JP2009157510A - System, method and program for identifying spam information - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To effectively identify spam descriptions included in blog articles. <P>SOLUTION: A system for identifying spam information includes a blog data storage unit 41 for storing blog articles to be posted and browsed via a communication network, and a spam determination means 2 for analyzing texts in the blog articles and determining whether or not preset spam description patterns are included in the blog articles. The spam determination means 2 has a function of specifying, as a spam description, text information set in different blog articles and guiding to common link information, and identifying and separating any blog article including the spam description from the blog articles. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for discriminating spam descriptions included in a blog (web log) article, and more particularly to a method for improving the accuracy of discriminating spam descriptions.

A blog (weblog: web log) article that is used as a personal diary or a bulletin board between specific members is published as a website, and a form of communication among many users via this blog article is widespread is doing.
The blog is uploaded as a website on the Internet, and a user who has viewed the blog can write a comment on the blog. Since viewers can easily write comments, it is used in various situations such as a place for communication between companies and consumers.

  However, on the other hand, unlike the normal use of blogs, information that is not necessary for viewers (users) in blog articles, automatic creation for the purpose of advertising, guidance to specific sites, number of clicks, etc. Etc., a lot of writing called “spam” (hereinafter referred to as “spam description”) is performed, and the inconvenience of burying the content of the blog article and the original trend tendency has arisen.

On the other hand, a related technique for discovering spam descriptions from blog articles has been disclosed (Patent Document 1).
This system is composed of a blog server for managing the blog operator and a lookup server for filtering. The blog server detected that a blog site published by the blog operator posted a comment or a trackback. In this case, the filtering lookup server determines whether or not the post is a spam description based on fixed pattern data such as a URL or a keyword. Performs post exclusion processing.

In addition, as a related technique of the above contents, an example of a spam description filtering update method is disclosed (Patent Document 2).
This related technology consists of a spam filter and an update component. When a spam filter is trained on a provider and additional updates occur, the update component does not distribute it all on the user device, Extract and distribute only incremental updates.

JP 2007-115173 A JP 2006-012165 A

  However, the spam description discovery method disclosed in Patent Document 1 cannot deal with spam description when one blog article is examined. For example, it deals with a spam description having a pattern described across different blog articles. There is an inconvenience that can not be done.

In addition, there are many types of spam descriptions such as multi-posts with the same description in the spam description. Therefore, the same description contents can be used to exchange spam descriptions that exist in different blog articles across dates and sites. Need to refer.
Similarly to Patent Document 1, the related art of Patent Document 2 cannot deal with spam descriptions other than URLs and keyword bases. In addition, because this related technology assumes email spam, it is a method that processes one data at a time, so it is a type that needs to be determined by comparison or reference with other data (blog articles) There is an inconvenience that it cannot cope with spam description.

  Furthermore, even if the rules and procedures for determining the spam description of the blog article are fixedly executed, there is a disadvantage that the accuracy of removing the spam description is quickly reduced. This is because spam descriptions are generated one after another, and the spam descriptions of these variants have new description patterns that are not simply word replacements. There is an inconvenience of increasing in a short period of time.

[Object of invention]
The present invention improves the inconvenience of the above-described conventional example, effectively discriminates and separates spam descriptions contained in a blog article, and sets a blog article so that a user can use it effectively, and spam information It is an object of the present invention to provide a discrimination method and a spam information discrimination program.

  To achieve the above object, a spam information determination system according to the present invention includes a blog article storage device that stores blog articles that are posted and viewed via a communication network, and performs text analysis on the blog articles. A spam information determination system comprising: a spam determination unit that determines whether or not each blog article includes a preset spam description pattern, wherein the spam determination unit is set in the different blog article The text information to be guided to the common link destination information is specified as a spam description, and the blog article including the spam description is discriminated and separated from the blog article.

  In addition, the spam information determination method according to the present invention includes a blog article storage device that stores blog articles that are posted and viewed via a communication network, and performs text analysis on the blog articles and sets the blog articles in advance. Spam determination means for determining whether or not a specified spam description pattern is included, and a spam information determination method for determining a blog article including a spam description from the blog article as a spam article, the blog article storage A text analysis step for performing text analysis on a blog article input from the apparatus, and spam for determining whether or not a predetermined spam description pattern is included in the text information of each blog article based on the analysis result Text information that is set in different blog posts and leads to common link destination information And spam description specifying step of specifying as Pam description, is characterized in that the posts that contains spam written from the posts with spam articles discriminating step for discriminating isolated as spam articles, the.

  Further, the spam information determination program according to the present invention includes a blog article storage device that stores blog articles that are posted and browsed via a communication network, and performs text analysis on the blog articles and sets the blog articles in advance. A spam information discriminating program for discriminating a blog article including a spam description from the blog article as a spam article, comprising: a spam judging means for judging whether or not the spam description pattern is included. A text analysis function for performing text analysis on a blog article input from an article storage device, and determining whether or not a predetermined spam description pattern is included in the text information of each blog article based on the analysis result Spam pattern judgment function, set in different blog articles and lead to common link information A spam description identifying function for identifying text information as a spam description, and a spam article determination function for determining and separating a blog article including a spam description from the blog article as a spam article, and causing the computer provided with the spam determination means to execute. It is characterized by.

  Since the present invention is configured and functions as described above, according to this, a blog article storage device that stores blog articles posted and viewed via a communication network and a common link destination set in different blog articles. It is possible to discriminate spam descriptions of various patterns by identifying the text information that leads to information as a spam description, and having a spam judging means for discriminating a blog article including the spam description. You can effectively reduce blog articles that contain descriptions.

[Embodiment 1]
Next, the basic configuration content of Embodiment 1 of the present invention will be described.

  As shown in FIG. 1, the spam information discrimination system of this embodiment is sent from a data storage device 4, blog data input means 1 for reading blog data from the data storage device 4, and the blog data input means. Spam determination means 2 for determining the presence / absence of spam description in the blog data, and data output means 3 for storing the blog data determined by the spam determination means 2 based on the presence / absence of spam description in the data storage device 4, respectively. It has become.

  Here, the spam determination means 2 includes a text analysis means 21 for text-processing the input blog data, and a spam pattern determination for referring to the text information of the blog data based on the spam definition information stored in the data storage device 4 in advance. Means 22 and spam description determination means 23 for determining whether the text information included in the blog data is a spam description, and discriminates and outputs a spam article including the spam description and an article data not including the spam description.

  Further, the data storage device 4 stores a blog data storage unit 41 for storing a blog article posted via a communication network, and spam definition information (file) including preset spam description pattern information (file). A spam definition information storage unit 42, a spam specific information storage unit 43 that stores spam specific information that lists spam descriptions corresponding to the spam patterns, and a blog article (data) that is determined not to contain a spam description. An article data storage unit 44 to be stored and a spam data storage unit 45 to store data determined to be a blog article including a spam description are configured.

This will be described in detail below.
The blog data input unit 1 has a data acquisition function for acquiring a blog article (blog data) from the blog data storage unit 41 and a blog data input function for inputting (transmitting) the acquired blog article to the spam determination unit 2. FIG. 2 shows an example of the input blog data.

The text analysis unit 21 of the spam determination unit 2 includes a data analysis function for performing morphological analysis on the blog data input from the blog data input unit 1, and a word break for text information included in the input blog data. An output information generation function for generating output information by performing sentence breaks, topic breaks, etc., and adding used special symbol information such as used words, phrases, sentence cutting (sentence) information, and link destination URLs for each blog article I have.
FIG. 3 is an example of output information from the text analysis means 21.

  Further, the text analysis means 21 has a sort processing function for performing sort processing such as alphabetical order or aitou order on the generated output information.

  The spam pattern determination unit 22 refers to the output information reading function for reading the output information from the text analysis unit 21 and the spam definition information (spam pattern) stored in advance in the spam definition information storage unit 42 based on the output information. And a spam pattern determination function for determining whether or not the output information matches the spam definition information. Here, FIG. 4 shows an example of spam definition information.

  Thereby, the spam pattern determination means 22 can determine whether or not the text information in each blog article is spam data, and the spam pattern determination means 22 examines the text information in a single blog article. It is possible to determine whether or not a spam pattern that can be determined is included in the text information of each blog article.

Here, the spam pattern determination means 22 performs the spam determination on the spam pattern having “single” in the determination category of FIG.
For example, in the P1 word enumeration type spam pattern, it is determined whether or not the word (W) and the word in the rule model are enumerated by separating them with commas.
Here, an example of output information including a spam article is shown in FIG. J1 in FIG. 5 which is the output information of the text analysis means 21 is that the result of the morphological analysis is a repetition of noun / symbol (comma) / noun / symbol /... Determined.

  Further, the spam pattern determination means 22 stores link destination data (Web page, PDF file, etc.) of links set in the blog article based on words and phrases registered in the spam unique information storage unit 43 as spam unique information in advance. It has a link destination data determination function that refers to text information and determines whether or not a word or phrase that matches spam-specific information is included in the link destination data.

  Further, the spam pattern determination unit 22 has a spam description specifying function for specifying a spam description pattern that appears redundantly by comparing and referring to text information (output information) in a plurality of different blog articles.

  Further, for example, the spam pattern determination means 22 includes a link extraction function for extracting links and URL descriptions from different blog articles to the same site, such as J2 and J3 in FIG. 5, and text information in a plurality of blog articles. By comparison, the same description extraction function is provided for extracting a multi-post description or an automatically generated description in which the same description is written in a plurality of different blog articles as shown by J4 and J5 in FIG.

Further, the spam pattern determination unit 22 extracts the above-mentioned automatically generated description and multi-post description extraction processing in addition to a common description written in a plurality of blog articles, an approximate description that is not completely identical, or similar. It also extracts expressions.
The spam pattern determination means 22 calculates a spam degree indicating the possibility of being a spam description for a description or expression that matches a predetermined percentage of text information included in a plurality of different blog articles. In addition, a comparison spam degree calculation storage function for storing the extracted description and the spam degree of each description in association with each other is provided.

  Here, the spam pattern determination unit 22 extracts common or approximate expressions and descriptions appearing in different blog articles, and may be a spam description based on the frequency, the number of descriptions, the approximate ratio, and the like. A higher numerical value may be assigned as a spam degree (weighting) for high expressions and descriptions.

  Spam description determination means 23 associates the spam degree and the spam description with the spam degree calculation function for calculating the spam degree based on the rate of coincidence with the characteristic word / phrase pattern for the spam determination result of the spam pattern determination means 22. And a spam description storage function for storing in a preset spam degree storage means 24.

Further, the spam description determination unit 23 has a processing completion determination function for determining whether or not the above processing has been performed on all output information from the text analysis unit 21.
Further, the spam description determination means 23 has a spam degree total notifying function for totalizing the spam degrees calculated for each blog article and notifying the result to the data output means 3.

  The data output means 3 sends a blog article whose total spam degree calculated by the spam judgment means 2 exceeds a spam article judgment threshold set in advance for each blog article to the spam data storage unit 45, or spam article judgment. A spam article discrimination storage function is provided for storing a blog article below the threshold value in the article data storage unit 44 as a normal article.

The blog data storage unit 41 has a blog article storage function for storing blog data (blog articles) including a title, posted content, date, and the like.
Note that the blog data storage unit 41 may be connected to the Internet and the stored blog articles may be set to be writable via the Internet.

The spam definition information storage unit 42 has a spam definition information storage function for storing a preset spam pattern and a rule model of spam description as spam definition information.
The spam unique information storage unit 43 is a unique information storage function for storing spam unique information in which spam descriptions are listed for words, phrases, and the like specified in advance as specific examples corresponding to the spam definition information and the spam pattern. Have

An article data storage unit 44 that stores a blog article (data) determined not to include a spam description, and a spam data storage unit 45 that stores data determined to be a blog article including a spam description Is done.
[Description of Operation of First Embodiment]

Next, the overall operation of this embodiment will be described.
First, text analysis is performed on the blog article input from the blog data storage unit 41 (text analysis process), and a spam description pattern set in advance is included in the text information of each blog article based on the analysis result. (Spam pattern determination step).
Next, text information that is set in different blog articles and leads to common link destination information is specified as a spam description (spam description specifying step), and a blog article that includes the spam description is determined as a spam article from the blog article. (Spam article discrimination process).

  Here, regarding the text analysis step, spam pattern determination step, spam description identification step, and spam article determination step, the execution contents may be programmed and executed by a computer.

  Next, the operation of the system of the above embodiment will be described based on the flowchart of FIG.

  First, the blog data input means 1 inputs the blog article (blog data) acquired from the blog data storage unit 41 to the spam determination means 2, and the text analysis means 21 performs morphological analysis on the blog data (step) S101).

  Here, the text analysis means 21 performs word breaks, sentence breaks, topic breaks, etc. on the text information included in the sent blog data, and also uses words, phrases, sentences (phrases) for each blog article. Information and use special symbol information such as a link destination URL are added to generate output information (step S102).

  Next, the spam pattern determination unit 22 reads the output information of the text analysis unit 21 and refers to the spam definition information (spam description pattern) stored in advance in the spam definition information storage unit 42 based on the output information. It is determined whether or not the output information matches the spam definition information (step S103).

Thereby, the spam pattern determination means 22 determines whether or not the text information in each blog article is spam data, that is, whether or not it corresponds to a spam description.
Thereby, the spam pattern determination means 22 determines whether or not a spam pattern that can be determined by examining text information in a single blog article is included in the text information of each blog article.

Here, in the determination of the presence / absence of the spam description in the single blog article described above, the spam determination is performed for the spam pattern having “single” in the determination category of FIG.
For example, in the P1 word enumeration type spam pattern, it is determined whether or not the word (W) and the word in the rule model are enumerated by separating them with commas.
In J1 of FIG. 5, the result of the morpheme analysis is the repetition of noun / symbol (comma) / noun / symbol /... As the output of the text analysis means 21, which matches the P1 pattern of FIG. Then, it is determined.

  Further, the spam pattern determination means 22 stores link destination data (Web page, PDF file, etc.) of links set in the blog article based on words and phrases registered in the spam unique information storage unit 43 as spam unique information in advance. With reference to the text information, it is determined whether or not a word or phrase that matches the spam specific information is included in the link destination data (step S104).

Next, the spam description determination unit 23 calculates and stores the spam degree based on the ratio of matching with the characteristic word / phrase pattern for the determination result (step S105).
Further, the spam description determination unit 23 determines whether or not the above data processing (spam degree calculation processing) has been performed on all output information from the text analysis unit 21 (step S106), and all output information is processed. Steps S103 to S106 are performed.

  Next, the text analysis unit 21 performs sort processing such as alphabetical order and aiueo order on the output information (step S107), and the spam pattern determination unit 22 performs text information (output information) in different blog articles. ) To identify spam description patterns that appear in duplicate (step S108).

Here, for example, the spam pattern determination means 22 extracts links and URL descriptions that lead to the same site from different blog articles such as J2 and J3 in FIG.
Further, by comparing text information in a plurality of blog articles, a multi-post description or an automatically generated description in which the same description is written in a plurality of different blog articles as shown in J4 and J5 in FIG. 5 is extracted.
In addition, regarding the above-described automatic generation description and multi-post description extraction processing, in addition to a common description written in a plurality of blog articles, an approximate description and a similar expression that are not completely matched are also extracted.

Next, the spam description determination unit 23 determines a spam degree indicating the possibility of being a spam description for a description or expression that matches a predetermined percentage of text information included in a plurality of different blog articles. In addition to the calculation, the extracted description and the spam level of each description are associated with each other and stored in the preset spam level storage unit 24 (step S109).
Here, the spam description determination unit 23 extracts common or approximate expressions and descriptions appearing in different blog articles, and may be a spam description based on the frequency, the number of descriptions, the approximate ratio, and the like. A higher numerical value may be assigned as a spam degree (weighting) for high expressions and descriptions.

  Next, the spam description determination unit 23 aggregates the calculated spam degree for each blog article unit and notifies the data output unit 3 of the result (step S110).

  Next, the blog article in which the data output means 3 lowers the blog article (spam article) whose calculated spam degree exceeds the spam article determination threshold set in advance for each blog article to the spam data storage unit 45. Is stored in the article data storage unit 44 as a normal article (step S111).

As described above, the present embodiment is configured to refer to the spam definition information storage unit 42 and the spam specific information storage unit 43, determine spam descriptions existing in large quantities and various patterns, and classify them as excluded. Therefore, the accuracy of analysis / reference / information collection can be improved.
In the present embodiment, the spam pattern determination unit 22 is configured to perform spam pattern determination executed by comparing and referring to a plurality of blog articles in addition to the spam pattern determination in a single blog article. It can also deal with types of spam that cannot be dealt with by simple word or rule removal.

[Embodiment 2]
Next, the spam information discrimination system according to the second embodiment of the present invention will be described. Here, the same reference numerals are assigned to the same portions as those of the first embodiment described above.
In the second embodiment, the components of the system have almost the same configuration as that of the first embodiment (FIG. 1), and the new spam learning means 5 and the spam definition updating means 6 are newly provided. This is different from the case of 1.
This will be described in detail below.

  In the second embodiment, as in the first embodiment (FIG. 1) described above, the data storage device 4, the blog data input means 1 for reading blog data from the data storage device 4, and the blog data input means. Spam determination means 2 for determining the presence or absence of spam description in the blog data sent from, and data output means 3 for storing the blog data determined by the spam determination means 2 based on the presence or absence of spam description in the data storage device 4, respectively. It has.

  Here, the present embodiment is connected to the data output means 3 and reads out the article data information to learn new spam description information and rules, and is connected to the new spam learning means 5 and the learning result described above. The spam definition updating means 6 for updating the information registered in the spam definition information storage unit 42 and the spam specific information storage unit 43 is provided.

The new spam learning means 5 is the article data reading function for reading the article data information stored in the article data information storage unit 45 of the data storage device 4 and the read article data. It has a spam article feature extraction function that extracts features of descriptions that are not spam and have a high degree of spam.
Here, the article data information read by the new spam learning means 5 is article data information determined by the spam determination means 2 as not spam based on the spam definition information in the spam definition information storage unit 42. However, this article data information does not fall under any spam description pattern, but includes descriptions that have the characteristics of spam descriptions and blog articles that contain spam descriptions that are lower than the spam level threshold. Yes.

Also included are descriptions that have been discovered as users browse blog articles and have been declared as spam descriptions, and spam descriptions that have been set up in advance according to the spam description declaration.
Here, the new spam learning means 5 uses, for example, a spam degree higher than a preset value as a positive example, a spam degree lower than a preset value or '0' as a negative example, and a spam degree Extract features specific to high blog posts.
Further, the new spam learning means 5 has a registered description selecting function for selecting a description having a high feature degree, a description declared by the user, or a description for setting rule description of the description among the descriptions extracted from the features. ing.

The spam definition updating means 6 has a new description pattern determination function for determining whether or not the selected description is a new spam description pattern based on the result of the description selection by the new spam learning means 5.
The spam definition updating means 6 corresponds to a definition information adding function for additionally registering a description or description pattern determined as a new spam description pattern in the spam definition information storage unit 42, and the additionally registered description or description pattern. A unique information additional registration function for additionally registering spam unique information to the spam unique information storage unit 43.
As a result, the spam definition updating unit 6 can update the description information registered in the spam definition information storage unit 42, the spam description information registered in the spam specific information storage unit 43, and the spam description rule.

[Description of Operation of Embodiment 2]

Next, the overall operation of the second embodiment will be described.
First, text analysis is performed on the blog article input from the blog data storage unit 41 (text analysis step), and whether or not a spam description pattern set in advance is included in the text information of each blog article based on the analysis result. It is determined whether or not (spam pattern determination step).
Next, text information that is set in different blog articles and leads to common link destination information is specified as a spam description (spam description specifying step), and a blog article that includes the spam description is determined as a spam article from the blog article. (Spam article discrimination process).
Next, the spam description feature characterizing the spam description is extracted from the blog article, and the spam description pattern is updated based on the spam description feature (spam description pattern updating step).

Next, the operations of the new spam learning unit 5 and the spam definition updating unit 6 of the second embodiment will be described based on the flowchart of FIG.
The operation of the first embodiment (FIG. 6) described above is assumed to operate similarly in the second embodiment.

The new spam learning means 5 reads a normal article (hereinafter referred to as “article data information”) stored in the article data information storage unit 45 of the data storage device 4 (step S201: article data information reading step).
Here, the new spam learning means 5 starts the article data information reading process in response to the data storage notification that the data output means 3 has stored data in the article data storage unit 44 or the spam data storage unit 45. It is good also as a setting.

Here, the article data information read by the new spam learning means 5 is article data information determined by the spam determination means 2 as not spam based on the spam definition information. However, this article data information does not fall under any spam description pattern, but includes descriptions that have the characteristics of spam descriptions and blog articles that contain spam descriptions that are lower than the spam level threshold. Yes.
Also included are descriptions that have been discovered while the user browses blog articles and have been declared as spam descriptions, and spam descriptions that have been set to a predetermined spam level according to the spam description declaration.

Next, the new spam learning means 5 takes the spam degree as a positive example when the spam degree is higher than a preset value, and as a negative example when the spam degree is lower than the preset value or '0'. Feature extraction specific to a high blog article is performed (step S202).
Here, the new spam learning means 5 performs the feature extraction of the blog article having a high spam degree by setting the ranking in the order of the characteristic degree such as words, phrases, URLs, and symbols.

  Next, the new spam learning means 5 selects a description having a high characteristic degree, a description declared by the user, or a description for which a description content is set to rule from among the descriptions extracted from the features (step S203). .

Next, the spam definition updating means 6 refers to the spam definition information in the spam definition information storage unit 42 based on the selected description (step S204), and whether each selected description is a new spam description pattern or not. Is determined (step S205).
If it is not a new spam description pattern or spam-specific information, the process is terminated.

Next, the spam definition updating means 6 additionally registers the description or description pattern determined as the new spam description pattern in the spam definition information storage unit 42 (step S206).
Further, the spam definition updating means 6 extracts spam specific information corresponding to the additionally registered description or description pattern from the read article data information and stores (registers) it in the spam specific information storage unit 43. (Step S207).

As described above, in the second embodiment, the new spam learning means 5 is used for the variant different from the spam rule and the spam word / spam phrase, and for the new spam description and spam article information discovered by the user. Based on this, the text description characterizing the spam description is extracted, and the spam definition updating means 6 is configured to update a preset spam description pattern, spam specific information, spam description rules, and the like.
Thereby, in this embodiment, it is possible to flexibly follow a spam description and its variants that are developed one after another, and to improve the accuracy of spam description discrimination.

The present invention can be applied to an information processing system that extracts trend information from blog information and a system that analyzes a reaction to advertisement information such as a commercial or a campaign as an effect measurement.
Further, it can be applied to spam description discrimination of a system that prevents a spam description from being posted in a blog article, trackback, and comment on a blog management site.
Furthermore, the present invention can also be applied to spam description discovery methods and spam description filter update services used by blog management sites and blog information collection and utilization companies.

It is a schematic block diagram which shows one Example of the spam information discrimination | determination system by this invention. It is explanatory drawing which shows an example of the blog article in the spam information discrimination | determination system disclosed in FIG. It is explanatory drawing which shows an example of the output information in the spam information discrimination | determination system disclosed in FIG. It is explanatory drawing which shows an example of the spam definition information in the spam information discrimination system disclosed in FIG. It is explanatory drawing which shows an example of the output information containing the spam article in the spam information discrimination system disclosed in FIG. 5 is a flowchart showing overall operation processing steps of the spam information discrimination system according to the present invention. It is a schematic block diagram which shows one Example of the spam information discrimination | determination system by this invention. 5 is a flowchart showing overall operation processing steps of the spam information discrimination system according to the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Blog data input means 2 Spam judgment means 3 Data output means 4 Data storage device 5 New spam learning means 6 Spam definition update means 21 Text analysis means 22 Spam pattern judgment means 23 Spam description judgment means 24 Spam degree memory means 41 Blog data storage Section 42 Spam definition information storage section 43 Spam specific information storage section 44 Article data storage section 45 Spam data storage section

Claims (9)

A blog article storage device that stores blog articles that are posted and viewed via a communication network, and whether or not a text description analysis is performed on each blog article and a preset spam description pattern is included in each blog article A spam information discriminating system comprising a spam judging means for judging
The spam determination means has a function of identifying text information that is set in the different blog articles and that leads to common link destination information as a spam description, and distinguishing and separating a blog article that includes the spam description from the blog article. A spam information discrimination system characterized by comprising. In the spam information determination system according to claim 1,
A common text discriminating function for the spam judging means to extract common text information included in each of the different blog articles and determine whether the text information includes a description characterizing a spam description based on the spam description pattern; Spam information discrimination system characterized by having. In the spam information determination system according to claim 1,
A spam information discrimination system comprising a link destination spam judgment function in which the spam judgment unit acquires the link destination information and judges whether or not the spam description pattern is included in the link destination information. In the spam information determination system according to claim 1,
Spam degree indicating the possibility that the text description is a spam description based on an approximate text extraction function for the spam determination means to extract approximate text information from the different blog articles and the degree of approximation of the extracted text information And a spam article discrimination function for discriminating the blog article as a spam article when the spam degree of the blog article is greater than or equal to a preset value. Information discrimination system. In the spam information determination system according to claim 4,
A link destination information acquisition function for the spam determination means to acquire link destination information of the extracted text information;
A link destination information spam degree calculation function for calculating a spam degree indicating the possibility that the text description is a spam description based on a degree of coincidence between the acquired link destination information and the text description is provided. Spam information discrimination system. In the spam information determination system according to claim 4,
Extracting a spam description feature having a high spam degree description from a blog article having a spam degree less than a preset value among the blog articles, and extracting the spam description pattern as an index for determining a description characterizing the spam description A spam information discriminating system, characterized in that a spam description pattern updating means for updating based on the described spam description characteristics is provided in the spam judging means. A blog article storage device that stores blog articles that are posted and viewed via a communication network, and whether or not each blog article includes a predetermined spam description pattern by performing text analysis on the blog article A spam information determining method for determining a blog article including a spam description from the blog article as a spam article, comprising:
A text analysis step for performing text analysis on the blog article input from the blog article storage device, and whether or not a predetermined spam description pattern is included in the text information of each blog article based on the analysis result A spam pattern determination step for determining a spam description, a spam description specifying step for specifying text information set in different blog articles and leading to common link destination information as a spam description, and a blog article including a spam description from the blog article And a spam article discriminating step for discriminating and separating them as spam articles. The spam information determination method according to claim 7,
A spam description pattern update step of extracting a spam description feature characterizing a spam description from the blog article and updating the spam description pattern based on the spam description feature is provided after the spam article determination step. Spam information discrimination method. A blog article storage device that stores blog articles that are posted and viewed via a communication network, and whether or not each blog article includes a predetermined spam description pattern by performing text analysis on the blog article A spam information determination program for determining a blog article including a spam description from the blog article as a spam article, comprising:
A text analysis function for performing text analysis on a blog article input from the blog article storage device, and whether or not a predetermined spam description pattern is included in the text information of each blog article based on the analysis result Spam pattern determination function to determine, spam description specification function to specify text information that is set in different blog articles and leads to common link destination information as spam description, blog articles that contain spam descriptions from the blog articles are spam articles A spam information determination program that causes a computer provided with the spam determination means to execute a spam article determination function that determines and separates

Images