. 6 Figure).Interactive flexibility is provided in this way, allows user that can freely explore a certain layer or certain is several layers of, and can support Neatly carry out lower probing rope.

Time view (referring to b views in Fig. 6) allows the specific Annual distribution of the keyword of user's understanding extraction.At this In view, there is the Time alignment of peak value according to from top to bottom by them in the keyword of extraction.Keyword is equally also according to its peak The value corresponding period is ranked up, user to be facilitated to be compared.B views are illustrated according to there is peak value earliest in Fig. 6 Keyword above, but direction can also customize sequence.X-axis is time shaft, and each keyword accounts for a line, and displaying includes The Annual distribution of its social media information.The color lump color of distribution is identical with keyword, while represents information with transparency Quantity.User can do different brush choosings above simultaneously, and b views illustrate five different event stages in Fig. 6.Have Such overview, user can to the appearance of the keyword of every time, wither away, the rules such as reappear and explore.

Multi-level space-time is visually explored shown in Fig. 7

One of advantage of space representation of the similar true map of construction is that user can be allowed to be visited using space-time filtering The multi-level map element of rope.In our design, city and cities and towns are the spatial organizations of two levels, are shown on map The information of hierarchical structure.In order to which space-time multi-level in this way is supported to explore, the interactive mode of three types is we provided.

First, user can select the selection and navigation of time shaft progress time by brush.Choose the social matchmaker in the period Volume data can regenerate corresponding E-Map.This new map is the subset of original all period maps.User can in this way Relatively easily to judge region corresponding in map, certain period social media constructions form.Further, user Animation exploration (referring to f views in Fig. 6) can be carried out to the construction process of map.In order to further keep, user's is haunting Image, our design provide a series of transition effect, highlight the region newly selected, and the gradually dim area to disappear Domain.User can be allowed to understand differentiation semantically in this way.

Second, user can use the interactive form of known similar Internet map on map.We provide including The operations such as navigate, scale and click.User can also zoom in and out explores interested city, region and river with displacement.When When rope is drilled under user, the cities and towns meeting automatic Display of details comes out.These space physics can be clicked.In order to further support Multi-level exploration, we additionally provide two views, i.e. information list (referring to b views in Fig. 6) and key person's list (ginseng See c views in Fig. 6).User can select a keyword city to highlight all relevant information.They equally can also Select primary data information (pdi) from the point of view of the cities and towns of a details, including content, time and user (referring to e views-S1 in Fig. 6, S2, S3).It is linked with map, information list also can choose information all to update, highlight by relevant.Key person's list simultaneously It can show the relevant statistical information for participating in personage, issue how many social media informations and his information altogether including him It is forwarded by how many people.

Third, user can select one group of cities and towns to carry out detail analysis on map using own regional choice tool. Because each region only represents its city with its most popular keyword, other more keywords are hidden.I Provide such brush and select function, a word cloud can be shown in by the map area of brush choosing.The size of word illustrates It chooses and occurs its frequency in information.Meanwhile information list can also update simultaneously with key person's list.Multi-level in this way Space-time explore in, user can explore the dynamic evolution and city of map, cities and towns and region detailed information.

Dynamic event evolution analysis shown in Fig. 7

It further seeks how each event stage map changes, we provides the event analysis side based on map Method, mainly analyzes user behavior and key content is analyzed.As we know that promotion of the behavior of user to event Play the role of with influence very big.Behavior for user is presented as two aspects on map, when social media user Track and their contact.On the one hand, a people may issue social media information in different times, the body on map It is exactly now that he has accessed different cities and towns.These cities and towns may be in a city, it is also possible in multiple cities, such track with And the distribution accessed reflects the feature that a people participates on theme.By summarizing and arranging these features, we can send out The trend that existing key person or user group change in different themes.On the other hand, a social media information may quilt People forwards, we pay close attention to the information with high-impact, i.e., are forwarded by a large amount of people.This is presented as on map in two cities and towns Contact, if there are forwarding relationships between them.High-level, the forwarding between a large amount of region is summarized as river by us The form of stream, in details and interaction are explored, user can explore the related forwarding of each cities and towns or a piece of downtown areas Relationship.Based on above 2 points, in E-Map, we come the track of map user using the curve of black, with the straight line of grey come Characterize forwarding relationship.The thickness of line all represents the sum of forwarding (referring to a views in Fig. 6).

Both of which is we provided to be visualized to track and forwarding relationship：Personality frame and accumulation mode. In personality frame, the two kinds of lines in track and relationship are shown in cities and towns rank, it illustrates the relationship between information, embodiment It is details distribution.In accumulation mode, line is organized to city rank, in this way we have seen that be between different zones Track precedence relationship and forwarding relationship.Specific mode is that the line in cities and towns all in city all polymerize by we, The starting point and emphasis of track and line are shown in key city (referring to a views in Fig. 6).Both lines are to exploring thing There is important role in the variation of part and the influence power of personage.We can be ranked up forwarding relationship, and user can be with See the information of high-impact and its correlation are embodied in which aspect of map.Two on the one hand, and user can brush and select the time Axis, user is it can also be seen that the dynamic evolution in the region that track is passed through, the analysis that this event develops provide strong card According to.

Further, in order to support to propagate the exploration developed with theme to information, a details keyword is we provided View carries out assistant analysis (referring to e views in Fig. 6).User can enclose on map and select a region, select some city or The track of key person is selected, the social media information of selection and its relationship can be imported keyword forwarding view by we In, as its input.In the view, a total of three row, are that the keyword chosen is placed on centre respectively, their forwarding father The left and right sides is individually placed to child's keyword node, is connected with curve.Size of node is represented comprising selected in the keyword The quantity for the information selected.The forwarding is highly represented in all ratios for choosing forwarding.By this view, we can explore choosing In certain keywords, which (referring to e view in Fig. 6) word of its strong correlation have.

It is developed to sum up, E-Map visual analysis systems allow user to explore multi-level event, by non-structured letter Breath is configured to the information for including semanteme of structuring, and dynamic interaction is supported to explore different levels, the details in different event stage.

Beneficial effects of the present invention are summarized as follows：

In order to which user is preferably allowed to understand the relevant behavioral characteristics of social media dynamic event, we have proposed E-Map, one A novel visualization based on map is presented.The summary of social media dynamic event that it provides a dynamic, can explore Form visualizes information communication space, allow people in familiar manner-form of true map understood and explored.In E- In Map, the keyword extracted in user social contact media information is mapped to city, is each reflected with the relevant information of the keyword The cities and towns of surrounding city are penetrated into, which reflects theme and the information distributions in this region.Contact between region represents theme Between the tight association degree that contacts, be mainly reflected in association in time and forwarding association.River Crossing on map is different Region, illustrate starting and terminal point region have significant forwarding relationship.All regions, associated formation continent, otherwise It is then isolated island one by one.Key point information, the behavior to release news with forwarding information including user, just fusion is on map Track and cities and towns between association on.User can freely explore in map space, scaling, carry out interaction at many levels Analysis, to obtain the semantic information that event develops train of thought.

It is specific embodiment below.

Case 1- gauze doors：Event analysis

In this case, it we show the process that E-Map is used to be explored, is persistently sent out to analyze a burst The event of ferment.Wherein, we have found different event evolving stages, and key person forwards the differentiation of relationship and theme Process.This is an event occurred in Shandong Province of China, and the origin of thing is that there is a woman in Shandong in August, 2016 After performing an operation, one piece of gauze has been left in her internal.This is in Shandong TV Station in October, 2016 locality program Report comes out in " life hack ".The beginning of report, the masses one after another work as thing hospital and as thing doctor by condemnation, but as event is drilled Become, more people take part in comment, and public opinion gradually develops toward opposite direction.

In our system, we use " gauze door " keyword to carry out data and crawl first, are collected into comprising gauze Door and its totally 9626 microbloggings of all forwardings, are related to 6963 people.Fig. 6 illustrates us and uses the analytic process and knot of E-Map Fruit.We look in each period 10 keys at most (TF-IDF values highest) occur the periods different to 10 first Word, then by merging after, we have found 60 keywords, and have constructed E-Map according to their spatial and temporal distributions.We It is constantly highlighted it can be found that some main keywords include " gauze, life, Shandong, media, hospital, doctor, broadcast ", It is commented within the entire period always.These keywords have sketched the contours of the beginning of event, and further we carry out event It explores, has been cut into 5 event stages and has analyzed one by one.

In the incipient stage (S1) of event, from 30 to 31 October, " army of flying fish army ", local " life hack " TV station One TV announcer has issued several microbloggings about " gauze door " first, his content includes the " yarn in puerpera's tripe Cloth can be removed as to when”.Only have a minority to be forwarded at the beginning, it will be seen that the region master on map It will be about gauze door, program, hospital etc..

At second stage (S2), on November 1 to 2, " army of flying fish army " constantly sends out his information, and emphasize this It is " fact ", we can be from emerging region " fact " and a curve from " gauze " direction " fact " it can be seen that coming. However, other people, stand out including " a somewhat ideal reporter " and have said another viewpoint, he represents hospital and doctor Victim is only, is because patient is arbitrary just so that gauze is not removed slowly.The evidence about " signature " is he provided, To illustrate " doctor being allowed to take out gauze this is because puerpera be unwilling to slowly sign letter of consent ".He has caused theme to be oriented to " signature ", As the grey straight line in map is shown.Further, because he issued it is more locality visit video, more people The camp of " a somewhat ideal reporter " is added, criticizes " army of flying fish army " and " life hack " " shameless " together, this also exists A new region is produced on map.However on the other hand, also there is the supporter of " army of flying fish army ", two sides are striven on the net By.

At three phases (S3), on November 2 to 3, very big reverse has occurred in event.Reason is CCTV, this Most influential TV station of China intervenes and has investigated this event.They finally found that tort-feasor essentially consists in patient, and It is not hospital.In this stage, it will be seen that a new region " CCTV " extends out from the west of map, simultaneously There is a connection forwarded from " gauze " direction " CCTV ".More and more people start investigation of the comment from " CCTV " simultaneously, And appeal " to punish " that patient.In this stage, the great effect from media of health care is engaged in, including " fourth Fragrant garden ", " one the ideal reporter of advantage " etc., their constantly photos and sending messages, forwardings allow more people to know truth.

In next event stage (S4), from 3 to 6 November, the comment about gauze door was still continuing, and It has been proposed that local TV station's " life hack " with " Putian system " hospital (Chinese Civilian Hospital) collude with by interests in fact.

The last event stage (S5) from 6 to 8 November, although comment is also continuing, touches without new event Hair, and the comment theme of people is extended in such as medical treatment such as " woman ", " uterus " correlation, the relevant theme of the rights and interests of women It goes.

In this event, we have understood the train of thought of gauze door development, and find by analyzing each event stage Different key person of intermediate each stage, they affect the comment and development of whole event.The extension of its mid-continental is closed The trail change of key personage and new linking relationship occur, all let us more fully understood event in developing it is important because Element.

To sum up, the present invention proposes a kind of method for visualizing of novelty, it provides a physical map that can be explored Design allows the variation of customer analysis major event.To comment on the social media information of event as input, we by information, The relevance map of keyword, forwarding behavior and keyword is cities and towns, city, river and the connected region on map. This is a kind of in shape similar to the visual presentation of physical map, at the same we provide it is a series of support interaction explore space-times can Depending on analysis operation.User can explore event according further to map feature and develop details.System can help user to find Event Characteristics of Evolution, key person and Exploring Analysis they be how to influence a step by a step this event differentiation.

The present invention includes：

One based on social media data it is novel to event develop summarize visual pattern it is proposed that it is expansible Map Design, contain the semantic key factor summarized, developed corresponding to event of complicated event.Map is by these because being known as It is merged to machine, and is presented in a manner of electronic map known to user.

A kind of space-time visual analysis method for exploring social media user behavior is based on the map constructed, user behavior quilt The Spatial elements being mapped as on map.The interaction flow of a set of visual analysis is we provided, how the behavior for exploring user is The evolution of influence event.

It is corresponding with the method shown in Fig. 1, a kind of social media dynamic event is additionally provided in embodiment of the present invention Visual analysis system is developed, as shown in Fig. 2, the system includes：

Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art God and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technology Within, then the present invention is also intended to include these modifications and variations.

Claims

1. a kind of social media dynamic event develops visual analysis method, include the following steps：

For certain event, keyword is extracted from social media, then according to keyword, social media is obtained from social media Data；

According to based on map metaphor processing rule, by the way of map maps, the social media data of input are converted into thing Part develops map E-Map；

It is described to be included based on map metaphor processing rule：City, cities and towns and river, formed structuring and include semanteme Projector space, user can by interaction go to explore the projector space；

Keyword,

Releasing news comprising keyword and forwarding information,

User information dissemination.

2. a kind of social media dynamic event according to claim 1 develops visual analysis method, it is characterised in that：In base In map metaphor processing rule, semantic map and vision mapping are preset in the following manner, is specifically included：

The semanteme map includes：

Keyword is mapped as city,

3. a kind of social media dynamic event according to claim 2 develops visual analysis method, it is characterised in that：It is semantic Map is mapped using unified color mapping, and the time-varying attribute of affair character, the unified face are characterized using color Color mapping includes but not limited to " inferno " color mapping of perceptibility equilibrium.

4. a kind of social media dynamic event according to claim 1 develops visual analysis method, it is characterised in that：It is described It is converted to event and develops map, including：Multiple entity figure constructs and event develops map generation.

5. a kind of social media dynamic event according to claim 4 develops visual analysis method, it is characterised in that：It is mostly real Body figure is the graph structure extracted from social media information, is described in social media about the keyword of some event, information With the correlation of time；

In multiple entity figure construction, input is social media information M=(m1, m2 ..., mi ..., mnm), and output is real more than one Body figure G=(N, E), wherein：

N=(K, M, T) includes keyword K, information M and virtual time point T,

E=(KK, KT, KM) includes the company side KT and keyword of the company side KK of keyword and keyword, keyword and time point With the company side KM of information.

6. a kind of social media dynamic event according to claim 5 develops visual analysis method, it is characterised in that：It is mostly real In body figure construction, including two-wheeled layout processing, respectively initial layout and second stage layout.

7. a kind of social media dynamic event according to claim 4 develops visual analysis method, it is characterised in that：Event It develops in map generation, based on multiple entity figure G as inputting, is ultimately generated according to the following steps with city, cities and towns, region, river Stream, continent and the event on island develop map：

Node is placed, node includes：Seed point, city and cities and towns；

Construct city, cities and towns and region；

Construct river；

Merging/separation land constructs continent and island；

Map tuning.

8. a kind of social media dynamic event develops visual analysis system, including：

Event develops map E-Map modular converters, for handling rule according to based on map metaphor, using the side of map maps The social media data of input are converted to event and develop map E-Map by formula.

9. a kind of social media dynamic event according to claim 8 develops visual analysis system, it is characterised in that：It is described Event develops map E-Map modular converters, including：

Multiple entity figure constructs submodule, for constructing multiple entity figure；Multiple entity figure is the figure knot extracted from social media information Structure is described in social media about the keyword of some event, information and the correlation of time；

Event develops map generation submodule, and city, cities and towns, region, river, continent and island are carried for ultimately generating Event develops map E-Map.

10. a kind of social media dynamic event according to claim 9 develops visual analysis system, it is characterised in that：It is more Sterogram construction submodule specifically includes：

Initial layout unit, for generating forwarding relationship and moving the keyword occurred in same time period to similar position It puts；

Second stage layout units for adjusting the position of node, allow node to be loosened as possible while state is kept building up Come, further layout calculation is carried out for node M and side KiMj.

11. a kind of social media dynamic event according to claim 9 develops visual analysis system, it is characterised in that：Thing Part develops map generation submodule and specifically includes：

Node unit is placed, node includes：Seed point, city and cities and towns, seed point are used to that the edge of map to be more clear, So that the Voronoi grids of generation are more uniform, city is directly projected on cities and towns position in Voronoi grids；

City, cities and towns and territory element are constructed, the position in city is placed for the position based on figure, for traversing in Voronoi Then they are assigned to nearest city by all polygonal mesh, to obtain the regional extent in each city, for finding Adjacent area belongs to the node of different cities, connects into the boundary in region, for cities and towns to be placed from inside to outside according to distance The information issued earliest, to the information issued the latest；

River unit is constructed, it, will be in the Voronoi grids where corresponding city for source-purpose keyword to be carried out even side Regional location be connected, for all region connections to be ranked up according to quantity, river is filtered out according to a parameter Number only shows and connects mostly concerned river, for the river with same paths to be merged；

Continent and island unit are constructed, for being based on Density Distribution, region of the density value more than 0 is attached, the company of bulk Region is connect as continent, independent join domain is ocean, for extracting the edge in all continents and island；

Map tuning unit, it is corrosion simulated for being carried out to the edge in continent and island, it is upper for river each to be put With subsequent point into row interpolation, entire river is become into relative smooth.