CN104536956A - A Microblog platform based event visualization method and system - Google Patents

Abstract

The present invention discloses a Microblog platform based event visualization method and system. The present invention relates to an information extracting and visualizing technology. The method includes: retrieving Microblogs in a time period related to an event through an event search interface of the Microblog platform according to a key word of the event and the time period; sorting the Microblogs according to time to generate a set of the Microblogs; generating a plurality of cluster subsets from the set of the Microblogs through a clustering algorithm; extracting a key word from the plurality of cluster subsets, to generate a plurality of word clouds, and endowing the key word, which repeatedly appears in the plurality of word clouds, with the same color, position and rotational mode; and performing a visualization presentation of the event by means of presenting each of the cluster subsets with one corresponding word cloud. Depending on the Microblog platform, Microblog information on a certain event can be fully obtained by collecting related Microblogs using the key word of the event.

Description

A method and system for event visualization based on Weibo platform

technical field

The invention relates to information extraction and visualization technology, in particular to an event visualization method and system based on a microblog platform.

Background technique

With the rapid development of the Internet, various social media have emerged in recent years, the common ones are Facebook (Facebook), Twitter (Twitter), Sina Weibo, and Renren, among which Twitter and Sina Weibo are the representatives. The microblog platform has become a popular Internet application due to its open information sharing and dissemination characteristics.

Microblog, short for Microblog, users can publish text, pictures, videos and other information within 140 characters anytime and anywhere on the platform. Weibo has the characteristics of originality, timeliness, fragmentation and repetition. In the Weibo platform, users can search and view topics they are interested in, browse the content related to the topic and participate in the discussion of the topic content. However, because the Weibo platform is filled with a large number of Weibo related to a certain event, and because of the short text characteristics of Weibo, this brings problems such as fragmentation and difficulty in understanding the published information. Uneven information release is a very significant phenomenon on the Weibo platform. It is precisely because of the above-mentioned various reasons that it is difficult for users to quickly understand the development of an event in a short period of time, which makes the user interaction experience worse.

Among the existing microblog event visualization technologies, it is generally simple to sort the event-related microblogs by time, display the microblogs in the latest time period to the user, and also sort by the popularity of the microblogs, and list the popular microblogs To display to users, there is another way to display by selecting Weibo within a certain time range in order of time or popularity. The above display methods are all direct displays of the original microblog content, and have many deficiencies. First, due to the explosive growth of network information, the traditional method of visually displaying the original Weibo is difficult for users to quickly obtain event-related information; second, due to the short text characteristics of Weibo, Weibo release The quality of information is uneven, and the problem of colloquial Weibo makes it difficult for users to quickly understand the content of Weibo, and it is even more difficult to find important information about events from Weibo texts.

Among the event visualization methods, one is to extract keywords from all the text information of the event, and then display the extracted keywords through a word cloud. In this way, microblog users can understand the main topic of the event from the main keywords, but microblog users cannot have an intuitive understanding of each sub-event of the event and the development and evolution of the event.

Some other visualization methods extract the characters, places, and event summary sentences in the event, use them as the node information of the event development, and use the relationship between them as the edge to visually display the event. However, this visual display method based on people, places, and event summary sentences has great limitations for Weibo events, because Weibo does not have standardized people, places, organizations, etc. information like formal news reports, so It is difficult to obtain such information from Weibo. Therefore, this visualization method has great limitations for Weibo.

Invention patent "microblog word cloud generation method and access support system based on user interest mining", the invention discloses a microblog word cloud generation method and microblog message access support system based on user interest mining, the method includes: given The newly released microblog message set of the user concerned by the currently logged-in user is extracted from the set of keywords; the interest degree of the currently logged-in user to the keywords in the keyword set is calculated based on the user relationship and the similarity of the keywords respectively, and the two The degree of interest obtained from the calculations is fused to calculate the final degree of interest; the k keywords with the highest degree of interest are selected from the keyword set; the selected k keywords are displayed in an area. The system includes key modules such as user information acquisition module and word cloud generator. The invention can enable users to obtain information they are interested in from microblog messages more efficiently. However, the research object of the present invention is different from that of the invention: the invention takes microblog users as the research object, analyzes microblog content of microblog users, and performs word cloud display by extracting keywords. However, the present invention takes news events as the research object; the difference in visualization: this invention only extracts keywords from microblogs and displays them in word cloud form. However, the present invention uses sub-events of events to extract keywords and perform multi-dimensional display of combined word clouds.

Invention patent "Microblog-based event feature evolution mining method and system", the invention discloses a microblog-based event feature evolution mining method, including: selecting the evolution starting document set in the microblog time series, and Based on the co-occurrence features of vocabulary on the document collection, the document graph model is constructed to obtain the knowledge network structure of the event; according to the literal features of the vocabulary and the compatibility features of the vocabulary tendency, the microblog graph model is merged to construct the micro-evolution of event features Graph; pruning, segmentation and transformation are performed on the micro-evolution graph of the event to form a macro-evolution graph of event characteristics. This method adopts the event-based knowledge network graph mining method in the process of mining the evolution rules of event features, which improves the inheritance of knowledge in the entire event feature evolution mining method, and the interpretability of mining results is stronger. However, the present invention is different from the feature extraction of this invention: this invention mainly performs feature extraction from the vocabulary structure, and performs evolution display of events by building a knowledge network structure. The present invention mainly clusters events and performs evolution display of subtopic characteristic information of mining events.

Contents of the invention

Aiming at the deficiencies of the prior art, the present invention proposes an event visualization method and system based on a microblog platform to solve the above technical problems.

The present invention proposes an event visualization method based on a microblog platform, including:

Step 1, according to the keywords and time range of the event, retrieve the microblogs related to the event within the time range through the event search interface of the microblog platform;

Step 2, sort the microblogs according to time to generate a microblog collection;

Step 3, the microblog collection generates multiple cluster subsets through a clustering algorithm;

Step 4, performing keyword extraction on the multiple clustering subsets, generating multiple word clouds, and assigning the same color, position, and rotation mode to the keywords repeatedly appearing in the multiple word clouds;

Step 5: Visually display the event by displaying each cluster subset and the word cloud corresponding to it.

The described event visualization method based on the Weibo platform, before the step 2, also includes:

Step 21, filtering the microblogs whose word count is less than a certain threshold in the microblogs within the time range;

Step 22, filtering the microblogs with popularity less than a certain threshold in the microblogs within the time range;

Step 23, filtering information in non-text format in the microblog within the time range;

Step 24, filter the "username" in the microblog within the time range.

In the event visualization method based on the microblog platform, the calculation formula of the popularity in step 22 is:

$\mathrm{<span\; class="notranslate">\; heat</span>}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; retweets</span>}}<span\; class="notranslate">\; +</span>\sqrt[\mathrm{<span\; class="notranslate">\; 3</span>}]{\mathrm{<span\; class="notranslate">\; comments</span>}}$

Among them, retweets represents the number of retweets on Weibo, comments represents the number of comments on Weibo, and Heat represents the popularity of Weibo.

Described event visualization method based on microblogging platform, carries out keyword extraction to each this clustering subset in this step 4, and the specific steps of generating combined word cloud include:

Step 41, performing word segmentation processing on each of the clustering subsets to generate a word set;

Step 42, merging the word set through Wikipedia entries and Internet hot words to generate the combined word cloud.

The event visualization method based on the microblog platform is characterized in that step 4 further includes: according to the inverse document frequency, assigning high transparency to the word.

The present invention also proposes a microblog platform-based event visualization system, including:

A retrieval module, configured to retrieve microblogs related to the event within the time range through the event search interface of the microblog platform according to the keywords and time range of the event;

The sorting module is used to sort the microblogs according to time to generate a set of microblogs;

The clustering module is used for the microblog collection to generate multiple clustering subsets through a clustering algorithm;

Generate a combined word cloud module, which is used to extract keywords from the multiple cluster subsets, generate multiple word clouds, and assign the same color, position, and rotation to the keywords that repeatedly appear in the multiple word clouds Way;

The display module is configured to display the event visually by displaying each cluster subset and the corresponding word cloud.

The event visualization system based on the microblog platform also includes a filtering module, which is used to filter the microblogs whose word count in the microblogs within the time range is less than a certain threshold; Threshold microblogs; filter information in non-text format in the microblogs within the time range; filter "username" in the microblogs in the time range.

In the event visualization system based on the microblog platform, the calculation formula of the popularity in the filtering module is:

$\mathrm{<span\; class="notranslate">\; heat</span>}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; retweets</span>}}<span\; class="notranslate">\; +</span>\sqrt[\mathrm{<span\; class="notranslate">\; 3</span>}]{\mathrm{<span\; class="notranslate">\; comments</span>}}$

Among them, retweets represents the number of retweets on Weibo, comments represents the number of comments on Weibo, and Heat represents the popularity of Weibo.

In the event visualization system based on the microblog platform, keyword extraction is performed on each of the clustering subsets in the generating combined word cloud module, and the specific steps of generating the combined word cloud include: performing a keyword extraction on each of the clustering subsets Word segmentation processing to generate a word set; combine the word set through Wikipedia entries and Internet hot words to generate the combined word cloud.

In the event visualization system based on the Weibo platform, the display module is also used to: according to the inverse document frequency, give the word a high degree of transparency.

As can be seen from the above scheme, the present invention has the advantages of:

Relying on the microblog platform, collecting related microblogs through event keywords can comprehensively obtain microblog information about a certain event; adopting microblog information filtering technology can obtain high-quality and meaningful microblog information; The microblog data set of the event is clustered based on the time dimension, and the obtained event cluster subset has the information of the time dimension. These data subsets can not only represent a certain topic of the event, but also can be seen in general. The development and evolution process; through the keyword extraction technology, representative microblog keywords can be extracted from a group of microblogs, and the keywords of a group of events can allow microblog users to have an intuitive understanding of the microblog content; through control The color and position information of the same word in multiple word clouds makes them highly consistent in the visual display process of the combined word cloud, so that Weibo users can easily see the main topic of the entire event through the combined word cloud. and the topics in each sub-event, and it is very convenient to compare and analyze each sub-event.

Description of drawings

Fig. 1 is the flow chart of the time visualization method based on microblog platform;

Fig. 2 is the flow chart of combined word cloud visualization display;

Fig. 3 is an example diagram of visual display of events.

Wherein reference sign is:

Step 100 is based on the specific steps of the microblog platform event visualization method, including:

Step 101/102/103/104/105/106/107.

Detailed ways

The specific implementation manners of the present invention will be described in detail below in conjunction with the drawings and embodiments.

Concrete flow process of the present invention comprises the following steps, as shown in Figure 1:

Step 101, simulating login to Weibo platform;

Since the present invention is aimed at visually displaying news events on the microblog platform, it is necessary to simulate the process of a user logging in to the microblog website before obtaining event information.

In the process of simulating the login to the Weibo platform, firstly, a batch of Weibo accounts are registered, and the user information table of the simulated login is formed by using these account information. During the simulated login, the request link of the login page is sent to the Weibo site next, Using the local registered user information form, you can provide the site with parameters such as user name, password, and encryption method required for login, and realize the user's simulated login operation.

Since the Weibo platform has restrictions on the number of visits that users can perform within a certain period of time, excessively frequent visits may cause the account to be blocked. Therefore, after a user successfully logs in, when the number of pages visited by the user exceeds a certain number of times, Just select another user from the local user information table to perform a simulated login operation. In this way, you can request access to various services on the Weibo platform and obtain the required news event data information.

Step 102, retrieving relevant microblogs according to event keywords;

An event usually consists of keywords and time. By filtering within a certain time range, microblogs within a specified time range can be obtained through the microblog platform.

In this step, rely on the event search interface provided by the microblog platform, and obtain relevant microblog pages through the event keywords and time range input by the user.

Step 103, microblog information preprocessing;

In this step, the microblog information is preprocessed to obtain the data set to be analyzed. The specific processing includes the following parts:

Filter the short text in the data set, and filter the microblogs whose word count is less than a certain threshold;

Filter out the less influential and unpopular microblogs in the data set (that is, microblogs whose popularity is less than a certain threshold), and the microblog popularity is calculated according to the following formula:

$\mathrm{<span\; class="notranslate">\; heat</span>}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; retweets</span>}}<span\; class="notranslate">\; +</span>\sqrt[\mathrm{<span\; class="notranslate">\; 3</span>}]{\mathrm{<span\; class="notranslate">\; comments</span>}}$

Among them, retweets represents the number of retweets on Weibo, and comments represent the number of comments on Weibo;

Filter non-text-formatted content information such as emoticons and web page link addresses in Weibo;

Filter the unique "username" in Weibo;

Sorting is performed according to the time information of microblogs to obtain a set of microblogs continuous in time.

Step 104, microblog event clustering;

In this step, the sorted microblog data sets are clustered to obtain time-continuous cluster subsets. In order to make each clustering subset represent a class of topics, a hierarchical clustering algorithm or a single-pass clustering algorithm (Single-Pass Clustering) is used. The first microblog of the document is used as an initial cluster subset. In each subsequent step, the document is divided into the cluster subset most similar to the document. If the similarity between the document and all current documents is If they are all smaller than the set threshold, it will be regarded as a new cluster subset, and the document similarity calculation is measured by the following formula:

$\mathrm{<span\; class="notranslate">\; sim</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; )</span>$

$\mathrm{<span\; class="notranslate">\; sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; \&times;</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}}{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&times;</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}$

Among them, m represents the number of documents in the time window before the time of document d, and i represents the position of the document in cluster c with the closest time distance to document d in the time window. Through the calculation in the above method, the closer the document is to the clustering time, the higher the similarity. When calculating the similarity of documents, a vector space model is established for the documents, and each document is represented as a vector in the space, each item in the vector is a word in the document, and the weight of each item is normalized in the present invention Calculated using the simplified TF-IDF (termfrequency–inverse document frequency), the formula is as follows:

${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; TF</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; IDF</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\sqrt{\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; TF</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; IDF</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}{<span\; class="notranslate">\; ]</span>}^{\text{<span class="notranslate"> 2</span>}}}}$

${\mathrm{<span\; class="notranslate">\; TF</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}}{\underset{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; IDF</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; log</span>}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; D.</span>}}{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; :</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; \}</span><span\; class="notranslate">\; |</span>}<span\; class="notranslate">\; )</span>$

Among them, n _i,j is the number of occurrences of the word in document d, is the sum of the occurrences of all words in the document. D represents the number of all documents, and |{j:t _i ∈ d _j }| represents the number of documents containing word t _i .

Step 105, sub-event dataset keyword extraction;

By clustering all related microblogs, sub-data sets with a certain continuity in time are obtained, where each sub-data set represents a sub-topic of the event. By extracting keywords from microblogs for each sub-event, A set of candidate keywords that need to be visually displayed in the combined word cloud can be obtained. Keyword extraction is performed in the following ways:

First, word segmentation is performed on each document in the document collection. The present invention adopts the ICTCLAS word segmentation tool (Institute of Computing Technology, Chinese Lexical Analysis System, whose main functions include Chinese word segmentation; part-of-speech tagging; named entity recognition; new word recognition; simultaneously supports user dictionary) to perform document word segmentation processing to obtain the processed word set. In order to make the semantic information of words richer, two dictionaries of Wikipedia entries and hot words on the Internet are used to phrase merge the original word set to obtain a more meaningful word set. In the above phrase merge process, an algorithm based on maximum matching is used The original set of words is processed and each word is weighted according to:

w _i ＝tf _i ×df _i ×Heat×|w _i |

Among them, tf _i represents the frequency of word i appearing in the document, df _i represents the number of documents containing word i in the document collection, Heat represents the popularity of Weibo, |w _i | represents the length of word i, that is, the number of words.

In order to highlight words in popular microblogs, the present invention combines the weight of each word with the popularity of microblogs, and multiplies the heat of microblogs with the frequency of words to be used as the weight of words, so that the words selected are more meaning, and based on the fact that long words have richer semantic information than short words, so the present invention introduces the word length item, so that the weight of long words is relatively increased.

Step 106, event visualization based on combined word cloud;

A simple way to generate a combined word cloud is to use tag cloud technology to generate a word cloud for each sub-event in the event, but the word cloud generated by this method is not good in visualization, because even if two topics discuss content is very similar, their word clouds will be very different. Therefore, when performing combined word cloud display, the present invention needs to optimize the generated word cloud, and the specific processing method is as follows, as shown in Figure 2:

Words that appear in multiple word clouds are given the same color, position, and rotation method, so that they maintain the consistency of attributes in the visualization effect, and it is convenient for readers to quickly browse and find commonalities between topics.

The transparency of words is controlled by the idf (inverse document frequency) of words, so that words that appear together in multiple word clouds are given higher transparency, while words with low document frequency are given lower transparency. In this way, the unique words in each word cloud are highlighted and the high-frequency words that appear in multiple documents are downplayed, so that readers can quickly grasp the content discussed in the topic.

Step 107, event visual display;

Take the time node as the ordinate (take the average time of all events in each cluster as the time node), and display the cluster subset through text information and combined word cloud to show the evolution process of the specific event , and at the same time allow readers to understand the details of each sub-event while quickly grasping the topic of the event.

Figure 3 shows a specific implementation of event visualization. The entire visualization is run through by a time axis, and the origin on the left represents each time node. There are two columns of display boxes on the right side of the time axis, one is the sub-event Weibo clustering result display box, and the other is the word cloud of the sub-event. In the embodiment, the event "Shenzhen heavy rain" is taken as an example, and three sub-events are taken for visual display: the first is that the heavy rain in Shenzhen caused road accumulation and inconvenience for people to travel; the second is that the heavy rain in Shenzhen cannot prevent Shenzhen people from buying houses ; The third is that the rainstorm in Shenzhen has been lowered from red to yellow. First of all, we can see the development trend of the event over a period of time as a whole. By observing the three word clouds, we can see that words such as "Shenzhen" and "storm" appear in the three word clouds, indicating that each sub-event has a common The topic characteristics of , are endowed with higher transparency due to their higher document frequency.

From another aspect, it can be seen that each word cloud has representative words, and these words usually appear frequently in this data set but appear less or do not appear in other data subsets. For example, words such as "accumulated water" and "travel" in word cloud 1, "buying a house" and "opening" in word cloud 2, and "whole city" and "warning" in word cloud 3. Due to their low document frequency, these terms tend to have less transparency and thus stand out more in the word cloud. Based on these words, readers can quickly understand the main content of each topic, and compare the differences between two document topics by observing whether a word appears in the same position in other word clouds.

This example embodies the characteristics of the event visual display method provided by the present invention, which can assist readers to quickly and comprehensively understand the main content of the event and the process of evolution with the event, and at the same time enable the reader to quickly compare the sub-event word clouds. Know the difference between events.

Claims (10)

1. A kind of event visualization method based on microblog platform, it is characterized in that, comprising: Step 1, according to the keywords and time range of the event, retrieve the microblogs related to the event within the time range through the event search interface of the microblog platform; Step 2, sort the microblogs according to time to generate a microblog collection; Step 3, the microblog collection generates multiple cluster subsets through a clustering algorithm; Step 4, performing keyword extraction on the multiple clustering subsets, generating multiple word clouds, and assigning the same color, position, and rotation mode to the keywords repeatedly appearing in the multiple word clouds; Step 5: Visually display the event by displaying each cluster subset and the word cloud corresponding to it. 2. the event visualization method based on microblog platform as claimed in claim 1, is characterized in that, also comprises before this step 2: Step 21, filtering the microblogs whose word count is less than a certain threshold in the microblogs within the time range; Step 22, filtering the microblogs with popularity less than a certain threshold in the microblogs within the time range; Step 23, filtering information in non-text format in the microblog within the time range; Step 24, filter the "username" in the microblog within the time range. 3. the event visualization method based on microblog platform as claimed in claim 2, is characterized in that, the computing formula of this popularity in this step 22 is: $\mathrm{<span\; class="notranslate">\; heat</span>}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; retweets</span>}}<span\; class="notranslate">\; +</span>\sqrt[\mathrm{<span\; class="notranslate">\; 3</span>}]{\mathrm{<span\; class="notranslate">\; comments</span>}}$ Among them, retweets represents the number of retweets on Weibo, comments represents the number of comments on Weibo, and Heat represents the popularity of Weibo. 4. the event visualization method based on microblog platform as claimed in claim 1, is characterized in that, carries out keyword extraction to each this clustering subset in this step 4, and the concrete steps of generating combination word cloud comprise: Step 41, performing word segmentation processing on each of the clustering subsets to generate a word set; Step 42, merging the word set through Wikipedia entries and Internet hot words to generate the combined word cloud. 5. The microblog platform-based event visualization method according to claim 1, characterized in that, step 4 further comprises: according to the inverse document frequency, assigning high transparency to the word. 6. A microblog-based event visualization system, characterized in that it comprises: A retrieval module, configured to retrieve microblogs related to the event within the time range through the event search interface of the microblog platform according to the keywords and time range of the event; The sorting module is used to sort the microblogs according to time to generate a set of microblogs; The clustering module is used for the microblog collection to generate multiple clustering subsets through a clustering algorithm; Generate a combined word cloud module, which is used to extract keywords from the multiple cluster subsets, generate multiple word clouds, and assign the same color, position, and rotation to the keywords that repeatedly appear in the multiple word clouds Way; The display module is configured to display the event visually by displaying each cluster subset and the corresponding word cloud. 7. the event visualization system based on microblog platform as claimed in claim 6, is characterized in that, also comprises filtering module, is used for filtering the microblog of word count in this microblog in this time range less than certain threshold value; Filter this time Among the microblogs within the range, the microblogs whose popularity is less than a certain threshold; filter the non-text format information in the microblogs within the time range; filter the "username" in the microblogs within the time range. 8. the event visualization system based on microblog platform as claimed in claim 7, is characterized in that, the computing formula of this popularity in this filtering module is: $\mathrm{<span\; class="notranslate">\; heat</span>}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; retweets</span>}}<span\; class="notranslate">\; +</span>\sqrt[\mathrm{<span\; class="notranslate">\; 3</span>}]{\mathrm{<span\; class="notranslate">\; comments</span>}}$ Among them, retweets represents the number of retweets on Weibo, comments represents the number of comments on Weibo, and Heat represents the popularity of Weibo. 9. the event visualization system based on microblog platform as claimed in claim 6, is characterized in that, carries out keyword extraction to each this clustering subset in this generation combination word cloud module, the concrete step of generating combination word cloud comprises : Carry out word segmentation processing for each cluster subset to generate a word set; merge the word set through Wikipedia entries and Internet hot words to generate the combined word cloud. 10. The event visualization system based on the microblog platform as claimed in claim 6, wherein the display module is further used for: according to the inverse document frequency, endow the word with high transparency.