KR101324459B1 - System and method for e-mail categorization - Google Patents

Abstract

Disclosed are an email category classification system and method. The category classification system generates a decision tree based on the morphological analysis unit for generating morphological analysis results by analyzing the morphemes of the contents included in the input e-mail, and based on the morphological analysis results for the e-mails corresponding to the categories previously classified in the e-mail system. Decision tree generation unit, a category selection unit for selecting one of the pre-categorized category based on the stemming analysis result and the decision tree for the newly received e-mail, the newly received e-mail in the e-mail contained in each of the pre-categorized categories And a category ranking calculator configured to calculate a weight for each category according to the number of mail addresses associated with the category and a category ranking determiner for determining a category ranking of the classified category based on the selected category and the weight for each category.

Laser, traffic, travel, mean shift

Description

Email category classification system and method {SYSTEM AND METHOD FOR E-MAIL CATEGORIZATION}

Embodiments of the present invention relate to an email categorization system and method.

Currently, most companies and other research institutes often e-mail communication between employees and report on their work in order to improve work efficiency. At this point, each user sends and receives dozens and hundreds of emails a day, which are needed later in business. Therefore, it is necessary to classify into a predefined category so that a large amount of emails can be easily viewed later without deleting a large amount of emails. However, with dozens and hundreds of emails sent and received each day, it's not easy to manage them by putting them in categories.

For example, existing email systems, such as Microsoft Outlook, provide a technique for managing email using email addresses or subjects, and categorizing emails into categories based on these rules. However, once you set this rule, your email will be classified only according to that rule, so that your email can be classified in a way you didn't intend. For example, when categorized by e-mail address, e-mail addressed to that address is unconditionally categorized into a category that you specify. In addition, if you categorize a specific word in the subject, the email containing the word is also categorized only into a specific category specified by the user. If the email address of a sender or a specific word is always classified into a specific category, emails that should be classified into other categories may be misclassified into a wrong category, thereby impairing user convenience.

In this specification, an efficient email category classification system and method are proposed.

A category classification system and method are proposed that can select categories for newly received emails using stemming and decision trees, and recommend categories to users based on mail addresses and selected categories of newly received emails.

A category classification system and method for selecting categories for newly received emails using stemming and decision trees and determining categories of newly received emails based on the newly received email addresses and selected categories are proposed. do.

A category classification system according to an embodiment of the present invention includes a morpheme analysis unit configured to generate a morpheme analysis result by analyzing a morpheme of contents included in an input email, and a morpheme analysis result for an e-mail corresponding to a category previously classified in an e-mail system. Decision tree generation unit for generating a decision tree based on the morphological analysis results for the newly received e-mail, a category selection unit for selecting one of the pre-categorized category based on the decision tree, Category ranking that calculates the category ranking of the classified categories based on the selected category and the category and weight based on the category weight calculation unit for calculating the weight for each category according to the number of included email addresses associated with the newly received email It includes a decision unit.

According to an aspect of the present invention, the category classification system may further include a category recommending unit for recommending at least one category for newly received e-mail based on the category ranking.

According to an aspect of the present invention, the category classification system may further include a category determination unit that determines a category of a newly received e-mail as a category corresponding to the highest ranking of the category ranking.

According to an aspect of the present invention, the morpheme analysis unit analyzes the morphemes of the contents included in the email, checks the words of the pre-set parts of the extracted words, and selects the main words according to the number of occurrences of each of the confirmed words. Can be generated.

According to an aspect of the present invention, the decision tree generation unit sets a word corresponding to a stemming analysis result of an email corresponding to a pre-categorized category as an attribute that is a non-terminal node capable of making a decision, and sets a pre-categorized category as a pseudo As a result of the determination, the decision tree can be generated by setting the class as a terminal node.

According to an aspect of the present invention, the category selector determines a class that is a terminal node by comparing a word corresponding to a stemming analysis result of a newly received email with an attribute that is a non-terminal node of a decision tree, and a category corresponding to the class. Can be selected.

According to an aspect of the present invention, the mail address may include a sender address and a reference mail address, and the weight calculator for each category may calculate weights for each category based on the number of sender addresses and reference mail addresses included in the category. Can be.

According to an aspect of the present invention, the mail address may include a sender address and a reference mail address, and the weight calculation unit for each category gives different weights to the number of each of the sender address and the reference mail address included in the category. In addition, the sum of the values given different weights may be calculated as the weight for each category.

According to an aspect of the present invention, the category ranking determiner determines the category ranking in the order of the weight for each category, the category ranking by adjusting the rank of the selected category.

According to an aspect of the present invention, the selected category may be adjusted to have the highest rank when the weight for each category is equal to or greater than a predetermined value.

According to an aspect of the present invention, the selected category may be additionally assigned a weight that is pre-selected to the weight for each category to adjust the ranking.

In the category classification method according to an embodiment of the present invention, a morpheme analysis result is generated by analyzing a morpheme of contents included in an e-mail inputted by a morpheme analysis unit, and a decision tree generation unit corresponds to a category previously classified in an e-mail system. Generating a decision tree based on the stemming result of the e-mail, the category selection unit selecting one of pre-categorized categories based on the stemming analysis result and the decision tree of the newly received e-mail, and the weight for each category The calculating unit calculates the weight for each category according to the number of emails associated with the newly received email in the email included in each of the pre-categorized categories, and the category ranking determiner is classified based on the selected category and the weight for each category. Determine the category ranking of It includes the steps:

The stemming and decision tree may be used to select a category for the newly received email and to recommend the category to the user based on the mail address and the selected category of the newly received email.

The stemming and decision tree may be used to select a category for the newly received email and to automatically determine the category of the newly received email based on the mail address and the selected category of the newly received email.

By further using weights based on mail addresses as well as decision trees based on the contents of emails, categories for newly entered emails can be flexibly recommended to the user.

By determining the ranking of the categories and recommending a predetermined number of categories to the user, it is possible to flexibly determine the category for the newly input email.

Hereinafter, various embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

1 is a block diagram illustrating an internal configuration of a category classification system according to an embodiment of the present invention. As shown in FIG. 1, the category classification system 100 according to the present embodiment includes a morpheme analysis unit 110, a decision tree determination unit 120, a category selection unit 130, and a weight calculation unit 140 for each category. , The category ranking determiner 150 and the category recommender 160 may be included.

The morpheme analysis unit 110 analyzes the morphemes of the contents included in the input email to generate a morpheme analysis result. In this case, one of the conventionally known morphological analysis methods or morphological analyzers may be selectively used to analyze the morpheme with respect to the contents of the email, and the morphological analysis unit 110 uses the analyzed morpheme to determine the morpheme analysis result. Can be generated. Here, as an example, the morpheme analysis unit 110 checks a word of a pre-set part of the words extracted by analyzing the morpheme of the contents included in the email, and selects a main word according to the number of occurrences of each identified word to analyze the morpheme. Can produce a result.

The decision tree determiner 120 generates a decision tree based on the result of the morpheme analysis of the e-mail corresponding to the category previously classified in the e-mail system. Here, the decision tree generation unit 120 sets a word corresponding to a stemming analysis result of an email corresponding to a pre-categorized category as an attribute that is a non-terminal node capable of making a decision, and sets the pre-categorized category as a decision. As a result, the decision tree can be generated by setting the class as a terminal node. In this case, the generated decision tree may be stored in the form of a file. That is, instead of generating a decision tree every time the category classification system 100 is driven, the decision tree is stored in the form of a file, and when the category classification system 100 is driven, the pseudo tree is stored in the file. Can invoke and use decision tree.

The category selector 130 selects one of the pre-categorized categories based on the stemming analysis result and the decision tree for the newly received e-mail. At this time, the category selector 130 determines a class that is a terminal node by comparing a word corresponding to a result of stemming analysis of a newly received email with an attribute that is a non-terminal node of a decision tree, and a category corresponding to the class. Can be selected. For example, there are branches in the decision tree that link to attribute 'report', attribute 'draft' and class 'draft review', and 'report' and 'draft' among the key words extracted from newly received email. If present, you can select the category of the newly received email as 'Draft Review'.

The category weight calculation unit 140 calculates the weight for each category according to the number of email addresses associated with the newly received email in the email included in each of the previously classified categories. That is, in the present embodiment, the weights for each category calculated using the mail address as well as the category determined based on the decision tree may be further used to determine the category of the newly received e-mail.

In this case, the mail address may include a sender address and a reference mail address, and the weight calculator 140 for each category may calculate a weight for each category based on the number of sender addresses and reference mail addresses included in the category. . For example, if the sender address and reference email address of a newly received email A appear five times and six times in category B and three and two times in category C, respectively, the weighting of category B in category B is 11 A value of (5 + 6) and a category-specific weight of category C may each have a value of 5 (3 + 2).

Also, differently, the weight calculator 140 for each category assigns different weights to the number of sender addresses and reference email addresses included in the category, and weights the sum of the weighted values to different categories. It can also be calculated as For example, a weight of 1 may be assigned to the number of appearances of the sender address and a weight of 0.5 may be assigned to the number of appearances of the reference mail address. In this case, in the above-described example, the category-weighted category B has a value of 8 (5 X 1 + 6 X 0.5), and the category-weighted category C has a value of 4 (3 X 1 + 2 X 0.5). Can be.

The category ranking determiner 150 determines the category ranking of the classified category based on the selected category and the weight for each category. For example, the category ranking determination unit 150 may determine the rank of the category in the order of the high weight for each category, and determine the category ranking by adjusting the rank of the selected category. In this case, when the weight of each category is greater than or equal to a predetermined value, the selected category may be adjusted to have the highest rank, or the selected category may be additionally assigned to the weight of the corresponding category and the ranking may be adjusted.

For example, the weight for each category is 10, 6, 3, 2, 1 for category A, B, C, D, and E, and category C is selected by category selector 130, and is pre-selected. Suppose the value is '1'. At this time, the category ranking is determined in the order of [A (10), B (6), C (3), D (2), E (1)] based on the weight of each category, and then the selected category C is ranked highest. It can be adjusted to have. In this case, the order of category ranking finally determined may appear as [C (3), A (10), B (6), D (2), E (1)].

Alternatively, the preselected weight '5' may be added to the selected category C. In this case, the order of category ranking finally determined may appear as [A (10), C (3 + 5), B (6), D (2), E (1)].

The category recommender 160 recommends at least one category for newly received emails based on the category ranking. That is, the category recommendation unit 160 may recommend a category for a newly received email to the user based on the category ranking. For example, the top '3' categories of the categories included in the category ranking may be recommended as categories for the new email.

At this time, according to another embodiment of the present invention, the category classification system 100, instead of the category recommending unit 160, a category determination unit (not shown) to determine the category of the newly received e-mail as a category corresponding to the highest ranking of the category ranking (not shown) May be included). In this case, the category determining unit may automatically classify the newly received email by determining a category corresponding to the highest ranking among the categories included in the category ranking as the category of the newly received email.

In addition, the category classification system 100 includes both the category recommendation unit 160 and the category determination unit, and allows the user to determine a desired mode among the category recommendation unit 160 and the category determination unit, and newly receives according to the determined mode. May automatically determine the category of a given email or recommend at least one category to the user.

2 is a flowchart illustrating a category classification method according to an embodiment of the present invention. The category classification method according to the present embodiment may be performed by the category classification system 100 described with reference to FIG. 1. In FIG. 2, a category classification method will be described by describing a process in which each step is performed by the category classification system 100.

In step 210, the category classification system 100 analyzes the morphemes of the contents included in the input e-mail to generate a morpheme analysis result. In this case, one of the conventionally known morphological analysis methods or morphological analyzers may be selectively used to analyze the morpheme with respect to the content of the email, and the category classification system 100 may use the analyzed morpheme to determine the morpheme analysis result. Can be generated. Here, as an example, the category classification system 100 analyzes the morphemes of the contents included in the email, checks the words of the pre-set parts of the extracted words, and selects the main words according to the number of occurrences of the identified words to analyze the morphemes. Can produce a result. Here, the input e-mail may include an e-mail corresponding to a category previously classified in the e-mail system and a newly received e-mail.

In step 220, the category classification system 100 generates a decision tree based on the morphological analysis result of the email corresponding to the category previously classified in the email system. Here, the category classification system 100 sets a word corresponding to a stemming result of an email corresponding to a pre-categorized category as an attribute that is a non-terminal node capable of making a decision, and sets the pre-categorized category as a result of the decision. In addition, the decision tree may be generated by setting the class as a terminal node. In this case, the generated decision tree may be stored in the form of a file. That is, instead of generating a decision tree every time the category classification system 100 is driven, the decision tree is stored in the form of a file, and when the category classification system 100 is driven, the pseudo tree is stored in the file. Can invoke and use decision tree. That is, this step 220 may be a preprocessing step for classifying a category of newly received email.

In operation 230, the category classification system 100 selects one of pre-classified categories based on the stemming analysis result and the decision tree for the newly received e-mail. At this time, the category classification system 100 compares the word corresponding to the stemming result of the newly received e-mail with an attribute that is a non-terminal node of the decision tree to determine a class that is a terminal node and select a category corresponding to the class. Can be. For example, there are branches in the decision tree that link to attribute 'report', attribute 'draft' and class 'draft review', and 'report' and 'draft' among the key words extracted from newly received email. If present, you can select the category of the newly received email as 'Draft Review'.

In operation 240, the category classification system 100 calculates a weight for each category according to the number of e-mail addresses associated with newly received e-mails included in e-mails included in each of the pre-categorized categories. That is, in the present embodiment, the weights for each category calculated using the mail address as well as the category determined based on the decision tree may be further used to determine the category of the newly received e-mail.

In this case, the mail address may include a sender address and a reference mail address, and the category classification system 100 may calculate a weight for each category based on the number of sender addresses and reference mail addresses included in the category. For example, if the sender address and the reference email address of a newly received email A appear five times and six times in category B, and three and two times in category C, respectively, the weighting of category B in category B is 11 A value of (5 + 6) and a category-specific weight of category C may each have a value of 5 (3 + 2).

Alternatively, the category classification system 100 may assign different weights to the number of sender addresses and reference email addresses included in the category, and calculate the sum of the weighted values as the weight for each category. have. For example, a weight of 1 may be assigned to the number of appearances of the sender address and a weight of 0.5 may be assigned to the number of appearances of the reference mail address. In this case, in the above-described example, the category-weighted category B has a value of 8 (5 X 1 + 6 X 0.5), and the category-weighted category C has a value of 4 (3 X 1 + 2 X 0.5). Can be.

In operation 250, the category classification system 100 determines a category ranking of the classified category based on the selected category and the weight for each category. For example, the category classification system 100 may determine the rank of the category in the order of the high weight for each category, and determine the category ranking by adjusting the rank of the selected category. In this case, when the weight of each category is greater than or equal to a predetermined value, the selected category may be adjusted to have the highest rank, or the selected category may be additionally assigned to the weight of the corresponding category and the ranking may be adjusted.

For example, for each of categories A, B, C, D, and E, the weight for each category is 10, 6, 3, 2, 1, category C is selected in step 230, and the predetermined value is' Assume 1 '. At this time, the category ranking is determined in the order of [A (10), B (6), C (3), D (2), E (1)] based on the weight of each category, and then the selected category C is ranked highest. It can be adjusted to have. In this case, the order of category ranking finally determined may appear as [C (3), A (10), B (6), D (2), E (1)].

Alternatively, the preselected weight '5' may be added to the selected category C. In this case, the order of category ranking finally determined may appear as [A (10), C (3 + 5), B (6), D (2), E (1)].

In step 260, the category classification system 100 recommends at least one category for newly received email based on the category ranking. That is, the category classification system 100 may recommend a category for the newly received email to the user based on the category ranking. For example, the top '3' categories of the categories included in the category ranking may be recommended as categories for the new email.

At this time, according to another embodiment of the present invention, instead of recommending a category in step 260, the category classification system 100 may determine the category of the newly received e-mail as a category corresponding to the highest ranking of the category ranking. In this case, the category classification system 100 may automatically classify the newly received email by determining a category corresponding to the highest ranking among the categories included in the category ranking as a category of newly received email.

In addition, the category classification system 100 allows a user to determine a mode for recommending a category and a mode for determining a category, and automatically determines or at least automatically determines a category of a newly received e-mail according to the mode determined by the user. I can recommend my category to users.

3 is an example for explaining the decision tree generation and the morpheme analysis according to an embodiment of the present invention. Decision tree generation screen 310 is an example of the appearance of generating a decision tree based on the main word stemming the email of the categories included in the email system. Here, the main word may be determined based on the number of repetitions of words whose parts of speech are nouns in morphemes, and the threshold number of repetitions for determining a noun word as a main word is a category classification system and a category according to the present embodiment. It may be determined by the user of the classification system, the email system or the user of the email system. 3 shows an example in which the user of the category classification system determines the threshold of the repetition number. That is, when a user inputs a threshold of repetition counts through the 'Threshold' field 320, the category classification system may determine a main word among words that are nouns according to the threshold of the input repetition count. Here, the main word field 330 is a field in which words having a repetition frequency of five or more times are displayed. That is, among the words that are nouns extracted from the emails of the categories included in the email system, words having a repetition number of five or more times may be displayed in the main word field 330. As such, when the main word is extracted, the category classification system may generate a decision tree that sets the main word as an attribute and the categories classified in advance in the e-mail system as classes.

4 illustrates an example of a category recommendation screen according to an embodiment of the present invention. The user screen 410 shows an example of an email management screen displayed to the user. In this case, the first list 420 may be displayed through the user event for the area where the second email is displayed, and the categories recommended through the category classification system may be displayed through the category recommendation list 430. In this case, when the user selects one of the recommended categories, the second email may be moved to the selected category. That is, the user can classify the category corresponding to each email more conveniently by checking the recommendation category. In this case, the category recommendation list 430 may include a predetermined number of categories among categories included in the category ranking.

Also, in another embodiment, when the user selects to automatically sort the newly received email, the category may be automatically classified after the second email is confirmed by the user on the user screen 410. In this case, the second e-mail may be automatically classified into a category ranked at the top of the category ranking.

FIG. 5 is a table illustrating a frequency of appearance of a sender address and a reference mail address and corresponding weights according to an embodiment of the present invention. The first table 510 shows an example of calculating a weight for each category based on the number of appearances of a sender address and a reference mail address included in one newly received email. That is, as described above, the weight for each category may be determined according to the number of times that the sender address and the reference mail address corresponding to the newly received email appear for each category. In this case, the weight for each category may be determined simply by the number including the sender address and the reference mail address, or may be determined by giving different weights to the sender address and the reference mail address, respectively. That is, referring to FIG. 5, when 'a' and 'b' are each 1, it may mean that the weight for each category is simply determined based on the number of mail addresses, and one of 'a' and 'b'. If not 1, it may mean that different weights are assigned.

FIG. 6 illustrates an example of determining a category ranking based on a weight for each category and whether a category is selected. Here, the second table 610 shows seven categories of the first to seventh categories, weights for each category for each category, and whether a category is selected through the decision tree. That is, in the second table 610, it can be seen that the fifth category is selected through the decision tree. The third table 620 shows an example of determining the ranking of the seven categories through the weight of each category and whether a category is selected. In this case, after the seven categories are sorted by the weight for each category, the ranking may be adjusted according to whether the category is selected. In the third table 620, the fifth category, which is the category selected through the decision tree, is adjusted to have the highest rank. At this time, even if the category selected through the decision tree, if the weight for each category is smaller than the predetermined value, the ranking may not be adjusted. That is, when the predetermined value is larger than '4', which is the weight for each category of the fifth category, the ranking of the fifth category is not adjusted and remains in the ranking 7 as it is.

FIG. 7 is another example of determining a category ranking based on a weight for each category and whether a category is selected. The fourth table 710 shows seven categories of the first to seventh categories, weights for each category for each category, and additional weights according to category selection for the category selected through the decision tree. This additional weight may be added to the category-specific weights of the corresponding categories as shown in the fifth table 720, and the respective categories may be sorted by the weight of the final category. In this case, as shown in the fifth table 720, the weight for each category of the fifth category may be 19 (4 + 15), and may be ranked after the fourth category having a weight for each category of 20.

As such, when using a category classification system or a category classification method according to the present embodiments, a category for a newly received email is selected by using morphological analysis and a decision tree, and the mail address and the selected category of the newly received email. May recommend a category to the user or automatically determine the category of the newly received email. In addition, by using weights based on e-mail addresses as well as decision trees based on e-mail contents, categories for newly input e-mails can be flexibly recommended to users, and rankings of categories are selected and selected among them. By recommending a number of categories to the user, it is possible to flexibly determine the categories for newly entered emails.

Embodiments according to the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. Computer-readable media may include, alone or in combination with the program instructions, file data, data structures, and the like. Program instructions recorded on the media may be those specially designed and constructed for the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of the computer-readable recording medium include magnetic media such as a hard disk, a floppy disk, and a magnetic tape; optical media such as CD-ROM and DVD; magnetic recording media such as a floppy disk; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. Hardware devices may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

In the present invention as described above has been described by the specific embodiments, such as specific components and limited embodiments and drawings, but this is provided to help a more general understanding of the present invention, the present invention is not limited to the, Various modifications and variations can be made by those skilled in the art to which the present invention pertains.

Accordingly, the spirit of the present invention should not be construed as being limited to the embodiments described, and all of the equivalents or equivalents of the claims, as well as the following claims, belong to the scope of the present invention .

1 is a block diagram illustrating an internal configuration of a category classification system according to an embodiment of the present invention.

2 is a flowchart illustrating a category classification method according to an embodiment of the present invention.

3 is an example for explaining the decision tree generation and the morpheme analysis according to an embodiment of the present invention.

4 illustrates an example of a category recommendation screen according to an embodiment of the present invention.

FIG. 5 is a table illustrating a frequency of appearance of a sender address and a reference mail address and corresponding weights according to an embodiment of the present invention.

FIG. 6 illustrates an example of determining a category ranking based on a weight for each category and whether a category is selected.

FIG. 7 is another example of determining a category ranking based on a weight for each category and whether a category is selected.

<Explanation of symbols for the main parts of the drawings>

100: categorization system

110: morphological analysis

120: decision tree generation unit

130: category selection unit

140: weight calculation unit for each category

150: category ranking determination unit

Claims (12)

A morpheme analysis unit configured to generate a morpheme analysis result by analyzing a morpheme of contents included in an input email; A decision tree generation unit generating a decision tree based on a result of stemming analysis of an email corresponding to a category classified in advance in the email system; A category selector which selects one of pre-categorized categories based on the stemming analysis result and the decision tree of the newly received e-mail; A category weight calculator configured to calculate weights for each category according to the number of email addresses associated with newly received emails included in the emails included in each of the pre-categorized categories; And A category ranking determination unit that determines a category ranking of categories classified based on the selected category and the weight for each category. Including, The category ranking determination unit, The category ranking is determined in order of the weight of each category being high, and the category ranking is determined by adjusting the ranking of the selected category. Category Classification System. The method of claim 1, Category recommending unit for recommending at least one category for the newly received email based on the category ranking Category classification system further comprising. The method of claim 1, A category determination unit that determines a category of a newly received email as a category corresponding to the highest ranking of the category ranking Category classification system further comprising. The method of claim 1, The morpheme analysis unit, A category classification system for generating a morphological analysis result by checking a word of a part of speech that is set in advance among words extracted by analyzing the morpheme of contents included in an email, and selecting a main word according to the number of occurrences of each identified word. The method of claim 1, The decision tree generation unit, The word corresponding to the stemming result of the e-mail corresponding to the pre-categorized category is set to an attribute which is a non-terminal node capable of making a decision, and the pre-categorized category is set to a class which is a terminal node as a result of the decision. Category classification system for generating decision trees. The method of claim 1, The category selection unit, And classifying the word corresponding to the stemming result of the newly received e-mail with an attribute that is a non-terminal node of the decision tree to determine a class that is a terminal node and selecting a category corresponding to the class. The method of claim 1, The mail address includes a sender address and a reference mail address. The weight calculation unit for each category, A category classification system that calculates a weight for each category based on the number of sender addresses and reference email addresses included in the category. The method of claim 1, The mail address includes a sender address and a reference mail address. The weight calculation unit for each category, A different weight is assigned to each of the number of the sender address and the reference mail address included in the category, and the sum of the different weighted values is calculated as the weight for each category. delete The method of claim 1, And the selected category is adjusted to have the highest rank when the weight for each category is equal to or greater than a predetermined value. The method of claim 1, And the selected category is further assigned with a predetermined weight to a corresponding weight for each category to adjust the ranking. In the method for recommending a category in a category classification system comprising a morpheme analysis unit, decision tree generation unit, category selection unit, weight category calculation unit for each category, category ranking determination unit and category recommendation unit, Generating a morpheme analysis result by analyzing a morpheme of contents included in the e-mail inputted by the morpheme analyzer; Generating, by the decision tree generation unit, a decision tree based on a stemming analysis result of an e-mail corresponding to a category classified in advance in the e-mail system; Selecting one of the categories classified based on the stemming analysis result and the decision tree of the newly received e-mail; Calculating, by the category weighting unit, weights for each category according to the number of e-mail addresses associated with newly received e-mails included in e-mails included in each of the pre-categorized categories; And Determining the category ranking of the classified category based on the selected category and the weight for each category by the category ranking determination unit Including, The category ranking determination unit may determine the category ranking of the category classified based on the selected category and the weight for each category, Determining a ranking of categories in order of increasing weight for each category, and adjusting a ranking of the selected category to determine category ranking Category classification method comprising a.

Images