JP2008123062A - Device, method, and program for classifying content - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device for classifying Web pages such as blog for every region from the details of the content of the Web page. <P>SOLUTION: This device 10 for classifying the content of a Web page or the like is provided with: a means 110 for performing the morphemic analysis of training data 30 including descriptions related with a residence area in the content; a means 120 for extracting predetermined morphemes from the result of the morphemic analysis; a means 130 for calculating mean mutual information quantity between the extracted morphemes and residence categories; a means 107 for storing data acquired by associating the residence categories, the extracted morphemes and the mean mutual information quantity between the residence categories and the morphemes with each other; and a means 140 for classifying the input predetermined content into residence categories based on the data stored in the means 107. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an apparatus, a method, and a program for classifying content.

  For example, a diary Web page called a blog that is updated daily is known. This blog uses a “trackback function” that automatically sends the comments you wrote to the other party, while also quoting the article when using comments about other people's blogs as the story of your site's diary. It is known as a Web page that can be expected to exchange a wide range of opinions.

  In recent years, the number of Web pages providing this blog has increased, and there are many blogs with different tastes. For this reason, it is desirable for businesses that provide blog services to be able to classify blogs according to predetermined criteria. For example, blogs can be classified by region as a category to be classified.

On the other hand, a method for classifying Web pages by feature is known (for example, Patent Document 1). According to Patent Document 1, it is possible to estimate and categorize the role and personality of the Web page on the network based on the statistical values of the actions of creators and viewers performed on the Web page. Become.
JP 2006-163997 A

  However, even if Patent Document 1 is used, information used as statistical information is limited to basic statistical information in browsing a Web page such as the number of accesses and the number of comments. It is not always possible. That is, as described above, it is difficult to classify blogs by region based on the content of the content.

  An object of the present invention is to provide a method, an apparatus, and a program for classifying a web page such as a blog by region from the content of the content of the web page and inferring the residence area of the creator of the content.

(1) An apparatus for analogizing the residence area of a content creator,
Among the contents, means for morphological analysis of training data including a description related to a residential area;
Means for extracting a predetermined morpheme from the result of the morpheme analysis;
Means for calculating an average mutual information amount between the extracted morpheme and the residence category;
Means for storing data in which the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme are related;
Means for classifying the inputted predetermined content into the residence category based on the data stored in the means for storing;
A device comprising:

  The device of (1) performs morphological analysis on training data including descriptions related to the residential area in the content, extracts a predetermined morpheme from the result of the morphological analysis, and averages the average morpheme between the extracted morpheme and the residential category. Calculate the amount of information, store the data related to the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme, and input based on the data stored in the storage means The predetermined content is classified into the residence category.

  Therefore, it is possible to classify the input predetermined content for each residence category based on the average mutual information stored in association with the training data. Therefore, for example, by classifying a Web page such as a blog by region from the content of the content of the blog, it is possible to infer the residence area of the creator of the content.

MI(w,c)：形態素となる単語wとカテゴリｃの間の平均相互情報量
により、平均相互情報量を算出することを特徴とする装置。 (2) A device for classifying the content described in (1),
The means for calculating the average mutual information amount has P as a probability,

MI (w, c): an apparatus for calculating an average mutual information amount from an average mutual information amount between a word w as a morpheme and a category c.

MI(w,c)：形態素となる単語wとカテゴリｃの間の平均相互情報量
ｅ：カテゴリｃの語を有し、かつ、形態素の単語を有するものの集合の数
ｆ：カテゴリｃの語を有し、かつ、形態素の単語を有しないものの集合の数
ｇ：カテゴリｃの語を有さず、かつ、形態素の単語を有するものの集合の数
ｈ：カテゴリｃの語を有さず、かつ、形態素の単語を有しないものの集合の数
Ｎ＝ｅ＋ｆ＋ｇ＋ｈ
により、平均相互情報量を算出することを特徴とする装置。 (3) A device for classifying the content described in (1) or (2),
The means for calculating the average mutual information amount is:

MI (w, c): Average mutual information between morpheme word w and category c e: Number of sets having category c words and morpheme words f: Category c words And the number of sets of those having no morpheme words g: the number of sets having no words of category c and having the words of morphemes h: having no words of category c, and Number of sets of words that do not have morpheme words N = e + f + g + h
An apparatus for calculating an average mutual information amount by

(4) A method in which a device classifies content,
Morphological analysis of training data including a description related to a residential area of the content;
Extracting a predetermined morpheme from the result of the morpheme analysis;
Calculating an average mutual information amount between the extracted morpheme and a residence category;
Storing the data in which the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme are related;
Classifying the input predetermined content into the residence category based on the data stored in the storing step;
A method comprising the steps of:

  In the method (4), morphological analysis is performed on training data including descriptions related to the residential area in the content, a predetermined morpheme is extracted from the result of the morphological analysis, and the average mutual between the extracted morpheme and the residential category is extracted. Calculate the amount of information, store the data related to the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme, and input based on the data stored in the storage means The predetermined content is classified into the residence category.

  Therefore, it is possible to classify the input predetermined content for each residence category based on the average mutual information stored in association with the training data. Therefore, for example, a web page such as a blog can be classified by region from the content of the content of the blog.

MI(w,c)：形態素となる単語wとカテゴリｃの間の平均相互情報量
により、平均相互情報量を算出することを特徴とする方法。 (5) A method for classifying the content described in (4),
In the step of calculating the average mutual information amount, P is a probability,

MI (w, c): A method of calculating an average mutual information amount from an average mutual information amount between a word w as a morpheme and a category c.

MI(w,c)：形態素となる単語wとカテゴリｃの間の平均相互情報量
ｅ：カテゴリｃの語を有し、かつ、形態素の単語を有するものの集合の数
ｆ：カテゴリｃの語を有し、かつ、形態素の単語を有しないものの集合の数
ｇ：カテゴリｃの語を有さず、かつ、形態素の単語を有するものの集合の数
ｈ：カテゴリｃの語を有さず、かつ、形態素の単語を有しないものの集合の数
Ｎ＝ｅ＋ｆ＋ｇ＋ｈ
により、平均相互情報量を算出することを特徴とする方法。 (6) A method for classifying content described in (4) or (5),
In the step of calculating the average mutual information amount,

MI (w, c): Average mutual information between morpheme word w and category c e: Number of sets having category c words and morpheme words f: Category c words And the number of sets of those having no morpheme words g: the number of sets having no words of category c and having the words of morphemes h: having no words of category c, and Number of sets of words that do not have morpheme words N = e + f + g + h
A method of calculating an average mutual information amount by:

(7) For devices that classify content,
Morphological analysis of training data including a description related to a residential area of the content;
Extracting a predetermined morpheme from the result of the morpheme analysis;
Calculating an average mutual information amount between the extracted morpheme and a residence category;
Storing the data relating the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme;
Classifying the input predetermined content into the residence category based on the data stored in the storing step;
A program for running

  The program of (7) performs morphological analysis on training data including descriptions related to the residential area in the content, extracts a predetermined morpheme from the result of the morphological analysis, and calculates the average mutual between the extracted morpheme and the residential category. Calculate the amount of information, store the data related to the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme, and input based on the data stored in the storage means The predetermined content is classified into the residence category.

  Therefore, it is possible to classify the input predetermined content for each residence category based on the average mutual information stored in association with the training data. Therefore, for example, it is possible to infer the residence area of the creator of the content by classifying Web pages such as a blog by region from the content of the content of the blog.

(8) A device for classifying Web pages related to blogs,
A means for morphological analysis of training data including a description related to a living area in the web page;
Means for extracting a predetermined morpheme from the result of the morpheme analysis;
Means for calculating an average mutual information amount between the extracted morpheme and the residence category;
Means for storing data in which the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme are related;
Means for classifying the inputted predetermined web page into the residence category based on the data stored in the means for storing;
A device comprising:

  The apparatus of (8) performs morphological analysis on training data including descriptions related to the residential area in the Web page, extracts a predetermined morpheme from the result of the morphological analysis, and averages between the extracted morpheme and the residential category Based on the data stored in the means for calculating the mutual information, storing the relationship between the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme, The inputted predetermined web page is classified into the residence category.

  Therefore, it is possible to classify the input predetermined web pages for each residence category based on the average mutual information stored in association with the training data. Therefore, for example, by classifying a Web page such as a blog by region from the content of the content of the blog, it is possible to infer the residence area of the creator of the content.

  According to the present invention, it is possible to classify input predetermined content for each residence category based on the average mutual information stored in relation to the training data. Therefore, for example, a web page such as a blog can be classified by region from the content of the content of the blog.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram showing functional blocks and a processing flow of an apparatus 10 according to a preferred embodiment of the present invention. The apparatus 10 includes at least a control unit 101 that controls data and a data storage unit 107 that stores data. The apparatus 10 includes a communication unit and an input unit, which will be described later, and data related to content such as a web page 40 and training data 30 is input.

  As the learning unit 105, the control unit 101 performs a morpheme analysis unit 110 that performs morpheme analysis of the input training data 30, a morpheme extraction unit 120 that extracts a morpheme from the result of the morpheme analysis, and the extracted morpheme and residence category. An average mutual information amount calculation unit 130 that calculates an average mutual information amount between them, and a classification unit 140 that classifies the input Web page 40 into a residence category based on the stored data.

  The training data 30 and the web page 40 may be content and data related to a blog, and are pages and data that include at least one information related to a residential area such as a region. In particular, the training data may be training data for causing the device 10 to store data for classifying the living area in the web page.

  Next, processing executed by these configurations will be described. The input training data 30 includes text including character data relating to a residential area such as a region as described above. The morphological analysis unit 110 performs morphological analysis on the sentence, divides the sentence (character data composed of a plurality of words) for each word, and classifies the divided words for each part of speech (step S01).

  The morphological analysis unit 110 performs stop word processing (processing for removing web pages including words such as tourism, travel, stay, etc.) as preprocessing for performing morphological analysis, and only carefully selected web pages are used as training data. It may be processed.

  Next, the morpheme extraction unit 120 filters the words analyzed by the morpheme analysis unit 110 using the part of speech (step S02). Specifically, the morpheme extraction unit 120 extracts only nouns.

  Next, the average mutual information amount calculation unit 130 calculates an average mutual information amount between the morpheme extracted by the morpheme extraction unit 120 and the residence category (step S03).

  The residence category is a word typically used to indicate a residence area, and may be, for example, a name such as a prefecture (for example, Toyama, Kanagawa, etc.).

MI(w,c)：形態素となる単語wと居住カテゴリｃの間の平均相互情報量
Ｐを確率として、Ｐ（Ｘ，Ｙ）は、「Ｘ」と「Ｙ」とが共起する確率であり、Ｐ（Ｘ）、Ｐ（Ｙ）は、それぞれ個別に生起する確率である。 For example, the following formula is used to calculate the average information amount.

MI (w, c): P (X, Y) is the probability of co-occurrence of “X” and “Y”, with the average mutual information P between the morpheme word w and the residence category c as the probability. Yes, P (X) and P (Y) are probabilities of occurrence individually.

  As a characteristic of the average mutual information MI, the value of the word (w) appearing on the Web page with high frequency is biased toward a specific residence category (c).

MI(w,c)：形態素となる単語wと居住カテゴリｃの間の平均相互情報量
ｅ：居住カテゴリｃの語を有し、かつ、形態素の単語を有するＷｅｂページの数
ｆ：居住カテゴリｃの語を有し、かつ、形態素の単語を有しないＷｅｂページの数
ｇ：居住カテゴリｃの語を有さず、かつ、形態素の単語を有するＷｅｂページの数
ｈ：居住カテゴリｃの語を有さず、かつ、形態素の単語を有しないＷｅｂページの数
これらの関係を表にしたものが、表１である。

Ｎ＝ｅ＋ｆ＋ｇ＋ｈと表すことが可能である。 This equation is, for example,

MI (w, c): Average mutual information between morpheme word w and residence category c e: Number of Web pages having residence category c and morpheme words f: Residence category c The number of Web pages that have the same words and no morpheme words g: The number of Web pages that do not have residence category c words and have morpheme words h: Have residence category c words The number of Web pages that do not have morpheme words. Table 1 shows these relationships.

N = e + f + g + h can be expressed.

これらの値を式２に代入すると、

と、算出される。 Using Table 2, for example, when the residence category is “Toyama” and the word “Yao” appears 100 times on the Web page, the MI is calculated. Here, e is the number of Web pages in which “Yao” appears in the “Toyama” residence category Web page, and f is the number of Web pages in the “Toyama” residence category and “Yao” does not appear. The number of Web pages, g is the number of Web pages that are not in the residence category of “Toyama” and “Yao” appears, and h is the number of Web pages that are not in the residence category of “Toyama” and “Yao” Is the number of Web pages that do not appear.

Substituting these values into Equation 2,

And calculated.

これらの値を式２に代入すると、

と、算出される。 As another example, using Table 3, when the residence category is “Toyama” and the word “Nuchu Town” appears 20 times on the Web page, the MI is calculated.

Substituting these values into Equation 2,

And calculated.

In this way, whether the residence category of “Toyama” is related to a predetermined word can be determined by comparing the values of MI.

  The control unit 101 calculates the average mutual information amount calculated by the average mutual information amount calculation unit 130 with a plurality of words for one residence category, and as shown in Equation 5, the residence category (Toyama) On the other hand, the average mutual information amount of each word (Yao, Nuchucho, Toyama City, etc.) is stored in the data storage unit 107 in a table or the like.

  Next, the classification unit 140 that the predetermined Web page 40 classifies into any residence category will be described. The classification unit 140 receives input of the web page 40 such as a blog (step S04), refers to the data storage unit 107 (steps S05 and S06), and outputs the residence category of the web page 40.

  For example, the classification unit 140 performs morphological analysis on the character data described in the Web page 40, divides sentences (character data composed of a plurality of words) into words, and classifies the divided words into parts of speech. Then, only nouns (for example, Owara) are extracted. Then, the classification unit 140 determines whether there is a match with the extracted noun (wara) among the words stored in the data storage unit 107, and if there is a match, the classification unit 140 relates to the matched word. When the average mutual information amount (0.04) with the attached residence category (Toyama) is equal to or greater than a predetermined value (for example, 0.035 or more), the residence category is set to the Web The residence category (Toyama) on page 40 is assumed.

  As another aspect, the classification unit 140 extracts a plurality of nouns (Nakachucho, Owara) from one Web page, and determines whether there is one or more that matches the word stored in the data storage unit 107. If there is one or more matches, the residence category (Toyama) related to the matched word and the average mutual information amount of the word are compared for each word. Then, the classification unit 140 may determine the residence category of the Web page 40 by comprehensively comparing the average mutual information amount of all the matched words. This will be described later with reference to FIG.

[Hardware Configuration of Device 10]
FIG. 2 is a diagram illustrating an example of a hardware configuration of the apparatus 10 according to the preferred embodiment of the present invention described in FIG. The apparatus 10 includes a central processing unit (CPU) 1010 (a plurality of CPUs such as a CPU 1012 may be added in a multiprocessor configuration) constituting a control unit 101, a bus line 1005, a communication I / F 1040, a main memory 1050, a BIOS ( (Basic Input Output System) 1060, USB port 1090, I / O controller 1070, keyboard, mouse 1100, and other input means and a display device 1022 may be used.

  Storage means such as a tape drive 1072, a hard disk 1074, an optical disk drive 1076, and a semiconductor memory 1078 can be connected to the I / O controller 1070.

  The BIOS 1060 stores a boot program executed by the CPU 1010 when the apparatus 10 is started up, a program depending on the hardware of the apparatus 10, and the like.

  The hard disk 1074 constituting the storage unit 107 stores various programs for the device 10 to function as a server and programs for executing the functions of the present invention, and can configure various databases as necessary.

  As the optical disc drive 1076, for example, a DVD-ROM drive, a CD-ROM drive, a DVD-RAM drive, or a CD-RAM drive can be used. In this case, the optical disk 1077 corresponding to each drive is used. A program or data may be read from the optical disk 1077 by the optical disk drive 1076 and provided to the main memory 1050 or the hard disk 1074 via the I / O controller 1070. Similarly, the tape medium 1071 corresponding to the tape drive 1072 can be used mainly for backup.

  The program provided to the apparatus 10 is provided by being stored in a recording medium such as the hard disk 1074, the optical disk 1077, or a memory card. This program may be installed in the apparatus 10 and executed by being read from the recording medium via the I / O controller 1070 or downloaded via the communication I / F 1040.

  The aforementioned program may be stored in an internal or external storage medium. Here, in addition to the hard disk 1074, the optical disk 1077, or the memory card, a magneto-optical recording medium such as an MD or a tape medium can be used as a storage medium constituting the data storage unit 107. Further, a storage device such as a hard disk 1074 or an optical disk library provided in a server system connected to a dedicated communication line or the Internet may be used as a recording medium, and the program may be provided to the apparatus 10 via the communication line.

  Here, the display device 1022 displays a screen for accepting data input to the user or displays a screen of a calculation processing result by the device 10, and is a cathode ray tube display device (CRT), a liquid crystal display device (LCD). ) And the like.

  Here, the input means accepts input by the user, and may be configured by a keyboard, a mouse 1100, and the like.

  The communication I / F 1040 is a network adapter that enables the apparatus 10 to be connected to a terminal via a dedicated network or a public network. The communication I / F 1040 may include a modem, a cable modem, and an Ethernet (registered trademark) adapter.

  In the above example, the apparatus 10 has been mainly described. However, the functions described above can be realized by installing a program in the apparatus and operating the apparatus as a server apparatus. Therefore, the functions realized by the server described as an embodiment in the present invention can be realized by executing the above-described method by the apparatus, or by introducing the above-described program into the apparatus and executing it. It is.

  FIG. 3 shows another example in which the apparatus 10 receives the training data 30 and stores data in which the residence category, the extracted morpheme, and the average mutual information amount are associated with each other in the data storage unit 107. It is a conceptual diagram which shows an embodiment.

  First, a URL (Uniform Resource Locator) list 210 of Web pages related to a predetermined residence category (for example, “Toyama”) is input from a predetermined device to the crawler 220 that periodically visits the Web server. (Step S10). Here, the process of extracting the residence category and the Web page related thereto may be performed by the user.

  Based on the URL list of the Web page related to this residence category (“Toyama”), the crawler 220 makes a request for the URL to the Web server 230 on the Internet (Step S11), and the contents of the Web server 230 240a is acquired (step S12). The crawler 220 that has acquired the content 240a stores the content 240a in the case database with residence category 250 (step S13). At this time, the content 240a may be processed (data conversion) into the content 240b. That is, only a predetermined word that is frequently used may be extracted from the Web page that is the content 240a to be the content 240b.

  For example, in the residence category “Toyama”, when one collected web page has five words “Nakachu-cho”, three words “Toyama City”, and three words “Yao” As shown in FIG. 3, these are stored in the case database with residence category 250 as data related to the residence category.

  By storing a large number of such data, conversely, in the database 250 with a residence category, the number of Web pages that do not include “Nchuchu-cho” in the residence category “Toyama” and the category “Toyama”. Without belonging, the number of Web pages including “Nakachu Town” is stored together. For this reason, for example, in the residence category “Toyama”, it is possible to prepare all data for calculating the average mutual information amount of the word “Nuchucho” and the like.

  When data with a residence category is input to the device 10 from the case database with residence category 250 (step S14), the learning unit 105 of the device 10 performs processing, and the data storage unit 107 stores the residence category, Data in which the extracted morpheme is associated with the average mutual information is stored (step S15).

  Here, as in the case database with residence category 250, the apparatus 10 has already extracted a predetermined word from a Web page and stored data that can calculate the average mutual information amount, the morpheme analysis unit 110. And the processing of the morpheme extraction unit 120 may not be performed.

  Finally, as shown in FIG. 3, the data storage unit 107 stores the words in association with the average mutual information amount.

  Next, processing of the classification unit 140 of the device 10 will be described based on FIG. It is assumed that a case where the residence category is not determined among the Web pages is input to the device 10 (step S20). For example, in the case of a Web page in which “Movie” is described three times, “Cinema” is described once, and “Kawasaki” is described twice, it is desired to determine which residence category the Web page is classified into. In this case, the classification unit 140 of the device 10 calculates the average mutual information amount of each word based on the data stored in the data storage unit 107 (steps S21 and S22), compares them, Determine the category.

  For example, the classification unit 140 extracts a plurality of nouns (movies, Kawasaki) from the Web page, determines whether there is at least one word that matches the word stored in the data storage unit 107, and matches at least one If there is, the living category (Kanagawa, Yokohama, etc.) related to the matched word and the average mutual information amount of the word are compared for each word. For example, it is assumed that the average mutual information amount of the residence category “Kanagawa” and the word “movie” is 0.01, and the average mutual information amount of the residence category “Kanagawa” and the word “Kawasaki” is 0.05. When the average mutual information amount between “Yokohama” and the word “movie” is 0.001, and the average mutual information amount between the residence category “Yokohama” and the word “Kawasaki” is 0.01, as a result The classification unit 140 may determine the residence category “Kanagawa” having a large average mutual information amount.

  As mentioned above, although embodiment of this invention was described, it only showed the specific example and does not specifically limit this invention. Further, the effects described in the embodiments of the present invention only list the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to the effects described in the embodiments of the present invention.

It is a figure which shows the functional block of the apparatus 10 which concerns on suitable embodiment of this invention, and the flow of a process. It is a figure which shows an example of the hardware constitutions of the apparatus 10 which concerns on suitable embodiment of this invention. It is a figure explaining the process of the apparatus 10 which concerns on an example of suitable other embodiment of this invention. It is a figure explaining the process of the apparatus 10 which concerns on an example of suitable other embodiment of this invention.

Explanation of symbols

10 apparatus 30 training data 40 web page 101 control unit 105 learning unit 107 data storage unit 110 morpheme analysis unit 120 morpheme extraction unit 130 average mutual information amount calculation unit 140 classification unit 210 URL list 220 crawler 240a, b content 250 case with residence category Database 1005 Bus line 1010, 1012 CPU
1022 Display device 1050 Main memory 1070 Controller 1071 Tape medium 1072 Tape drive 1074 Hard disk 1076 Optical disk drive 1077 Optical disk 1078 Semiconductor memory 1090 USB port 1100 Keyboard and mouse

Claims (8)

A device that infers the residence area of the content creator,
Among the contents, means for morphological analysis of training data including a description related to a residential area;
Means for extracting a predetermined morpheme from the result of the morpheme analysis;
Means for calculating an average mutual information amount between the extracted morpheme and the residence category;
Means for storing data in which the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme are related;
Means for classifying the inputted predetermined content into the residence category based on the data stored in the means for storing;
A device comprising: An apparatus for classifying content according to claim 1,
The means for calculating the average mutual information amount has P as a probability,

MI (w, c): An apparatus that calculates an average mutual information amount from an average mutual information amount between a word w as a morpheme and a category c. An apparatus for classifying the content according to claim 1 or 2,
The means for calculating the average mutual information amount is:

MI (w, c): Average mutual information between morpheme word w and category c e: Number of sets having category c words and morpheme words f: Category c words And the number of sets of those having no morpheme words g: the number of sets having no words of category c and having the words of morphemes h: having no words of category c, and Number of sets of words that do not have morpheme words N = e + f + g + h
An apparatus for calculating an average mutual information amount by A device is a method of classifying content,
Morphological analysis of training data including a description related to a residential area of the content;
Extracting a predetermined morpheme from the result of the morpheme analysis;
Calculating an average mutual information amount between the extracted morpheme and a residence category;
Storing the data relating the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme;
Classifying the input predetermined content into the residence category based on the data stored in the storing step;
A method comprising the steps of: The method of classifying content according to claim 4, wherein P is a probability,
In the step of calculating the average mutual information amount,

A method of calculating an average mutual information amount by: A method for classifying content according to claim 4 or claim 5, comprising:
In the step of calculating the average mutual information amount,

MI (w, c): Average mutual information between morpheme word w and category c e: Number of sets having category c words and morpheme words f: Category c words And the number of sets of those having no morpheme words g: the number of sets having no words of category c and having the words of morphemes h: having no words of category c, and Number of sets of words that do not have morpheme words N = e + f + g + h
A method of calculating an average mutual information amount by: For devices that classify content,
Morphological analysis of training data including a description related to a residential area of the content;
Extracting a predetermined morpheme from the result of the morpheme analysis;
Calculating an average mutual information amount between the extracted morpheme and a residence category;
Storing the data relating the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme;
Classifying the input predetermined content into the residence category based on the data stored in the storing step;
A program for running An apparatus for classifying web pages related to a blog,
A means for morphological analysis of training data including a description related to a living area in the web page;
Means for extracting a predetermined morpheme from the result of the morpheme analysis;
Means for calculating an average mutual information amount between the extracted morpheme and the residence category;
Means for storing data in which the residence category, the extracted morpheme, and the average mutual information amount of the residence category and the morpheme are related;
Means for classifying the inputted predetermined web page into the residence category based on the data stored in the means for storing;
A device comprising:

Images