CN104615715A - Social network event analyzing method and system based on geographic positions - Google Patents

Abstract

The present invention provides a social network event analysis method and system based on geographic location, including: performing word segmentation processing on each social network data text to obtain the words of the social network data text; establishing the geographical location corresponding to the social network data text The mapping relationship between the position and the text of the social network data, the geographical position corresponding to the text of the social network data is a word related to the geographical position in the words of the text of the social network data; Social network data text; for each target geographic location, carry out weight calculation to the words of the social network data text corresponding to the target geographic location, obtain and use the keywords of the social network data text corresponding to the target geographic location as the Push the popular events in the target geographic location. Through the solution provided by the present invention, users can be helped to intuitively acquire popular events related to geographic location.

Description

Location-based social network event analysis method and system

technical field

The present invention relates to the field of data mining, in particular to a geographical location-based social network event analysis method and system.

Background technique

The social characteristics and rapid and timely dissemination of social networks have attracted a large number of users who have a high demand for real-time information, enabling everyone in the world to become a source of information and spread it globally. The amount of information carried by the network incident itself has greatly increased. Social network events are a collection of massive news, events and information, and are updated and circulated every day, and have a huge impact on the real society. Especially in the information dissemination of emergencies, it has surpassed the traditional media and become a channel for the rapid dissemination of information. Information on social networks is not only released in a timely manner, but also a microcosm of real social life. Mining information in social network events is conducive to analyzing the real world situation from different angles.

With the explosive development of the mobile Internet, devices with positioning functions are becoming more and more popular, and users can easily obtain more accurate geographic location information, which makes more and more data with geographic attributes. At the same time, in the application of urban planning, tourism, security and other fields, the demand for analysis of this kind of data with geographic location information is also increasing.

The social network represented by Weibo has become the fastest growing Internet application in China. It is a platform for information sharing, dissemination and information acquisition based on user relationships. At present, geographical locations can be marked while publishing information, but these geographical locations are relatively isolated, that is, they are only related to this Weibo. Although massive microblog information can be connected through comments, forwarding, and friends, it cannot be connected in a real spatial range, and it lacks location-based service (Location Based Service, LBS for short) related factors.

Contents of the invention

The present invention provides a method and system for analyzing social network events based on geographic location, which are used for analyzing related social network events based on geographic location.

A first aspect of the present invention provides a method for analyzing social network events based on geographic location, including:

Perform word segmentation processing on each social network data text to obtain words of the social network data text;

Establishing a mapping relationship between the geographic location corresponding to the social network data text and the social network data text, where the geographic location corresponding to the social network data text is a word related to the geographic location among the words in the social network data text;

According to the geographic location corresponding to each social network data text, determine the social network data text corresponding to each preset target geographic location;

For each target geographic location, the words of the social network data text corresponding to the target geographic location are weighted, and the keywords of the social network data text corresponding to the target geographic location are obtained as the target geographic location Popular events are pushed.

Another aspect of the present invention provides a social network event analysis system based on geographic location, comprising:

A word segmentation module, configured to perform word segmentation processing on each social network data text, to obtain words of the social network data text;

A geographic location acquisition module, configured to establish a mapping relationship between a geographic location corresponding to the social network data text and the social network data text, where the geographic location corresponding to the social network data text is a word in the social network data text, words related to geographic location;

The geographic location analysis module is used to determine the preset social network data text corresponding to each target geographic location according to the geographic location corresponding to each social network data text;

The event analysis module is used to calculate the weight of the words of the social network data text corresponding to the target geographic location for each target geographic location, and obtain and use the keywords of the social network data text corresponding to the target geographic location as The popular events in the target geographic location are pushed.

The geographical location-based social network event analysis method and system provided by the present invention analyze the social network data text associated with the geographical location by studying the social network data text, and the key points of the social network data text corresponding to each geographical location are analyzed. Words are pushed as popular events in the geographic location, which can help users intuitively obtain popular events related to the geographic location.

Description of drawings

FIG. 1 is a schematic flow diagram of a method for analyzing social network events based on geographic location provided by Embodiment 1 of the present invention;

FIG. 2 is a schematic structural diagram of a social network event analysis system based on geographic location provided by Embodiment 2 of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention.

FIG. 1 is a schematic flowchart of a method for analyzing social network events based on geographic location provided in Embodiment 1 of the present invention. As shown in FIG. 1 , the method includes:

101. Perform word segmentation processing on each social network data text to obtain words of the social network data text.

In practical applications, a certain amount of social network data texts within a certain period of time can be obtained from the big data analysis platform, for example, Weibo, and correspondingly, before 101, the method can also include: obtaining A preset number of said social networking data texts are posted within a segment.

Specifically, the social network data text in this embodiment may come from the ElasticSearch search engine of the big data analysis platform. Taking microblogs as an example, all microblogs may be original microblogs, excluding those forwarded by users.

Correspondingly, after obtaining a certain amount of social network data texts within a certain period of time, these social network data texts need to be preprocessed, and the preprocessing is mainly to segment the social network data texts. Still taking Weibo as an example, it is necessary to perform word segmentation processing on Weibo text information. Specifically, the text information here does not include information such as pictures posted by users.

In practical applications, the word segmentation processing may be implemented through various implementation manners. For example, a word segmenter may be used to perform word segmentation processing on social network data text. Optionally, 101 may specifically include: using an IKAnalyzer tokenizer to perform word segmentation processing on the social network data text. Specifically, take the word segmentation of microblog text as an example. The word segmenter first loads the dictionary, analyzes the microblog text, and intercepts a token. The search keyword is segmented by iterative retrieval from the largest word to the smallest word, and retrieved in the dictionary. The largest segmented word in the search word, and so on, continue to perform iterative retrieval until the end.

102. Establish a mapping relationship between the geographic location corresponding to the social network data text and the social network data text, where the geographic location corresponding to the social network data text is the word related to the geographic location among the words in the social network data text word.

In practical applications, the geographical location corresponding to the social network data text can be obtained by filtering words related to the geographical location from the words in the social network data text. Specifically, Sogou’s three-level administrative division geographical names dictionary can be used for screening. Correspondingly, taking Weibo as an example, if the geographical location information in the above dictionary is detected in the Weibo text, combined with the context of the Weibo, extract output the geographic location information. Correspondingly, after obtaining the geographical location corresponding to the social network data text, associate the social network data text with these geographical location-related words. Specifically, the geographical text analysis method of named entity recognition can be used, and Sogou’s three-level administrative division gazetteer can be used. If there is geographic location information in the administrative division gazetteer in the microblog, the geographic location can be extracted in combination with the context of the microblog. information, and associated with the Weibo.

Optionally, in order to save processing resources and improve processing efficiency, for social network data texts that do not contain words related to geographical locations, it is determined that the social network data texts do not contain geographical location information, and correspondingly, they can be discarded without processing.

103. According to the geographic location corresponding to each social network data text, determine the social network data text corresponding to each preset target geographic location.

In practical applications, according to the mapping relationship between each social network data text and the geographic location, the social network data text corresponding to each target geographic location can be determined.

Specifically, the target geographic locations can be determined according to actual needs. For example, the popular event analysis can be performed on the geographic locations in the visualized map. Correspondingly, before step 103, the method can also include: The geographic location in the visualized map is used as the target geographic location. For example, the visualization of geographical location analysis can be based on the Web map application implemented by Baidu Map API.

104. For each target geographic location, perform weight calculation on the words of the social network data text corresponding to the target geographic location, and obtain and use the keywords of the social network data text corresponding to the target geographic location as the target geographic location Push the popular events of the location.

Specifically, after determining the social network data text corresponding to each target geographic location, the weight calculation of each word of these social network data texts can be performed using the TF-IDF method, and the keywords are extracted according to the calculation results, and the keyword It is pushed as a popular event of the target geographic location, for example, marked on the target geographic location. Correspondingly, in step 104, the weight calculation of the words of the social network data text corresponding to the target geographic location may specifically include: using the TF-IDF method to calculate the weight of the words of the social network data text corresponding to the target geographic location Do weight calculations.

Specifically, obtaining keywords may specifically include the following steps: normalized calculation of word frequency; reverse word frequency calculation; calculation of entry weights to extract popular words.

To be more specific, Term Frequency (TF for short) refers to the frequency with which a given word appears in a document. This number is normalized to the number of words to prevent it from being biased towards long documents. Inverse Document Frequency (Inverse Document Frequency, referred to as IDF), used to represent that if there are fewer documents containing an entry, the larger the IDF, which means that the entry has a good category discrimination ability. After that, the TF-IDF value of the entry is obtained, and the TF-IDF value is equal to the product of the TF value and the IDF value. Finally, the TF-IDF values of each word are sorted from large to small, and the first few keywords are selected as the results of popular event analysis.

In practical applications, when the analysis object is short text, take microblog as an example, if the usual TF-IDF algorithm is used to directly extract keywords, because microblog text is usually short, if all microblogs are regarded as a document, it will IDF information is lost, and if a single microblog is regarded as a document, a microblog has very few words, and the frequency of each word is basically 1, and the TF information will be lost, which will affect the final selection of keywords. accuracy.

In order to improve the accuracy of the analysis results, the TF-IDF method described in 104 is used to calculate the weight of the words of the social network data text corresponding to the target geographic location, and obtain the key of the social network data text corresponding to the target geographic location words, including:

For each word t _i of the social network data text corresponding to the target geographic location, the social network data text corresponding to the target geographic location is used as the first document, and the word frequency tf _i of the word t _i is calculated according to the first formula _,j , the first formula is: Wherein, n _i,j is the number of occurrences of the word t _i in the first document, ∑ _k n _k,j is the sum of the number of occurrences of all words in the first document;

Taking the social network data text to which the word belongs as the second document, calculate the reverse document frequency idf _i of the word t _i according to the second formula, and the first formula is: Wherein, |D| is the number of files in the corpus, and |{j:t _i ∈ d _j }| is the number of files including the word t _i in the corpus;

According to the third formula, the weight tfidf _i,j of the word t _i is calculated and obtained, and the third formula is: tfidf _i,j =tf _i,j ×idf _i ;

According to the weight of each word of the social network data text corresponding to the target geographic location, the words are sorted from large to small, and the words ranked in the top k positions are used as the social network corresponding to the target geographic location The keyword of the data text, where k is a preset value.

In this embodiment, when calculating the IDF, a single social network data text is regarded as a document, and when the TF value is calculated, all social network data texts corresponding to the target geographic location are regarded as a document, so that there will be different Word frequency also includes IDF information, which can effectively improve the accuracy of keyword results, thereby improving the accuracy of popular event analysis.

Optionally, you can use the visual map to push popular events corresponding to each target geographic location. For example, you can call the Baidu Map API for visual push. Specifically, it can be divided into the following steps: (1) Register Baidu Map API , load the API JS file; (2) create a map container, instantiate and initialize the map; (3) integrate and display map interactive operation data. Specifically, each target geographic location and its corresponding popular events are encapsulated into a JSON format and passed to the front desk for parsing and processing, and the parsed coordinates and corresponding keywords, that is, popular events, are marked on the visual map.

In practical applications, with the popularization of mobile devices and the advancement of wireless communication technology, more and more data have geospatial attributes. Taking Weibo as an example, due to the real-time characteristics of Weibo and the large amount of information, a large number of Weibo contains geographical location information, but a geographical location may be associated with thousands of Weibo, and it is impossible for users to view all of them. If you only read one or two microblogs, you will not have a comprehensive understanding of geographically related microblog events.

Most of the usual social network analysis tools are for comprehensive analysis of microblog text, including the analysis of microblog characteristics, topic mining and emergency detection, etc. Social network data contains a large amount of geographical location information, including user However, the analysis functions of the geographical location information provided by these social network analysis tools are relatively weak, mainly focusing on the analysis of the user’s location information, while the analysis of the event’s The application of location information analysis is seldom, and the geographic location characteristics related to Weibo events are not highlighted.

The geographical location-based social network event analysis method provided in this embodiment analyzes the social network data text associated with the geographical location by studying the social network data text, and uses the keywords of the social network data text corresponding to each geographical location , which is pushed as a popular event in the geographic location, which can help users intuitively obtain popular events related to the geographic location.

FIG. 2 is a schematic structural diagram of a geographical location-based social network event analysis system provided in Embodiment 2 of the present invention. As shown in FIG. 2 , the system includes:

The word segmentation module 21 is used to carry out word segmentation processing to each social network data text, and obtains the words of the social network data text;

The geographic location acquisition module 22 is configured to establish a mapping relationship between the geographic location corresponding to the social network data text and the social network data text, and the geographic location corresponding to the social network data text is in the words of the social network data text , words related to geographic location;

The geographic location analysis module 23 is used to determine the social network data text corresponding to each preset target geographic location according to the corresponding geographic location of each social network data text;

The event analysis module 24 is used for, for each target geographic location, carries out weight calculation to the word of the social network data text corresponding to the target geographic location, obtains and the keyword of the social network data text corresponding to the target geographic location, push as a popular event in the target geographic location.

In practical applications, a certain amount of social network data texts within a certain period of time can be obtained from the big data analysis platform, for example, microblogs, and correspondingly, the system also includes: an acquisition module, which is used in the word segmentation module 21. Before word segmentation processing is performed on each social network data text, a preset number of social network data texts published within a preset time period are obtained.

Specifically, the social network data text in this embodiment may come from the ElasticSearch search engine of the big data analysis platform. Taking microblogs as an example, all microblogs may be original microblogs, excluding those forwarded by users.

Correspondingly, after obtaining a certain amount of social network data texts within a certain period of time, these social network data texts need to be preprocessed, and the preprocessing is mainly to segment the social network data texts. Still taking Weibo as an example, it is necessary to perform word segmentation processing on Weibo text information. Specifically, the text information here does not include information such as pictures posted by users.

In practical applications, the word segmentation processing may be implemented through various implementation manners. For example, a word segmenter may be used to perform word segmentation processing on social network data text. Optionally, the word segmentation module 21 may be specifically configured to use the IKAnalyzer tokenizer to perform word segmentation processing on the social network data text. Specifically, taking the word segmentation of microblog text as an example, the word segmentation module 21 first loads the dictionary, analyzes the microblog text, intercepts a token, and uses the iterative retrieval method from the largest word to the smallest word to segment the search keywords into the dictionary Retrieve the largest segmented word in the search term, and so on to continue iterative retrieval until the end.

In practical applications, the geographic location acquisition module 22 can obtain the geographic location corresponding to the social network data text by filtering words related to the geographic location from the words in the social network data text. Specifically, the geographic location acquisition module 22 can use Sogou's three-level administrative division gazetteer for screening. Correspondingly, taking Weibo as an example, if it is detected that there is geographic location information in the above-mentioned dictionary in the Weibo text, it will be combined with the Weibo The context of the blog is used to extract the geographic location information. Correspondingly, after obtaining the geographic location corresponding to the social network data text, the geographic location acquisition module 22 associates the social network data text with these geographical location-related words. Specifically, the geographical text analysis method of named entity recognition can be used, and Sogou’s three-level administrative division gazetteer can be used. If there is geographic location information in the administrative division gazetteer in the microblog, the geographic location can be extracted in combination with the context of the microblog. information, and associated with the Weibo.

Optionally, in order to save processing resources and improve processing efficiency, the geographic location acquisition module 22 determines that the social network data text does not contain geographic location information for a social network data text that does not contain a word related to a geographic location. It is discarded without processing.

In practical applications, according to the mapping relationship between each social network data text and the geographic location, the social network data text corresponding to each target geographic location can be determined.

Specifically, the target geographic locations can be determined according to actual needs. For example, the popular event analysis can be performed on each geographic location in the visualized map. Correspondingly, in the geographic location analysis module 23, it is also used to For the geographic location corresponding to each social network data text, before determining the social network data text corresponding to each preset target geographic location, the geographic location in the visualized map is used as the target geographic location. For example, the visualization of geographical location analysis can be based on the Web map application implemented by Baidu Map API.

Specifically, after the geographic location analysis module 23 determines the social network data text corresponding to each target geographic location, the event analysis module 24 can use the TF-IDF method to perform weight calculations on each word of these social network data texts, and extract Keyword is generated, and the keyword is pushed as a popular event of the target location, for example, marked on the target location. Correspondingly, the event analysis module 24 may be specifically configured to use the TF-IDF method to perform weight calculation on the words of the social network data text corresponding to the target geographic location.

Specifically, the event analysis module 24 can specifically be used for normalized calculation of word frequency; reverse word frequency calculation; calculation of entry weights to extract popular words.

To be more specific, Term Frequency (TF for short) refers to the frequency with which a given word appears in a document. This number is normalized to the number of words to prevent it from being biased towards long documents. Inverse Document Frequency (Inverse Document Frequency, referred to as IDF), used to represent that if there are fewer documents containing an entry, the larger the IDF, which means that the entry has a good category discrimination ability. After that, the TF-IDF value of the entry is obtained, and the TF-IDF value is equal to the product of the TF value and the IDF value. Finally, the TF-IDF values of each word are sorted from large to small, and the first few keywords are selected as the results of popular event analysis.

In practical applications, when the analysis object is short text, take microblog as an example, if the usual TF-IDF algorithm is used to directly extract keywords, because microblog text is usually short, if all microblogs are regarded as a document, it will IDF information is lost, and if a single microblog is regarded as a document, a microblog has very few words, and the frequency of each word is basically 1, and the TF information will be lost, which will affect the final selection of keywords. accuracy.

In order to improve the accuracy of the analysis results, the event analysis module 24 may specifically include:

The first computing unit is configured to use the social network data text corresponding to the target geographic location as the first document for each word t _i of the social network data text corresponding to the target geographic location, and calculate the word frequency tf _i,j of word t _i , the first formula is: Wherein, n _i,j is the number of occurrences of the word t _i in the first document, ∑ _k n _k,j is the sum of the number of occurrences of all words in the first document;

The second calculation unit is used to use the social network data text to which the word belongs as the second document, and calculate the inverse document frequency idf _i of the word t _i according to the second formula, and the first formula is: Wherein, |D| is the number of files in the corpus, and |{j:t _i ∈ d _j }| is the number of files including the word t _i in the corpus;

The third calculation unit is used to calculate and obtain the weight tfidf _i,j of the word t _i according to a third formula, the third formula is: tfidf _i,j =tf _i,j ×idf _i ;

A processing unit, configured to sort the words from large to small according to the weight of each word in the social network data text corresponding to the target geographic location, and use the top k words as the target geographic location A keyword of the social network data text corresponding to the location, wherein k is a preset value.

Wherein, the value of k may be determined according to the actual required number of keywords, which is not limited in this embodiment.

In this embodiment, when the first calculation unit calculates the IDF, a single social network data text is regarded as a document, and when the second calculation unit calculates the TF value, all social network data texts corresponding to the target geographic location are regarded as a document. Documents, so that there will be different word frequencies and IDF information, so as to effectively improve the accuracy of keyword results, and then improve the accuracy of popular event analysis.

Optionally, a visual map may be used to push popular events corresponding to each target geographic location. Specifically, each target geographic location and its corresponding popular events can be packaged into a JSON format and passed to the front desk for parsing and processing, and the parsed coordinates and corresponding keywords, that is, popular events, can be marked on the visual map.

The social network event analysis system based on the geographic location provided in this embodiment analyzes the social network data text associated with the geographic location by studying the social network data text, and uses the keywords of the social network data text corresponding to each geographic location , which is pushed as a popular event in the geographic location, which can help users intuitively obtain popular events related to the geographic location.

Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the system described above can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

Those of ordinary skill in the art can understand that all or part of the steps for implementing the above method embodiments can be completed by program instructions and related hardware. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above-mentioned method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other various media that can store program codes.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the various embodiments of the present invention. scope.

Claims (10)

1., based on the social networks affair analytical method in geographic position, it is characterized in that, comprising:

Word segmentation processing is carried out to each social network data text, obtains the word of described social network data text;

Set up the mapping relations of geographic position corresponding to described social network data text and described social network data text, the geographic position that described social network data text is corresponding is in the word of described social network data text, the word relevant to geographic position;

The geographic position corresponding according to each social network data text, determines the social network data text that each target geographic position of presetting is corresponding;

For each target geographic position, weight calculation is carried out to the word of social network data text corresponding to described target geographic position, obtain and by the keyword of social network data text corresponding for described target geographic position, the hot ticket as described target geographic position pushes.

2. method according to claim 1, is characterized in that, the word of the described social network data text corresponding to described target geographic position carries out weight calculation, comprising:

Utilize TF-IDF method, weight calculation is carried out to the word of social network data text corresponding to described target geographic position.

3. method according to claim 2, it is characterized in that, describedly utilize TF-IDF method, weight calculation is carried out to the word of social network data text corresponding to described target geographic position, obtain the keyword of social network data text corresponding to described target geographic position, specifically comprise:

For each word t of social network data text corresponding to described target geographic position _i, using social network data text corresponding to described target geographic position as the first document, according to the first formulae discovery institute predicate t _iword frequency tf _i,j, described first formula is: wherein, n _i,jfor institute predicate t _ioccurrence number in described first document, Σ _kn _k,jfor the occurrence number sum of all words in described first document;

Using the social network data text belonging to institute's predicate as the second document, according to the second formulae discovery institute predicate t _ireverse document-frequency idf _i, described first formula is: wherein, | D| is the number of files in corpus, | { j:t _i∈ d _j| for described corpus comprises institute predicate t _inumber of files;

According to the 3rd formula, calculate and obtain institute predicate t _iweights tfidf _i,j, described 3rd formula is: tfidf _i,j=tf _i,j× idf _i;

According to the weights of each word of social network data text corresponding to described target geographic position, descending sequence is carried out to described each word, using the word that the comes front k position keyword as social network data text corresponding to described target geographic position, wherein, k is default value.

4. the method according to any one of claim 1-3, is characterized in that, describedly carries out word segmentation processing to social network data text, comprising:

Utilize IKAnalyzer segmenter, word segmentation processing is carried out to described social network data text.

5. the method according to any one of claim 1-3, is characterized in that, the described geographic position corresponding according to each social network data text, before determining the social network data text that each target geographic position of presetting is corresponding, also comprises:

Using the geographic position in visualized map as described target geographic position.

6. the method according to any one of claim 1-3, is characterized in that, described word segmentation processing is carried out to each social network data text before, also comprise:

Obtain the described social network data text of the predetermined number issued in preset time period.

7., based on the social networks event analysis system in geographic position, it is characterized in that, comprising:

Word-dividing mode, for carrying out word segmentation processing to each social network data text, obtains the word of described social network data text;

Geographic position acquisition module, for setting up the mapping relations in geographic position corresponding to described social network data text and described social network data text, the geographic position that described social network data text is corresponding is in the word of described social network data text, the word relevant to geographic position;

Geolocation analysis module, for the geographic position corresponding according to each social network data text, determines the social network data text that each target geographic position of presetting is corresponding;

Event analysis module, for for each target geographic position, weight calculation is carried out to the word of social network data text corresponding to described target geographic position, obtain and by the keyword of social network data text corresponding for described target geographic position, the hot ticket as described target geographic position pushes.

8. system according to claim 7, is characterized in that,

Described event analysis module, specifically for utilizing TF-IDF method, carries out weight calculation to the word of social network data text corresponding to described target geographic position.

9. system according to claim 8, is characterized in that, described event analysis module comprises:

First computing unit, for each word t for social network data text corresponding to described target geographic position _i, using social network data text corresponding to described target geographic position as the first document, according to the first formulae discovery institute predicate t _iword frequency tf _i,j, described first formula is: wherein, n _i,jfor institute predicate t _ioccurrence number in described first document, Σ _kn _k,jfor the occurrence number sum of all words in described first document;

Second computing unit, for using the social network data text belonging to institute's predicate as the second document, according to the second formulae discovery institute predicate t _ireverse document-frequency idf _i, described first formula is: wherein, | D| is the number of files in corpus, | { j:t _i∈ d _j| for described corpus comprises institute predicate t _inumber of files;

3rd computing unit, for according to the 3rd formula, calculates and obtains institute predicate t _iweights tfidf _i,j, described 3rd formula is: tfidf _i,j=tf _i,j× idf _i;

Processing unit, for the weights of each word according to social network data text corresponding to described target geographic position, descending sequence is carried out to described each word, using the word that the comes front k position keyword as social network data text corresponding to described target geographic position, wherein, k is default value.

10. the system according to any one of claim 7-9, is characterized in that,

Described word-dividing mode, specifically for utilizing IKAnalyzer segmenter, carries out word segmentation processing to described social network data text.